HAZARD COMMUNICATION: 

AN UPDATED REVIEW OF THE SCIENCE OF HEALTH AND SAFETY COMMUNICATION

Contract No. GS-10F-0215P

Order No. DOLQ059622303

Task Order No. 15

Submitted To:

Directorate of Standards and Guidance

Occupational Safety and Health Administration

Washington, DC 20210

Submitted By:

Eastern Research Group, Inc.

110 Hartwell Avenue

Lexington, MA 02421

July 17, 2007

CONTENTS

Page

  TOC \o "1-5" \h \z \u    HYPERLINK \l "_Toc172373275"  1.	Executive
Summary	  PAGEREF _Toc172373275 \h  2  

  HYPERLINK \l "_Toc172373276"  2.	Introduction	  PAGEREF _Toc172373276
\h  2  

  HYPERLINK \l "_Toc172373277"  3.	Sources and Methods	  PAGEREF
_Toc172373277 \h  2  

  HYPERLINK \l "_Toc172373278"  3.1	Information sources	  PAGEREF
_Toc172373278 \h  2  

  HYPERLINK \l "_Toc172373279"  3.2	Search methods	  PAGEREF
_Toc172373279 \h  2  

  HYPERLINK \l "_Toc172373280"  4.	Overview of the GHS	  PAGEREF
_Toc172373280 \h  2  

  HYPERLINK \l "_Toc172373281"  5.	Safety Data Sheets	  PAGEREF
_Toc172373281 \h  2  

  HYPERLINK \l "_Toc172373282"  5.1	Background	  PAGEREF _Toc172373282
\h  2  

  HYPERLINK \l "_Toc172373283"  5.2	State of the science	  PAGEREF
_Toc172373283 \h  2  

  HYPERLINK \l "_Toc172373284"  5.2.1	Literature reviews	  PAGEREF
_Toc172373284 \h  2  

  HYPERLINK \l "_Toc172373285"  5.2.2	Utility	  PAGEREF _Toc172373285 \h
 2  

  HYPERLINK \l "_Toc172373286"  5.2.3	Accuracy and completeness	 
PAGEREF _Toc172373286 \h  2  

  HYPERLINK \l "_Toc172373287"  5.2.4	Comprehensibility	  PAGEREF
_Toc172373287 \h  2  

  HYPERLINK \l "_Toc172373288"  5.2.5	Structure and format	  PAGEREF
_Toc172373288 \h  2  

  HYPERLINK \l "_Toc172373289"  5.2.6	Research needs	  PAGEREF
_Toc172373289 \h  2  

  HYPERLINK \l "_Toc172373290"  6.	Labels and Warnings	  PAGEREF
_Toc172373290 \h  2  

  HYPERLINK \l "_Toc172373291"  6.1	Background	  PAGEREF _Toc172373291
\h  2  

  HYPERLINK \l "_Toc172373292"  6.2	State of the science	  PAGEREF
_Toc172373292 \h  2  

  HYPERLINK \l "_Toc172373293"  6.2.1	Literature reviews	  PAGEREF
_Toc172373293 \h  2  

  HYPERLINK \l "_Toc172373294"  6.2.2	Warning theory	  PAGEREF
_Toc172373294 \h  2  

  HYPERLINK \l "_Toc172373295"  6.2.3	Utility	  PAGEREF _Toc172373295 \h
 2  

  HYPERLINK \l "_Toc172373296"  6.2.4	Evaluation of current labels and
warnings	  PAGEREF _Toc172373296 \h  2  

  HYPERLINK \l "_Toc172373297"  6.2.5	Warning content	  PAGEREF
_Toc172373297 \h  2  

  HYPERLINK \l "_Toc172373298"  6.2.5.1	Signal words	  PAGEREF
_Toc172373298 \h  2  

  HYPERLINK \l "_Toc172373299"  6.2.5.2	Symbols	  PAGEREF _Toc172373299
\h  2  

  HYPERLINK \l "_Toc172373300"  6.2.5.3	Hazard statements	  PAGEREF
_Toc172373300 \h  2  

  HYPERLINK \l "_Toc172373301"  6.2.5.4	Text	  PAGEREF _Toc172373301 \h 
2  

  HYPERLINK \l "_Toc172373302"  6.2.5.4.1	Length	  PAGEREF _Toc172373302
\h  2  

  HYPERLINK \l "_Toc172373303"  6.2.5.4.2	Difficulty	  PAGEREF
_Toc172373303 \h  2  

  HYPERLINK \l "_Toc172373304"  6.2.5.4.3	Explicitness	  PAGEREF
_Toc172373304 \h  2  

  HYPERLINK \l "_Toc172373305"  6.2.5.4.4	Tone	  PAGEREF _Toc172373305
\h  2  

  HYPERLINK \l "_Toc172373306"  6.2.5.4.5	Bilingual concerns	  PAGEREF
_Toc172373306 \h  2  

  HYPERLINK \l "_Toc172373307"  6.2.5.5	Combinations of elements	 
PAGEREF _Toc172373307 \h  2  

  HYPERLINK \l "_Toc172373308"  6.2.6	Warning design	  PAGEREF
_Toc172373308 \h  2  

  HYPERLINK \l "_Toc172373309"  6.2.6.1	General principles	  PAGEREF
_Toc172373309 \h  2  

  HYPERLINK \l "_Toc172373310"  6.2.6.2	Color	  PAGEREF _Toc172373310 \h
 2  

  HYPERLINK \l "_Toc172373311"  6.2.6.3	Layout and formatting	  PAGEREF
_Toc172373311 \h  2  

  HYPERLINK \l "_Toc172373312"  6.2.6.4	Order	  PAGEREF _Toc172373312 \h
 2  

  HYPERLINK \l "_Toc172373313"  6.2.6.5	Location	  PAGEREF _Toc172373313
\h  2  

  HYPERLINK \l "_Toc172373314"  6.2.7	Receiver characteristics	  PAGEREF
_Toc172373314 \h  2  

  HYPERLINK \l "_Toc172373315"  6.2.7.1	Gender	  PAGEREF _Toc172373315
\h  2  

  HYPERLINK \l "_Toc172373316"  6.2.7.2	Age	  PAGEREF _Toc172373316 \h 
2  

  HYPERLINK \l "_Toc172373317"  6.2.7.3	Product familiarity	  PAGEREF
_Toc172373317 \h  2  

  HYPERLINK \l "_Toc172373318"  6.2.7.4	Cost of compliance	  PAGEREF
_Toc172373318 \h  2  

  HYPERLINK \l "_Toc172373319"  6.2.8	Related topics	  PAGEREF
_Toc172373319 \h  2  

  HYPERLINK \l "_Toc172373320"  6.2.9	Research needs	  PAGEREF
_Toc172373320 \h  2  

  HYPERLINK \l "_Toc172373321"  7.	Training	  PAGEREF _Toc172373321 \h 
2  

  HYPERLINK \l "_Toc172373322"  7.1	Background	  PAGEREF _Toc172373322
\h  2  

  HYPERLINK \l "_Toc172373323"  7.2	State of the science	  PAGEREF
_Toc172373323 \h  2  

  HYPERLINK \l "_Toc172373324"  7.2.1	Literature reviews	  PAGEREF
_Toc172373324 \h  2  

  HYPERLINK \l "_Toc172373325"  7.2.2	The need for training	  PAGEREF
_Toc172373325 \h  2  

  HYPERLINK \l "_Toc172373326"  7.2.3	Training intervention studies	 
PAGEREF _Toc172373326 \h  2  

  HYPERLINK \l "_Toc172373327"  7.2.4	Models and recommendations	 
PAGEREF _Toc172373327 \h  2  

  HYPERLINK \l "_Toc172373328"  7.2.5	Research needs	  PAGEREF
_Toc172373328 \h  2  

  HYPERLINK \l "_Toc172373329"  8.	Related Considerations	  PAGEREF
_Toc172373329 \h  2  

  HYPERLINK \l "_Toc172373330"  8.1	Legal considerations	  PAGEREF
_Toc172373330 \h  2  

  HYPERLINK \l "_Toc172373331"  8.2	Emerging approaches for hazard
communication and control	  PAGEREF _Toc172373331 \h  2  

  HYPERLINK \l "_Toc172373332"  9.	References	  PAGEREF _Toc172373332 \h
 2  

 

TABLES

  TOC \h \z \c "Table"    HYPERLINK \l "_Toc172083306"  Table 1. Keyword
search terms used for this report	  PAGEREF _Toc172083306 \h  2  

  HYPERLINK \l "_Toc172083307"  Table 2. Author search terms used for
this report	  PAGEREF _Toc172083307 \h  2  

 

1.	Executive Summary

In 1997, the United States Occupational Safety and Health Administration
(OSHA) commissioned a report reviewing the state of the science
regarding key elements of chemical hazard communication: labels and
warnings, safety data sheets (SDSs), and worker training (Sattler et
al., 1997). That report was based on a comprehensive literature search.

Several developments in hazard communication programs have taken place
since 1997. Chief among these developments is the emergence of the
Globally Harmonized System of Classification and Labeling of Chemicals
(GHS) (United Nations, 2005). In 2006, OSHA announced its intent to
modify the Hazard Communication Standard (HCS) (29 CFR 1910.1200) to
include the GHS’s provisions for labeling and SDSs (OSHA, 2006a).

In light of these proposed changes, OSHA commissioned the present report
as an update to the 1997 review. This report reflects a search of the
literature published since 1997. 

Overall, the current literature search produced over 100 relevant
journal articles, as well as books, dissertations, key consensus
standards and regulations, and general reviews. In some topic areas, the
literature included much new research—particularly studies related to
general principles of warnings. The recent literature on SDSs and
training was less extensive, but still offered some useful evidence to
confirm or expand upon the findings reported in 1997. While some of the
findings may not appear to be particularly profound, in many cases they
offer important empirical evidence to back up common assertions about
critical aspects of hazard communication. Key findings of this report
are as follows:

Overview of GHS

The GHS is not a regulation or standard, but a uniform system for
classifying and labeling chemicals with respect to their health,
physical, and environmental hazards. A standard GHS label includes a
pictogram that represents the hazard, a signal word, a hazard statement,
product and supplier identification, and supplemental information.

In addition to labels, the GHS also establishes the structure and
content for an SDS. The GHS does not describe how hazard communication
training should take place; however, it does note that worker training
is an integral part of hazard communication. The GHS includes provisions
for classifying mixtures, as well as several annexes to guide
implementation (United Nations, 2005). 

Safety Data Sheets

Findings on the utility of SDSs:

Workers in certain fields report that they find SDSs useful overall.

Case studies suggest that it can be difficult for people to find needed
information on an SDS in an emergency.

SDSs are used for a variety of purposes. Employers use SDSs to identify
less hazardous products for substitution, and SDSs are also used by both
sides in workers compensation claims.

Findings on the accuracy and completeness of SDSs:

Audits of SDSs frequently find missing or incorrect data on ingredients,
health effects, and other concerns.

Under current standards, SDSs do not provide sufficient information on
the explosion hazard of combustible dusts.

Suggestions for improving the quality of SDSs include more frequent
updates, better auditing, and increased requirements for SDS writers.

Findings on the comprehensibility of SDSs:

In most cases, surveys find that a sizable portion of the audience for
SDSs finds them difficult to read and understand.

Studies suggest that users absorb roughly two-thirds of the information
on an SDS.

The literature contains many ideas on how to make an SDS more
comprehensible without sacrificing comprehensiveness.

Findings related to the structure and format of SDSs:

Studies have found a few differences in how the various SDS formats in
use today address certain topic areas, but overall efficacy is similar. 

It may be helpful to consider ordering the SDS in a way that is most
intuitive to users.

SDS designers should consider layout, easy-to-read text, and effective
use of color (contrast, color coding, etc.).

Labels and Warnings

General findings on the utility of warnings:

Warnings influence attention and behavioral compliance.

Labels are particularly important because they are one of the first
places people are inclined to look when they want information on a
particular chemical.

General findings related to current labels and warnings:

Current labels on industrial chemicals vary in their effectiveness. One
frequent limitation is the lack of explicit instructions—for example,
what constitutes “adequate” ventilation?

Labels from consumer products and pharmaceuticals may offer lessons that
can be applied to workplace chemical labels. Pharmaceutical labels have
been researched extensively.

Findings on signal words:

Signal words are generally recognized as a key alerting element of
warning labels.

Signal words also convey a particular level of hazard.

Studies of English-speaking people around the world have found a
generally consistent hierarchy of signal words with respect to perceived
hazard. DEADLY, DANGER, and WARNING seem to connote different levels of
hazard, while the perceived difference between WARNING and CAUTION is
often insignificant.

Findings on symbols:

Symbols generally make warning labels more noticeable and easier to
comprehend. They are often the most easily recalled elements of labels.

Symbols can improve label comprehension among children, people with low
literacy, and those who do not understand the language.

Many GHS pictograms are widely recognized, but some others are more
obscure.

Symbols are most effective when there is a direct relationship between
the image and the meaning, requiring minimal inference.

Symbols tend to be most effective when paired with redundant or
reinforcing text, which is a form of dual coding. GHS labels do this.

Findings on hazard statements:

Users generally like to have information that explains why they need to
take precautions. A clear hazard statement can do this.

Hazard statements should be clear and should minimize the need for
inference.

Users appreciate hazard statements in which consequences are paired with
actions they can take to reduce or avoid the risk.

Findings on warning text:

Too little text may not provide enough information for safe product use,
while too much text can crowd out important messages and decrease
recall.

Some current labels are too difficult to understand clearly,
particularly among users with low education levels.

Labels that are overly complex can elicit negative reactions from both
low-educated and highly educated users.

Many studies show that explicit terminology can improve understanding of
warning labels, as well as overall perceptions.

Current labels use terms that are too subjective for some users—for
example, “copious,” “adequate,” “repeated,” and
“prolonged.”

Explicit warning information does not appear to make people less likely
to use a product, contrary to some manufacturers’ fears.

Definitive statements may improve believability and recall, while the
use of personal pronouns in instructions may improve compliance.

The most effective format for bilingual labels appears to be
side-by-side text with equal font size and flags (e.g., U.S. and Mexican
flags) to draw the reader’s attention at a glance.

Findings related to combinations of warning elements:

Using all of the key label elements together can generally improve
warning performance, compared with labels that only contain a subset of
these elements. However, in some situations, too much information may be
detrimental.

Among the various label elements, some studies suggest that pictorials
have the greatest influence on recall.

Findings related to general principles of warning design:

Many resources are available to advise people on basic principles of
warning design.

In general, labels are more effective when they include enhancements
such as color, borders, and symbols. 

Proper font and spacing are important, as is color contrast. Some common
recommendations include black text on white background, a 10-point
sans-serif font, and bold type for the most important items.

Findings on color in warnings:

Color can make warnings more noticeable and can help organize the
content (i.e., through color coding).

Red is generally perceived to reflect the greatest degree of hazard.
Yellow, orange, and black reflect a lesser degree of hazard, but studies
disagree on the relative ranking among these three colors. 

When using color, label designers need to consider other factors such as
color blindness and lighting conditions.

Findings on warning layout and formatting:

Logical headings convey information more effectively.

Font size and white space are important variables that affect
readability. 

Wide, colorful, and jagged borders are generally perceived to attract
attention better than thin or missing borders, but borders may not be
the most important factor in attracting attention overall.

Findings on the order of warning components:

Presenting a hazard and the corresponding precautions together may
improve understanding.

Warnings related to immediate, specific hazards are generally considered
most important. 

The most effective order for warnings may depend on audience
characteristics such as familiarity with the product.

Findings on warning location:

Warnings tend to be noticed more quickly when they are located in the
upper left portion of an instruction sheet.

Safety precautions are more likely to be followed when they are
incorporated into a product’s directions for use, although warnings
may be better noticed in general if they are included on posters or
signs, rather than on the product.

Cluttered locations make warnings less noticeable.

Findings related to gender effects on warning efficacy:

Research has found very few gender differences related to warning
comprehension or compliance.

Warning designers probably do not need to consider gender factors unless
they are creating labels for gender-specific products.

Findings related to age effects on warning efficacy:

In general, children and adults have similar perceptions of the relative
hazard conveyed by colors and signal words.

Most studies have found that younger adults comprehend warning text and
symbols better than older adults do.

Many cognitive functions decline with age. Understanding these
age-related declines in ability can help warning designers create labels
that are better understood by older adults.

Findings related to product familiarity effects on warning efficacy:

Familiarity with a product can make users more familiar with the proper
precautions, and therefore more likely to comply with warnings. In other
cases, however, familiarity might have the opposite effect.

Familiarity with a label might make users less likely to read it
carefully.

Familiarity with a particular type of container can affect hazard
perceptions.

Findings on cost of compliance:

Cost of compliance is widely recognized as a factor that affects
behavioral compliance with warning labels. Important considerations
include the amount of time, money, and effort required to comply with an
instruction.

Training

Findings related to the need for training:

Training will continue to be a need under the GHS, considering that even
some common label elements may not be widely understood.

Worker surveys confirm that training can lead to safer practices.

Findings from training intervention studies:

Overall, training has been shown to improve safety attitudes and safety
behavior in the workplace.

In studies on symbol comprehension, training generally leads to greater
understanding. A simple explanation of each symbol is just as effective
as a longer explanation.

Findings related to training models and recommendations:

In one case, a mandatory training program was more cost-effective and
reached a wider audience than a voluntary alternative. 

Training should consider the needs and characteristics of the audience.
Training that accounts for cultural and literacy differences can not
only reduce injuries, but also improve productivity and reduce turnover.

Recent literature includes a number of recommendations for more
effective training.

2.	Introduction

In 1997, the United States Occupational Safety and Health Administration
(OSHA) commissioned a report by Sattler, Lippy, and Jordan reviewing the
state of the science regarding key elements of chemical hazard
communication: labels and warnings, safety data sheets (SDSs), and
worker training (Sattler et al., 1997). That report included a
comprehensive literature search.

Several developments in hazard communication programs have taken place
since 1997. Chief among these developments is the emergence of the
Globally Harmonized System of Classification and Labeling of Chemicals
(GHS), a system created to standardize and harmonize the classification
and labeling of chemicals worldwide (United Nations, 2005). In 2006,
OSHA published an Advance Notice of Proposed Rulemaking concerning
adoption of the GHS by the Agency (OSHA, 2006a).

In light of these developments, OSHA commissioned the present report as
an update to the 1997 review. This report does not intend to replace the
previous review, but to supplement it by presenting findings from the
literature since 1997, including findings related to provisions of the
GHS.

As Sattler et al. (1997) note, “There is a strong consensus that a
fully harmonized hazard communication system must take into account the
scientific findings on comprehensibility, readability, and other human
factors regarding the use of labels, warning placards, and safety data
sheets.” In updating the Hazard Communication Standard (HCS) (29 CFR
1910.1200), it will be important to consider the latest findings from
case studies and laboratory studies, as well as “lessons learned”
from other hazard communication systems around the world.

To produce this report, an extensive search was conducted for literature
published from 1997 to early 2007. In addition, this search looked for
earlier literature that may not have been cited in the 1997 report, but
may still be relevant to the discussion. The section on “Sources and
Methods” describes how this literature search was conducted.

Following a brief overview of the GHS, findings from the literature are
presented in three areas: SDSs, warning labels, and hazard communication
training. In cases where the literature is extensive, findings are
organized into logical subtopics.

This report closes with a brief discussion of related issues, including
emerging approaches such as control banding. The “References”
section lists all documents that are cited in the report.

3.	Sources and Methods

3.1	Information sources

The field of occupational safety and health is broad and
multidisciplinary. Thus, gaining a complete understanding of the state
of the science requires a variety of approaches and tools. This report
was produced with help from an extensive literature search, existing
bibliographies, and personal communications.

Much of the information used in this report was obtained through a
multidimensional literature search. The following types of literature
were considered:

Articles from scholarly journals in a variety of fields, such as
psychology and human factors

Articles from popular media

Graduate dissertations and theses

Textbooks on hazard communication

Literature search techniques are described in detail in the section
below. Other information sources included OSHA’s current HCS, hazard
communication regulations from other countries, and U.S. and
international consensus standards. The search considered comments posted
to the docket in response to OSHA’s 2006 Advance Notice of Proposed
Rulemaking for HCS revisions, including comments from the American
Society of Safety Engineers (ASSE, 2004). In addition, two
bibliographies were informative:

Bibliography on Chemical Hazard Communication, published by the
International Labor Organization (ILO) (ILO, 1999)

Published Research on the Comprehensibility of Chemical Hazard Warning
Labels and Material Safety Data Sheets, published by the Society for
Chemical Hazard Communication (SCHC) (SCHC, 1999).

A thorough report would not have been possible without assistance from
experts in the field. For this review, the authors received guidance and
suggestions from current and former OSHA employees, as well as experts
from the SCHC.

3.2	Search methods

The scientific literature was searched using six comprehensive
databases:

ABI/INFORM (ProQuest, 2007)

COMPENDEX (Engineering Information, 2007)

International Pharmaceutical Abstracts (IPA) (Thomson Scientific, 2007)

NIOSHTIC-2 (NIOSH, 2007)

PsycINFO (APA, 2007)

PubMed (NLM, 2007)

These six databases were searched using a wide range of keywords, which
are shown in Table 1. Most of these terms were adopted from the earlier
literature search conducted by Sattler et al. (1997), with a few
additional terms identified through consultation with OSHA. Additional
searches were conducted using the names of authors cited most frequently
in the report by Sattler et al. (1997); these search terms are provided
in Table 2. Searches were not date-limited, except where indicated in
Table 2. Four of the databases produced a large number of citations
(ABI/INFORM, COMPENDEX, PsycINFO, and PubMed); results from the other
two databases were less extensive due to the more limited nature of
their coverage (IPA covers the pharmaceutical industry, while NIOSHTIC-2
is limited to documents produced or funded by the National Institute for
Occupational Safety and Health [NIOSH]).

To supplement and verify the six database searches, several of the
search terms in Table 1 and Table 2 were employed in a Google Scholar
web search (Google, 2007). This search found numerous results but most
already appeared in the database results, suggesting that the database
search was relatively comprehensive.

Once each database search was conducted, the resulting titles and
abstracts were screened to remove articles with no relation to hazard
communication. Remaining articles were evaluated on two criteria: topic
and relevance. The topic scale was used to group articles for review;
topics were non-exclusive (i.e., an article could be assigned to more
than one topic) and included SDSs, labels, training, general articles,
and various specialized areas such as bilingual concerns. The relevance
score consisted of a rating from 1 to 5, with 5 being the most relevant
to this report. Articles scoring a 5 for relevance were obtained in full
text (if possible) so they could be reviewed and cited in the report.
Because Sattler et al. (1997) represents a comprehensive search of the
literature up to 1997, only articles published in 1997 or later were
obtained for this report, with the exception of a few earlier articles
that appeared to be relevant yet were not cited by Sattler et al.
(1997).

Once the database search was complete, the search for relevant
literature continued by following available leads. In the process of
reviewing articles, textbooks, and consensus standards, it became
apparent that a few additional references could prove useful. These
references were obtained in full text (if possible) for review and
citation.

Overall, the literature search led to over 100 articles cited in this
report, as well as two textbooks (Wogalter et al., 1999; Wogalter,
2006). Although several other books related to hazard communication have
been published in recent years, these particular books were selected for
further review because they offer an extremely comprehensive overview
from some of the most prominent researchers in the field.

Table   SEQ Table \* ARABIC  1 . Keyword search terms used for this
report

Keyword search

COMPENDEX	ABI/INFORM	IPA	NIOSHTIC-2	PsycINFO	PubMed

behavior AND communication AND workplace 	behavior AND communication AND
safety	behavior AND communication AND safety	behavior AND communication
AND safety	behavior AND communication AND safety	behavior AND
communication AND workplace

compliance AND safety AND workplace	chemical AND label	chemical AND
warning label	chemical AND warning	chemical AND warning label	compliance
AND safety AND workplace

compliance AND training AND workplace	compliance AND training AND
workplace	chemical AND warning	compliance AND training	compliance AND
training AND workplace	compliance AND training AND workplace

comprehensibility AND warning AND label	comprehensibility	compliance AND
training (AND workplace)	comprehensibility (AND warning)
comprehensibility AND warning	comprehensibility AND warning

control banding	comprehension AND label	comprehensibility (AND warning)
(AND label)	comprehension AND label	comprehension AND label	control
banding

cost of compliance	control banding	comprehension AND label	control
banding	control banding	cost of compliance AND warning

hazard communication	cost of compliance	control banding	cost AND
compliance	cost of compliance	hazard communication

human factor AND communication	hazard communication AND training	cost of
compliance	hazard AND communication AND training	hazard communication
AND training	human factor AND communication

(material) safety data sheet	human factor AND communication	hazard
communication AND training	human factor AND communication	human factor
AND communication	(material) safety data sheet

pictogram	(material) safety data sheet	human factor AND communication
(material) safety data sheet	(material) safety data sheet	pictogram

pictograph	pictogram	(material) safety data sheet	pictogram	pictogram
pictograph

risk management AND label	pictograph	pictogram	pictograph	pictograph
risk management AND label

safety AND training AND language	risk management AND chemical	pictograph
risk management AND chemical	risk management AND chemical	safety AND
training AND language

safety management AND workplace 	safety AND training AND language	risk
management AND chemical	risk management AND workplace	risk management
AND workplace	safety management AND workplace

signal word	safety management AND workplace	risk management AND
workplace	safety AND training AND language	safety AND training AND
language	signal word

warning AND comprehension 	signal word	safety AND training AND language
safety management AND workplace	safety management AND workplace	warning
AND comprehension

warning AND memory	warning AND comprehension	safety management AND
workplace	signal word	signal word	warning AND memory

warning label	warning AND memory	signal word	warning AND comprehension
warning AND comprehension	warning label

	warning label (1997 and later)	warning AND comprehension	warning AND
memory	warning AND memory



	warning AND memory	warning label	warning label



	warning label



	

Table   SEQ Table \* ARABIC  2 . Author search terms used for this
report

Author search

COMPENDEX	ABI/INFORM	IPA	NIOSHTIC-2	PsycINFO	PubMed

Braun, C (CC, C.C.)	Braun, C	Braun, C	Braun	Braun, C	Braun, C

Brelsford	Brelsford	Brelsford, J	Brelsford	Brelsford	Brelsford

Lehto AND warning	Lehto AND warning	Lehto AND warning	Lehto	Lehto AND
warning	Lehto AND warning

Silver AND warning	Silver AND warning	Silver AND warning 	Silver	Silver
AND warning	Silver AND warning

Wogalter (1997 and later)	Wogalter (1997 and later)	Wogalter	Wogalter
Wogalter (1997 and later)	Wogalter (1997 and later)



4.	Overview of the GHS

The Globally Harmonized System of Classification and Labeling of
Chemicals (GHS) is not a regulation or standard, but a system for
classifying and labeling chemicals. This system was developed in order
to harmonize chemical labeling and classification internationally. 

At the present time, many different hazard communication systems are in
use worldwide. While the existing hazard communication laws and
regulations are similar in general—for example, many countries require
that a hazardous chemical be accompanied by a safety data sheet
(SDS)—there are numerous differences in hazard definitions and label
and SDS requirements. As a result, some chemicals are classified
differently in different jurisdictions, and manufacturers have to
produce multiple versions of their labels and SDSs for international
trade. OSHA’s Guide to the Globally Harmonized System of
Classification and Labeling of Chemicals (OSHA, 2006b) provides several
practical examples of how hazard definitions vary across countries—and
even across different regulatory domains within the U.S.

Recognizing the complex nature of the present system for managing
hazardous chemicals, the 1992 United Nations Conference on Environment
and Development established an international mandate to create a
globally harmonized system for classification and labeling. Over the
ensuing years, the GHS was developed on a consensus basis with support
from major stakeholders such as national governments, industry, and
workers. This effort was coordinated by a group working under the
Inter-organizational Program for the Sound Management of Chemicals
(IOMC). (More details about the development of the GHS can be found in
OSHA’s guide [OSHA, 2006b].) The GHS was formally adopted by the
United Nations Committee of Experts on the Transport of Dangerous Goods
and the Globally Harmonized System of Classification and Labelling of
Chemicals in December 2002, and was endorsed by the The United Nations
Economic and Social Council in July 2003. The GHS is not binding;
rather, it represents a set of “building blocks” that each nation
can elect to implement within its own regulatory framework.

The GHS document, also referred to as “The Purple Book” (United
Nations, 2005), establishes a uniform system for classifying chemicals
with respect to their health, physical, and environmental hazards. For
each type of hazard, the GHS defines thresholds for different levels of
hazard severity, based on standard tests. For example, substances with
an oral LD50 (lethal dose 50) of 5 mg/kg or less are assigned to Acute
Toxicity Category 1; an LD50 between 5 and 50 mg/kg is Category 2; etc.

The GHS includes provisions for labeling chemicals according to their
hazards. A standard GHS label includes a pictogram that represents the
hazard, a signal word (either DANGER or WARNING), and a hazard
statement—a phrase describing the nature of the hazard. The
appropriate pictogram, signal word, and hazard statement are all
determined by the hazard classification. Other GHS label elements
include product identification, supplier identification, and
supplemental information such as precautionary pictograms and
statements.

In addition to labels, the GHS also establishes the structure and
content for an SDS. Under the GHS, an SDS should be organized under 16
specific headings, which are in a prescribed sequence. The GHS describes
information that should be included under each heading.

The GHS does not describe how hazard communication training should take
place; however, it does note that worker training is an integral part of
hazard communication, and it suggests that training requirements be
“appropriate for and commensurate with the nature of the work or
exposure” (United Nations, 2005, § 1.4.9). This view echoes what many
others have noted—that hazard communication needs to be a complete
process. Redundancy is an important part of this process, with signs and
labels ideally serving to reinforce knowledge already instilled through
training and exposure to the SDS (Rousseau and Wogalter, 2006).

In addition to standardized classifications, label elements, and SDS
elements, the GHS “Purple Book” contains several annexes with
guidance to help with implementation. The GHS also includes provisions
for classifying mixtures (United Nations, 2005). 

5.	Safety Data Sheets

5.1	Background

The HCS currently states that “Chemical manufacturers and importers
shall obtain or develop a material safety data sheet [MSDS] for each
hazardous chemical they produce or import. Employers shall have a
material safety data sheet in the workplace for each hazardous chemical
which they use.” The HCS includes a list of basic information that
must be included in the MSDS, but does not prescribe a specific format
(OSHA, 1994).

In the GHS and other hazard communication systems used around the world,
the MSDS is referred to as simply the Safety Data Sheet (SDS). For
consistency, this report will use the term “SDS.”

 

Under the GHS, the SDS is organized into a 16-section structure, with
specific information required in each section. This scheme is designed
to follow a logical progression:

 

What is the material and what do I need to know immediately in an
emergency? (Sections 1-3)

What should I do if a hazardous situation occurs? (Sections 4-5)

How can I prevent hazardous situations from occurring? (Sections 7-10)

Is there other useful information about this material? (Sections 11-16)

Several countries prescribe a structure similar to the GHS, as do many
of the key consensus standards. For example, Canada’s Workplace
Hazardous Materials Information System (WHMIS) prescribes a 9-section
SDS format, but will also accept the 16-section format as long as all of
the information required by Canada is included (WorkSafeBC, 2006).
European Commission Directive 2001/60/EC—an amendment to Directive
91/155/EEC—provides Europe’s most up-to-date guidelines for SDSs
(European Commission, 2001). These European guidelines also include a
16-section structure, which is identical to the GHS format except that
two headers are reversed. In Australia, the National Code of Practice
for the Preparation of Material Safety Data Sheets, 2nd Edition requires
the same 16 sections as the GHS. This document also lists information
resources for SDS preparers (NOSHC, 2003).

Reflecting the growing consensus around the GHS, ANSI Z400.1, the
American National Standard for Hazardous Industrial Chemicals—Material
Safety Data Sheets—Preparation, was updated in 2004 to include the
same 16 SDS sections as the GHS. Within each of these sections, ANSI’s
list of required data elements is similar, though not always identical,
to the GHS requirements (ANSI, 2004).

The International Organization for Standardization (ISO) has published
its own standard for SDS preparation. This standard, ISO 11014-1, is
currently being revised to bring it closer to the GHS (United Nations,
2002b). The draft standard includes the same 16 sections as the GHS, as
well as similar data requirements in each section.

Another related standard is ASTM E 1628-94, Standard Practice for
Preparing Material Safety Data Sheets to Include Transportation and
Disposal Data for the General Services Administration (ASTM, 2000). This
standard recommends using the ANSI Z400.1 SDS format as a baseline, then
supplementing the data required by HCS with additional information
related to transport. This standard illustrates one of the complexities
of the current system of SDSs—the fact that different user groups have
different needs, and hence they may have different SDS requirements.

Annex A of ANSI Z400.1 contains a helpful comparison of required SDS
sections and data elements across different systems, including
Canada’s WHMIS, the EU, Australia, Mexico, Japan, ANSI, the HCS, and
GHS (ANSI 2004). Another document from the United Nations provides a
comparison between the GHS SDS and the revised draft ISO 11014-1 SDS
standard (United Nations, 2002a).

Note that ANSI is in the process of publishing a new standard, the
American National Standard for Product Safety Information in Product
Manuals, Instructions, and Other Collateral Materials. This standard
explicitly excludes SDSs (Frantz and Hall, 2005; Hall et al., 2006). 

People from a range of disciplines have offered perspectives on the
present system of SDSs in the U.S., as well as prospects for future
improvements. Although some of these views reflect apparent
misunderstandings of current requirements, they consistently indicate
concerns about the adequacy of information currently supplied on SDSs.
Examples include the following published opinions—the vast majority of
which suggest a need for improvement:

Bernstein (2002) lists several limitations of current SDSs, including:

Omission of ingredients deemed non-hazardous or trade secrets;

Omission of potential respiratory and skin sensitizing agents, some of
which are not typically classified as toxic or irritant substances; and,

Failure to document clinical information about specific diseases that
are known to occur as a result of exposure.

Among other things, the author recommends more standard formatting and
wording, as well as specific criteria for determining the hazard status
of an ingredient. Bernstein also recommends adding respiratory tract and
cutaneous sensitization data requirements for the SDS, acknowledging
that some chemicals can induce occupational asthma at levels below the
permissible exposure limit (PEL), differentiating between irritants and
sensitizers (the latter have irreversible effects), and developing
alternative strategies to convey information about components that are
trade secrets but may have irritation/sensitization potential. 

Pullen (2003) complains that SDSs contain complex terminology, and users
currently have to work hard to decipher the format and find the
appropriate section. Workers can have trouble interpreting toxicological
findings or ascertaining that the SDS is for the correct concentration
or variation of the chemical—particularly in the chaos of an actual
emergency, or if the worker is not well educated. Because scanning and
faxing can reduce document quality, Pullen recommends using a web-based
SDS format that is searchable, easy to read, connected to a glossary,
and can be linked to laptops, wireless devices, and speech-enabling
software for people with reading disabilities.

Describing a deadly accident related to a resin that workers did not
know was explosive, Rooney (2005) calls on OSHA to improve SDSs to
prevent such incidents from occurring in the future. The author notes
that SDSs in the U.S. presently have no standard format, contain highly
technical language, and often end up being maintained in an “unwieldy
series of binders.” Without more oversight, Rooney suggests that
“Expecting MSDS currency and access to maximize workplace safety is
like expecting an infant to be able to swim because a lifeguard is on
duty.” The author encourages the use of standard hazard classes,
standardized phrases, minimum required testing for acute hazards, and
using the GHS as a “springboard” for a more accountable and reliable
framework for classifying and labeling chemicals.

Walker (1997) discusses a 1997 letter from the National Paint and
Coatings Association (NPCA) to the State Department’s Office of
Environmental Policy, stating that harmonizing the HCS, labels, and SDSs
is “sorely needed.” In its letter, the NPCA notes that importers
frequently have to redo Asian or European labels and SDSs to meet U.S.
regulations, while exporters often find it difficult to obtain other
countries’ laws and regulations. The author also cites the need for
domestic consistency at a time when some states, perceiving inadequacies
in the federal standard, have taken it upon themselves to enact more
stringent regulations.

Fagotto and Fung (2002) state that the ANSI standard SDS format is
helpful, but more standardization is still needed. The authors note that
technical vocabulary is not well suited to a broad range of user groups,
and different manufacturers can still end up with different hazard
evaluations or descriptions of effects. 

Logsdon (2004) notes that SDSs are difficult to prepare because hazard
information must address the worst-case scenario, and if the
manufacturer does not list every potential physical or health hazard,
the company could be found liable for failure to warn. Technical
information can be difficult for users to understand.

From the perspective of health care providers, Mitchell and Schwartz
(2001) describe three limitations to relying on SDSs: They may not have
all the information a health care provider needs, like chronic health
effects data; they cannot describe the actual magnitude of exposure; and
some physicians may not be well trained in occupational and
environmental medicine.

Beach (2002) relates experiences from the perspective of a doctor
treating occupationally exposed patients. The author notes that some
chemicals in mixtures may be below reporting thresholds, but still
present at levels that can affect certain susceptible individuals (e.g.,
the case of sensitizers). In addition, different suppliers use different
risk phrases for the same chemical, making it hard for users to compare
relative risks.

Packham (2002) lists several concerns about SDSs in the United Kingdom
(UK):

SDSs may have incorrect information.

SDSs cover the pure product, ignoring possible contaminants.

The 1 percent threshold for including ingredients represents 10,000
parts per million (ppm), while some people might be sensitive to just a
few hundred ppm of a sensitizer or allergen.

SDSs cover the chemical “as supplied,” but the chemical may react
during certain processes. For example: “Oxidization of d-limonene in
some modern degreasants can result in the presence of potent dermal
sensitizers.”

Bentley (2006) urges SDS writers to be accurate and honest, but not
provide too much information. The author, an expert from the polymer
industry, suggests a variation on the “golden rule:” provide
whatever information you would like to have if you had to clean up a
spill, treat someone who ingested the material, or handle any other
mishap. Bentley notes that the people using an SDS in emergency
situations are not necessarily trained chemists, toxicologists, or
medical personnel.

Testifying before the U.S. Senate Subcommittee on Employment, Safety,
and Training in 2004, several individuals indicated a need for updating
the present system:

Jon Hanson, Director of Safety at Wyoming Medical Center, discussed
several concerns with the present system of SDSs—particularly
inconsistency, inaccuracy, and complex language. Citing several case
studies in which poor hazard information caused injury or put workers at
risk, Hanson urged Congress to work with OSHA to revise the HCS to
create a simplified national framework for chemical hazard
determination, training, and document preparation. Hanson also suggested
that current SDSs are written defensively to protect against lawsuit,
resulting in language that is too technical for the average user
(Hanson, 2004).

Anne Jackson testified on behalf of the American Bakers Association. She
noted that companies in her industry routinely deal with SDSs that
appear to be written for either lawyers or chemical engineers, and are
therefore incomprehensible to most safety professionals and production
workers. Jackson also observed that many SDSs are inaccurate. Among
other things, Jackson recommended that OSHA establish a standard SDS
format that is accessible to the intended audience, and suggested that
OSHA consult that audience when developing its new standard (Jackson,
2004).

Thomas Grumbles, president of the American Industrial Hygiene
Association (AIHA), recommended updating the HCS to include such items
as “a non-mandatory appendix… that addresses training guidelines for
MSDS authors” (Grumbles, 2004).

Michael Wright of the United Steelworkers of America spoke about the
need for better SDS quality and a more effective HCS training
requirement. He urged the U.S. to adopt the GHS (Wright, 2004).

5.2	State of the science

5.2.1	Literature reviews

Sattler et al. (1997) provides a comprehensive review of the literature
on SDSs, including international use of SDSs; studies on the utility,
quality, and comprehensibility of SDSs; and recommendations for
improving this aspect of hazard communication. 

A search of the literature published since 1997 found no comprehensive
new reviews of research on SDSs. That is not to say that additional
research has not been conducted since 1997, however. People from a range
of disciplines have continued to study the effectiveness of SDSs; their
recent findings are summarized in the sections that follow.

One recent paper did include a literature search component, although
most of the studies reviewed were already cited in the review by Sattler
et al. (1997). Fung et al. (2004) reviewed the literature to determine
the effectiveness of several “transparency systems,” including
OSHA’s HCS. Overall, Fung et al. rate the HCS as moderately effective,
meaning that “The transparency system has changed behavior of a
substantial portion of users and disclosers in the intended direction
but has also left gaps in behavior change and/or generated unintended
consequences.” The authors note that improved access to information
has helped some employers switch to safer alternatives, but cite studies
showing that this information is not well embedded in employees’
decision-making, and that SDSs are of varying quality and are often too
complex. Fung et al. sum up their assessment as follows:

“Overall, the hazard communication system appears to function better
as a tool to exchange information among chemical producers and chemical
users than as a device to help employees to protect themselves at work,
avoid dangerous workplaces, or demand higher pay in light of increased
risk.”

5.2.2	Utility



Key Findings:



Workers in certain fields report that they find SDSs useful overall.



Case studies suggest that it can be difficult for people to find
needed information on an SDS in an emergency.



SDSs are used for a variety of purposes. Employers use SDSs to identify
less hazardous products for substitution, and SDSs are also used by both
employees and employers in workers compensation claims.





After reviewing the available literature on SDSs, Sattler et al. (1997)
concluded that “MSDSs, by themselves, are a poor means of informing
workers of hazards to which they may be exposed,” which the authors
attribute to the technical nature of SDS content, the generic nature of
the situations described, and information that is too difficult, brief,
or vague to allow meaningful understanding. As described in the sections
that follow, studies since 1997 have continued to raise concerns about
the quality and comprehensibility of SDSs currently in use.

Despite the many limitations to current SDSs, a few studies indicate
that the SDS can still serve a useful function in hazard
communication—particularly among workers who are generally familiar
with the format and content of this tool. For example, in an experiment
with engineering students, Lehto (1998a) provided labels and SDSs for
chemicals, then asked the subjects to find specific health and safety
information. When subjects could not find the information on the label,
they consulted the SDS 64 percent of the time. People with more
experience with workplace chemicals were more likely to use the SDS,
suggesting that overall the SDS plays a positive role in hazard
communication.

Conklin (2003) demonstrated the utility of SDSs among employees of a
multinational petrochemical company. Across three countries (the U.S.,
Canada, and the UK), 46 percent of workers said that SDSs are too long.
However, 98 percent still felt that the SDS is a satisfactory
information source (the percentage was similar across all three
countries), and 72 percent said they would request an SDS all or most of
the time when introduced to a new chemical. The author notes one
limitation, however, which is that this sample did not include any
workers with low literacy.

Case studies suggest more limitations to the utility of SDSs. In one
example, a hospital safety director describes a recent situation in
which an SDS did not serve its intended function:

“Now, fast forward to July of 2000 when two gallons of the chemical
Xylene spilled in the lab of my hospital. By the time an employee had
noticed the spill, the ventilation had already sucked most of the vapors
into the HVAC. This, in turn, became suspended in the ceiling tile over
our radiology department. Twelve employees were sent to the emergency
room. To make the matter worse, the lab employee was frantically
searching through the MSDS binder in her area for the Xylene MSDS. Once
she found it, she had difficulty locating the spill response
section. After notifying our engineering department, she began to clean
up the spill with solid waste rags, known for spontaneous combustion,
and placing the rags into a clear plastic bag for disposal. She did not
know that Xylene has a flash point of 75 degrees Fahrenheit. She then
walked the bag down to our incinerator room and left it there, basically
creating a live bomb. Twelve people were treated from this exposure. The
lab employee was very upset and concerned about the safety of the
affected employees and visitors, and hysterically kept stating that she
could not find the necessary spill response information.” (Hanson,
2004)

In testimony before the U.S. Senate Subcommittee on Employment, Safety,
and Training, this hospital safety director noted that SDSs can be
clumsy to use, and their thoroughness varies widely. SDSs at this
particular hospital range from one page to 65 pages in length. With
2,500 chemicals in use, the result is more than 20,000 pages of SDSs,
which have to be manually archived in 26 four-inch binders (Hanson,
2004).

SDSs are used for a variety of purposes. Fagotto and Fung (2002) suggest
that the primary users are actually employers, who have used SDS
information to substitute less hazardous chemicals in their processes.
Doing so not only helps to manage liability, but also helps the company
respond to market demands for transparency and “greener” products.
Others note that SDSs play a role in workers compensation claims. In a
review of claims for neurotoxic chemical exposures in Canada, Baldwin et
al. (2003) found that SDSs contributed evidence in one-third of the
cases studied. The authors note that an accurate SDS can work in favor
of either side, as it might establish a link between exposures and
effects, or it might reveal that the alleged link is not plausible.

5.2.3	Accuracy and completeness



Key Findings:



Audits of SDSs frequently find missing or incorrect data on ingredients,
health effects, and other concerns.



Under current standards, SDSs do not provide sufficient information on
the explosion hazard of combustible dusts.



Suggestions for improving the quality of SDSs include more frequent
updates, better auditing, and increased requirements for SDS writers.





In their assessment of the literature available at the time, Sattler et
al. (1997) cite a pair of studies showing problems with the quality of
SDSs in use in the U.S., including inaccuracies and omissions. Studies
conducted since 1997 continue to show deficiencies.

Frazier et al. (2001) evaluated the quality of 30 SDSs from different
manufacturers of products containing toluene diisocyanate (TDI)—a
chemical well known to cause or exacerbate asthma. Approximately 27
percent of the SDSs did not mention the word "asthma" or include
language pertaining to allergic or sensitizing respiratory reactions.
One manufacturer mentioned no respiratory effects at all. Other problems
involved the date of last review. Some SDSs listed the date when the
study authors requested the document, 10 percent were missing a date,
and 20 percent had a date more than 6 years old, before several
significant studies with new information on TDI were published. The
authors recommend that a new literature review be conducted periodically
for each SDS (e.g., every two years), and that a bibliography be
provided.

In a survey of 160 workers at a large national laboratory, more than 90
percent of respondents said that SDSs are satisfactory or very
satisfactory in providing protective information and answering questions
(Phillips et al., 1999). However, others have cited serious deficiencies
in SDS content. In testimony before the U.S. Senate Subcommittee on
Employment, Safety and Training, Carolyn Merritt, the chairwoman of the
U.S. Chemical Safety and Hazard Investigation Board (CSB), stated that
out of 19 chemical accident investigations conducted by the CSB between
1998 and 2004, nine investigations found that deficiencies in hazard
communication were a root cause, contributing cause, or major causal
factor in the accident (Merritt, 2004). In a previous paper, OSHA
reviewed all of the CSB investigations that had been completed as of the
end of 2003 (OSHA, 2004). Although none of the CSB reports reviewed by
OSHA specifically cite SDS inaccuracy as a root or contributing cause
(many of the hazard communication deficiencies actually reflected
failure to comply with the HCS), several of the reports note SDS
deficiencies. For example, CSB found that in three incidents, SDSs
provided inadequate or incorrect data on the reactive potential of
chemicals. In two other cases, SDSs did not mention that a product was
flammable (Merritt, 2004; OSHA, 2004). 

In her testimony, Merritt (2004) expressed particular concern about the
lack of information on combustible dust hazards, which have caused
several explosions in recent years. She noted that the current HCS and
ANSI standards do not define “combustible dust.” In investigations
published since OSHA’s 2004 review (OSHA, 2004), CSB attributed two
2003 dust explosions at least in part to SDS deficiencies:

  

An explosion at West Pharmaceutical Services in Kinston, North Carolina
killed six people and injured 38. The explosion was caused by a
polyethylene slurry which released combustible dust when it dried; this
dust accumulated above a suspended ceiling and ultimately ignited. The
SDS for the slurry did not identify combustible dust as an end-use
hazard; it only described the hazards of the aqueous slurry itself (CSB,
2004).

Seven workers at the CTA Acoustics plant in Corbin, Kentucky were killed
and 37 were injured when a powdered resin dust accumulated in a
production area and exploded during a fire. The SDS indicated that the
resin was combustible, but did not describe the potential for a
catastrophic explosion if the dust were allowed to accumulate. CSB
compared the SDS from this resin manufacturer with SDSs from seven other
manufacturers of the same resin and found notable differences. This
particular SDS lacked information that other manufacturers provided
about explosive risks, and it listed an incorrect flammability rating
and inappropriate handling measures (e.g., dry sweeping, which the NFPA
advises against) (CSB, 2005).

As a result of these investigations and industry-wide concerns, CSB
conducted a broader study of combustible dust hazards, which it
completed in 2006 (CSB, 2006). CSB reviewed the SDSs of 140 known
combustible dusts or powders and published the following findings:

Of the 140 SDSs reviewed, 83 included some form of dust hazard warning,
but much of this language was unspecific. Only 10 percent of these SDSs
addressed combustibility and explosion potential in the most logical
section of the SDS, “Hazard Identification.”

Only seven SDSs referenced the NFPA standard for managing dust hazards.

None of the SDSs listed the key physical properties necessary to
determine explosive potential. Critical properties include the
deflagration index, minimum ignition energy, minimum explosive
concentration, and volume resistivity (Merritt, 2004).

Most of the SDSs did not identify combustible dust hazards that could be
reasonably anticipated to arise through processing and handling, such as
evaporation of water or another solvent.

In light of these results, CSB recommended that the HCS, GHS, and ANSI
Z-400.1 be modified “so that the explosion hazard of combustible dusts
is clearly communicated and included in companies’ hazard
communication programs, and that the type and location of information
about this hazard and how to prevent it are clearly specified and appear
on MSDSs” (CSB, 2006).

Studies in Canada have found similar inadequacies in SDS content. When a
supplier files a trade secret claim, Canada’s Hazardous Materials
Information Review Commission (HMIRC) audits the full SDS. From 2000 to
2001, HMIRC found 1,450 violations of the WHMIS SDS regulations, most
frequently in the sections on “Hazardous Materials” (31 percent) and
“Toxicological Properties” (21 percent) (Baldwin et al., 2003).
Welsh et al. (2000) describe a more detailed audit in which a few
controlled substances were chemically analyzed for comparison with the
corresponding SDS. The authors present three cases in which undisclosed
hazardous ingredients were present at concentrations requiring
disclosure. In at least one case, other sections of the SDS did not
adequately describe the hazards and protective measures associated with
the true composition of the product. To reduce these types of errors,
the authors suggest improved education for employers, inspectors, and
users; improved SDS writer qualifications; and incorporating chemical
analysis in active auditing programs.

5.2.4	Comprehensibility



Key Findings:



In most cases, surveys find that a sizable portion of the audience for
SDSs finds them difficult to read and understand.



Studies suggest that users absorb roughly two-thirds of the information
on an SDS.



The literature contains many ideas on how to make an SDS more
comprehensible without sacrificing comprehensiveness.





As the ANSI Z400.1 standard explains, “One of the greatest challenges
in preparing a material safety data sheet is writing so that various
audiences can read and understand the information. Reading levels of
users vary widely. The target audiences range from an untrained person
needing general information to a highly trained professional. The
information being conveyed is often very technical. It must be complete
enough for the specialist yet understandable for the inexperienced MSDS
user” (ANSI, 2004).

Sattler et al. (1997) cite numerous studies showing problems with the
comprehensibility of SDSs. Since that review, additional studies have
been published on comprehensibility; results are summarized below.
Recent studies also include recommendations for developing an SDS that
is both comprehensive and comprehensible.

Smith-Jackson and Wogalter (1998) asked 60 undergraduates and community
volunteers to sort SDS data into a logical order. After completing the
task, subjects were asked for their opinions on the difficulty of the
content. Overall, 43 percent found the information easy to understand,
42 percent said it was not easy, and the remaining 15 percent felt that
only scientists, experts, or very experienced workers would be able to
understand the information.

Among workers themselves, Phillips et al. (1999) found greater perceived
comprehensibility, but still noted some concerns. Overall, a group of
160 workers at a large U.S. national laboratory felt that SDSs are easy
to read and understand and not too long or confusing. Most workers also
agreed that the use of health hazard ratings on SDSs is not confusing.
However, one-fifth to one-third of workers indicated that SDSs were
confusing or difficult to read and understand, and a comprehension test
found that one-third of SDS information was not absorbed. A similar
result is reported by Conklin (2003), who found that after viewing a
variety of SDS formats, a group of petrochemical workers answered 65
percent of post-test questions correctly. In general, Conklin found that
health-related information was least well comprehended.

Frazier et al. (2001) give a specific example of an SDS phrase that
might fail to convey an important risk. In a study of SDSs from 30
manufacturers of products containing toluene diisocyanate (TDI), one of
the most well-established causes of occupational asthma, the authors
found that only 50 percent of manufacturers list the effect as
“asthma,” while others describe the effect as “allergic
respiratory sensitization.” This technical phrase may not communicate
clearly that the risk is in fact asthma. “Asthma” was more likely to
be specified if TDI was present at a higher concentration in the
product; however, toxicological evidence shows that adverse effects can
occur even at low concentrations. In some cases, health effects were
described in an unspecific way—e.g., “shortness of breath.” 

As part of a study for South Africa’s National Economic Development &
Labour Council (NEDLAC), London (2003) tested workers’ ability to
identify and understand information on an SDS. Although this study does
not represent the U.S. workforce, the results do suggest some reasons to
be concerned with the comprehensibility of current SDSs. In a
comprehension test, subjects from the industrial, transport,
agricultural, and consumer sectors clearly identified some hazards
(e.g., “flammable”), but identified other hazards as little as 29
percent of the time. Other than “first aid,” many key phrases and
headings were not well understood. For example, almost half of those
tested did not know the difference between a trade name and an active
ingredient. When surveyed about their opinions, 18 percent of
respondents said the SDS was “not easy at all,” and an additional 38
percent found it moderately difficult. About two-thirds said the SDS
could be made easier, citing complicated words and sentences as the
greatest problem.

Among the most common methods for assessing comprehensibility of SDSs
are formulas that calculate reading level by considering the length of
sentences, the number of polysyllabic words, and other characteristics
of the text. However, Hochhauser (1997) suggests that readability
formulas are an imperfect way to assess SDS comprehensibility because
the reader’s familiarity with the subject matter is also important.

The review by Sattler et al. (1997) includes several recommendations to
make SDSs more comprehensible. Recommendations from more recent studies
include the following:

Among other things, Frazier et al. (2001) suggest that standardized
language templates could help reduce convoluted phrasing.

An abbreviated version of the SDS could be developed for use on the shop
floor (London, 2003).

SDSs should use simpler English and provide a glossary for any complex
terms. Unless the SDS language is simplified dramatically, however, this
glossary could end up just as long as the SDS itself (London, 2003).

5.2.5	Structure and format



Key Findings:



Studies have found a few differences in how the various SDS formats in
use today address certain topic areas, but overall efficacy is similar. 



It may be helpful to consider ordering the SDS in a way that is most
intuitive to users.



SDS designers should consider layout, easy-to-read text, and effective
use of color (contrast, color coding, etc.).





The way an SDS is organized and formatted can affect its usability.
Important factors include both the structure of the document (order of
topics, etc.) and the way it is physically laid out.

Sattler et al. (1997) describe a study by Phillips (1997) in which
workers at a large national laboratory were tested on their knowledge
before and after viewing an SDS. Three standard SDS designs were tested:
a 9-section OSHA form, the 16-section ANSI Z400.1 format, and the
9-section International Chemical Safety Card (ICSC). In a paper
published after the 1997 review by Sattler et al., Phillips et al.
(1999) describe the final results of this study. All three SDS formats
led to significant improvements in subjects’ knowledge, and there was
no statistically significant difference among the three formats in terms
of total test score. However, there were a few significant differences
in how well readers of each SDS format answered specific types of
questions:

The ICSC performed better than the OSHA form regarding chronic and
immediate health effects.

The ANSI format performed worse than the other two formats on
fire-related questions.

The OSHA form performed better than the other two formats on spill
response questions.

The OSHA form was better than the ANSI format regarding carcinogenic
potential.

In a separate comparison, Conklin (2003) also found similarities in the
overall performance of several standard SDS formats. In this study,
employees of a multinational petrochemical company were given one of
three versions of an SDS for an unfamiliar chemical: a U.S. version
(OSHA’s required content within an ANSI Z400.1-1998 16-part
structure); Canadian WHMIS content in a 9-part structure; and the
European Union’s content and 16-part structure. SDSs were controlled
for font, layout, and reading level. Overall, Conklin found no
statistically significant difference in mean post-test scores using the
three different formats, although there were significant differences on
5 out of 10 questions (no one format was consistently better). 

Because extensive searching can be a barrier to SDS use, some have
wondered whether there is a preferred order of information that more
closely matches users’ cognitive expectations. Smith-Jackson and
Wogalter (1998) asked 60 undergraduates and community volunteers to
arrange portions of six SDSs in the order they considered most usable.
The authors found a few consistent results:

Information about health hazards, protective equipment, and fire and
explosion data tended to be placed toward the beginning.

Physical and reactivity data tended to be placed near the end.

Spill or leak procedures were placed near the beginning or the middle,
depending on the type of chemical.

A majority of subjects reported that they had attempted to prioritize
the hazard information that needed to be communicated. The
participants’ suggested order of information generally did not match
either the original SDS order or the order listed in the
HCS—particularly the subjects’ emphasis on health hazard information
near the beginning. However, the authors note that preferences may
depend on the user’s occupation, so it may be hard to develop one
optimal order for all users (Smith-Jackson and Wogalter, 1998).

In a follow-up survey, Smith-Jackson and Wogalter (1998) found that the
most common suggestion for improving SDSs was to add icons and color
coding. Others cite past research noting that poor color contrast and
small text can make it harder to read an SDS, as can text in all capital
letters (Hochhauser, 1997).

5.2.6	Research needs

Recent publications have suggested a few areas for further research
related to SDSs. For example, Phillips et al. (1999) recommend further
study on the effects of age on SDS efficacy, particularly in light of
the growing research on how working memory changes with age. Others note
that comprehensibility research should not stop with adoption of the
GHS. For instance, Fagotto and Fung (2002) urge OSHA to test the
comprehensibility of harmonized labels and SDSs and examine the extent
to which they impact worker behavior. If standardized materials are not
found to be effective, OSHA should consider a stronger emphasis on
training.

6.	Labels and Warnings

6.1	Background

Unlike “open and obvious” physical safety hazards, which may be
obvious to the professional, the hazards, consequences, and appropriate
procedures for safe use of chemicals cannot be easily determined by
looking at the chemical or smelling it (Laughery and Smith, 2006). Thus,
proper labeling is critical for even the most experienced users.

The HCS currently requires a label on any container of a hazardous
chemical under OSHA jurisdiction. “Label” is officially defined as
“any written, printed, or graphic material displayed on or affixed to
containers of hazardous chemicals” (OSHA, 1994). 29 CFR 1910.1200(f)
covers label requirements, mandating that the label include the
following information:

Identity of the hazardous chemical(s).

Appropriate hazard warnings which provide at least general information
regarding the hazards of the chemical(s).

If the product leaves the workplace, it must also include the name and
address of the chemical manufacturer, importer, or other responsible
party.

According to the HCS, signs or other written materials can be used in
lieu of labels on individual stationary process containers, so long as
the information is accessible, etc. Labels must be legible and in
English, although employers may add information in another language if
they wish.

The HCS does not specify a standard format or design elements for
chemical warning labels. Several organizations have developed voluntary
standards for labels and warnings, however, addressing considerations
such as content, order, format, color, signal word, and pictorials. Some
of these standards apply to warnings in general, while others apply
specifically to the labels that accompany hazardous chemicals.

Key consensus standards related to labels and warnings include the
following:

ANSI Z129.1

The American National Standard for Hazardous Industrial
Chemicals—Precautionary Labeling (ANSI Z129.1) was updated in 2006.
Sponsored by the American Chemistry Council, it covers warning labels
for hazardous industrial chemicals, outlining classification criteria
for physical hazards, health hazards, and environmental hazards. Some of
the criteria have different thresholds or a different number of
categories from the GHS system, but they generally cover the same types
of hazards. Annex B of this ANSI standard provides a side-by-side
comparison with GHS criteria (ANSI, 2006a).

ANSI Z129.1 provides a list of elements that must be included in a
label. This list reflects HCS requirements but also advises the designer
to include additional elements, many of which are listed in the GHS. The
standard provides specific content that should be included, based on
hazard classification. Label elements include:

Identification of the chemical product

Identification of the hazardous component(s)

Statements of hazard, which can be drawn from a specific set based on
hazard classification. For highly toxic substances, this statement may
be accompanied by the skull-and-crossbones icon (which also appears in
the GHS).

Name, address, and telephone number

Three possible signal words (DANGER, WARNING, and CAUTION)

Precautionary measures

First aid instructions

Instructions in case of fire

Instructions in case of spill or leak

Instructions for handling and storage

References to additional labeling or other documents

These label elements are largely the same as what would be required
under the GHS, with a few exceptions; for example, the GHS has only two
tiers of signal word (DANGER and WARNING) and the GHS hazard phrases are
worded slightly differently. Further, as noted in Annex C of this ANSI
standard, the GHS includes pictograms for each hazard type, while ANSI
does not.

Like the GHS, ANSI Z129.1 has a protocol for dealing with mixtures. The
standard also provides general advice on designing labels to communicate
information effectively, including considerations such as color, font,
layout, and durability.

ANSI Z535 series

This set of six American National Standards does not specifically
address labels for chemicals, but it does offer general principles for
warning labels. These standards were all updated in 2006 or 2007,
although some of the updated versions were not available in final form
at the time of this report. This series includes the following
standards:

ANSI Z535.1: American National Standard for Safety Colors (ANSI, 2006b)

ANSI Z535.2: American National Standard for Environmental and Facility
Safety Signs (ANSI, 2002b)

ANSI Z535.3: American National Standard Criteria for Safety Symbols
(ANSI, 2002a)

ANSI Z535.4: American National Standard for Product Safety Signs and
Labels (ANSI, 2002c)

ANSI Z535.5: American National Standard for Safety Tags and Barricade
Tapes (for Temporary Hazards) (ANSI, 2002d)

ANSI Z535.6: American National Standard For Product Safety Information
In Product Manuals, Instructions, And Other Collateral Materials (ANSI,
in press)

Note that ANSI Z535.2 and ANSI Z535.4 specifically exclude chemicals,
instead directing readers to ANSI Z129.1. ANSI Z535.6 is a new standard
designed to address the need for consistency in instruction manuals and
other materials that accompany a product, but it specifically excludes
SDSs (Frantz and Hall, 2005; Hall et al., 2006). Some messages are still
relevant to chemical labels, however. For example, ANSI Z535.5 advises
the use of headline style in labels, and discusses the importance of
order of information. Annex B of ANSI Z535.4 provides general guidance
on label design, while ANSI Z535.3 suggests that a safety symbol should
be understood by 85 percent of the audience, with no more than 5 percent
critical confusion (e.g., deriving the opposite meaning from that which
is intended).

These updated ANSI standards define four signal words. In order from
most to least severe, these words are: DANGER, WARNING, CAUTION, and
NOTICE. The latter should be used only for messages unrelated to
personal injury (e.g., potential property damage). The updates also
allow for use of the safety alert symbol (triangle with an exclamation
point), which brings the standards more in line with the corresponding
ISO standards (see below). Other aspects of the standards have also been
harmonized with ISO (Frantz and Hall, 2005; Hall et al., 2006). 

ISO standards

The International Organization for Standardization (ISO) has published
several standards related to the design of warnings and
labels—particularly graphical symbols. Possible resources for chemical
warning label design include the following standards:

ISO 3864-1 covers design principles for safety signs in the workplace
and public areas. Topics include color standards, luminescence
requirements, and relative dimensions of sign elements (ISO, 2002).

ISO 3864-3 covers design principles for the graphical symbols used in
safety signs. Annex A emphasizes simplicity and efficiency in conveying
the message (ISO, 2006). This standard suggests that symbols be used
only if they are comprehended by at least 67 percent of the audience.

ISO 17724 provides standard definitions for the various elements found
in labels and warning signs (ISO, 2003).

ISO 9186-1 and ISO/DIS 9186-2 describe methods for testing the
comprehensibility and perceptual quality of graphical symbols,
respectively (ISO, 2007a, 2007b).

ISO/IEC Guide 74 offers technical guidelines for considering users’
needs when designing graphic symbols (ISO/IEC, 2004).

Although the ISO system includes many symbols, they are not identical to
the symbology in the GHS, which adopted its graphical elements from the
UN Recommendations on the Transport of Dangerous Goods, Model
Regulations.

NFPA 704

This standard describes the National Fire Protection Association’s
system for identifying hazards of materials from an emergency response
perspective. The standard covers criteria for the NFPA rating system,
techniques for evaluating hazards, and recommended label design (NFPA,
2007).

ASTM E 1445-03

This standard from ASTM International provides a glossary of terms
related to the hazard potential of chemicals (ASTM, 2003). 

For more insight into the practical considerations of labeling hazardous
chemicals, it can also be useful to look into how companies have
actually applied the various standards and regulations to the workplace.
For example, Sussman et al. (1997) describe how one large pharmaceutical
company approaches health hazard labeling for the workplace. This
company developed its own system for labeling acute and chronic health
hazards, borrowing from various regulatory and standard-setting
approaches, including the EU and Canada’s WHMIS. To ensure that labels
focus on the most important information, the company established
criteria for significant effects, based on risk considerations. This
system of decision criteria can be used to select appropriate risk
phrases from the regulations, based on what is known or can be inferred
about a new chemical in the workplace.

6.2	State of the science

6.2.1	Literature reviews

Sattler et al. (1997) describe several literature searches that had been
published as of 1997. Since then, additional authors have compiled
reviews of the scientific literature on labels and warnings. Two of the
more comprehensive reviews can be found in textbooks on warnings:
Warnings and Risk Communication (Wogalter et al., 1997) and Handbook of
Warnings (Wogalter, 2006a). In the latter book, the chapter by Kalsher
and Williams (2006) offers a particularly helpful set of tables that
summarize findings from previous studies.

Parsons et al. (1999) reviewed more than 150 laboratory and field
studies, grouping the results into alphabetical categories. The end
product is a handy reference for anyone who is designing or evaluating a
warning. Wogalter et al. (2002) offer a similar review of empirical
literature on warnings, along with a discussion of approaches for
evaluating warning effectiveness. Williams and Noyes (2007) provide a
more recent overview of research on factors that influence risk
perceptions, including signal words, color, color-word combinations,
surround shape, positive versus negative framing, warning source, and
warning target (self, family, general public, etc.).

Others have conducted literature reviews as part of an effort to fit
empirical results into theoretical models of the warning process. For
example, Conzola and Wogalter (2001) reviewed the relevant literature in
order to describe how research has addressed each of the stages of the
Communication–Human Information Processing (C–HIP) model. Rogers et
al. (2000) sought to develop their own integrative model of the warning
process, based on a comprehensive review of warnings literature from
1980 to 1998.

In late 1996, EPA’s Consumer Labeling Initiative produced a review of
research on consumer knowledge, perceptions, and reaction to labels. EPA
focused on chemicals—indoor pesticides, outdoor pesticides, and
household cleaners—so even though this research project specifically
covered consumers, some of the findings may be transferable to
occupational chemical use (Abt Associates, 1996).

6.2.2	Warning theory

To get to the heart of what makes labels and warnings effective,
researchers have tried to view warning messages through the lens of
psychological theory. Models of cognition and persuasion have proven
particularly useful in this endeavor. This section summarizes recent
advances in warning theory, along with findings that may assist in
evaluating warnings in the future.

One of the most widely cited models of warning processing is the
Communication–Human Information Processing (C–HIP) Model. This model
includes four “steps:” (1) Source; (2) Channel; (3) Receiver
(attention, comprehension, attitudes and beliefs, and motivation); and
(4) Behavior. Recently, some authors have found this model to be a
useful way to explain results from the literature; for example, see
Conzola and Wogalter (2001). Wogalter et al. (1999) used this scheme to
organize a textbook on warnings.

Other models have been used to describe warnings as well. Some of these
models are described in chapters 6 and 7 of Wogalter (2006a). 

Rogers et al. (2000) propose an integrative model of the warning process
consisting of four steps: (1) notice; (2) encode; (3) comprehend; and
(4) comply. The authors reviewed recent literature and found many
variables that affect the warning process during these four stages. The
authors also recommend several research techniques that would reflect
these distinct stages of the warning process.

Zuckerman and Chaiken (1998) present the Heuristic-Systematic Model of
persuasion as a theoretical framework for research on product warning
labels. Heuristic and systematic represent two modes by which people
process information to reach judgments. The former uses learned
knowledge structures (simple decision rules) to reach judgments, while
the latter relies on accessing, scrutinizing, and integrating all useful
information to reach a judgment. Applying this model, the authors are
able to explain empirical findings from the literature and suggest
implications: namely the need to tailor warnings accordingly. For
example, if the user’s systematic processing ability is reduced (e.g.,
through diminished motivation or capacity to reach a reasoned decision),
a warning should present only essential information that requires the
least amount of processing.

Lehto and Papastavrou (1998b) describe the Distributed Signal-Detection
Theory Model, noting the importance of selectivity in warning design in
order to avoid information overload. The authors suggest that
selectivity always increases effectiveness—which may have implications
for warning against intermittent versus continuous hazards.

Other models have proven less adept at describing the warning process.
For example, Grotjohn (2000) tested the Elaboration Likelihood Model
(ELM), which predicts that if individuals thoroughly process a
persuasive communication, they will be more likely to be influenced by
content (central clues). Without thorough processing, individuals will
be influenced more heavily by peripheral cues. To test this model, the
author evaluated the influence of a power sander warning on users, only
some of whom thought they would actually be using the product. Results
show that the amount of elaboration (presumed thoroughness of
processing) did not affect perceived hazardousness, suggesting that the
ELM is not a good model for warnings.

Recent textbooks present the most accepted methods of warnings research
(e.g., Wogalter, 2006a), while studies continue to uncover useful
information for study design. For example, Adams et al. (1998) found no
difference in reaction and behavioral intentions between a group of
undergraduates and a group of industrial workers, suggesting that in
some cases, students may be an accurate surrogate for other groups.
Similarly, Drake et al. (1998) found that students and community
volunteers ranked signal words in a fairly consistent manner.

6.2.3	Utility



Key Findings:



Warnings influence attention and behavioral compliance.



Labels are particularly important because they are one of the first
places people are inclined to look when they want information on a
particular chemical.





Before stepping into the fine details of label design, some researchers
have ventured to answer the more fundamental question of whether
warnings work. Overall, analyses conducted since 1997 continue to say
“Yes.” As Kalsher and Williams (2006) note in their summary of
behavioral compliance with warnings, studies generally show that “the
presence of a warning reliably leads to greater compliance than with the
absence of a warning.”

 Other literature supports this claim. In an extension of Cox et al.
(1997), Argo and Main (2004) present a meta-analysis of 48 studies
conducted between 1975 and 1997. The authors consider five dimensions of
warning effectiveness (attention, reading and comprehension, recall,
judgments, and behavioral compliance), and determine that warnings do
influence attention and behavior.

In occupations that use chemicals, labels are a particularly important
form of warning because they are often located on the product with which
the user is interacting. London (2003) revealed the importance of labels
through a study of label and SDS comprehensibility in South Africa. When
asked where they would go to get information about a chemical,
respondents most often said “the label”—rather than an SDS,
another person, or another resource. These findings held true for the
industrial sector (52%), transport sector (61%), agricultural sector
(60%), and household consumers (64%).

6.2.4	Evaluation of current labels and warnings



Key Findings:



Current labels on industrial chemicals vary in their effectiveness. One
frequent limitation is the lack of explicit instructions—for example,
what constitutes “adequate” ventilation?



Labels from consumer products and pharmaceuticals may offer lessons that
can be applied to workplace chemical labels. Pharmaceutical labels have
been researched extensively.





The literature contains many studies that have evaluated actual product
warnings in the context of noticeability, comprehensibility, or
effectiveness at influencing behavior. Riley and Fischhoff (2000) is one
example of a recent study. By interviewing workers about their habits
and work environment, the authors developed a risk model for assessing
how effectively different labels would protect workers’ health. Next,
they applied this model to evaluate the labels found on 14 containers of
methylene chloride paint stripper from six different manufacturers.
Certain health effects were listed on some labels but not others. All
but one had instructions regarding ventilation, but the directions
always provided a non-explicit instruction like “use adequate
ventilation,” rather than specific actions to be taken. Finally, the
authors modeled the inhalation exposures that would result from
following the instructions in a typical workplace. Results were as
follows:

Four of the six manufacturers provided useful information for reducing
exposures. For the other two, however, even reading the full label would
result in a peak exposure well above 500 ppm (the levels associated with
acute health effects).

In all cases, users who read only the “directions for usage” or
statements in bold would also face peak exposures well above 500 ppm.

Recognizing that research has already identified deficiencies in
current labeling practices, much of the recent study has gone into
testing the effects of potential enhancements to label design. Thus,
many of the studies conducted since 1997 use current chemical labels not
as the experimental condition, but as a control condition (i.e., the
status quo). These studies are outlined in the sections that follow.

Other researchers have focused on the effectiveness of consumer product
labels. While the user groups differ, some of the results may be
applicable to workplace labels as well—particularly findings related
to household chemical products. 

In an in-depth survey of hundreds of consumers regarding labels for
household hard surface cleaners and indoor and outdoor pesticides,
EPA’s Consumer Labeling Initiative (Abt Associates, 1999) found that
“Consumers are generally satisfied with current labels and are able to
find the information they want on the label,” but that certain
improvements would encourage more reading and use of product labels.
Respondents felt that different types of products merit different types
of labels. In particular, respondents felt that household cleaner labels
should be simpler, with “exceptional” (i.e., very important or
unexpected) information highlighted. Consumers found indoor insecticide
labels to be quite effective, but outdoor pesticide labels were
confusing and should be “simplified and arranged for easier
reading.” Respondents generally noted that a laundry list of
ingredients does not provide all the information they would like, and
suggested communicating this information by type, category, or purpose
of ingredient. The study concludes with several recommendations for
consumer labeling. 

Another type of consumer product that is commonly studied is
prescription or over-the-counter medication. Pharmaceutical labels are
similar to chemical labels in that they often have explicit instructions
for use which, if not followed, can cause adverse health effects or
death. Designers of pharmaceutical labels also encounter many of the
same challenges faced by those who design chemical labels, such as
container space limitations and the need for standardized pictograms
that convey concepts to low-literate or non-English literate users. For
these and other reasons, many of the findings on pharmaceutical labeling
may be applicable to chemical labeling. Indeed, for certain sensitive
populations like the elderly, pharmaceutical studies represent the most
thorough body of research on label effectiveness. Relevant findings are
reported in the sections that follow.

For reference, Wolf et al. (2006) provide an overview of pharmaceutical
label effectiveness. The authors asked 74 low-literate patients (at or
below a grade 6 reading level) to interpret and comment upon 8 commonly
used prescription warning labels, Only one label—“take with
food”—was correctly interpreted by more than 50 percent of patients,
and one label was not understood by any. Overall, the study found five
major causes for misunderstanding:

Multiple-step instructions

Reading difficulty of text

Icons that may be confusing or discordant with text

Use of color—people interpreted red to mean danger and green to
represent a recommendation, even though this was not the intended effect

Message clarity

Wolf et al. suggest addressing these causes for misunderstanding by
creating standard color schemes and simplifying text, among other
things.

6.2.5	Warning content

6.2.5.1	Signal words



Key Findings:



Signal words are generally recognized as a key alerting element of
warning labels.



Signal words also convey a particular level of hazard.



Studies of English-speaking people around the world have found a
generally consistent hierarchy of signal words with respect to perceived
hazard. DEADLY, DANGER, and WARNING seem to connote different levels of
hazard, while the perceived difference between WARNING and CAUTION is
often insignificant. 





The signal word is a word that typically appears near the top of a
warning, sometimes in all capital letters. Common examples include
DANGER and WARNING. The signal word is generally understood to serve a
dual purpose: alerting the user to a hazard and indicating a particular
level of hazard. For example, users generally perceive the word DEADLY
to indicate a far greater degree of hazard than a term like NOTICE.

The GHS prescribes one of two signal words for labels—DANGER and
WARNING—depending on the hazard classification of the substance in
question. DANGER is used for the more severe hazard categories. The
least severe hazards have no signal word at all. GHS signal words are
similar to those in other established hazard communication systems,
except that some other systems have three or more tiers—for example,
ANSI uses DANGER, WARNING, and CAUTION, in order of descending severity.
Recently, a great deal of research has focused on how people perceive
signal words, and in particular, how they perceive words to be different
from one another. Overall, this research supports the use of signal
words in labels, demonstrating that they can attract attention and help
people clearly distinguish between certain levels of hazard. The
research also tends to support the decision to use only two tiers in the
GHS, as many recent studies have found clear differences between DANGER
and WARNING but little perceived difference between WARNING and CAUTION.
These findings are summarized by Rousseau and Wogalter (2006).

A search of the literature published since 1997 (or not included in
Sattler et al. [1997]) found numerous studies on signal words. Key
findings of these studies are as follows:

Seeking to test warning signs in realistic settings rather than in a
laboratory, Adams et al. (1998) tested five industrial warning signs on
a group of 40 blue-collar workers employed in heavy industry, as well as
a group of students. Signs were manipulated to include all four key
elements (signal word, hazard statement, consequences statement, and
instructions statement) or one or more of these elements missing.
Participants were asked questions to gauge their reaction and behavioral
intentions. Overall, 77 percent (66 percent of the worker group)
recognized DANGER as the key word when it appeared, and more than 80
percent recognized BEWARE and CAUTION, suggesting that the signal word
was generally noticed, and it was recognized as the key alerting
element. DANGER was significantly more likely than other words to
influence behavioral intentions.

Laughery et al. (1993) also demonstrate the usefulness of signal words
in context. The authors tested the noticeability of warnings on
alcoholic beverage containers in the U.S., and found that a signal word
(WARNING) was one of several factors that decreased the amount of time
it took for participants to locate the warning label.

Several studies have tested the arousal strength or perceived hazard of
different signal words. For example, Griffith and Leonard (1997) asked
80 female undergraduates (who were unlikely to have already received
industrial safety training) to rate signal words on nine scales related
to three factors: evaluation, activity, and potency. Results included a
list of terms in order of “meaningfulness,” representing conceptual
“distance” from the neutral term NOTICE. From most to least
meaningful, these terms are DANGER, URGENT, BEWARE, WARNING, STOP,
CAUTION, and IMPORTANT.

Wogalter et al. (1998a) asked over 100 undergraduates and community
volunteers to rank signal words. DEADLY was perceived as most hazardous,
followed by DANGER, WARNING, and CAUTION. All differences were
statistically significant except the difference between WARNING and
CAUTION. In a follow-up experiment using labels produced in the ANSI
Z535.2, ANSI Z535.4, and alternative formats, the authors found a
similar rank order for signal words with all labeling systems. Finally,
the authors tested the same terms on workers from manufacturing and
assembly plants and found the same general order: DEADLY, then DANGER,
then WARNING and CAUTION with no significant difference between them.

In more of a free-form experiment, Young (1998) asked 30 subjects to
produce warning signs for a set of scenarios, using different sign
components available on a computer screen. In roughly 80 percent of the
signs, the participant chose to use a signal word. DANGER, DEADLY, and
LETHAL were more likely to be used for scenarios with severe hazards;
CAUTION and NOTICE for non-severe scenarios. WARNING was used equally in
both types of scenarios. The author suggests that these results support
a two-tiered system of signal words (like the GHS). In a separate task,
users ranked the perceived hazard of signal words, resulting in the
following list from most to least severe: DEADLY, LETHAL, DANGER,
WARNING, CAUTION, and NOTICE.

While many have focused on the relative perceptions of signal words,
others have sought to evaluate how the absolute meaning of common signal
words is perceived. Drake et al. (1998) asked a group of students and
community volunteers to match signal words with definitions borrowed
from consensus standards and other sources. Participants matched DANGER
to a correct definition 64 percent of the time, while NOTICE was matched
correctly 68 percent of the time. WARNING and CAUTION were matched
correctly less than half of the time, suggesting confusion. These two
terms also appeared to connote less hazard than the actual definitions
assigned to them, suggesting that “people tend to underestimate the
degree of hazard that the words are intended to convey relative to their
assigned definitions.” According to people’s perceptions, the most
accurate definitions for GHS signal words were ANSI’s (for DANGER) and
Webster’s dictionary (for WARNING). The authors recommend using
WARNING and CAUTION interchangeably and using DEADLY for extreme
hazards, as this word is perceived to be significantly stronger than
DANGER. The authors also suggest that a standard set of signal words is
helpful for users with limited English skills, who can be trained to
recognize a few key words (but not synonyms).

Signal word perceptions generally hold true among non-American
populations, too, a testament to the universality of some of the basic
concepts. For example, Hellier et al. (2000a) asked 984 adults and 547
children (age less than 15) in the UK to rate DANGER, WARNING, and
CAUTION on a hazard scale from 1 to 10. Adults and children ranked
DANGER as 8.5 and 8.4, respectively. Both groups ranked WARNING as 7.8,
while adults and children rated CAUTION as 7.25 and 7.4, respectively.
These results match what previous studies in the U.S. have found
regarding order, while supporting the idea that WARNING and CAUTION may
be too close to serve distinct purposes.

Hellier et al. (2000b) asked a mixed-age group of participants in the UK
to rate the arousal strength of 84 signal words commonly used in the
U.S. The authors found that 72 of these words were at least 90 percent
understandable, and confirmed that DANGER is stronger than WARNING,
while WARNING and CAUTION are not significantly different from each
other. Next, the authors tried three different experimental methods of
rating arousal strength for a subset of terms. They found high
correlation among the rankings generated by the three different methods,
suggesting that the rankings established by this study and by others are
relatively robust. In general, the results support using signal words to
facilitate hazard matching, and support matching hazards and signal
words consistently and appropriately.

Others have extended the analysis to developing countries, and found
similar results. Banda and Sichilongo (2006) tested GHS-style labels
among workers in Zambia, using four different signal words (as well as
other variables, which are discussed later). Among workers in the
industrial and transport sectors, DANGER was generally perceived as the
most hazardous signal word. WARNING was one of a group of terms that
were largely indistinguishable from one another, but distinct from
DANGER. The authors support adoption of the GHS, suggesting that having
just two possible signal words will lead to “more impact and less
confusion about the extent of hazard.” 

Similarly, in a large study on SDS and label comprehensibility conducted
for South Africa’s National Economic Development & Labour Council
(NEDLAC), London (2003) found that DANGER was generally ranked as more
hazardous than WARNING by participants in the four sectors tested:
industry, transport, agriculture, and consumers. 

Obviously, perceptions of English signal words may not hold true among
non-English-speaking populations. However, European authorities have
developed mechanisms for translating hazard information, and as Parsons
et al. (1999) suggest, for example, it would be possible to develop a
pair of words in Spanish that effectively convey the same hazard levels
as DANGER and WARNING.

6.2.5.2	Symbols



Key Findings:



Symbols generally make warning labels more noticeable and easier to
comprehend. They are often the most easily recalled elements of labels.



Symbols can improve label comprehension among children, people with low
literacy, and those who do not understand the language.



Many GHS pictograms are widely recognized, but some others are more
obscure.



Symbols are most effective when there is a direct relationship between
the image and the meaning, requiring minimal inference.



Symbols tend to be most effective when paired with redundant or
reinforcing text, which is a form of dual coding. GHS labels do this.





Symbols serve several important functions in warning labels. As Wogalter
et al. (2006b) explain, symbols may alert the user to a hazard more
effectively than text alone: “Symbols may be more salient than text
because of visual differentiations of shape, size, and color. Usually
symbols have unique details and possess more differences in appearance
than do the letters of the alphabet. Letters are highly familiar and are
more similar to one another than most graphical symbols.” Symbols also
can bolster a text message and improve label comprehension among
children, people with low literacy, and those who do not understand the
language in which the label text is written (Parsons et al., 1999).

All symbols are not created equal, however. As Wogalter et al. (2006b)
note, some studies have found slower processing, poorer recognition, and
greater learning difficulties with symbols versus with
text—particularly if the symbols are complex or non-intuitive. These
results speak to the need to choose symbols carefully and to train users
on what they mean. Wogalter et al. (2006b) discuss methods and criteria
for testing comprehension of symbols, as do DeJoy et al. (2006),
Smith-Jackson and Wogalter (2006), Wogalter and Dingus (1999), and Young
and Lovvoll (1999).

The GHS uses two types of symbols on chemical labels: a hazard pictogram
and a precautionary pictogram. The hazard pictogram is taken from a
standard set, based on the health or physical hazard class of the
chemical. The precautionary pictogram depicts a measure to be taken to
minimize or prevent adverse effects; suggested precautionary pictograms
are included in Annex 3 of the GHS. 

Note that some sources offer distinct definitions for “pictogram,”
“pictorial,” and “symbol.” For example, Rogers et al. (2000)
state that “Pictorials refer to pictures that represent the concept of
interest (e.g., a picture of a fire extinguisher). Symbols are more
abstract representations of a concept, the meaning of which must be
learned (e.g., the use of a skull-and-crossbones to denote poison).”
ANSI and others combine these terms in the definition of “symbol,”
however, and for the purposes of discussing the literature on this
subject, this report will use the words “pictogram,”
“pictorial,” and “symbol” interchangeably.

In the literature published since 1997, several researchers have sought
to evaluate how people comprehend symbols currently in use, including
those that are part of the GHS. Key findings are described below.

As part of an experiment to see how people comprehend warnings on
household chemical labels, Akerboom and Trommelen (1998) asked 60
university students whether they understood the meaning of several
European pictograms—including five that are part of the GHS. Results
for the GHS pictograms are as follows:

Flammable: 93 percent comprehension

Environmental toxicity: 93 percent comprehension (a separate experiment
said 79 percent)

Acute toxicity: 85 percent comprehension

Corrosive: 20 percent comprehension

Oxidizer: 13 percent comprehension

Only the first three symbols met the 85 percent comprehension criteria
suggested by ANSI Z535.3 (ANSI, 2002a). The authors recommend that
labels present the hazard phrase and symbol together, along with
corresponding precautions. The GHS does so, and standard GHS hazard
statements mention the type of hazard, so symbol comprehension will not
be a prerequisite for understanding a GHS label.

Banda and Sichilongo (2006) tested comprehension of GHS labels among 364
workers in four sectors in Zambia (transport, agriculture, industrial,
and household consumers). Within this population, the
skull-and-crossbones symbol was widely understood, as were
“flammable” and “environmental hazard.” The simple “X”
inside a diamond was not well understood. Based on these results, the
authors suggest a preference for symbols that depict familiar,
meaningful, and recognizable images. London (2003) performed a similar
study among the same four sectors in South Africa, finding that the
skull-and-crossbones was understood by at least 96 percent of each
sector and “flammable” by at least 89 percent. “Explosive” and
“environmental toxicity” were correctly comprehended by 44 to 71
percent and 50 to 58 percent of each sector, respectively. Many
health-related symbols did not fare well, and six GHS symbols had less
than 50 percent comprehension across all four sectors. Outside the
transport sector, “compressed gas” was the least well comprehended
GHS symbol. Advisory pictograms were generally well comprehended,
particularly those pertaining to personal protective equipment (PPE).

Schroeder et al. (2001) asked a group of 104 adults—over a wide range
of ages—to rate their familiarity with 40 ANSI safety symbols and then
generate phrases about each. These symbols included mandatory actions,
prohibitory actions, information symbols, and hazard alerting symbols.
None of these four groups of symbols were comprehended above the 85
percent criterion set forth in ANSI Z535.3 (ANSI, 2002a), suggesting
that safety symbols may not be as well understood as some would expect.

A few studies have found that the skull-and-crossbones icon—used for
certain hazard classes in the GHS—is among the most recognizable
safety symbols. For example, Wogalter et al. (1998a) asked 112
undergraduates and community volunteers to rank various label elements.
Among shapes and icons, the skull symbol (in this case, without the
crossbones) was rated most hazardous and most noticeable. The skull
connoted the greatest hazard among industrial workers, too.
Smith-Jackson and Wogalter (2000) asked 48 English speakers to rate the
perceived hazards of six alerting symbols. The skull was rated
significantly higher than all others, followed by the prohibition and
electrical shock symbols. These three symbols also were ranked highest
in an earlier study with Spanish speakers (Wogalter, 1997b). English
speakers rated the alert symbol (exclamation point in a triangle)
second-lowest in terms of perceived hazard (Smith-Jackson and Wogalter,
2000). 

Swindell (1999) found that the type of pictorial does not strongly
influence noticeability, however. The author measured subjects’
reaction time when searching for labels with different elements,
including five different icons (none, asterisk, alert, Mr. Yuk, and
skull-and-crossbones). Having no icon present led to significantly
slower reaction time, but there were no significant differences between
different types of icons, although respondents suggested that the
asterisk connoted less hazard than the other icons. 

Others have examined the effectiveness and comprehensibility of
pictograms among special populations. For example, Vaillancourt et al.
(2004) tested medical label pictograms among three non-English-speaking
ethnic groups. The study found that some pictograms were understood by
all groups, but others had to be redesigned to address cultural values
or allow greater comprehension. These results illustrate the importance
of testing symbols on the target population. Kassam et al. (2004) report
a similar study with Cantonese, Punjabi, and Somali immigrants in
Canada. Without verbal instructions, participants understood only 56
percent of the 16 pharmaceutical pictograms, on average. Next,
participants were allowed to suggest modifications, and these new
pictograms were shown to a new group of participants. Modification
improved comprehension by 22 percent, again indicating the importance of
testing and feedback. Participants’ education level was the only
external factor with a significant effect on comprehension; the authors
surmise that education is related to “visual literacy.” Participants
noted that they prefer to receive pictograms with verbal instructions,
not alone.

Hypothesizing that symbols can improve communication with the mentally
disabled, Hoonhout (2000) tested symbol comprehension among mildly
retarded (IQ 50-70) and non-retarded adults. Symbols were taken from a
coffee bar setting, generally showing different types of food or
beverage. Among the mildly retarded group, only 1 of 24 symbols met the
67 percent criterion for comprehension set forth in ISO 3864-3 (ISO,
2006); 7 of 24 symbols were adequately understood in the non-retarded
group. Retarded adults were more likely to describe the symbol but not
extrapolate its meaning, suggesting that for this group, it is best to
denote concrete, familiar, and simple objects rather than relying on
complex associations. Training appeared to improve symbol comprehension
among the retarded adults.

Other recent research has focused on the variables that determine how
effective a particular symbol will be. These studies offer numerous
recommendations on how to design better symbols for warnings and labels.

Braun and Shaver (1999) asked 178 subjects to evaluate the perceived
hazard of warning signs containing combinations of color, signal word,
symbol, and text. Symbols and text varied in terms of explicitness. The
study found that the explicitness of symbols accounts for a portion of
the variability in hazard ratings. The authors recommend that “When
creating warnings, proposed warning designers should select symbols that
convey the type of injury resulting from noncompliance. Similarly, the
warning text should state the severity of possible injuries.”

Dowse and Ehlers (2003) tested comprehension of 46 pharmaceutical
pictograms among a group of 130 Xhosa-speakers in South Africa, who
ranged from illiterate to college-educated. This study found that
education level had a significant influence on the interpretation of 24
of the 46 pictograms. People with low education had the most trouble
with abstract representations like heat. The authors theorize that less
skilled readers are less able to focus on key visual elements. “Poor
readers are slow to interpret perceptual information and tend to
interpret literally. They often find it difficult to match the logic
contained in the message with their own experience, and unless the
message can be rapidly understood, they lose interest easily, as this
inability to understand is a more common occurrence in this group than
in skilled readers.” 

Houts et al. (2001) studied long-term recall of spoken medical
instructions when accompanied by a handout with pictograms. Nearly 200
pictograms were tested with 21 low-literate adults (less than grade 5
reading level). Immediately after training, participants recalled the
meaning of 85 percent of the pictograms, and they recalled 71 percent
after 4 weeks. This study found that recall was better for simple
pictograms where there is a direct relationship between the image and
its meaning—no inference required.

Another body of literature focuses on the utility of symbols in general.
Ganier (2001) found that people generally construct mental
representations faster with pictures than they do with text, supporting
earlier findings on the usefulness of symbols (Wogalter et al., 2006b).
Evans et al. (2002) found similar results with a task in which
undergraduates were asked to sort items into categories using either
text clues, visual clues, or a combination of pictures and text. When
categories were fixed (i.e., sorting instructions were specific), people
sorted the cards more similarly to one another when presented with
pictures than when presented with text alone. The authors suggest that
“pictorially represented stimuli can add some value to the meaning of
a concept thereby leading to higher overlap among participants.”

In a follow-up article on the South African study described previously,
Dowse and Ehlers (2005) found that Xhosa-speaking patients receiving
antibiotics adhered to instructions much better when the instructions
included pictograms (54 percent with high adherence, versus 2 percent
when given text-only instructions). Pictograms also contributed to
understanding.

To examine factors that influence the effectiveness of pharmaceutical
labels, Kalsher et al. (1996) asked subjects to rate the noticeability,
ease of reading, and overall appeal of labels with or without
pictorials. A group of 84 undergraduates gave consistently higher
ratings to labels with pictorials, particularly on tag and fold-out
labels. A group of elderly subjects had similar preferences, rating
labels with pictorials as significantly more noticeable and likely to be
read.

Laughery et al. (1993) found similar results with a timed test on
alcoholic beverage labels. When a pictorial was present to the left of
the warning showing what not to do when drinking, the amount of time it
took to find the label was significantly reduced. An icon consisting of
the alert symbol and the signal word WARNING also decreased response
time. The fastest response time came when four different enhancements
(including the pictorial and the icon) were included. In a follow-up
exercise, an eye scan test found that the pictorial had a particularly
strong influence on reaction time, compared with other enhancements.

As far as chemical labels are concerned, London (2003) found that
symbols tend to be the most easily recalled label elements. In a
comprehensibility test of GHS labels among South African workers,
symbols were the most commonly recalled elements—particularly the
skull-and-crossbones—and people recalled looking at symbols first.
Symbols were also cited as by far the most important factor in
determining hazard perception. Overall, the author concludes that
“Symbols are therefore key to attracting attention, and informing risk
perception regarding a chemical.”

Wogalter et al. (1993) found less encouraging evidence on pictorials,
however. The authors tested the influence of various warning variables
on whether subjects wore proper protective equipment during a task
involving measuring and mixing chemicals. Warning location and the
amount of clutter around the warning had significant effects on
compliance, but the presence or absence of pictorials did not.

Meingast (2001) asked subjects to recall warning content after viewing
labels that were either high quality (conforming to the ANSI Z535.4
standard with color signal icons, pictorials, and organized text) or low
quality (text only). Pictorials were the items remembered most often,
accounting for 48 percent of what viewers of high quality labels
recalled. The author suggests that these pictorials also served the role
of dual coding, meaning that they help to improve the retention of
corresponding text.

Other recent studies support this dual-coding function of pictorials,
finding that symbols tend to be most effective when paired with
redundant or reinforcing text. For example:

Sojourner and Wogalter (1997) asked 35 participants to rate several
prescription label formats in terms of ease of reading, ease of
understanding, overall effectiveness, likelihood of reading, overall
preference, pictorial understanding, and how helpful pictorials are in
helping to remember the instructions. The authors found that people
prefer fully redundant text and pictorials, which they judged easiest to
read, most effective, and preferred overall. Dual-coded pictorials aided
understanding and memory more than labels with pictorials only (no text)
or pictorials for only certain parts of the text. Users preferred
text-only labels to pictorials-only labels, and they found that adding
just a few pictorials didn’t improve the label—perhaps because they
were more likely to ignore text that did not have an associated
pictorial. In a follow-up study, Sojourner and Wogalter (1998) gave
undergraduates, young adults, and older adults a free recall test after
viewing medication labels. Fully redundant text and pictorials led to
significantly greater recall than other formats, and were rated most
effective by all age groups.

Similarly, Sansgiry et al. (1997) found that pictograms on
over-the-counter drug labels improved comprehension, but only when they
were congruent with the corresponding text. A group of 96 young and old
adults were less confused, more satisfied, more certain about their
knowledge, and understood more when shown labels that contained
congruent pictures and verbal instructions, versus verbal instructions
alone. The results were significantly better with congruent pictures and
text than with either pictures alone or incongruent pictures and text. 

Young (1997) asked participants to create warning signs on a computer in
response to 30 different scenarios. Participants were asked to choose
from sign components available on the screen. In some cases there were
no pictorials available; in other cases, there were generic symbols
(e.g., the “alert” symbol) or hazard-specific pictorials. The author
found that generic pictorials were used infrequently—mostly in
high-severity scenarios—but hazard-specific pictorials were used most
of the time they were available. Specific pictorials were typically
redundant with text, not used in place of text. These results suggest
that people perceive the value of pictorials in conveying information,
not just attracting attention.

Sansgiry et al. (1997) found that even an incongruent picture-text
design seemed to create a more positive product evaluation and increase
purchase intentions in older adults. This result provides evidence of
the “picture superiority effect.” Lotto et al. (1999), however,
found that pictures are not always superior. In a categorization task,
Lotto et al. found that when pictures are visually similar, they require
more processing to enable discrimination than do the corresponding
words. Thus, pictures may not be the best way to convey subtle
differences.

6.2.5.3	Hazard statements



Key Findings:



Users generally like to have information that explains why they need to
take precautions. A clear hazard statement can do this.



Hazard statements should be clear and should minimize the need for
inference.



Users appreciate hazard statements in which consequences are paired with
actions they can take to reduce or avoid the risk.





Hazard statements are a fundamental element of the chemical warning
label because they explain the nature of the hazard. In the GHS, labels
contain hazard statements that are assigned based on hazard
classification. Research conducted since the review by Sattler et al.
(1997) generally supports the role of the hazard statement, and offers
some advice on how this element can be used most effectively.

Some justification for the role of the hazard statement comes from
EPA’s Consumer Labeling Initiative (Abt Associates, 1999). As
previously noted, this study asked consumers about their attitudes
toward labels on household chemical products. Overall, consumers
indicated that they like to have information that clearly connects
consequences with actions, and they like to know why they are being
instructed to take a particular precaution. A clear hazard statement can
provide this information.

Others have suggested ways to improve hazard statements. In a study on
inference, Bowles et al. (2002) asked subjects to review product
warnings, then either decide what actions they should take or evaluate
whether someone else’s actions were safe, based on the warning. Some
labels were typical (e.g., “chipper shredder can draw body parts
in…”), some contained similes (e.g., “chipper shredder is like a
vacuum…”), and others contained metaphors (e.g., “chipper shredder
is a vacuum…”). In general, situations that required the user to
make inferences about a hazard—particularly when they had to come up
with their own ideas for protective actions—led to decreased intent to
comply. Similes did not improve compliance over more typical warnings,
and metaphors improved compliance only marginally. The authors conclude
that hazard statements should be written in a way that reduces the need
for inference.

Braun et al. (1995) examined consequence statements in product
instructions for a pool treatment chemical and a polyvinyl chloride
(PVC) adhesive, asking subjects to rate the injury risk posed by each
product. The experimenters manipulated the instructions to include
either recommended actions only, actions followed by consequences,
consequences followed by actions, or a simple restatement of the product
label. The authors found that actions paired with consequences led to
significantly higher risk perception than a restatement of the label or
actions alone. The very highest risk perception came when consequences
preceded actions. Although the preferred wording was longer than the
alternatives, subjects did not feel that the instructions were too
complex, suggesting that they appreciate having actions and consequences
paired together. Freeman (2001) echoes these findings in a discussion on
communicating health risks to fishers and farmers, noting that to be
useful, risk statements should be balanced with equally strong
statements of ways to reduce or avoid the risk. In addition, Freeman
points out that “To be effective, health and medical information for a
specialized audience must address risks and remedies specific to that
audience.”

Additional studies have looked at specific characteristics of the text
in labels—including hazard statements. These studies are outlined in
the sections that follow.

6.2.5.4	Text



Key Findings:



Too little text may not provide enough information for safe product use,
while too much text can crowd out important messages and decrease
recall.



Some current labels are too difficult to understand clearly,
particularly among users with low education levels.



Labels that are overly complex can elicit negative reactions from both
low-educated and highly educated users.



Many studies show that explicit terminology can improve understanding of
warning labels, as well as overall perceptions.



Current labels use terms that are too subjective for some users—for
example, “copious,” “adequate,” “repeated,” and
“prolonged.”



Explicit warning information does not appear to make people less likely
to use a product, contrary to some manufacturers’ fears.



Definitive statements may improve believability and recall, while the
use of personal pronouns in instructions may improve compliance.



The most effective format for bilingual labels appears to be
side-by-side text with equal font size and flags (e.g., U.S. and Mexican
flags) to draw the reader’s attention at a glance.





Warning labels may contain several important pieces of text. Under the
GHS, a label will include a signal word, hazard statements,
precautionary statements, a list of ingredients, supplier
identification, and supplemental information. Recent studies on warning
text can be grouped into five main topic areas: length, difficulty,
explicitness, tone, and bilingual concerns. Key findings are outlined in
the sections below.

6.2.5.4.1	Length

Studies on the length of text have generally sought the “middle
ground” between two undesirable extremes. Without enough text, a label
may not provide adequate information for reasonably safe product use.
With too much text, however, a label can risk “overwarning” and
drowning out the most important information. As Laughery and Smith
(2006) explain, brevity is important because people are less likely to
remember information from long warnings.

Parsons et al. (1999) discuss the “overwarning” effect, citing a
study by Chen et al. (1997) in which perceived risk declined as more
non-critical messages were added to a product warning. Similarly, in a
study that asked subjects to recall information from chemical safety
labels, Akerboom and Trommelen (1998) found that increased length led to
decreased recall. Akerboom and Trommelen also found that including
environmental information in the label did not diminish perceived hazard
or recall of safety information, but it did decrease comprehension
because some subjects incorrectly interpreted safety messages as being
related to the environment.

To address concerns about label length, Krenek and Purswell (1998) offer
advice on deciding what content to include in a new product label. Here,
the authors suggest four variables to consider: hazard detectability,
user knowledge of the consequences of exposure to the hazard, level of
seriousness of injury, and likelihood of user exposure. Hazards can be
ranked in terms of these variables, and hazards rising above a certain
threshold (e.g., exposure above a certain duration or frequency) will be
more important to include in the label.

6.2.5.4.2	Difficulty

Most of the recent studies on difficulty of text in hazard
communications seem to have focused on SDSs, as discussed earlier.
However, a few studies offer new insights that are specifically relevant
to labels.

Lepkowska-White and Parsons (2001) asked 107 subjects to compare two
versions of a warning label: one rated at a 5th grade level according to
the Flesch-Kincaid Grade Level scale, and one at 10th grade level. Some
subjects had not completed high school, while others had earned
bachelor’s degrees. People who had not finished high school had
significantly more trouble comprehending the complex label than the
simple label, and they understood the complex label much less than
college-educated subjects did. In an attitudes survey, both groups
perceived products with simpler labels as safer, and they all reported
more negative attitudes towards warnings with complex words. The authors
suggest that this negative attitude arises because less educated people
get frustrated trying to understand a complex label, while more educated
people dislike what they perceive as unnecessary complexity. 

EPA’s Consumer Labeling Initiative (Abt Associates, 1999) offers the
perspective of a broad-based consumer survey. Overall, respondents felt
that household cleaner labels are unnecessarily complex, and should be
simplified to present exceptional information only. They also found
outdoor pesticide labels to be confusing, and generally suggested
simplifying the labels and arranging them for easier reading. In a
related finding, consumers indicated that a list of ingredients (often
with difficult names) does not provide useful information. The study
suggests that ingredient information be organized by type, category, or
purpose, and displayed in tabular form.

6.2.5.4.3	Explicitness 

Quite a few studies have demonstrated the need for text that is explicit
and objective, reducing the need for inference. As the overview by
Laughery and Smith (2006) suggests, while brevity is helpful,
explicitness is much more important because of its influence on hazard
perception, which forms the motivational basis for compliance. Research
also suggests that explicit information is easier to remember.
Highlights of recent studies are as follows:

In a study to determine how well people comprehend environmental
warnings on household chemical labels, Akerboom and Trommelen (1998)
found that subjects had difficulty determining the relative level of
hazard indicated by the words “toxic” and “harmful,” and did not
know which of these terms was more serious. As a result, the authors
recommend using a more straightforward labeling system with words that
are concrete, specific, and more frequently used.

London (2003) found similar concerns among South African workers who
were surveyed about the comprehensibility of GHS chemical labels. In
hazard statements, workers identified several words that are too
subjective to convey a clear meaning—namely “copious,”
“adequate,” “prolonged,” and “repeated.” London suggests
replacing these terms with more concrete, quantitative terms, or—if
that is not possible—at least using more lay terms such as “lots of
water.”

Hancock et al. (2001) compared how older and younger adults comprehend
explicit and implicit warnings on products such as cleaners, paints, and
polishes. Among other results, the authors found that subjects were
generally more certain about the truth or falsity of explicit
statements, and were less confident in interpreting implicit
information.

Some studies have found that the preference for explicit text applies to
ingredients as well. For example, Edworthy et al. (2004) found that
amateur and professional pesticide users in the UK preferred labels with
explicit hazard explanations (e.g., “dangerous because of X substance
which causes…”) versus statements that simply indicate that the
product “may cause” some effect. Similarly, Heaps and Henley (1999)
found that naming the hazard-causing agent in a chemical product led to
significantly greater believability, although it did not significantly
affect recall or ratings of dangerousness. The authors suggest that the
explicit mention of hazardous ingredients serves a largely symbolic
function (i.e., adding credibility). In a related experiment, Heaps and
Henley tested implicit versus explicit statements of the worst possible
consequences of misusing a chemical product. In this case, explicitness
led to significantly greater believability, perceived dangerousness, and
cued recall among undergraduate subjects (Heaps and Henley, 1999). 

Conzola and Wogalter (1998) found that embedding explicit injury
statistics has an effect on label perceptions, too. Warnings with either
numerical statistics or a verbal quantifier (e.g., “a great number of
injuries…”) were recalled better than warnings with no statistics.
Numerical statistics also led to higher ratings of warning importance
and vividness—even if the number was clearly invalid, which suggests
that the key factor is simply having a number. Explicit statistics did
not increase the hazard rating or the likelihood of compliance, however.

Other studies lend statistical support to the notion that explicit text
is more helpful. For example, Braun and Shaver (1999) asked 178 subjects
to rate the perceived hazard of warning signs with many combinations of
variables, including three levels of text explicitness. Explicitness of
text and explicitness of symbols were two variables that accounted for
the largest part of the variability in responses. Specifically, the
authors advise that “warning text should state the severity of
possible injuries,” which will help to justify the need for
precautionary measures.

In a study on baby carriers and feeding bottles, Trommelen (1997) tested
perceptions and comprehension with three warning conditions: no warning,
non-explicit warning (instruction only, such as “keep this plastic
cover away from your child”), and explicit warning (instruction plus
hazard plus consequence, as in “keep this plastic cover away from your
child to avoid suffocation”). Subjects perceived products with
explicit warnings as more hazardous, and when asked to determine whether
an actor in a video was using the product safely or not, the subjects
given explicit warnings were correct 88 percent of the
time—significantly better than those who saw non-explicit warnings.
Explicit warnings were better understood and remembered, although they
did not necessarily increase intent to comply (perhaps because all
groups already had good intentions). Explicit warnings did not take
significantly longer to read than the other conditions.

One frequent concern among manufacturers is that a strong, explicit
warning will scare consumers away. As Dain and Studley (2004) explain in
a legal journal, “[One should] be cognizant of an additional competing
interest held by all manufacturing clients: to actually sell products.
The trick is to create warnings that can help users avoid injuries and
meet an industry standard of care, but do not scare potential users from
wanting to buy the product in the first place.” However, Heaps and
Henley (1999) found that explicit information did not make people less
likely to use a product.

6.2.5.4.4	Tone

Two of the studies identified in the literature search deal specifically
with the way label text is worded. There are many considerations related
to tone and wording, including the use of active versus passive voice,
the use of personal pronouns, and the use of probabilistic versus
definitive statements. Examples of probabilistic statements include risk
statements that say “…may cause…” or precautionary statements
like “Avoid contact with eyes or skin. Avoid taking internally.” A
definitive alternative to the latter statement would be “Do not get in
eyes or on skin. Do not swallow” (Heaps and Henley, 1999).

In a study of amateur and professional pesticide users in the UK,
Edworthy et al. (2004) adjusted the wording of hazard, instruction, and
consequence statements and then asked subjects questions to gauge their
intent to comply. Personal pronouns (e.g., “You should do this…”)
were found to improve compliance among both amateurs and professionals,
and were generally preferred over instructions that were less personal. 

Heaps and Henley (1999) tested recall and user perceptions of labels
with definitive versus probabilistic consequence statements. Definitive
statements led to a significant increase in believability and cued
recall. Tone had no significant effect on likelihood of use, perceived
dangerousness, or perceived importance of following product
instructions. The authors suggest that better recall may have the
greatest impact on actual behavior, though.

6.2.5.4.5	Bilingual concerns 

While bilingual labels are not formally addressed in the GHS, they are
an issue that continues to grow as the U.S. population becomes more
diverse. One recent study provides a number of useful insights on the
best ways to present label text in more than one language. 

For his Ph.D. dissertation, Raymond Lim surveyed English- and
Spanish-speakers about their preferences related to bilingual labels
(Lim, 2005). English-speaking participants generally agreed that
bilingual labels are important because some people in the U.S. do not
speak English, and they did not feel overall that bilingual labels
suggest a lower-quality product—debunking one myth that may have led
some manufacturers to resist using bilingual labels. English- and
Spanish-speakers were both asked to rate several configurations of a
pesticide label; variables included the proportion (same size font or
Spanish smaller than English), top/bottom versus side-by-side, and boxes
versus cylinders (to determine the influence of container shape).
Overall, both groups preferred labels with equal font sizes and the two
languages side-by-side, rather than one above the other—perhaps
because it is more natural to scan from left to right. Each group
preferred to have its own language on the left—the primary
position—but this preference was much stronger among English speakers,
suggesting that Spanish speakers did not mind having their text in a
secondary position, as long as it is present. Boxes were preferred over
cylinders, perhaps because a flatter label offers more viewable surface
area, making it easier to see all the content at once.

In the second part of the study, Lim (2005) tested search time when he
varied the location of text, color of text, and whether the text was
signaled by a U.S. or Mexican flag. Not surprisingly, search time was
fastest when only one language was present. Among bilingual designs, the
most effective configuration had both flags, which allowed users to
locate their native language at a glance. When one language was printed
in blue and the other in black, results were mixed. Overall, color and
flags together may have offered too many distracting clues.

6.2.5.5	Combinations of elements



Key Findings:



Using all of the key label elements together can generally improve
warning performance, compared with labels that only contain a subset of
these elements. However, in some situations, too much information may be
detrimental.



Among the various label elements, some studies suggest that pictorials
have the greatest influence on recall.





Some studies could not be categorized easily in the previous sections
because they tested multiple label elements simultaneously. These
studies are important because they can offer insight into the way
different label elements interact with one another. Highlights of recent
research are described below.

Wogalter et al. (1998a) asked students and community volunteers to rate
the noticeability and hazard level of two existing warning
formats—ANSI Z535.4 product labels and ANSI Z535.2 environmental
signs—as well as possible alternatives. When the same signal word was
used, the ANSI Z535.2 format was generally rated as connoting a greater
hazard than the corresponding Z535.4 sign. Perceived differences in
signal word strength were consistent across all systems. The strongest
alternative was a sign that said DEADLY with a skull icon.

Some studies have examined the relative importance of label elements.
Which items have the greatest impact? To help answer this question,
Braun and Shaver (1999) asked 178 college students to rate the perceived
hazard of several three-part warning signs (signal word, hazard symbol,
and warning text) in which the color, signal word, and explicitness of
text and symbols were varied. Explicitness of text and symbols accounted
for some of the variability in perceived hazard, but color and signal
word did not. In other words, with all other elements being equal, a red
warning was not perceived as significantly stronger than a blue warning,
and a different signal word did not significantly sway the user’s
perceptions.

In a similar study, Meingast (2001) asked students to recall information
from two variations of warning labels: enhanced warnings with color,
signal icons, pictorials, and organized text (following the ANSI Z535.4
standard); and warnings with text only. Among subjects who viewed the
enhanced warnings, the pictorial was the element most frequently
recalled (48 percent of what was recalled). This study also validated
the overall utility of enhanced warnings, finding that the more complete
warnings were more noticeable, led to significantly greater recall, and
made people report a higher likelihood of compliance.

Other findings generally agree that improving all label elements can
improve warning performance. For example, Lehto (1998a) tested
information retrieval from three chemical label formats and found that
subjects generally did best with an “extensive” format that included
pictograms, paragraphs, and horizontal bars indicating the degree of
hazard. Subjects were able to answer more questions correctly when the
label included a range of well-organized content—particularly
information on first aid and spill procedures.

Wogalter et al. (1997a) reported similar results in a test of four
different signs that discouraged people from using an elevator for short
trips. Three signs were text-only, taken from a 1965 study. The fourth
sign had a signal word panel, icons, a pictorial, and more explicit
wording of the desired behavior (i.e., “use the stairs”). Subjects
rated the enhanced sign as more understandable, and a field test found
that it significantly increased compliance over the other options.

Adams et al. (1998) asked a group of students and industrial workers to
react to a variety of warning signs from industrial settings. Signs were
sometimes presented as full versions with signal word, hazard statement,
consequences statement, and instructions statement; other times, they
were presented with one or more of these elements missing. Both groups
recognized the signal word, consequences, and instructions most of the
time they appeared. Overall, though, when asked about the likelihood
that the sign would result in the desired behavior, subjects rated the
full sign as less effective than some of the variations with one or more
elements missing. Here, the authors suggest that instructions may be
less effective when part of a more complete presentation, and thus they
recommend simplicity and clarity, taking into account what the target
audience may already know. The review by Sattler et al. (1997) reached a
similar conclusion: 

“Widely encountered guidelines that warnings should comprise four
elements—a standard signal word, statements of the hazard, the
potential consequences, and how to avoid the hazard—are primarily
based on preference studies and may be in error to suggest that all of
this information must be provided in order to elicit the appropriate
response.”

6.2.6	Warning design

6.2.6.1	General principles



Key Findings:



Many resources are available to advise people on basic principles of
warning design.



In general, labels are more effective when they include enhancements
such as color, borders, and symbols. 



Proper font and spacing are important, as is color contrast. Some common
recommendations include black text on white background, a 10-point
sans-serif font, and bold type for the most important items.





The field of warning design is continually evolving. Recently, several
authors have reviewed the literature to develop guidance for warning
designers. Johnson (2006) describes general design concepts for
warnings, while Frascara (2006) provides advice on typography and visual
design. Wogalter et al. (2002) also offer design guidelines, based on a
review of empirical literature.

A few general principles are worth noting because they appear
consistently in the literature. In general, characteristics that enhance
“vividness”—font size, color, spacing, level of specificity, and
symbols—have been found to improve attention, even though some studies
have shown that these characteristics do not have a significant effect
on comprehension or recall (Argo and Main, 2004). In particular, reviews
of the literature suggest enhancing conspicuity through larger and
bolder print, greater color contrast, and highlighting (Parsons et al.,
1999). Wogalter et al. (2006b) provide similar advice, noting that
larger symbols tend to work better than smaller ones, and thick, bold
elements are more legible than thin small details. Researchers also note
that bullets in outline form appear to be preferred over continuous text
(Wogalter et al., 2002). 

Rudin-Brown et al. (2004) offer additional design recommendations, based
on a study that gauged the effectiveness of four different label
conditions for child restraint systems to be installed in a vehicle.
Some subjects saw no label, some saw the current manufacturer’s label,
some saw labels that reflected current and proposed Canadian and U.S.
regulations, and some saw “optimal” labels, which were designed
based on a hierarchical behavioral task analysis. The most effective
labels were those with a 10-point sans-serif font, black text on a white
background (for maximum contrast), color-coded arrows, and
illustrations. On some tasks, subjects were more successful with no
label than with the less optimal labels, suggesting that a poorly
designed label can actually interfere with performance.

In an earlier study that was not cited by Sattler et al. (1997),
Laughery et al. (1993) tested warnings on alcoholic beverage containers
to determine which elements influenced noticeability. The authors
manipulated labels by adding a pictorial, adding an alert symbol with a
signal word, making the text red, and/or adding a border around the
warning. The warning was located fastest when all four of these
modifications were present, suggesting that the best designs include a
combination of enhancements.

6.2.6.2	Color



Key Findings:



Color can make warnings more noticeable and can help organize the
content (i.e., through color coding).



Red is generally perceived to reflect the greatest degree of hazard.
Yellow, orange, and black reflect a lesser degree of hazard, but studies
disagree on the relative ranking among these three colors. 



When using color, label designers need to consider other factors such as
color blindness and lighting conditions.





In their review of the literature on warning effectiveness on behavioral
compliance, Kalsher and Williams (2006) summarize several studies that
examined the effects of adding color to warnings. Many of these studies
were cited in Sattler et al. (1997). Overall, Kalsher and Williams
suggest that adding color can influence the noticeability and
effectiveness of warnings—a claim that is generally supported by
studies conducted since 1997.

In a test on the noticeability of warnings, Swindell (1999) measured the
amount of time it took subjects to locate warning text that had been
embedded in medication instructions. Warnings were found significantly
faster when the icon and signal word were presented in either red or
blue, causing the warning to “pop out” from the black text. Colored
text had a greater impact on noticeability than the presence or absence
of an icon. Swindell’s findings echo the results reported by Laughery
et al. (1993), who found that alcoholic beverage labels were located
significantly faster when the text was red instead of black.

Color also can aid in organizing label information. In the child
restraint system study previously described, Rudin-Brown et al. (2004)
found that color-coded warnings improved compliance. In this particular
case, color was used to differentiate between instructions for
rear-facing installation and forward-facing installation, an important
organizational distinction.

In addition to its effects on noticeability and organization, color can
be used to connote a particular meaning. One of the most common uses of
color in warnings involves signal words. The ANSI Z535.4 standard for
product safety signs and labels prescribes the use of color in signal
word panels as follows: DANGER in white letters on red background,
WARNING in black letters on orange background, and CAUTION in black
letters on yellow background (ANSI, 2002c). Many other standards use
similar arrangements. The ANSI Z129.1 standard for chemical labels does
not require colored signal word panels, and the GHS does not either;
however, the GHS does allow for the use of color in signal words,
backgrounds, and other elements as required by the appropriate
regulatory authority. The GHS does prescribe the colors to be used in
the standard pictograms (United Nations, 2005).

Many researchers have investigated the hazard connotations of different
colors. In a review of the literature, Parsons et al. (1999) suggest
that the red-orange-yellow hierarchy used by the ANSI standard generally
matches people’s perceptions of risk, including perceptions among
native Spanish speakers and children as young as 8. 

Experimental results since 1997 confirm that red generally connotes the
highest degree of hazard, but among the remaining colors, there is more
disagreement. For example:

Smith-Jackson and Wogalter (2000) asked English-speaking community
members to rate the perceived hazard of ten ANSI safety colors. Red,
yellow, black, and orange were rated the highest (in descending order).
Differences were statistically significant except the difference between
yellow and black. 

Wogalter et al. (1997b) asked Spanish speakers to rank the perceived
hazard of ANSI safety colors. Red was ranked highest, followed by
orange, black, and yellow.

Wogalter et al. (1998a) asked undergraduates and community volunteers to
rank various warning components. Red connoted a significantly greater
hazard than other colors, followed by yellow, orange, and black (in that
order). A group of industrial workers ranked the colors from greatest to
least hazard as follows: red, yellow, black, orange.

London (2003) asked workers in four sectors in South Africa to rank the
colors red, yellow, green, and blue in terns of perceived hazard; 95
percent said red represents the greatest hazard, and 58 percent said
yellow is the second greatest hazard.

Banda and Sichilongo (2006) asked workers in Zambia to rate the
perceived hazard of various colors used in chemical labels. Red was
associated with the greatest hazard, followed by yellow.

In a related experiment, Griffith and Leonard (1997) asked college
students to rate colors on nine scales related to three factors:
evaluation, activity, and potency. Red was rated the most
“meaningful” color (i.e., most distinct in meaning from neutral
gray), followed by green, orange, black, white, blue, and yellow. 

Griffith and Leonard (1997) also conducted an association test to see
what signal words people associate with specific colors, and vice-versa.
The authors found that DANGER-red and CAUTION-yellow have the strongest
associations; in each case, the color evoked the correct signal word
nearly 80 percent of the time. The pairing of orange with WARNING was
much weaker. Black was most often associated with FATAL and POISON. The
authors theorize that red is a particularly strong warning color because
of its association with stop signs and blood.

Label designers should consider a few other factors when selecting
colors, too. For example, Dain and Studley (2004) advise designers to
consider color-blind or color-deficient users, who may not be able to
distinguish certain color combinations. Dain and Studley also urge
designers to consider environmental conditions—e.g., avoid using a
dark label if it will be read in a dimly lit setting. Other
considerations include fluorescence, glare, and the durability of
colored pigments.

6.2.6.3	Layout and formatting



Key Findings:



Logical headings convey information more effectively.



Font size and white space are important variables that affect
readability. 



Wide, colorful, and jagged borders are generally perceived to attract
attention better than thin or missing borders, but borders may not be
the most important factor in attracting attention overall.





Research on the layout and formatting of warning labels suggests that
the most effective labels are those that use logical headings and
minimize clutter. As Conzola and Wogalter (2001) state, “Visual
warnings that are formatted to be aesthetically pleasing, with plenty of
white space and coherent information groupings (Hartley, 1994), are more
likely to attract and hold attention than warnings without these
features (Vigilante and Wogalter, [1998a]).” Borders may aid in
recognition, too. Wogalter and Vigilante (2006c) summarize a number of
empirical findings, including the fact that white space helps to show
organization. The GHS does not prescribe a particular format or layout
for workplace chemical labels; just the set of elements that must be
included (OSHA, 2006b). Thus, the GHS offers regulatory agencies and/or
manufacturers the flexibility to adopt formatting that they feel is most
appropriate.

In a timed information retrieval test, Lehto (1998a) found that larger
labels decreased the accuracy of information retrieved, perhaps because
they discouraged subjects from consulting other sources (e.g., the SDS)
for information. Subjects did express a preference for labels with
logical headings. Wogalter and Vigilante (2003) measured information
recall and users’ preferences for labels on over-the-counter
medication bottles; variables included font size (4-, 7-, or 10-point),
amount of white space (none, section spacing, or line spacing), and
standard label size versus an extended/pull-out format. Although font
size did not affect recall among younger adults, older subjects
performed significantly better with 7- or 10-point text, versus 4-point.
The amount of white space did not have a substantial effect on knowledge
acquisition, but both groups rated the design with line spacing as the
most readable, followed by the design with spacing between sections. The
extended format helped older adults because it could accommodate larger
print.

In a review of literature on warnings, Parsons et al. (1999) note that
wide, colorful, and jagged borders are generally perceived to be most
effective. In one recent study in this area, Wogalter and Rashid (1998c)
observed how often (and for how long) people looked at warning signs
with different types of borders. Signs with thick red borders or thick
yellow/black diagonal striped borders were noticed more frequently and
examined more thoroughly than signs with a thin black or red border or
with no border at all. In an earlier study on alcoholic beverage
warnings, however, Laughery et al. (1993) found that a border led to
only a marginal decrease in the amount of time it took to locate the
warning on a container. Instead, the factors that seemed to affect
noticeability the most were heading length, the number of lines and
letters per line, and the amount of clutter around the warning (less
clutter is better). 

6.2.6.4	Order



Key Findings:



Presenting a hazard and the corresponding precautions together may
improve understanding.



Warnings related to immediate, specific hazards are generally considered
most important. 



The most effective order for warnings may depend on audience
characteristics such as familiarity with the product.





The GHS does not specify the order in which label elements should
appear. However, a few recent studies have found that the order of
presentation can affect comprehension and compliance. For example,
Akerboom and Trommelen (1998) found that combining environmental and
safety information on household chemical labels led to some confusion
during a comprehension exercise; thus, the authors recommend that each
potential hazard be presented together with the corresponding
precautions (and possibly pictograms), so it is clear why each
precaution must be taken. The authors also suggest presenting warning
messages in a “logical” order.

To find out what a “logical” order means, Vigilante and Wogalter
(1997) asked 25 subjects to look at a list of safety warnings for power
tools and place the warnings in order, starting with those that were
most critical. In this experiment, people tended to put specific actions
near the top of the list, particularly warnings that would be most
critical for the initial use of the product (e.g., correct way to
operate the tool, protective equipment, and electric shock hazard).
Warnings related to transport, storage, and maintenance—i.e., later
use of the product—were ranked lower. The preferred order also
corresponded with how another group of subjects ranked the warnings in
terms of importance, injury severity, and likelihood of injury.

As a follow-up to the previous study, Vigilante and Wogalter (1998b)
tested how well people recalled a set of warnings for a power tool,
depending on warning sequence. Males performed best when warnings were
presented with the least obvious hazards first; females did best when
the most obvious hazards were listed first (obviousness was determined
through a pilot study).The second most effective order was by
importance, as defined by Vigilante and Wogalter (1997). This study
suggests that the best ordering may depend on the audience. In this
case, the authors suggest that male users are probably more familiar
with power tools, and therefore only the non-obvious hazards need to be
displayed prominently.

6.2.6.5	Location



Key Findings:



Warnings tend to be noticed more quickly when they are located in the
upper left portion of an instruction sheet.



Safety precautions are more likely to be followed when they are
incorporated into a product’s directions for use, although warnings
may be better noticed in general if they are included on posters or
signs, rather than on the product.



Cluttered locations make warnings less noticeable.





The location of warning labels can have a great effect on whether they
are noticed or read. Studies on the influence of warning location are
summarized by Kalsher and Williams (2006); many of these studies were
also cited in Sattler et al. (1997). Since 1997, a few additional
studies—summarized below—have examined the location of warnings
within a product label or within the user’s total environment (e.g.,
signs versus instructions versus on-product labels).

Swindell (1999) studied several factors that may affect warning
noticeability, including where the warning appears within a two-column
product instruction sheet. Warnings in the left column were found
significantly faster than warnings on the right, and warnings in the top
or middle of a column were located faster than those at the bottom.
These findings are consistent with the way people read English or
Spanish—left to right, top to bottom. These results build on findings
reported in Laughery et al. (1993) regarding alcoholic beverage
warnings, which were significantly more noticeable when located on the
front of the container (as opposed to the sides or back) and when the
text was printed horizontally. Vertical text was significantly easier to
find when it read “up” rather than “down,” and overall label
clutter decreased noticeability—confirming a commonly held belief.

In a study of amateur and professional pesticide users in the UK,
Edworthy et al. (2004) found that both groups of users complied with
safety precautions most often when these warnings were included in the
directions for use. This location was found to be more effective than
warnings in a “precautions” section, a “statutory conditions”
section, or an additional leaflet. With professional users, compliance
was also higher when a supplemental directive was present, pointing them
to where safety instructions were located.

Argo and Main (2004) analyzed results from 48 warning studies published
between 1975 and 2001. Among other things, the authors found that
warnings are more effective in attracting attention when they are on
posters, signs, and/or advertisements, as opposed to on the product
itself. It is not clear whether this finding extends to situations
involving occupational chemical use, however. For example, in a
chemistry lab task, Wogalter et al. (1998b) found that subjects were
more likely to use PPE when a safety warning was included in an
instruction sheet, rather than a sign posted on the wall.

6.2.7	Receiver characteristics

Some of the factors that influence the effectiveness of warnings have to
do not with the warning itself, but with the audience. These audience
variables—or receiver characteristics—include gender, age,
familiarity with the product, and the cost of compliance. Although
receiver characteristics are out of the label designer’s control, they
are still worth considering because it may be possible to adapt the
label design to better accommodate the needs and characteristics of a
particular group of users. Recent research on receiver
characteristics—and how they may impact the way we think about label
design—are summarized in the sections that follow.

6.2.7.1	Gender



Key Findings:



Research has found very few gender differences related to warning
comprehension or compliance.



Warning designers probably do not need to consider gender factors unless
they are creating labels for gender-specific products.





Many studies on warning labels have looked at how gender might have
influenced the results. Overall, as Parsons et al. (1999) explain,
“Most research has failed to find (or report) gender differences on
warning related measures.” A review by Wogalter et al. (2002)
reiterates this assertion, noting that gender factors tend to be
relevant only when the warning is for gender-specific products. 

The search for studies from 1997 and later turned up two examples of
gender differences in warning perception. In a survey of 1,500 adults
and children in the UK, Hellier et al. (2000a) asked participants to
rate three signal words on a hazard scale from 1 to 10. Males and
females gave the same order (DANGER, WARNING, CAUTION), but for all
three words, females gave higher hazard ratings than males. The other
example of gender difference comes from Vigilante and Wogalter (1998b),
who found that when multiple warnings are present on a safety label, the
order that optimizes comprehension varies by gender. Males recalled more
information when hazards were ordered from least to most obvious, while
females performed better when the most obvious hazards were listed
first. The authors surmise that this effect may be due more to
familiarity than to gender, as male subjects were more likely to be
familiar with the product (a power tool) and would thus be more engaged
by a presentation that starts with the least recognizable hazards.

6.2.7.2	Age



Key Findings:



In general, children and adults have similar perceptions of the relative
hazard conveyed by colors and signal words.



Most studies have found that younger adults comprehend warning text and
symbols better than older adults do.



Many cognitive functions decline with age. Understanding these
age-related declines in ability can help warning designers create labels
that are better understood by older adults.





In a study that asked 1,500 adults and children in the UK to rate the
degree of hazard implied by the signal words DANGER, WARNING, and
CAUTION, Hellier et al. (2000a) found strong similarities among the
rankings from children and adults. When the data are broken down,
smaller age ranges demonstrate the same pattern as well, with the
exception of 15-to-24-year-olds, who showed more variability—perhaps,
as the authors surmise, because this age group has been confused or
desensitized by so many different warnings regarding drugs, alcohol,
safe sex, etc. Overall, though, the study shows that adults and children
agree on hazard associations. Other studies have shown that adults and
children have similar hazard perceptions for warning colors, too
(Parsons et al., 1999).

Far more research, however, has been devoted to differences in warning
effectiveness between younger adults and older adults. The effects of
aging are of particular concern as the American workforce—i.e., those
who the HCS is designed to protect—becomes older.

Studies generally show that label comprehension declines with age. For
example, Sansgiry et al. (1997) tested how well consumers process
information on over-the-counter drug labels, and found that older adults
scored significantly lower than younger adults on a comprehension test.
In another test on medication instructions, Sojourner and Wogalter
(1998) found that older adults (mean age 68.0) recalled only 50 percent
as much information as undergraduates did. Younger adults (mean age
33.6) performed better than older adults but not as well as
undergraduates. Wogalter and Vigilante (2003) found that younger adults
acquired information from a medication bottle significantly better than
older adults no matter what label format was used. Older people also
took longer to answer the comprehension questionnaire.

Schroeder et al. (2001) studied age effects in understanding 40 ANSI
safety symbols. For all four types of symbols (mandatory action,
prohibitory action, information, and hazard alerting), younger adults
had significantly higher average comprehension than older adults. In a
similar study, Hancock et al. (1999) tested 12 safety symbols and
discovered that older adults had poorer comprehension for the alert,
hand protection, no lit matches, no people, flammable, and slip symbols.
However, older adults still said that symbols were helpful in
understanding warnings.

In another study on symbols, Lesch (2004) examined how well younger (age
18-35) and older (age 50-65) individuals comprehended warning symbols
before and after training. Of 31 symbols, 15 had significant age-related
differences in comprehension before training—in all cases, older
adults understood less. Younger people were not more familiar with these
symbols, so the author suggests that the difference is more likely due
to how effectively the symbols cue other knowledge. These results build
upon previous studies which found that young people comprehended
chemical and physical safety symbols better than older adults both
before and after training (Lesch, 2002, 2003).

Among recent studies, at least one did not find a decline in
comprehension with age. To compare how older and younger adults
comprehend explicit and implicit information in consumer product
warnings, Hancock et al. (2001) asked two groups of subjects to read
warnings for household chemical products, then rate the truth or falsity
of test statements. Older and younger adults had similar comprehension
levels overall. One noticeable difference was that older people tended
to rate false statements as more “false” than younger people did,
perhaps because their greater life experience made them more certain
about their answer.

Although several studies have shown reduced label comprehension among
older adults, there may be some room for improvement through label
designs that account for what science has learned about the aging
process. Psychology literature provides a large and growing list of ways
in which cognitive processes change with age—changes that affect every
one of the steps in processing warnings (notice, encode, comprehend, and
comply) (Rousseau et al., 1998). Rousseau et al. (1998) provide the
following list of age effects from the literature, along with specific
recommendations for label designers:

Older adults have more difficulty discriminating between colors,
especially those with short wavelengths (green, blue, and violet).
Labels should avoid using these colors to differentiate items.

Older adults have more difficulty discriminating between contrasting
patterns or areas of darkness. Labels should offer high contrast between
text and background.

Older adults have increased susceptibility to glare. Avoid placing
warnings on surfaces that produce glare.

Older adults have reduced visual acuity (ability to resolve small
details). Label designers should use 12- or 14-point type with a
sans-serif, medium to bold font, and should avoid text in all capital
letters.

Older adults have more difficulty selecting relevant information from a
visual array, as a person’s useful field of view (UFOV) declines
rapidly with age. Designers should eliminate clutter, place important
information near the focal point, and use a 1:100 ratio of letter size
to distance for signs.

Older adults have reduced working memory, which can be a particular
problem if a warning requires multiple steps to comply safely. Warning
designers should provide cues to aid memory—for example, by putting
warning information on the product to cue recall.

Older adults have a decreased ability to process complicated text.
Labels should use simple, explicit sentences.

Older adults show a decline in prospective memory (i.e., remembering to
perform actions in the future). Warnings should be available when
needed.

Empirically, older adults have difficulty comprehending the information
in symbols, although the reasons why are unclear. This effect may be
related to working memory. Symbol designers should include older adults
in pilot testing.

Empirical evidence suggests that many of the age-related cognitive
factors listed by Rousseau et al. (1998) do indeed affect older
adults’ performance in reading and complying with warning labels. For
example, Parsons et al. (1999) reviewed the literature and found that
older adults may have difficulty establishing links between symbols, and
they are strongly affected by legibility. Wogalter and Vigilante (2003)
found that older adults recalled significantly more information from a
drug label when the font was 7- or 10-point, rather than 4-point. After
finding that older adults had more trouble than younger adults
comprehending abstract warning symbols, Lesch (2004) suggests that older
adults may have a greater need to trigger context-specific knowledge,
which they are more likely to have for symbols that are less abstract. 

6.2.7.3	Product familiarity



Key Findings:



Familiarity with a product can make users more familiar with the proper
precautions, and therefore more likely to comply with warnings. In other
cases, however, familiarity might have the opposite effect.



Familiarity with a label might make users less likely to read it
carefully.



Familiarity with a particular type of container can affect hazard
perceptions.





In some cases, familiarity with a product may increase compliance. In
other cases, however, familiarity may make a user less likely to read
instructions or safety labels, or more comfortable using the product
without taking the recommended precautions. Kalsher and Williams (2006)
summarize several pre-1997 studies on the effects of product familiarity
on compliance. Three more recent studies shed further light on the
relationship between warning labels and familiarity.

After conducting a meta-analysis of 48 warning studies to date, Argo and
Main (2004) conclude that familiarity has a positive effect on
compliance with safety instructions. When users have previously
interacted with a product, they suggest, those users have had more
opportunity to learn appropriate safety behaviors. Edworthy et al.
(2004) found a similar effect among pesticide users in the UK. Overall,
professional users complied with 60 percent of safety directives,
compared with 34 percent for amateurs. The authors theorize that people
who use the product every day may be more likely to use safe procedures
because they are subject to regulatory requirements and they may be more
concerned with their legal liability, personal safety, and job security.

Edworthy et al. (2004) draw a distinction between familiarity with the
product and familiarity with the label, however. In the pesticide user
study described above, compliance was lowest when safety information
appeared in the part of the label where people would most expect to find
it. Compliance was greatest when that information was moved to the
directions for use. The authors suggest that while product familiarity
increases safety knowledge, familiarity with the label makes people less
likely to read it carefully—a case in which “familiarity breeds
contempt.”

In a somewhat related study, Meingast (2001) asked subjects to recall
the contents of safety labels which were manipulated by placing them on
different types of containers, among other things. Container type did
not affect recall, but it did affect perceptions of hazard
severity—perhaps because of users’ familiarity with other products
in these types of containers (e.g., boxes versus bottles). Argo and Main
(2004) agree that product type can affect judgments.

6.2.7.4	Cost of compliance



Key Findings:



Cost of compliance is widely recognized as a factor that affects
behavioral compliance with warning labels. Important considerations
include the amount of time, money, and effort required to comply with an
instruction. 





Another factor that can affect compliance with safety instructions is
the cost of compliance. Kalsher and Williams (2006) explain this concept
well:

“Costs associated with compliance include expenditures of time, money,
or effort. When people believe that following the proscribed behavior
will take too long, cause too much discomfort, or is too expensive, they
will be less likely to heed the warning.”

Kalsher and Williams (2006) provide a useful table of studies on cost of
compliance, while Argo and Main’s (2004) meta-analysis of previous
studies agrees that cost of compliance can have a significant effect on
behavior. However, a search of the literature published since 1997
produced few recent studies on cost of compliance. In one study that was
available, Chunin (2002) conducted a safety attitude survey among 91
workers and 13 foremen in chemical factories in Thailand, and found that
likelihood of complying with warning signs (e.g., instructions to wear
PPE) depended on perceived cost of compliance. A follow-up survey found
that rewards and punishment also had a significant effect on compliance
with safety signs.

6.2.8	Related topics

Given the relatively robust body of literature on visual warnings, many
recent studies have ventured into the newer territory of multi-modal
warnings (i.e., warnings that appeal to multiple senses). Because the
HCS and GHS pertain to visual warnings, we will not attempt to describe
the literature on multi-modal warnings in great detail here. It may be
worth noting some key findings, however. 

Kalsher et al. (1997) examined the extent to which subjects complied
with safety warnings on a bottle of glue. One condition involved a
raised border around the safety warning, which was designed to attract
attention through vision and touch. The raised border led to higher
compliance than a safety label with no raised border, but the difference
was statistically insignificant. When asked their opinions, 94 percent
of subjects said the label with the raised border was most effective at
conveying information, and 55 percent thought it had the most consumer
appeal, compared with the nonraised border or the control product (no
warning).

Cohen et al. (2006) describe the latest research on interactive
warnings, audio warnings, and warnings that involve the other senses
(tactile, olfactory, and gustatory). For example, the authors note that
raised borders can help visually impaired users recognize a potential
hazard.

Other studies have examined the role of stress in warning compliance.
Research suggests that in general, compliance with precautionary
measures is reduced when the user is under pressure to complete a task
in a limited amount of time. For example, in a study that asked
participants to measure and mix chemicals in a laboratory setting,
Wogalter et al. (1998b) found that subjects given a 5 minute time limit
were significantly less likely to follow safety instructions (wearing
PPE) than subjects who were given no time limit to complete the task.
Wogalter et al. also tested another form of stress: social monitoring.
In this study, placing an evaluator with a clipboard next to the
participant (i.e., higher social stress) led to a small but
statistically non-significant increase in warning compliance. 

6.2.9	Research needs

In addition to reporting findings, the documents obtained through this
literature search contain a number of recommendations for future
research. The list below is by no means complete, but instead represents
a partial list of key suggestions:

Rousseau et al. (1998) suggest assessing the costs and benefits of
standardization, which may facilitate visual search but cause
habituation to the warning message. In particular, Drake et al. (1998)
suggest a need for more information about habituation to signal words.
This area of research will be increasingly important as more regulatory
bodies transition to the international standard set of signal words.

Fagotto and Fung (2002) recommend that after introducing harmonized
labels and SDSs, OSHA should test their comprehensibility and impact in
changing workers’ behavior. If the new materials are not effective,
stronger emphasis on training may be needed.

Based on a review of warnings research as of 1999, Parsons et al. (1999)
suggest additional research in the following areas:

In general, conduct more measurement of actual behavioral compliance.

For cases where actual behavior cannot be measured (e.g., because it
would put the subject in danger), confirm the utility of alternative
measures.

Consider virtual reality experiments to simulate actual behavior.

Consider effects of state (alcohol, drugs, fatigue, stress) and
personality factors on warning effectiveness.

Involve a broader range of participants, in terms of both age and
culture. 

Based on another literature review, Rogers et al. (2000) note that there
is little empirical evidence to date to evaluate potentially important
variables such as attention fatigue, working memory, vision, motivation,
reading comprehension ability, intelligence, technical knowledge,
ability to comply, decision making, physical quality of labels (e.g.,
degradation), legibility, and credibility of source. In addition, more
research is needed on the relative importance of different variables, as
well as interactions among variables.

7.	Training

7.1	Background

In the current HCS, OSHA requires employers to provide employees with
“effective information and training” on hazardous chemicals (OSHA,
1994). The HCS describes what elements must be addressed, but does not
tell employers exactly how to conduct the training. OSHA’s Training
Institute has prepared some voluntary guidelines that are applicable to
training in general (OSHA, 1998).

The GHS does not prescribe specific requirements for hazard
communication training. However, the GHS does note that worker training
is an integral part of the hazard communication, and it suggests that
training requirements be “appropriate for and commensurate with the
nature of the work or exposure” (United Nations, 2005, § 1.4.9). As
the next section explains, research on hazard communication continues to
support the role of training in promoting worker health and safety.
Consensus standards support this view as well. For example, Annex A of
ANSI Z535.2 recommends training on the meaning of standard safety
symbols and signal words; ANSI Z535.4 contains similar guidance.
Countries in Europe have adopted European Commission Directive 98/24/EC,
which includes general provisions for training workers who deal with
hazardous chemicals (European Commission, 1998).

For more specific guidance, employers can consult other resources. For
example, the American Society of Safety Engineers (ASSE) has developed
ANSI/ASSE Z490.1, the American National Standard on Criteria for
Accepted Practices in Safety, Health, and Environmental Training (ASSE,
2001). ANSI/ASSE Z490.1 provides requirements and guidance in the
following areas:

Program administration and management

Program development (e.g., course design)

Delivery (e.g., trainer criteria and the training environment)

Evaluation

General systems and procedures

7.2	State of the science

7.2.1	Literature reviews

Sattler et al. (1997) reviewed health and safety training literature
published up to 1997. Among other things, the studies reviewed in that
document included successful training intervention studies and research
on the many factors that can affect training performance, such as the
degree of support from plant management. Sattler et al. also suggest
that literacy and language issues require more investigation.

Since 1997, a few additional reviews of the training literature have
been published. In 1998, NIOSH published an update to a previous review
on occupational health and safety training, adding studies conducted
through 1996 (Cohen and Colligan, 1998). Bouchard (2007) reviewed nine
training intervention studies conducted between 1983 and 2005; however,
most of these studies were already cited by Sattler et al. (1997).

In a related study, Fung et al. (2004) reviewed the literature to
determine the effectiveness of several “transparency systems,”
including OSHA’s HCS. Overall, the authors rate the HCS as moderately
effective, meaning that “The transparency system has changed behavior
of a substantial portion of users and disclosers in the intended
direction but has also left gaps in behavior change and/or generated
unintended consequences.” Fung et al. note that improved access to
information has helped some employers switch to safer alternatives, but
cite studies showing that this information is not well embedded in
employees’ decision-making, and that SDSs are of varying quality and
are often too complex. As described previously in the section on SDSs,
Fung et al. sum up their assessment as follows:

“Overall, the hazard communication system appears to function better
as a tool to exchange information among chemical producers and chemical
users than as a device to help employees to protect themselves at work,
avoid dangerous workplaces, or demand higher pay in light of increased
risk.”

7.2.2	The need for training



Key Findings:



Training will continue to be a need under the GHS, considering that even
some common label elements may not be widely understood.



Worker surveys confirm that training can lead to safer practices.





Many of the recent studies on SDSs and warning labels have included some
type of training component—for example, testing comprehension of
pictograms both before and after a brief training session. Nearly all of
these studies demonstrate some type of improvement due to training;
indeed, several authors specifically recommend training as a tool to
increase comprehension and compliance. 

Training will continue to be a need under the GHS. For example, in a
test of GHS labels in Zambia, Banda and Sichilongo (2006) found that
“…correct responses to label elements were not a result of social
class and/or age but appeared to be influenced by extent of duration of
exposure either through specialized training or acquaintance.” In a
similar study in South Africa, Dowse and Ehlers (2003) found that some
subjects were unfamiliar with fundamental pictogram elements like the
prohibition (slash) symbol—particularly less educated users. Some
subjects exhibited critical confusion: that is, interpreting the
opposite of the intended meaning. As the authors explain, these results
show the importance of basic training to make people more familiar with
safety symbols.

Worker surveys have illustrated the importance of training, too. For
example, in a survey of 180 emergency responders who were trained at a
hazardous materials worker training center, Weidner et al. (1998) found
that 42 percent of respondents had experienced at least one incident
that would have resulted in injury or death without training.

7.2.3	Training intervention studies



Key Findings:



Overall, training has been shown to improve safety attitudes and safety
behavior in the workplace.



In studies on symbol comprehension, training generally leads to greater
understanding. A simple explanation of each symbol is just as effective
as a longer explanation.





Overall, a wide range of studies have shown that training is effective
at improving worker safety and health. NIOSH sums up these findings in
its 1998 literature review, stating, “…evidence that OS&H training
can reduce risks from workplace hazards remains strong. Indeed, the
issue is not so much whether OS&H training is worthwhile but what
factors both within and beyond the training process can produce the
greatest possible impact” (Cohen and Colligan, 1998). The NIOSH review
also provides examples showing the success of OSHA’s voluntary
training guidelines in action. In an appendix, the authors provide
results from training intervention studies, essentially representing the
state of the science as of 1998 (Cohen and Colligan, 1998). Many of
these studies were already described by Sattler et al. (1997).

As Cohen and Colligan (1998) note, the effectiveness of training depends
on many factors, including size of group, length/frequency of training,
training mode, transfer, motivation, trainer qualifications, management
role (apparent benefits from using supervisors or foremen as trainers),
etc. Unfortunately, the literature published in the last decade does not
contain much more information about these factors. There have been some
additional studies showing the effectiveness of training intervention,
however.

Tan-Wilhelm et al. (2000) compared attitudes and behavior among workers
at two plants that handle beryllium. At the intervention plant, a
presentation on beryllium hazards was given to all employees, along with
two information bulletins plus reminder stickers and posters. These
materials recommended that employees reduce exposure by washing their
hands, changing clothes before leaving the plant, and vacuuming their
cars regularly. The control plant did not receive these materials.
Although workers at both plants had high levels of perceived severity,
the intervention subjects reported stronger perceptions of
susceptibility, higher self-efficacy (believing they can take steps to
reduce exposures), and higher response efficacy (believing these steps
will be effective). They also had more positive attitudes toward safety
measures and stronger intentions to engage in safety behaviors. One
month after training, workers at the intervention plant reported
significantly higher rates of hand-washing and vacuuming cars, compared
with the control plant. These self-reported results were supported by a
significant increase in hand soap use. Workers reported that the
plant-wide presentations were the most influential part of the training.

Weidner et al. (1998) surveyed 180 emergency responders who attended a
hazardous materials worker training center funded by the National
Institute of Environmental Health Sciences (NIEHS). After training, at
least 90 percent of trainees felt confident in their ability to recall
specific critical concepts in a crisis. Respondents rated technical
topics (hazard recognition, decontamination, protective equipment) as
the most critical areas for training. Workers recommended that training
be extensive and uniform across jurisdictions, emphasize technical
aspects of health and safety, and include demonstration and hands-on
techniques. The survey results also suggest that integrated
organizational support is critical for implementing health and safety
practices.

Following a review of nine training intervention studies, Bouchard
(2007) identified three main problems with current hazard communication
training:

The lack of learner involvement to improve hazard communication.

The lack of employer assessment of employee understanding of training
provided.

The lack of studies assessing retention of the material taught and its
application at the worksite.

In addition to the studies described above, there have been other recent
investigations which, while not truly training intervention studies,
nonetheless provide useful observations about the role of training in
improving warning comprehension and compliance.

In one such study, Wogalter et al. (1997c) tested how well undergraduate
subjects comprehended a set of 40 pharmaceutical and industrial safety
pictorials before and after training. Training led to a significant
increase in pictorial comprehension, even up to 6 months later. The
improvement was greatest for the most complex symbols. Training was
equally effective whether the subject was given a simple printed label
(e.g., “Danger, cancer-causing substance”) or a label with
additional explanatory text.

Lesch (2002, 2003) conducted a similar study, testing how well workers
recognized a set of 31 chemical and physical safety symbols before and
after training. Training significantly improved comprehension, which
remained higher up to 8 weeks later. Younger and older adults showed
similar amounts of improvement due to training, although younger adults
performed better overall both before and after. As in the Wogalter et
al. study described above, Lesch found little difference in performance
whether training took the form of a written label assigned to each
symbol, a label plus explanatory text, or an accident scenario. Training
also improved response speed.

In a survey of South African workers, London (2003) examined the impact
of brief training on the meaning of symbols and hazard phrases. Here,
the author found no statistical difference in comprehensibility of four
familiar hazard symbols, but did find that training improved
comprehension of one “new” symbol (chronic hazard), and it also
reduced the overall incidence of critical confusion. This study also
demonstrated the benefits of prior workplace training. The author found
that workers with previous training were five times more likely than
others to recognize the SDS as a source of hazard information, and were
also more likely to understand label words and some pictograms and be
able to identify the active ingredient.

7.2.4	Models and recommendations



Key Findings:



In one case, a mandatory training program was more cost-effective and
reached a wider audience than a voluntary alternative. 



Training should consider the needs and characteristics of the audience.
Training that accounts for cultural and literacy differences can not
only reduce injuries, but also improve productivity and reduce turnover.



Recent literature includes a number of recommendations for more
effective training.





When designing a hazard communication training program, it helps to
consider what others have done to achieve success. The literature
published since 1997 contains several examples of effective training
programs, along with a host of recommendations on how training programs
can be improved.

In a three-year study, Wolford et al. (1997) examined differences in
training and self-protective practices between painters in
Alaska—which has mandatory health and safety training for
painters—and Oregon and Washington, where such training is voluntary.
Alaska’s regulations require painters who use organic, solvent-borne
coatings to take 16 hours of initial training plus 8 hours of
recertification training every 2 years. This training covers health
hazards, protective equipment, use of fans for temporary ventilation,
and how to obtain and use an SDS. Through observation and
questionnaires, the study found that painters with Alaska’s training
were 2.7 times more likely to wear respirators than untrained painters
were, and 1.65 times more likely to use fans. The authors found other
benefits to Alaska’s mandatory approach as well:

Alaska’s mandatory system tends to reach untrained painters better
than voluntary training programs in other states. In contrast, voluntary
training largely attracts people who have already been trained
before—a subset of “true believers” who are already more
safety-conscious than the average painter.

Mandatory training is more effective than voluntary programs at reaching
non-union painters and those working for companies with fewer than four
employees.

Alaska’s training program has lower recruitment costs than voluntary
programs—only one-tenth of the cost per painter trained. Thus, the
mandatory program is more cost-effective.

Because Alaska requires each painter to pay a portion of the training
fee, the cost to employers is lower per painter trained, compared with
the voluntary programs in other states. 

Language and cultural considerations are a growing concern in the field
of health and safety training. At the Dallas-Fort Worth airport, a large
expansion project began in 2002 involving 10,000 workers, roughly half
of whom speak Spanish. To train these workers, a local expert
interviewed over 100 Hispanic workers and designed a program that would
meet their needs better than a conventional training program. The
barriers were more than just a lack of English literacy; many of the
Spanish-speaking workers could not read or write Spanish either. Thus,
rather than relying on written materials, the training program
emphasized hands-on training until an employee could master a task. This
program also used Hispanic instructors, which made trainees feel more
comfortable asking questions. As a result of improved training, this
project experienced less than 10 percent of the lost time injuries that
would be expected for a project of its size and scope, along with many
fewer workers compensation claims (Boone, 2003). As Halcarz (2003)
explains, this training approach was successful because it was
equality-based, recognized the dignity of all employees, and minimized
literacy requirements. In addition to lowering injury rates and
insurance expenses, such an approach can also increase productivity and
promote loyalty. The Dallas-Fort Worth program has been recognized as an
OSHA “National Best Practice” (Halcarz, 2003).

In an example related to HCS, Dean et al. (2002) describe how the Kansas
City (MO) Street and Traffic Division updated its hazard communication
training after employees indicated that the existing program was
insufficient. In this case, the division created a CD-ROM-based training
with on-site photos and employees’ voices. This training is
self-paced, with a proficiency test at the end. At the time of the
article, however, the authors did not have data on how well this new
approach performed.

Other recommendations to improve training include the following ideas:

Weidner (2000) suggests that while many training programs use 70 percent
as the passing score on knowledge tests, this threshold is too low
because it offers a false sense of security and “reinforce[s]
mediocrity as a training objective.” The author recommends a more
“reasonably rigorous” standard of expectation; the exact value may
depend on the difficulty of the test. Two-way feedback also should be
encouraged.

Recognizing that symbols are the items most often recalled from a label,
London (2003) advises a strong emphasis on training for GHS symbols,
particularly the “oxidizer” and “flammable” symbols—which were
reported to be easily confused—and symbols that may generate critical
confusion (eliciting the opposite action—for example, interpreting the
“environmental toxicity” symbol to indicate an area where people can
catch fish).

Bouchard (1997) recommends that employers involve workers on the
workplace safety committee; determine how often they need to repeat
training (how long do people retain information?); and ask trainees to
explain concepts in their own words (e.g., explaining what they read on
an SDS), rather than assessing knowledge by simply asking “Do you
understand?”

To encourage employers to improve training, Fagotto and Fung (2002)
suggest that if companies use good training techniques like interactive
training, real life simulations, and evaluating workers’
understanding—and achieve a low rate of injuries and illnesses—they
should be rewarded with lower premiums for workers compensation
insurance.

7.2.5	Research needs

Studies published since 1997 suggest a continued need for research on
worker safety and health training. Specific recommendations come from
Cohen and Colligan (1998) and Lesch (2003).

After reviewing the literature available as of 1998, Cohen and Colligan
(1998) suggest additional research in the following areas:

Effects of group size and the length and frequency of training.

More information to correlate training with outcomes—e.g.,
retrospective studies of injuries and diseases in workers. 

More in-depth studies on hazard training practices and how they relate
to other hazard control measures.

The extent to which industry has implemented OSHA training approaches,
and the success of those programs.

Lesch (2003) suggests that “Future research should look more closely
at how warning, person, and situational factors interact to influence
warning processing and determine whether training programs can be
designed to help counteract factors that prevent adequate processing of
warnings.”

8.	Related Considerations

8.1	Legal considerations

In testimony before the U.S. Senate Subcommittee on Employment, Safety,
and Training, Michelle Sullivan, Chairman of the Society for Chemical
Hazard Communication (SCHC) Board of Directors, notes that the U.S. has
a unique set of issues around label development (Sullivan, 2004). Among
these unique issues are liability and the duty to provide an adequate
warning. Legal considerations have long been considered when developing
product labels and warnings, and have recently become a factor for SDS
development as well. Sullivan expresses support for adopting the GHS in
the U.S., but recommends that these unique issues be kept in mind.

Many textbooks on warnings include information about legal perspectives.
For example, Madden (2006) offers an overview of the manufacturer’s
duty to warn, as does Madden (1999). While much of the discussion and
legal precedent concerns consumer products, some of the basic principles
may be useful when evaluating occupational settings as well.

Another perspective on product liability comes from the insurance
industry. For instance, The Hartford has published a set of tips on
minimizing product liability. Among other things, these tips direct
companies to follow the ANSI Z535.4 standard for product safety signs
and to test their warning labels with real users (The Hartford, 2002).

One important extension of “duty to warn” concerns low-literate and
non-English-speaking populations. What is the manufacturer’s duty to
communicate safety information to these groups? Ross (1995) provides
some examples from relevant case law. In a 1962 decision, a U.S. federal
appellate court established two essential characteristics of a legally
adequate warning: “(l) It must be in such a form that it could
reasonably be expected to catch the attention of a reasonably prudent
man in the circumstances of its use; and (2) the content of the warning
must be of such a nature as to be comprehensible to the average user and
to convey a fair indication of the nature and extent of the danger to
the mind of a reasonably prudent person.” (Ross, 1995, citing Spruill
v. Boyle-Midway, Incorporated, 308 F.2d 79 [4th Cir. 1962]). Other court
decisions have varied. 

In some locations (e.g., areas like Miami), the average user may not be
English literate. Therefore, as a general principle, Ross (1995)
suggests that “if a manufacturer is selling a product in areas where
the average user likely will not speak English or possibly not read at
all, it should consider including at least a pictorial or symbol which
identifies the hazard, and, possibly, a word message in a foreign
language.” The author warns, however, that a manufacturer may run some
risk of liability if it includes foreign language labels and these
labels contain inadequate information or are not effectively
communicated. Pictograms may also be useful—particularly for users who
are not literate in any language—but designers must be careful not to
omit important information by relying on simple representations.

8.2	Emerging approaches for hazard communication and control 

In the last decade, a new approach to the management of hazardous
chemicals has emerged. This approach, called “control banding,”
involves classifying chemicals into general hazard groups or “bands”
(such as “skin and eye irritants”), then prescribing a specific set
of control measures for a given range of exposures to a chemical within
that group (CDC/NIOSH, 2005).

Control banding was pioneered by the pharmaceutical industry as a way to
manage exposure risks from an ever-increasing number of chemicals in the
workplace—a particular concern in an industry where workers frequently
interact with new substances that have yet to be studied extensively
(Naumann et al., 1996). More recently, the control banding approach was
adopted by the UK Health and Safety Executive (HSE) as a means to help
small- and medium-sized enterprises comply with the Control of
Substances Hazardous to Health (COSHH) safety regulations without
on-site technical experts and expensive exposure measurements (Jackson,
2002). In its COSHH Essentials package, the HSE provides a stepwise
approach to assessing risks, based on standard hazard phrases that the
manufacturer must assigned to a given chemical (HSE, 2005). Based on the
level of risk, the employer must follow a specific set of control
measures (e.g., ventilation, engineering control, or containment)
(Jackson, 2002).

Control banding also has been adopted for international use, targeted
mainly at developing countries where employers may lack technical
expertise in hazard evaluation and control. Together with the ILO, WHO,
and the United Nations Environment Programme, the International
Occupational Hygiene Association (IOHA) developed the International
Chemical Control Toolkit, which provides resources and guidance on
control banding (ILO, 2006).

Although control banding is not part of the GHS, some have suggested
that control banding and the GHS are complementary, and that both
strategies are worth considering when updating the HCS (e.g., Garman,
2005). Among other things, control banding could provide a means to
protect trade secrets without putting workers at risk.

9.	References

Abt Associates, Inc., 1996. Consumer labeling initiative: Phase I
report. U.S. Environmental Protection Agency.

Abt Associates, Inc., 1999. Consumer labeling initiative: Phase II
report. U.S. Environmental Protection Agency.

Adams, A., S. Bochner, and L. Bilik, 1998. The effectiveness of warning
signs in hazardous work places: Cognitive and social determinants.
Applied Ergonomics 29(4): 247-254.

Akerboom, S. and M. Trommelen, 1998. Environmental labeling on household
chemicals: Comprehensibility and impact on warning information.
International Journal of Cognitive Ergonomics 2(1-2): 107-122.

[ANSI] American National Standards Institute, 2002a. American National
Standard Criteria for Safety Symbols. ANSI Z535.3-2002.

[ANSI] American National Standards Institute, 2002b. American National
Standard for Environmental and Facility Safety Signs. ANSI Z535.2-2002.

[ANSI] American National Standards Institute, 2002c. American National
Standard for Product Safety Signs and Labels. ANSI Z535.4-2002.

[ANSI] American National Standards Institute, 2002d. American National
Standard for Safety Tags and Barricade Tapes (for Temporary Hazards).
ANSI Z535.5-2002.

[ANSI] American National Standard Institute, 2004. American National
Standard for Hazardous Industrial Chemicals—Material Safety Data
Sheets—Preparation. ANSI Z400.1-2004.

[ANSI] American National Standard Institute, 2006a. American National
Standard for Hazardous Industrial Chemicals—Precautionary Labeling.
ANSI Z129.1-2006.

[ANSI] American National Standards Institute, 2006b. American National
Standard for Safety Colors. ANSI Z535.1-2006.

[ANSI] American National Standards Institute, in press. American
National Standard for Product Safety Information in Product Manuals,
Instructions, and Other Collateral Materials. ANSI Z535.6.

[APA] American Psychological Association, 2007. PsycINFO research
database. Available online at www.psycinfo.apa.org/psycinfo/. Last
accessed 12 April 2007.

Argo, J.J., and K.J. Main, 2004. Meta-analyses of the effectiveness of
warning labels. Journal of Public Policy & Marketing 23(2): 193.

[ASSE] American Society of Safety Engineers, 2001. American National
Standard—Criteria for Accepted Practices in Safety, Health, and
Environmental Training. ANSI/ASSE Z490.1-2001.

[ASSE] American Society of Safety Engineers, 2004. ASSE comments on
draft OSHA guidance documents concerning hazard communication standard.
OSHA Hazard Communication Docket No. H022J. Document ID 1575253. Public
Comment. Exhibit No. 4-13. 17 May 2004.

[ASTM] American Society for Testing and Materials, 2000. Standard
Practice for Preparing Material Safety Data Sheets to Include
Transportation and Disposal Data for the General Services
Administration. ASTM E 1628-94.

[ASTM] American Society for Testing and Materials, 2003. Standard
Terminology Relating to Hazard Potential of Chemicals. ASTM E 1445-03.

Baldwin, M., K. Lippel, M. Lefèbvre, and D. Mergler, 2003. Material
safety data sheets: their use in Quebec workers' compensation appeals
involving neurotoxic chemical exposures. Applied Occupational and
Environmental Hygiene 18(6): 405-412.

Banda, S.F. and K. Sichilongo, 2006. Analysis of the level of
comprehension of chemical hazard labels: A case for Zambia. Science of
the Total Environment 363(1-3): 22-27.

Beach, J., 2002. The problem with material safety data sheets.
Occupational Medicine (London) 52(2): 67-68.

Bentley, D.J., 2006. What makes an MSDS useful? Paper, Film and Foil
Converter 80(4): 14.

Bernstein, J.A., 2002. Material safety data sheets: are they reliable in
identifying human hazards? Journal of Allergy and Clinical Immunology
110(1): 35-38.

Boone, E., 2003. The language of safety. Rough Notes 146(4): 24.

Bouchard, C., 2007. Literacy and hazard communication: ensuring workers
understand the information they receive. Journal of American Association
of Occupational Health Nurses 55(1): 18-25.

Bowles, C.T., A.D. Fisk, and W.A. Rogers, 2002. Inference and the use of
similes and metaphors in warnings. Proceedings of the Human Factors and
Ergonomics Society 46th Annual Meeting: 1703-1707.

Braun, C.C., S.A. Glusker, R.S. Holt, and N.C. Silver, 1995. Adding
consequence information to product instructions: Changes in hazard
perceptions. Proceedings of the Human Factors and Ergonomics Society
39th Annual Meeting: 346-350.

Braun, C.C. and E.F. Shaver, 1999. Warning sign components and hazard
perceptions. Proceedings of the Human Factors and Ergonomics Society
43rd Annual Meeting: 878-882.

[CDC/NIOSH] Centers for Disease Control and Prevention and the National
Institute for Occupational Safety and Health, 2005. FAQs about control
banding. Available online at
http://www.cdc.gov/niosh/topics/ctrlbanding/pdfs/CBFAQ.pdf. Last
accessed 6 June 2007.

Chen, J.Y.C., R.D. Gilson, and M. Mouloua, 1997. Perceived risk dilution
with multiple warnings. Proceedings of the Human Factors and Ergonomics
Society 41st Annual Meeting: 831-835. Cited in: Summary of warnings
research by S.O. Parsons, J.L. Seminara, and M.S. Wogalter, 1999.
Ergonomics in Design 7(1): 21-31.

Chunin, M., 2002. Factors determining compliance with safety signs in
industrial settings. Dissertation Abstracts International: Section B:
The Sciences and Engineering 63(3-B): 1550.

Cohen, A. and M.J. Colligan, 1998. Assessing occupational safety and
health training - a literature review. National Institute for
Occupational Safety and Health: 1-174.

Cohen, H.H., J. Cohen, C.C. Mendat, and M.S. Wogalter, 2006. Warning
channel: Modality and media. In M.S. Wogalter (ed.), Handbook of
Warnings (123-134). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Conklin, J., 2003. An international assessment of the comprehensibility
of material safety data sheets (MSDSs). DPH dissertation, University of
Texas Health Science Center at Houston, School of Public Health.

Conzola, V.C. and M.S. Wogalter, 1998. Consumer product warnings:
Effects of injury statistics on recall and subjective evaluations.
Proceedings of the Human Factors and Ergonomics Society 42nd Annual
Meeting: 559-563.

Conzola, V.C. and M.S. Wogalter, 2001. A communication-human information
processing (C-HIP) approach to warning effectiveness in the workplace.
Journal of Risk Research 4(4): 309-322.

Côté, R., H. Davis, C. Dimock, M. Korpan, K. Loewen, and L. Segal,
1998. The evaluation and hazard classification of toxicological
information for workplace hazardous materials information system
material safety data sheets. Regulatory Toxicology and Pharmacology 27:
61-74.

Cox, E.P., M.S. Wogalter, S.L. Stokes, and E.J. Tipton Murff, 1997. Do
product warnings increase safe behavior? A meta-analysis. Journal of
Public Policy & Marketing 16: 195-204.

[CSB] U.S. Chemical Safety and Hazard Investigation Board, 2004.
Investigation report: Dust explosion, West Pharmaceutical Services,
Inc., Kinston, North Carolina, January 29, 2003. Report No.
2003-07-I-NC. September 2004. Available online at
http://www.csb.gov/index.cfm?folder=completed_investigations&page=info&I
NV_ID=34. Last accessed 6 July 2007.

[CSB] U.S. Chemical Safety and Hazard Investigation Board, 2005.
Investigation report: Combustible dust fire and explosions, CTA
Acoustics, Inc., Corbin, Kentucky, February 20, 2003. Report No.
2003-09-I-KY. February 2005. Available online at
http://www.csb.gov/index.cfm?folder=completed_investigations&page=info&I
NV_ID=35. Last accessed 6 July 2007.

[CSB] U.S. Chemical Safety and Hazard Investigation Board, 2006.
Investigation report: Combustible dust hazard study. Report No.
2006-H-1. November 2006. Available online at
http://www.csb.gov/index.cfm?folder=completed_investigations&page=info&I
NV_ID=53. Last accessed 6 July 2007.

Dain, D.P. and A.L. Studley, 2004. Guidelines for practitioners:
Advising clients on how to develop product warnings. For the Defense:
10-11, 58-59.

Dean, M., G. Goodale, and G. Rogers, 2002. An innovative approach to
hazard communication training. American Public Works Association
Reporter 69(6): 16-17.

DeJoy, D.M., K.A. Cameron, and L.J. Della, 2006. Postexposure evaluation
of warning effectiveness: A review of field studies and population-based
research. In M.S. Wogalter (ed.), Handbook of Warnings (35-48). Mahwah,
NJ: Lawrence Erlbaum Associates, Inc.

Dowse, R. and M.S. Ehlers, 2003. The influence of education on the
interpretation of pharmaceutical pictograms for communicating medicine
instructions. International Journal of Pharmacy Practice 11(1): 11-18.

Dowse, R. and M. Ehlers, 2005. Medicine labels incorporating pictograms:
Do they influence understanding and adherence? Patient Education and
Counseling 58(1): 63-70.

Drake, K.L., V.C. Conzola, and M.S. Wogalter, 1998. Discrimination among
sign and label warning signal words. Human Factors and Ergonomics in
Manufacturing 8(4): 289-301.

Edworthy, J., E. Hellier, N. Morley, C. Grey, K. Aldrich, and A. Lee,
2004. Linguistic and location effects in compliance with pesticide
warning labels for amateur and professional users. Human Factors 46(1):
11-31.

Engineering Information, 2007. COMPENDEX research database. Available
online at www.ei.org/databases/compendex.html. Last accessed 5 April
2007.

European Commission, 1991. Commission Directive 91/155/EEC of 5 March
1991 defining and laying down the detailed arrangements for the system
of specific information relating to dangerous preparations. 

European Commission, 1998. Council Directive 98/24/EC of 7 April 1998 on
the protection of the health and safety of workers from the risks
related to chemical agents at work.

European Commission, 2001. Commission Directive 2001/60/EC of 7 August
2001 adapting to technical progress Directive 1999/45/EC of the European
Parliament and of the Council concerning the approximation of the laws,
regulations and administrative provisions of the Member States relating
to the classification, packaging, and labeling of dangerous
preparations.

European Commission, 2003. Downstream consequences on other community
legislation arising from the classification and labeling of dangerous
substances under directive 67/548/EEC. 

European Union, 2007. Activities of the European Union: Summaries of
legislation. Available online at
http://europa.eu/scadplus/info_scad_en.htm. Last accessed 6 June 2007. 

Evans, A.W., R.M. Hoeft, F.G. Jentsch, and C.A. Bowers, 2002. Is a
picture worth a thousand words? Investigating structural knowledge with
textual and pictorial stimuli. Proceedings of the Human Factors and
Ergonomics Society 46th Annual Meeting: 240-244.

Fagotto, E. and A. Fung, 2002. Improving workplace hazard communication.
Issues in Science and Technology 19(2): 63-68.

Frantz, J.P. and S.M. Hall, 2005. Development, status and implications
of ANSI Z535.6 - a new standard for safety information in
product-accompanying literature. Proceedings of the 2005 ASSE
Professional Development Conference, Session No. 728. Available online
at http://www.safetycouncil.com/pdf/728Frantz-Hall.pdf. Last accessed 6
June 2007. 

Frascara, J., 2006. Typography and the visual design of warnings. In
M.S. Wogalter (ed.), Handbook of Warnings (385-406). Mahwah, NJ:
Lawrence Erlbaum Associates, Inc.

Frazier, L.M., B.W. Beasley, G.K. Sharma, and A.A. Mohyuddin, 2001.
Health information in material safety data sheets for a chemical which
causes asthma. Journal of General Internal Medicine 16(2): 89-93.

Freeman, K.S., 2001. Health and safety information for specialized
vocational audiences. Proceedings/STC, Society for Technical
Communication Annual Conference, 2001: 537-542.

Fung, A., D. Weil, M. Graham, and E. Fagotto, 2004. The political
economy of transparency: What makes disclosure policies effective? Ash
Institute for Democratic Governance and Innovation, John F. Kennedy
School of Government, Harvard University: 1-46.

Ganier, F., 2001. Processing text and pictures in procedural
instructions. Information Design Journal 10: 146-153. Cited in: Warning
symbols by M.S. Wogalter, N.C. Silver, S.D. Leonard, and H. Zaikina,
2006b. In M.S. Wogalter (ed.), Handbook of Warnings (159-176). Mahwah,
NJ: Lawrence Erlbaum Associates, Inc., 2006.

Garman, D.D., 2005. Is chemical control banding the future of OSHA’s
hazard communication program? Proceedings of the 2005 ASSE Professional
Development Conference, Session No. 547. Available online at
http://www.safetycouncil.com/pdf/547Garman.pdf. Last accessed 6 June
2007. 

Google, 2007. Google Scholar Search Engine. Available online at
http://scholar.google.com/. Last accessed May 2007.

Griffith, L.J. and S.D. Leonard, 1997. Association of colors with
warning signal words. International Journal of Industrial Ergonomics
20(4): 317-325.

Grotjohn, D.K., 2000. An application of the elaboration likelihood model
to product safety warnings. Dissertation Abstracts International:
Section B: The Sciences and Engineering 61(6-B): 3299.

Grumbles, T.G., 2004. Statement before the U.S. Senate Subcommittee on
Employment, Safety, and Training during its hearing on “Hazard
Communication in the 21st Century Workforce.” 25 March 2004.

Halcarz, J.E., 2003. Safety training for diverse workers populations.
Occupational Hazards 65(2): 26.

Hall, S.M., J.P. Frantz, S.L. Young, T.P. Rhoades, J.J. Isaacson, and
C.G. Burhans, 2006. Update on ANSI Z535.6: A new standard for safety
information in product manuals, instructions, and other collateral
materials. Available online at
http://www.appliedsafety.com/ansi_z535dot6_article.pdf. Last accessed 6
June 2007. 

Hancock, H.E., W.A. Rogers, and A.D. Fisk, 1999. Understanding
age-related differences in the perception and comprehension of symbolic
warning information. Proceedings of the Human Factors and Ergonomics
Society 43rd Annual Meeting: 617-621.

Hancock, H.E., A.D. Fisk, and W.A. Rogers, 2001. Comprehension of
explicit and implicit warning information in younger and older adults.
Proceedings of the Human Factors and Ergonomics Society 45th Annual
Meeting: 1474-1478.

Hanson, J., 2004. Statement before the U.S. Senate Subcommittee on
Employment, Safety, and Training during its hearing on “Hazard
Communication in the 21st Century Workforce.” 25 March 2004.

Hartley, J., 1994. Designing Instructional Text, 3rd ed. London: Kogan
Page/East Brunswick, NJ: Nichols. Cited in: A communication-human
information processing (C-HIP) approach to warning effectiveness in the
workplace by V.C. Conzola and M.S. Wogalter, 2001. Journal of Risk
Research 4(4): 309-322.

Heaps, C.M. and T.B. Henley, 1999. Language matters: Wording
considerations in hazard perception and warning comprehension. Journal
of Psychology: Interdisciplinary and Applied 133(3): 341-351.

Hellier, E., D. Wright, and J. Edworthy, 2000a. Investigating the
perceived hazard of warning signal words. Risk, Decision and Policy 5:
39-48.

Hellier, E., D.B. Wright, J. Edworthy, and S. Newstead, 2000b. On the
stability of the arousal strength of warning signal words. Applied
Cognitive Psychology 14(6): 577-592.

Hochhauser, M., 1997. Why aren't material safety data sheets more
effective? Occupational Health & Safety 66(8): 58.

Hoonhout, H.C.M., 2000. Evaluating graphical messages for mentally
retarded adults; Do pictures get the intended information across?
Proceedings of the XIVth Triennial Congress of the International
Ergonomics Association and 44th Annual Meeting of the Human Factors and
Ergonomics Association, Ergonomics for the New Millennium: 789-801.

Houts, P.S., J.T. Witmer, H.E. Egeth, M.J. Loscalzo, and J.R. Zabora,
2001. Using pictographs to enhance recall of spoken medical instructions
II. Patient Education and Counseling 43(3): 231-242.

[HSE] UK Health and Safety Executive, 2005. COSHH: A brief guide to the
regulations. Available online at
http://www.coshh-essentials.org.uk/assets/live/indg136.pdf. Last
accessed 6 June 2007. 

[ILO] International Labour Organization, 1999. Bibliography on chemical
hazard communication. 

[ILO] International Labour Organization, 2006. International chemical
control toolkit. Last modified 2006. Available online at
http://www.ilo.org/public/english/protection/safework/ctrl_banding/toolk
it/icct/-index.htm. Last accessed 6 June 2007. 

[ISO] International Organization for Standardization, 2002. Graphical
Symbols—Safety Colours and Safety Signs—Part 1: Design Principles
for Safety Signs in Workplaces and Public Areas. ISO 3864-1.

[ISO] International Organization for Standardization, 2003. Graphical
Symbols—Vocabulary. ISO 17724.

[ISO] International Organization for Standardization, 2006. Graphical
Symbols—Safety Colours and Safety Signs—Part 3: Design Principles
for Graphical Symbols for Use in Safety Signs. ISO 3864-3.

[ISO] International Organization for Standardization, 2007a. Graphical
Symbols—Test Methods—Part 1: Methods for Testing Comprehensibility.
ISO 9186-1. 

[ISO] International Organization for Standardization, 2007b. Graphical
Symbols—Test Methods—Part 2: Methods for Testing Perceptual Quality.
Draft Standard: ISO/DIS 9186-2.

[ISO/IEC] International Organization for Standardization and
International Electrotechnical Commission, 2004. Graphical
Symbols—Technical Guidelines for the Consideration of Consumers’
Needs. ISO/IEC Guide 74. 

Jackson, A., 2004. Statement before the U.S. Senate Subcommittee on
Employment, Safety, and Training during its hearing on “Hazard
Communication in the 21st Century Workforce.” 25 March 2004.

Jackson, H., 2002. Control Banding – Practical Tools for Controlling
Exposure to Chemicals. Asian-Pacific Newsletter on Occupational Health
and Safety 9: 62-63. Available online at
http://www.ilo.org/public/english/protection/safework/ctrl_banding/pract
ools.pdf. Last accessed 6 June 2007.

Johnson, D.A., 2006. Practical aspects of graphics related to safety
instructions and warnings. In M.S. Wogalter (ed.), Handbook of Warnings
(463-476). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Kalsher, M.J., M.S. Wogalter, and B.M. Racicot, 1996. Pharmaceutical
container labels and warnings: Preference and perceived readability of
alternative designs and pictorials. International Journal of Industrial
Ergonomics 18: 83-90.

Kalsher, M.J., M.B. Cote, M.V. Champagne, and M.S. Wogalter, 1997.
Effects of a raised label border on warning effectiveness measures.
Proceedings of the Human Factors and Ergonomics Society 41st Annual
Meeting: 811-815.

Kalsher, M.J. and K.J. Williams, 2006. Behavioral compliance: Theory,
methodology, and results. In M.S. Wogalter (ed.), Handbook of Warnings
(313-332). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Kassam, R., L.R. Vaillancourt, and J.B. Collins, 2004. Pictographic
instructions for medications: Do different cultures interpret them
accurately? International Journal of Pharmacy Practice 12(4): 199-209.

Krenek, R.F. and J.L. Purswell, 1998. Warning label content strategies:
An approach to the integration of user knowledge and designer judgment.
International Journal of Cognitive Ergonomics 2(1-2): 53-60.

Laughery, K.R., S.L. Young, K.P. Vaubel, and J.W. Brelsford, 1993. The
noticeability of warnings on alcoholic beverage containers. Journal of
Public Policy & Marketing 12(1): 38.

Laughery, K.R. and D.P. Smith, 2006. Explicit information in warnings.
In M.S. Wogalter (ed.), Handbook of Warnings (419-428). Mahwah, NJ:
Lawrence Erlbaum Associates, Inc.

Lehto, M.R., 1998a. The influence of chemical warning label content and
format on information retrieval speed and accuracy. Journal of Safety
Research 29(1): 43-56.

Lehto, M.R. and J.D. Papastavrou, 1998b. A signal-detection theory based
perspective on design of warning. Perceptual and Motor Skills 86(2):
720-722.

Lepkowska-White, E. and A.L. Parsons, 2001. Comparison of warnings and
resulting attitudes. Journal of Consumer Affairs 35(2): 278-294.

Lesch, M.F., 2002. Signs of age in today's workforce. Occupational
Health & Safety 71(12): 45.

Lesch, M.F., 2003. Comprehension and memory for warning symbols:
Age-related differences and impact of training. Journal of Safety
Research 34(5): 495-505.

Lesch, M.F., 2004. Age-related effects on warning symbol comprehension.
Proceedings of the Human Factors and Ergonomics Society 48th Annual
Meeting: 233-237.

Lim, R., 2005. Perceptions and performance for bilingual labels as a
function of native users of English and Spanish. Ph.D. dissertation,
North Carolina State University.

Logsdon, R.K., 2004. Reading between the lines. Rock Products 107(8):
50.

London, L., 2003. FRIDGE/UNITAR study into the implications of
implementing the Globally Harmonised System of Classification and
Labelling of Chemicals and development of an implementation strategy for
South Africa. National Economic Development & Labour Council (NEDLAC),
Johannesburg, South Africa. Parts 2 and 3. Available online at
http://www.nedlac.org.za/. Last accessed 4 June 2007.

Lotto, L., R. Job, and R. Rumiati, 1999. Visual effects in picture and
word categorization. Memory & Cognition 27: 674-684.

Madden, M.S., 1999. The law relating to warnings. In M.S. Wogalter, D.M.
Dejoy, and K.R. Laughery (eds.), Warnings and Risk Communication
(315-330). Philadelphia: Taylor and Francis.

Madden, M.S., 2006. The duty to warn in products liability. In M.S.
Wogalter (ed.), Handbook of Warnings (583-588). Mahwah, NJ: Lawrence
Erlbaum Associates, Inc.

Meingast, M., 2001. Increasing attention and retention of warnings:
Effects of container hazardousness, warning quality, and severity of
injury. Proceedings of the Human Factors and Ergonomics Society 45th
Annual Meeting: 1482-1486.

Merritt, C.W., 2004. Written testimony submitted to the U.S. Senate
Subcommittee on Employment, Safety, and Training during its hearing on
“Hazard Communication in the 21st Century Workforce.” 25 March 2004.
Available online at
http://www.csb.gov/news_releases/docs/CSBFinalTestimonyMSDS32404.pdf.
Last accessed 6 July 2007.

Mitchell, C.S., and B.S. Schwartz, 2001. Limitations of information
about health effects of chemicals. Journal of General Internal Medicine
16(2): 135-136.

Naumann, B.D., E.V. Sargent, B.S. Starkman, W.J. Fraser, G.T. Becker,
and G.D. Kirk, 1996. Performance-based exposure control limits for
pharmaceutical active ingredients. American Industrial Hygiene
Association Journal 57(1): 33-42.

[NFPA] National Fire Protection Association, 2007. Standard System for
the Identification of the Hazards of Materials for Emergency Response.
NFPA 704. 

[NIOSH] National Institute for Occupational Safety and Health, 2007.
NIOSHTIC-2 research database. Available online at
http://www2a.cdc.gov/nioshtic-2/. Last accessed 23 April 2007.

[NLM] National Library of Medicine, 2007. PubMED research database.
Available online at http://www.ncbi.nlm.nih.gov/sites/entrez?db=pubmed.
Last accessed 12 April 2007.

[NOHSC] National Occupational Health and Safety Commission (Australia),
2003. National Code of Practice for the Preparation of Material Safety
Data Sheets, 2nd edition. 

[OSHA] U.S. Occupational Safety and Health Administration, date unknown.
GHS—OSHA HCS comparison: Comparison of hazard communication
requirements. Available online at
http://www.osha.gov/dsg/hazcom/GHSOSHAComparison.html. Last accessed 7
June 2007.

[OSHA] U.S. Occupational Safety and Health Administration, 1994. Hazard
Communication Standard—29 CFR 1910.1200. Available online at
http://www.osha.gov/pls/oshaweb/owadisp.show_document?p_table=STANDARDS&
p_id=10099. Last accessed 6 June 2007. 

[OSHA] U.S. Occupational Safety and Health Administration, 1998.
Training requirements in OSHA standards and training guidelines.
Available online at http://www.osha.gov/Publications/2254.html;
http://www.osha.gov/Publications/osha2254.pdf. Last accessed 6 June
2007. 

[OSHA] U.S. Occupational Safety and Health Administration, 2004. Hazard
Communication in the 21st Century Workplace. March 2004. Available
online at http://www.osha.gov/dsg/hazcom/finalmsdsreport.html. Last
accessed 6 July 2007. 

[OSHA] U.S. Occupational Safety and Health Administration, 2005. U.S.
implementation of the Globally Harmonized System: The GHS journey
continues. Presentation by J. Silk at the Society for Chemical Hazard
Communication 2005 Fall Meeting, 27 September 2005. Available online at
http://www.osha.gov/dcsp/alliances/schc/silk_schc_sept05/index.html.
Last accessed 7 June 2007.

[OSHA] U.S. Occupational Safety and Health Administration, 2006a.
Advance notice of proposed rulemaking. Federal Register 71(176):
53617-53627. 12 September 2006.

[OSHA] U.S. Occupational Safety and Health Administration, 2006b. A
guide to the Globally Harmonized System of Classification and Labeling
of Chemicals (GHS). Available online at
http://www.osha.gov/dsg/hazcom/ghs.html. Last accessed 6 June 2007.

Packham, C., 2002. The problem with material safety data sheets.
Occupational Medicine (London) 52(8): 513.

Parsons, S.O., J.L. Seminara, and M.S. Wogalter, 1999. Summary of
warnings research. Ergonomics in Design 7(1): 21-31.

Phillips, C.C., 1997. The efficacy of material safety data sheets and
worker acceptability. Ph.D. dissertation, University of Tennessee at
Knoxville. May 1997. Cited in: Hazard communication: A review of the
science underpinning the art of communication for health and safety by
B. Sattler, B. Lippy, and T.G. Jordan, 1997. Available online at
http://www.osha.gov/SLTC/hazardcommunications/hc2inf2.html. Last
accessed 6 June 2007.

Phillips, C.C., B.C. Wallace, C.B. Hamilton, R.T. Pursley, G.C. Petty,
and C.K. Bayne, 1999. The efficacy of material safety data sheets and
worker acceptability. Journal of Safety Research 30(2): 113-122.

ProQuest, 2007. ABI/INFORM research database. ProQuest Information and
Learning. Last accessed 9 April 2007.

Pullen, M., 2003. Making MSDSs universally understandable. Occupational
Health & Safety 72(7): 46.

Riley, D.M., and B. Fischhoff, 2000. An integrative approach for label
evaluation and design. Proceedings of the XIVth Triennial Congress of
the International Ergonomics Association and 44th Annual Meeting of the
Human Factors and Ergonomics Association, Ergonomics for the New
Millennium: 778-781.

Rogers, W.A., N. Lamson, and G.K. Rousseau, 2000. Warning research: an
integrative perspective. Human Factors 42: 102-139.

Rooney, J., 2005. Right-to-know vs. right-to-understand. Occupational
Health & Safety 74(9): 186.

Ross, K., 1995. The duty to warn illiterate and non-English-reading
product users. BNA Product Safety and Liability Reporter, 13 October
1995. Bowman and Brooke, LLP. Available online at
http://www.bowman-brooke.com/CM/Publications/publications54.asp. Last
accessed 6 June 2007. 

Rousseau, G.K., N. Lamson, and W.A. Rogers, 1998. Designing warnings to
compensate for age-related changes in perceptual and cognitive
abilities. Psychology & Marketing 15(7): 643.

Rousseau, G.K. and M.S. Wogalter, 2006. Research on warning signs. In
M.S. Wogalter (ed.), Handbook of Warnings (147-158). Mahwah, NJ:
Lawrence Erlbaum Associates, Inc.

Rudin-Brown, C.M., M.P. Greenley, A. Barone, J. Armstrong, A.F. Salway,
and B.J. Norris, 2004. The design of child restraint system (CRS) labels
and warnings affects overall CRS usability. Traffic Injury Preventions
5: 8-17.

Sansgiry, S.S., P.S. Cady, and B.A. Adamcik, 1997. Consumer
comprehension of information on over-the-counter medication labels:
effects of picture superiority and individual differences based on age.
Journal of Pharmaceutical Marketing & Management 11(3): 63-76.

Sattler, B., B. Lippy, and T.G. Jordan, 1997. Hazard communication: A
review of the science underpinning the art of communication for health
and safety. Available online at
http://www.osha.gov/SLTC/hazardcommunications/hc2inf2.html. Last
accessed 6 June 2007.

[SCHC] Society for Chemical Hazard Communication, 1999. Published
research on the comprehensibility of chemical hazard warning labels and
material safety data sheets. Available online at
http://www.schc.org/downloads/Comprehensibility-Rev2.PDF. Last accessed
6 June 2007.

Schroeder, D.G., H.E. Hancock, W.A. Rogers, and A.D. Fisk, 2001. Phrase
generation and symbol comprehension for 40 safety symbols. Proceedings
of the Human Factors and Ergonomics Society 45th Annual Meeting:
1479-1480.

Smith-Jackson, T. and M.S. Wogalter, 1998. Determining the preferred
order of material safety data sheets (MSDS): A user-centered approach.
Proceedings of the Human Factors and Ergonomics Society 42nd Annual
Meeting: 1073-1077.

Smith-Jackson, T.L. and M.S. Wogalter, 2000. Users’ hazard perceptions
of warning components: An examination of colors and symbols. Proceedings
of the XIVth Triennial Congress of the International Ergonomics
Association and 44th Annual Meeting of the Human Factors and Ergonomics
Association, Ergonomics for the New Millennium: 55-58.

Smith-Jackson, T.L. and M.S. Wogalter, 2006. Methods and procedures in
warning research. In M.S. Wogalter (ed.), Handbook of Warnings (23-34).
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Sojourner, R.J. and M.S. Wogalter, 1997. The influence of pictorials on
evaluations of prescription medication instructions. Drug Information
Journal 31: 963-972.

Sojourner, R.J. and M.S. Wogalter, 1998. The influence of pictorials on
the comprehension and recall of pharmaceutical safety and warning
information. International Journal of Cognitive Ergonomics 2(1-2):
93-106.

Spruill v. Boyle-Midway, Incorporated, 308 F.2d 79 (4th Cir. 1962).
Cited in: The duty to warn illiterate and non-English-reading product
users by K. Ross, 1995. BNA Product Safety and Liability Reporter, 13
October 1995. Bowman and Brooke, LLP. Available online at
http://www.bowman-brooke.com/CM/Publications/publications54.asp. Last
accessed 6 June 2007.

Sullivan, M.R., 2004. Statement before the U.S. Senate Subcommittee on
Employment, Safety, and Training during its hearing on “Hazard
Communication in the 21st Century Workforce.” 25 March 2004. 

Sussman, R.G., T.A. Kimmel, and F.W. Neske, 1997. Health hazard labeling
for hazard communication in the pharmaceutical industry. Occupational
Medicine 12(1): 107-117.

Swindell, J., 1999. Measuring visual search reaction time and accuracy
for a product label warning as a function of icon, color, column and
vertical placement. M.S. dissertation, North Carolina State University.

Tan-Wilhelm, D., K. Witte, W. Liu, L.S. Newman, A. Janssen, C. Ellison,
A. Yancey, W. Sanderson, and P.K. Henneberger, 2000. Impact of a worker
notification program: assessment of attitudinal and behavioral outcomes.
American Journal of Industrial Medicine 37(2): 205-213.

The Hartford, 2002. Warning and precautionary labels: Reducing product
liability exposure. Loss Control TIPS (Technical Information Paper
Series). Available online at
http://www.thehartford.com/corporate/losscontrol/TIPS/855-003.pdf. Last
accessed 6 June 2007. 

Thomson Scientific, Inc., 2007. International Pharmaceutical Abstracts
(IPA) research database. Last accessed 6 May 2007.

Trommelen, M., 1997. Effectiveness of explicit warnings. Safety Science
25(1-3): 79-88.

United Nations, 2002a. Comparison among draft of revision, GHS-SDS and
ISO 11014-1:1994. Sub-committee of experts on the GHS of classification
and labeling of chemicals. UN/GHS-SC/4/INF.14/Add.1. Available online at
http://www.unece.org/trans/doc/2002/ac10c4/UN-SCEGHS-04-inf14a1e.doc.
Last accessed 6 June 2007.

United Nations, 2002b. Safety Data Sheet for Chemical Products—Part 1:
Content and Order of Sections. Draft Standard: ISO/WD 11014-1.
Sub-committee of experts on the GHS of classification and labeling of
chemicals. UN/GHS-SC/4/INF.14. Available online at
http://www.unece.org/trans/doc/2002/ac10c4/UN-SCEGHS-04-inf14e.doc. Last
accessed 6 June 2007.

United Nations, 2005. Globally Harmonized System of Classification and
Labeling of Chemicals (GHS). Available online at
http://www.unece.org/trans/danger/publi/ghs/ghs_rev01/01files_e.html.
Last accessed 6 June 2007.

Vaillancourt, R., K. Ryan, G. Beckett, and S. de Silva, 2004.
Pre-testing of pictograms used in missions of humanitarian relief.
American Society of Health-System Pharmacists (ASHP) Midyear Clinical
Meeting 39(Dec): P201E. (Abstract)

Vigilante, W.J. and M.S. Wogalter, 1997. On the prioritization of safety
warnings in product manuals. International Journal of Industrial
Ergonomics 20(4): 277-285.

Vigilante, W.J. and M.S. Wogalter, 1998a. Older adults’ perceptions of
OTC drug labels: Print size, white space, and design type. In S. Kumar
(ed.), Advances in Occupational Ergonomics and Safety (599-602).
Louisville, KY: IOS Press and Ohmsha. Cited in: A communication-human
information processing (C-HIP) approach to warning effectiveness in the
workplace by V.C. Conzola and M.S. Wogalter, 2001. Journal of Risk
Research 4(4): 309-322.

Vigilante, W.J. and M.S. Wogalter, 1998b. Product manual safety
warnings: The effects of ordering. Proceedings of the Human Factors and
Ergonomics Society 42nd Annual Meeting: 593-597.

Walker, J.H., 1997. Harmony sought on global hazards communications.
Modern Paint and Coatings 87(7): 16.

Weidner, B.L., A.R. Gotsch, C.D. Delnevo, J.B. Newman, and B. McDonald,
1998. Worker health and safety training: assessing impact among
responders. American Journal of Industrial Medicine 33(3): 241-246.

Weidner, B.L., 2000. Testing as a measure of worker health and safety
training: perspectives from a hazardous materials program. American
Journal of Industrial Medicine 37(2): 221-228.

Welsh, M.S., M. Lamesse, and E. Karpinski, 2000. The verification of
hazardous ingredients disclosures in selected material safety data
sheets. Applied Occupational and Environmental Hygiene 15(5): 409-420.

Williams, D.J. and J.M. Noyes, 2007. How does our perception of risk
influence decision-making? Implications for the design of risk
information. Theoretical Issues in Ergonomics Science 8(1): 1-35.

Wogalter, M.S., M.J. Kalsher, and B. Racicot, 1993. Behavioral
compliance with warnings: Effects of voice, context, and location.
Safety Science 16(5-6): 637-654.

Wogalter, M.S., P.B. Begley, L.F. Scancorelli, and J.W. Brelsford,
1997a. Effectiveness of elevator service signs: Measurement of perceived
understandability, willingness to comply and behaviour. Applied
Ergonomics 28(3): 181-187.

Wogalter, M.S., L.J. Frederick, A.B. Magurno, and O.L. Herrera, 1997b.
Connoted hazard of Spanish and English warning signal words, colors, and
symbols by native Spanish language users. Proceedings of the 13th
Triennial Congress of the International Ergonomics Association 3:
353-355. Cited in: Users’ hazard perceptions of warning components: An
examination of colors and symbols by T.L. Smith-Jackson and M.S.
Wogalter, 2000. Proceedings of the XIVth Triennial Congress of the
International Ergonomics Association and 44th Annual Meeting of the
Human Factors and Ergonomics Association, Ergonomics for the New
Millennium: 55-58.

Wogalter, M.S., R.J. Sojourner, and J.W. Brelsford, 1997c. Comprehension
and retention of safety pictorials. Ergonomics 40(5): 531-542.

Wogalter, M.S., M.J. Kalsher, L.J. Frederick, A.B. Magurno, and B.M.
Brewster, 1998a. Hazard level perceptions of warning components and
configurations. International Journal of Cognitive Ergonomics 2(1-2):
123-143.

Wogalter, M.S., A.B. Magurno, R. Rashid, and K.W. Klein, 1998b. The
influence of time stress and location on behavioral warning compliance.
Safety Science 29: 143-158.

Wogalter, M.S. and R. Rashid, 1998c. A border surrounding a warning sign
affects looking behavior: A field observational study. Proceedings of
the Human Factors and Ergonomics Society 42nd Annual Meeting: 1628.

Wogalter, M.S., D.M. DeJoy, and K.R. Laughery (eds.), 1999. Warnings and
Risk Communication. Philadelphia: Taylor & Francis.

Wogalter, M.S. and T.A. Dingus, 1999. Methodological techniques for
evaluating behavioral intentions and compliance. In M.S. Wogalter, D.M.
Dejoy, and K.R. Laughery (eds.), Warnings and Risk Communication
(53-82). Philadelphia: Taylor and Francis.

Wogalter, M.S., V.C. Conzola, and T.L. Smith-Jackson, 2002.
Research-based guidelines for warning design and evaluation. Applied
Ergonomics 33(3): 219-230.

Wogalter, M.S. and W.J. Vigilante, 2003. Effects of label format on
knowledge acquisition and perceived readability by younger and older
adults. Ergonomics 46(4): 327-344.

Wogalter, M.S. (ed.), 2006a. Handbook of Warnings. Mahwah, NJ: Lawrence
Erlbaum Associates, Inc.

Wogalter, M.S., N.C. Silver, S.D. Leonard, and H. Zaikina, 2006b.
Warning symbols. In M.S. Wogalter (ed.), Handbook of Warnings (159-176).
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Wogalter, M.S. and W.J. Vigilante, 2006c. Attention switch and
maintenance. In M.S. Wogalter (ed.), Handbook of Warnings (245-266).
Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Wolf, M.S., T.C. Davis, H.H. Tilson, P.F. Bass, and R.M. Parker, 2006.
Misunderstanding of prescription drug warning labels among patients with
low literacy. American Journal of Health-System Pharmacy 63(11):
1048-1055.

Wolford, R., M. Larson, D. Merrick, M. Andrews, S. Tillett, S. Morris,
and M. Keiffer, 1997. A comparison of safety-and-health training of
painters in Alaska, Oregon, and Washington. Silver Spring, MD: The
Center to Protect Workers’ Rights. Available online at
http://www.cpwr.com/pdfs/pubs/research_pubs/krpainters.pdf. Last
accessed 8 June 2007.

WorkSafeBC, 2006. Suppliers’ guide to WHMIS: Preparing compliant
material safety data sheets and labels. Workers’ Compensation Board of
British Columbia. Available online at
http://www.worksafebc.com/publications/health_and_safety/whmis/assets/pd
f/whmis_suppliers_full.pdf. Last accessed 6 June 2007.

Wright, M.J., 2004. Statement before the U.S. Senate Subcommittee on
Employment, Safety, and Training during its hearing on “Hazard
Communication in the 21st Century Workforce.” 25 March 2004.

Young, S.L., 1997. The role of pictorials in environmental safety signs.
Proceedings of the Human Factors and Ergonomics Society 41st Annual
Meeting: 797-800.

Young, S.L., 1998. Connotation of hazard for signal words and their
associated panels. Applied Ergonomics 29(2): 101-110.

Young, S.L. and D.R. Lovvoll, 1999. Intermediate processing stages:
Methodological considerations for research on warnings. In M.S.
Wogalter, D.M. Dejoy, and K.R. Laughery (eds.), Warnings and Risk
Communication (27-52). Philadelphia: Taylor and Francis.

Zuckerman, A. and S. Chaiken, 1998. A heuristic-systematic processing
analysis of the effectiveness of product warning labels. Psychology &
Marketing 15(7): 621.

 Some research articles, standards, and regulations use the term
“symbol;” others use “pictorial,” “pictogram,” or
“pictograph.” Some sources provide distinct definitions for these
terms, as described further in Section 6.2.5.2. However, others
essentially use the terms interchangeably. This report follows the
latter approach.

 Studies differ in how they define “younger” and “older” adults.
For example, some define “older” adults as those over ago 50, while
others define this group as those over 65. For information on the age
ranges in specific studies, see Section 6.2.7.2 or consult the studies
that are cited.

 The ANSI SDS used in this experiment reflected an earlier version of
the standard; however, the same 16-section format was in place at the
time.

 ANSI Z535.6 is a new standard; thus, no older versions were available
to review.

 “Mr. Yuk” is a cartoon image created by the Pittsburgh Poison
Center at Children’s Hospital of Pittsburgh (  HYPERLINK
"http://www.chp.edu/mryuk/05a_mryuk.php" 
http://www.chp.edu/mryuk/05a_mryuk.php ). The image—a scowling green
face with its tongue hanging out—is commonly used on household
chemicals to deter children from ingesting the product.

 Tactile, olfactory, and gustatory refer to touch, smell, and taste,
respectively.

Final Report		July 17, 2007

 PAGE   

  PAGE  ii 

 PAGE   89 

