| <<< Back to Hazard Communication -
Additional Information |
Printing Instructions
|
Hazard Communication: A Review of the Science Underpinning
the Art of Communication for Health and Safety
|
Submitted to ToxaChemica, International
in a subcontract to
the Occupational Safety and Health Administration
Contract number: J-9-F-5-0050
Task Order No. 2, Option Year 1 Prepared by Barbara Sattler, R.N., Dr. P.H.,
Bruce Lippy, CIH, CSP, and Tyrone G. Jordan, B.S.
The Environmental Health Education Center
The University of Maryland Medical School
May 23, 1997
|
1 INTRODUCTION
This report was commissioned by the United States Occupational Safety and Health Administration to review the state
of scientific inquiry supporting our knowledge regarding key elements of chemical hazard communication programs:
labeling, warnings, material safety data sheets, and worker training. This endeavor supports the international effort
to harmonize laws, regulations, and consensus standards affecting the ways in which information about hazardous
chemicals is communicated. The international effort can be divided into three major functions: classifying health and
environmental hazards, classifying physical hazards; and communicating hazard information. This last component
involves the determination of what information will be communicated to users regarding the hazards and appropriate
protective measures, as well as the way in which it will be transmitted, i.e. through symbols, labels, standard
phrases, and training.
Internationally, six organizations – the World Heath Organization, the International Labour Organization, the
United Nations Environmental Program, the FAO, UNIDO, and OECD – participate in the Inter-Organization Programme
for the Sound Management of Chemicals (IOMC). The IOMC has compiled a bibliography of references and has administered
a survey sent to member countries asking for information about their respective requirements on hazard communication.
Additionally, the countries responding to the survey were asked to submit any studies and supporting documentation
for their country's policies, laws and regulations. The resulting draft report was produced by the Working Group on
Harmonization of Chemical Classification Systems which was established by the Coordinating Group of the
IOMC. The
full text of this important report is contained Appendix A and is referenced throughout this document as the IOMC
draft report. The following key chart is taken directly from the report.
| |
Australia |
Canada |
Japan |
Korea |
Mexico |
Portugal |
Sweden |
UK |
USA |
Zimbabwe |
| Labels |
yes |
yes |
yes |
yes |
yes |
yes |
yes |
yes |
yes |
? |
| Text |
yes |
yes |
yes |
yes |
yes |
yes |
yes |
yes |
yes |
? |
| Symbols |
no |
yes |
no |
yes |
yes |
yes |
yes |
yes |
no |
yes |
| Colors |
no
(no) |
no
(no) |
no
(no) |
yes
(yes) |
yes
(yes) |
no
(yes) |
no
(yes) |
no
(yes) |
no
(no) |
?
(yes) |
| Format |
no |
yes |
no |
yes |
yes |
yes |
yes |
yes |
no |
? |
| Safety data sheets |
yes |
yes |
yes |
yes |
? |
yes |
yes |
yes |
yes |
? |
| Headings |
yes |
yes |
yes |
? |
? |
yes |
yes |
yes |
yes |
? |
| Symbols |
no |
no |
no |
? |
? |
no |
no |
no |
no |
? |
| Training |
yes |
yes |
yes |
? |
? |
? |
yes |
no |
yes |
? |
Legend
| Labels: |
|
if labels are required by regulations or recommended. |
| Text: |
|
if any specific written information is required or recommended
on the label. |
| Symbols: |
|
if any specific symbols are required or recommended on the
label. |
| Colors: |
|
if the label is required or recommended to have specific colors;
parentheses indicate if symbols are required or recommended to have specific colors. |
| Format: |
|
if the label is required or recommended to have a specific size,
shape, or design. |
| Safety Data Sheets: |
|
if data sheets are required by regulation or
recommended |
| Headings: |
|
if standardized headings are required or recommended on the data
sheet. |
| Symbols: |
|
if specific symbols are required or recommended on the data
sheet. |
| Training: |
|
if training is required or recommended. |
| ? |
|
means there was no indication in the documents received.
|
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. Further, this effort must be built on previous work. It is
impractical and ill-advised to ignore the fact that there are existing systems, several used internationally, that
should be revised to make them more effective.
An extensive literature search was conducted on hazard communication and the information is summarized within this
report. Not all citations entered in the bibliography are cited within the body of the report. Several U.S. agencies
have done similar reviews of literature regarding effectiveness of labeling, warning symbols, and hazardous chemical
training: the Environmental Protection Agency (EPA), the Consumer Product Safety Commission
(CPSC), the National
Institute for Occupational Safety and Health (NIOSH), and the National Institute for Environmental Health Science
(NIEHS).
There is strong international interest in the comprehensibility of information. In addition to summarizing the
literature, deficiencies in our knowledge have been identified and recommendations for further studies have been
suggested.
2 SEARCH METHODS AND INFORMATION SOURCES
In December 1996, the ILO's Working Group on Harmonization of Chemical Hazard Communication issued a draft document
entitled "Report on the responses to the call on chemical hazard communication."
The calls for information were actually sent in May and July 1995 to about 270 relevant national and international
institutions with a formal questionnaire going to a few organizations selected from this larger body. The requests
went out through the ILO's International Occupational Safety and Health Information Centre. Background materials on
hazard communication were requested, particularly scientific studies, with an emphasis on comprehensibility studies
and procedures used to implement hazard communication systems at the national level as well as by industry. The
information was collected as background information for creating a harmonized chemical hazard communication systems.
The work in harmonization is based on a number of international resolutions, conventions, and reports. In June 1992,
the United Nations Conference on Environment and Development (UNCED) stipulated that "a globally harmonized
hazard classification and compatible labeling system including material safety data sheets and easily understandable
symbols, should be available, if feasible, by the year 2000." Part B of Chapter 19 of UNCED's Agenda 21 is
devoted to harmonization of classification and labeling.
The Working Group on Harmonization of Hazard Communication addresses issues regarding hazard symbols, colors, and
written information used on labels; the preparation of chemical safety data sheets and instructions; the
comprehensibility of precautionary statements used on both labels and in chemical safety data sheets; and training
related to these areas.
The Coordinating Group for the Harmonization of Chemical Classification Systems brings together experts representing
existing national, regional and international systems of classification and labeling of chemicals, as well as
representatives of interested international organizations of suppliers, employers, workers, consumers and
environmental groups. The objectives are a finalized proposal for harmonized classification criteria by 1997 and
possible implementation of a globally harmonized system at the national level by the year 2000.
The field of occupational safety and health is by nature multidisciplinary and the literature describing its varied
works must be searched using several tools. Several bibliographic databases were searched back to 1980. No single
database identified the necessary literature to accomplish the task at hand. The health and safety databases had
little information on labeling. This information was found in a variety of databases including PsycLIT which more
extensively identified the human factors and ergonomic literature. The health and safety databases were also
deficient in their coverage of training efficacy; that body of literature was found in educational databases and to
some degree in the NIOSH database (NIOSHTIC). The use of Internet searching tools uncovered additional information
from a variety of sources including international agencies, national agencies, universities, and private
organizations. The literature cited and reviewed was limited substantially to peer reviewed documents, such as peer
reviewed journals and government reports. Because the literature on the effectiveness of labeling,
MSDS, and training
specifically in the context of workplace hazard communication was surprisingly sparse, the search was extended to
look at other applications of hazard communication, such as the use of labels for hazard warnings on consumer
products and in other public safety situations such as transportation safety. In this extended search, a substantial
literature was identified regarding the effectiveness of labeling and warning signs.
Additionally, agency staff were interviewed in several of the U.S. agencies responsible for product and chemical
labeling, including the Environmental Protection Agency, the Consumer Product Safety Commission, and the Department
of Transportation. Staff at the National Fire Protection Association were also contacted. Chairmen and key committee
members were contacted from the two main American National Standards Institute (ANSI) standards committees on
labeling and MSDS. Members of the Society for Chemical Hazard Communication were also contacted as a means of
identifying and locating materials and studies that may have been done within companies or trade associations and
therefore not retrievable through bibliographic searches. Some members of the HazCom Workgroup within the National
Advisory Committee on Occupational Safety and Health were also contacted for the same reason.
OSHA provided a bibliographic search that had been done by the International Occupational Safety and Health
Information Centre at the ILO. Approximately half of the documents listed were written in languages other than
English. The majority of the documents listed in this bibliography are governmental reports, guidance, and
information documents. Few of the documents are articles or books describing research on the effectiveness of hazard
communication.
A substantial bibliographic search had been done as the first phase of a current, U.S., multi-agency project entitled
the "Consumer Labeling Initiative." This project has the stated goal of making the health and safety
information on labels easier for consumers to understand. This bibliography provided a rich source of literature
including agency reports and research on labeling. The second phase of this project will continue in 1997. The
Consumer Products Safety Commission received a report in May, 1995 entitled, "Product Labeling Guide" which
contained an excellent literature review. The report was prepared by Bari M. Kotwal and Neil D. Lerner of COMSIS
Corporation and has been quite valuable for this present report.
The Society for Chemical Hazard Communication produced a compendium of citations entitled, "Published Research
on the Comprehensibility of Chemical Hazard Warning Labels and Material Safety Data Sheets." Although this
listed a number of articles on labeling, the vast majority of articles described labels in non-workplace settings,
such as on consumer products and in transportation contexts. Further, only one article listed referred to Material
Safety Data Sheets.
In 1995, a literature review entitled "Assessing Occupational Safety and Health Training" was published by
the NIOSH that sought evidence from the literature bearing on two questions: Are OSHA training requirements effective
in reducing workplace injury and illnesses? Does the available evidence show certain training factors or practices as
being more important than others in having a positive effect on illness and injury prevention? Additionally, a set of
papers was collected for a 1996 workshop sponsored by the National Institute for Environmental Health Science,
entitled "Measuring and Evaluating the Outcomes of Training."
All of the above-mentioned searches, reviews, and listings are included in this report either as independent
documents or subsumed into the comprehensive bibliography.
The following databases were searched for this report using the key words identified for each database:
| 1. DIALOG, ABI/INFORM and COMPENDEX |
|
KEY WORDS: |
risk management, safety management, human factors and
communication, behavior and communication, compliance/non-compliance, chemical, toxic, worker, employee,
occupational health occupational injury, health, injury |
| 2. MEDLINE (total of three searches) |
|
KEY WORDS: |
material safety data sheets, hazard communication, risk
communication, hazardous substances, hazardous chemicals, international, product labeling, label* |
| 3. IPA (International Pharmaceutical Abstracts) |
|
KEY WORDS: |
package inserts, labels, safety, hazard, caution,
compliance/non-compliance, hazard communication, risk communication |
| 4. PsychLIT |
|
KEY WORDS: |
decision*,
working conditions, compliance/non-compliance, workplace, employ*,
safety, health, warn*, human factors, behavior,
material safety data sheet |
| 5. NIOSHTIC |
|
KEY WORDS: |
training, material safety data sheets, MSDS, hazard
communication |
* Denotes any variation on the root word.
3 REVIEW OF LITERATURE
3.1 LABELS AND WARNINGS
3.1.1 SUMMARY OF RESEARCH WITH RECOMMENDATIONS
The following information represents a summary of the general areas of consensus in the literature on labels and
warnings. The summary also includes what the authors of this report consider key findings and recommendations.
- It is difficult to find comprehensive, usable and definitive guidance for the design or evaluation of labels, in
spite of the existence of several label standards such as ANSI Z535 standards and the ISO standards.
- Label effectiveness is often judged on the basis of subjective opinion, particularly what would be done upon
encountering the label. The relative validity and ultimate safety consequences of labeling alternatives are poorly
understood, however.
- Much of the literature has also been characterized by the use of college students as research subjects, rather
than workers, the key target audience.
- Perceived hazard, familiarity with a product, and gender influence the consumer's decision to look for a warning
message on the labels of potentially hazardous household products.
- Warnings laid out in outline form and organized by type of hazard were ranked as having greater eye appeal,
easier to process, and more effective than alternative approaches such as paragraph layouts.
- Educational efforts are essential if symbols or colors are to communicate effectively. Even brief training –
as little as giving the pictorials' verbal meaning once – can have a large impact in facilitating comprehension
for pictorials that would otherwise not be understood by many people.
- Flame and poison warning symbols prescribed by the Canadian government are well understood generically but
people have difficulty inferring the specific safety precautions most necessary for a particular product.
- The addition of pictographs to a verbal warning will increase compliance rates but at least one key study showed
that enclosure shapes made no difference in compliance rates.
- Conducting behavioral studies is preferable to relying on preference data but is rarely conducted.
- Studies of the ANSI Z535.3-1991 protocol for testing found that the standard slash recommended by ANSI was not
as effective as an "X" for warning consumers of prohibited uses.
- Warnings containing a pictorial, color, or an icon elicited significantly faster response times among subjects
than warnings without them. However, the addition of a border did not improve response times.
- The shape of a label may enhance the conspicuousness of a warning, even if the targeted user is not actively
searching for safety information. This may be because the public may become more sensitive to certain shape stimuli
which they have come to associate with danger much as they do with color.
- Even with reasonable attempts to mandate legibility there have been a number of factors not specified that can
have significant effects.
- One of the key variables on warning effectiveness is where the information is placed.
- Warnings printed horizontally were found more quickly than warnings printed vertically.
- If safety information is perceived as familiar or redundant, it may simply be dropped from short term memory,
and have no further effect on behavior. Alternatively, if the information is perceived as inconsistent with
existing knowledge and beliefs, it may be rejected as not credible. Even if the information is previously unknown
and is accepted as true, the recipient may be insufficiently motivated to alter behavior.
- 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.
- Women report a greater likelihood than men to look for and read warnings. They are also more likely to comply
with warnings.
- It may be necessary to apply relatively strong signal words for older users: they exhibit a lower level of
comprehension. Those over 40 are more likely than younger subjects to take precautions in response to warnings they
understand, however.
- Perceived severity of consequences has a strong influence on behavioral intentions.
- Higher stress produce significantly lower compliance.
- Hazard information communicated by different colors followed a consistent pattern across language groups. Red
resulted in the highest hazard ratings followed by orange, yellow, blue, green, and white, respectively.
- The importance that ANSI and the U.S. military attaches to signal words may not be shared by the population at
large. Most studies indicated that DANGER connoted greater strength than WARNING and CAUTION. But generally the
results failed to show a difference between WARNING and CAUTION.
3.1.2 INTRODUCTION
One of the most comprehensive review of literature on labels and warnings was performed by COMSIS Corporation under
contract to the U.S. Consumer Product Safety Commission (Kotwal and Lerner, 1995). The authors described the state of
knowledge on labels and warnings with the following excellent summary:
"Although warning label design may superficially appear to be a simple matter, proper design and use is, in
fact, rather complex. It is clear that inappropriately designed or used warnings simply will not be effective, by
any standard of evaluation. Yet, it is difficult to find comprehensive, usable and definitive guidance for the
design or evaluation of labels, in spite of the fact that there are numerous label standards (e.g., ANSI Z535
standards, ISO standard), label design guides (e.g., Westinghouse and FMC handbooks), general human factors guidance
for warnings and labels, and various literature review articles... Unfortunately, the findings of the research
literature on labels are often contradictory and frequently weak in terms of face validity or experimental
procedures. Label effectiveness is often judged on the basis of subjective opinion or people's reports of what they
think they would do if they encountered the label. However, the relative validity and ultimate safety consequences
of labeling alternatives is poorly understood.
The literature does contain a number of experimental studies that did directly measure some behavior, such as
reading a label or complying with some action, although these studies have usually been rather narrow in approach.
Furthermore, there have been many examples of poor research design, or inappropriate conclusions drawn from data.
Much of the literature has also been characterized by the use of samples of convenience for the population of
research subjects. Very often the subjects are comprised of college students. This is, at best, a narrow sample and,
quite often, an inappropriate one, as these people may differ from typical users of the product in terms of product
knowledge, hazard perception, perceptual abilities, or safety motivation. Also, the research literature has
sometimes been characterized by a sort of naive absolutism. For example, a study may ask the question whether it is
most effective to present the message element of the hazard, the required action, the injury outcome, or the level
of danger, or some combination of these. While such research is certainly useful, when designing an actual label,
the question very likely becomes what information is most effective for this application (considering the hazard,
public understanding of the hazard and related actions, space considerations, the obviousness of the hazard on the
product, competition with other messages, and many other factors). Thus, the limitations of existing research make
it difficult for a designer to find definitive guidance for preparing an effective label or for an analyst to make
consistent and defensible evaluations of product labeling practices."
3.1.3 COMPREHENSIBILITY
Comprehensibility refers to the ability of the individual reading a label, warning, or material safety data sheet to
understand the information sufficiently to take the desired action. Comprehensibility is different from readability
because the latter is simply a measure of the grade level of the written material while the former is a measure of
how well the receiver of the information understood it. A warning about incompatible chemicals may be written at the
correct reading level for a specific audience but may do such a poor job explaining the concept of incompatibility
that the warning isn't understand by most of the intended audience. Additionally, the same warning may be highly
comprehensible to a population of chemical workers but poorly understood by firefighters with the same educational
level but different work experiences. Finally, achieving comprehensibility does not ensure that the informed
individual will take the actions prescribed in the warning or label. This final, behavioral step is affected by a
complex mix of attitudes, experiences, motivations, and potential consequences that are specific to each individual
in a particular situation.
Two experiments involving warning labels on household products (Godfrey and
Laughery, 1993b) indicated that subjects
are able to discriminate among products based on overall hazard. The more hazardous they perceived a product, the
more likely it was they would look for a warning. Further, females are more likely to look for warnings than males.
In summary, perceived hazard, familiarity, and sex influence the consumer's decision to look for a warning message on
the labels of potentially hazardous household products.
Other researchers looked at the variables related to people's perceptions of hazard for various types of household
pest-control products (Silver and Wogalter, 1991a). Seventy college-aged students, 20 older subjects, and four
pesticide experts judged 26 pest-control products on hazard. Results showed that fumigators and foggers were the
products perceived as the most hazardous followed by sprays, systems, and traps, respectively. Although several
misperceptions of hazard for certain product categories were evident, the judgments of the students and older adults
were consistent with those of the experts. Perceived hazard was also found to be positively correlated with a number
of objective characteristics of the product labels, including the quantity of chemical ingredients, effectiveness
against pests, number of words and sentences on the label, readability, and the presence and location of certain
statements on the back label. These finding suggest that people can discriminate the hazard levels of different types
of pest-control products, and the presence of various cues on the label may aid this determination.
Desaulniers (1980) examined the effects of warning layout and semantic organization on the readability and recall of
warning information. In general, warnings in outline layout and type of hazard organization were ranked as having
greater eye appeal, easier to process, and more effective than alternative organization layout conditions (i.e.,
paragraph layouts). The results indicated that warnings in outline layout were read and complied with by a larger
proportion of subjects than warnings in paragraph layout. Collins and Lerner (1982b) also noted other important
criteria in addition to understandability in fire safety alerting measures. Variables such as visual range,
detectability in smoke, and legibility all were critical.
3.1.3.1 Use of Symbols and Pictograms
There has been a great interest for years in depicting critical safety and health information as succinctly as
possible for immediate response by individuals in dangerous situations. The focus has been to avoid the numerous
shortcomings of written language, particularly the length of time it takes to read a warning and the proximity the
reader must have with the information in order to see it. Additionally, the increasing number of American workers who
do not speak English as a first language or who are illiterate argues well for the use of symbols and pictograms.
This is also essential for any effort towards international harmonization.
Symbols and pictograms are widely used in warning labels. However, this trend, at least in the United States, is
relatively recent. Research on the use and effectiveness of pictorial signage began in the early 1970's and continued
into the early 1980's. At the same time, the use of symbols burgeoned for applications such as the highway, public
facilities, workplace signs, and product labels. The sizable body of research on symbol effectiveness from this
decade of activity primarily focused on signs, for highways, workplaces, and public facilities, rather than labels.
Collins' 1982 review of literature through the early 1980's reported that "there has been little if any research
on the effectiveness of symbols for product warnings." According to Kotwal and Lerner (1995) although there has
been subsequent research specifically on labels, most research still has been based on signs; much of current
practice is an extrapolation from sign-related research or is based on logical argument.
One study showed college students 16 symbols from the 1981 Westinghouse Product Safety Label Handbook. For each
graphic, the subject was asked to describe the hazard. Recognition rates ranged from nearly 100% to almost completely
unrecognizable (Laux, Mayer and Thompson, 1989). Results indicated that there are two questions which must be
addressed in the development of symbols and pictorials: 1) will the user recognize the hazard? and 2) will the
recognition of the hazard suggest proper precautionary behavior?
The U.S. EPA (1986) received universal agreement from their advisory committee that educational efforts are essential
if symbols or colors are to communicate effectively. They cited the experience of the Canadian government which
instituted a system of symbols to convey degree of hazard in the mid-seventies. They since undertook an educational
effort which resulted in 95% of Canadians understanding the meaning of the symbols. A survey conducted in the U.S.
where no similar educational campaign was held revealed comprehension below 25%.
Because of their relatively universal information transmission potential, pictorials have been suggested as common
means of safety communication across heterogeneous groups of users and uses (Brelsford,
Wogalter, and Scoggins,
1994). The researchers conducted a study and found that easy pictorials were comprehended (both initially and
following training) better than difficult pictorials, although the latter showed the most dramatic increase in
understandability after training. In addition, they found that the substantial gains measured after training suggest
that even brief training - as little as giving the pictorials' verbal meaning once - can have a large impact in
facilitating comprehension for pictorials that would otherwise not be understood by many people.
Boersema and Zwaga (1989) identified five warning messages that were used to reduce the hazardous behavior of
swimming pool slide users. For each warning, nine symbols were developed. The comprehensibility of the symbols was
tested using an evaluation procedure based to a large extent on the International Standards Organization testing
procedure to determine the comprehensibility of public information symbols (ISO 9186, Procedures
for the Development and Testing of Public Information Symbols). Respondents were swimming pool users between 7
and 19 years of age. Seven acceptable symbols were found referring to four of the five warnings.
Collins and Lerner (1982b) evaluated U.S. participants' understandability of twenty-five internationally proposed
symbols for fire-safety alerting. The poor performance of some critical symbols such as "exit" was noted,
and some potentially dangerous confusions in meaning were revealed. As a result, the researchers advise incorporating
testing procedures as integral parts of the symbol development and standardization process before widespread
adoption.
Frantz, Miller, and Lehto (1994) examined the impact of the flame and poison warning symbols prescribed by the
Canadian government. The results suggest that although the generic meanings of these two symbols are well understood,
people have difficulty inferring the specific safety precautions most necessary for a particular product. In the test
case of extremely flammable floor adhesive, users apparently did not realize that adhesive vapors, rather than the
adhesive itself, pose the fire hazard.
The authors of the FMC Corporation Product Safety Sign and Label System manual believe that communication
effectiveness between a greater cross-section of the population, both nationally and internationally, could be
increased by using pictorial or symbolic language in place of, or as a supplement to, written words. The pictorials
are combined with words and colors in specifically designed formats intended to present comprehensive hazard
information in an orderly and understandable manner.
Jayne and Boles (1993) investigated whether different warning designs, specifically those with symbols, affect
compliance rates. Five conditions were tested: a verbal warning, a pictographs warning with a circle enclosing each
graphic, a pictographs warning with a triangle on its vertex enclosing each graphic, a warning with both words and
pictographs, and a control (no warning). Participants performed a chemistry laboratory task using a set of
instructions that contained one of the five conditions. The warnings instructed them to wear safety goggles, mask and
gloves. All four warning conditions had significantly greater compliance than the no-warning condition. A significant
main effect was found for the "presence of pictographs" variable, suggesting to the authors that the
addition of pictographs to a verbal warning will increase compliance rates. The enclosure shape made no difference in
compliance rates, despite research that indicates that unstable shapes are preferred as warning enclosures. The
authors echoed the recommendations of Horst that conducting behavioral studies was preferable to relying on
preference data.
3.1.3.2 Shape of labels and warnings
The proper shape of warning signs has been evaluated. The triangle arranged on its vertex was the preferred shape,
with the circle – despite its common use – scoring among the least preferred warning shapes (Riley, Cochran, and
Ballard, 1982a). Similarly, the Consumer Products Safety Commission (Federal Register, May 3, 1996) tested several
pictograms for effectiveness in warning about the hazards of carbon monoxide poisoning from burning charcoal briquettes
in confined spaces. They followed the ANSI Z535.3-1991 protocol for testing and found that the standard slash
recommended by ANSI was not as effective as an "X" for warning consumers of prohibited uses.
Work by Young (1993) focused on increasing the noticeability of warnings by manipulating four variables: pictorial,
color, signal icon and border. Subjects viewed 96 simulated alcohol labels on a computer, half with a warning and
half without. The results showed that warnings containing a pictorial, color, or an icon had significantly faster
response times than warnings without them. However, the addition of a border did not improve response times. More
detailed analyses demonstrated that pictorials, color and icons can enhance the noticeability of warning information
and interact with each other in such a way to that they should not be used indiscriminately without adequate
knowledge of the interactions.
Warning labels are most often rectangular in shape. The orientation of the rectangular shape is determined by the
format of the label and the number of elements contained within the label (i.e., signal word, pictograms, message
text). The shape of any symbols contained within the label, however, are drawn from the pool of shapes commonly
associated with safety, such as octagons, diamonds, triangles, and circles. While the shapes of the elements within
labels are somewhat standardized and commonly chosen for prompt user recognition, the effects of the shape of the
label itself have not been scrutinized as closely Kotwal and Lerner, 1995). Rectangular shapes are predominantly used
because of their simplicity.
Besides allowing for efficient use of space, the shape of a label may enhance the conspicuousness of a warning, even
if the targeted user is not actively searching for safety information. This may be because the public may become more
sensitive to certain shape stimuli which they have come to associate with danger much as they do with color. As a
result of this perceptual learning phenomenon, Cunitz (1992) suggested that standardized shapes (e.g., octagon shaped
stop sign, triangle shaped yield sign) can take on greater attention-getting capabilities than other shapes.
ISO 9186, Procedures for the Development and Testing of Public Information Symbols
specifically recommends the use of surround shapes. The defined shapes are used to distinguish different alerting
categories. However, these shapes only surround the pictogram portion of the warning. If a verbal message is included
in the warning, it is not included in the surround shape. ANSI Z535.3 (1991) does state that the perimeter of the
label panels may take on the surround shape of safety symbols, although justifications for this allowance are not
specified. Safety shapes recommended by ANSI Z535.3 (1991) and ISO 9186, along with their associated meanings are as
follows:
- Equilateral triangle resting on base: Warning and hazard alerting
- Circle: Mandatory action
- Circle with 45 degree slash from upper left to bottom right: Prohibited action
- Square or oblong as necessary to accommodate text: Information
All other recommendations in current guidelines regarding shapes of elements are actually
referring to shapes of symbols which are incorporated as elements within the rectangular borders of the label
(Kotwal
and Lerner, 1995).
3.1.3.3 Use of Multiple Media
There have been several important studies that point to the value of presenting messages in several media
simultaneously. For instance, a symbol may be combined with written information or a signal word to reinforce the
effect. In one study, fifty-eight subjects were shown randomly-ordered facsimiles of standard danger signs required
by OSHA and rated the signs on 13 dimensions related to perceived effectiveness. Signs containing a hazard label and
instructions (e.g., GASOLINE - NO SMOKING) were rated as least likely to be recalled at a later time; however, they
were rated as easiest to understand, most informative, and most likely to be complied with. Signs containing a hazard
label only (e.g., POISON) were rated as least informative and most difficult to understand; however, they were rated
as most likely to be recalled, as depicting a high degree of danger, and likely to be complied with. Signs containing
instructions only (DO NOT ENTER) were rated as generally least effective.
Dingus, Hunn and Wreggit (1993) found that the inclusion of protective devices in the packaging of hazardous consumer
products (e.g. gloves in hair coloring products) decreased the cost of complying with warnings specifying their use
and led, consequently, to more compliance. See the discussion in Section 3.1.5.8, Cost of Compliance.
3.1.4 READABILITY
3.1.4.1 Legibility
Legibility involves both the typographical elements and the surface upon which the message is printed (Hale, 1991).
This was abundantly illustrated when Congress required a health warning on tobacco packages but provided little in
the way of specifications. The labels produced could scarcely be read in many instances, due to type size and the
contrast between the ink and paper chosen. In 1988, Congress passed a similar requirement for beer and wine but this
time included specific requirements, including that the first two words must be "GOVERNMENT
WARNING," set in boldface capitals.
A study commissioned by the Public Health Service (1990) looked at the effectiveness of these proscribed warnings and
found that when the color of the typography and the color of the background were measured, a Contrast Ratio (CR)
could be derived, which proved to be a valid measure of legibility. Specimens with high CR values were less legible,
despite containing the same warning in the same size type. Hale (1991) concluded that even with reasonable attempts
to mandate legibility there have been a number of factors not specified that can have significant effects. "If
regulators and others who wish to formulate rules for legibility are to enjoy even modest success, it is clear that
they will need the assistance of well-defined standards covering all the attributes described (Hale, 1991)."
Howett (1983) derived a formula giving the letter stroke-width needed for legibility of words on a sign at any given
distance by an observer with any given visual acuity. The stroke width, in turn, determines the letter size,
depending upon the characteristics of the type face used. The derivation is strictly mathematical and is based on the
assumption that beyond a distance of a few meters, a person's visual acuity is specifiable by a fixed visual angle,
independent of the distance. The author does provide graphs for correcting the critical stroke width for nonstandard
contrast or background luminance based on a body of data on visual acuity as a function of contrast and background
luminance.
3.1.4.2 Reading level evaluations
Dangerous accidents may occur when people misread written materials, consequently materials must match the reading
level of the intended audience. There are several internationally recognized instruments for measuring the reading
level of written materials but none of them deal with the comprehensibility of the materials, as indicated earlier.
Kreindler and Luchsinger (1978) suggest determining the reading-grade level of all safety materials using
standardized methods such as the FORCAST formula or the Flesch Reading Ease Index. Laughery
and Brelsford (1996) reported commonly finding product warnings written at significantly higher grade level than the
grade 4 to 6 range that is usually recommended for general target audiences. A discussion of reading level measures
and their application to the design of instructions and warnings can be found in Duffy (1985).
3.1.4.3 Type Fonts and Type Size
Braun and Silver (1992) looked at over-the-counter and prescription drugs and examined certain variables associated
with legibility, namely font type, font weight, point size contrast between the signal word and the main body of the
warning. The results indicated that participants were more likely to read the warning in Helvetica type (a sans serif
face) than in Times or Goudy (both serif faces). Times was more likely to be read than
Goudy. Bold type was more
likely to be read than Roman type. There was a greater likelihood of reading the warning when the main body was in
10-point size as compared to 8-point size. A 2-point size difference between the signal word and the main body of the
warning produced a greater likelihood of reading the warning over a 4-point size difference. One possibility for this
result is that the 4-point size difference minimizes the importance of the main body of the warning, therefore making
only the signal word salient.
Looking at pesticide labeling, Silver, Kline and Braun (1994) obtained different results from their earlier work with
prescription drugs. The variables used were the same: font type, point size contrast between the signal word and the
main body of the warning (signal word-text size difference), and point size of the signal word. College students
rated 36 insecticide labels that contained a warning which varied across all levels of the three variables. Results
showed that Century Schoolbook was perceived as more readable than Bookman or Helvetica. Moreover, greater perceived
readability was obtained when there was no difference between the point sizes of the signal word and the body of the
warning. Perceived readability and perceived hazardousness decreased as the signal word-text size difference
increased. There was greater perceived readability and perceived hazardousness when the signal word was printed in
14-point type than in 12- or 10- point types.
3.1.5 PLACEMENT OF INFORMATION
One of the key variables on warning effectiveness is where the information is placed. This includes placement of
information within a label as well as the physical placement of a warning sign in proximity to the hazard the warning
describes. In tests on the effectiveness of three different warning label designs for a portable electric heater,
researchers Gill, Barbera, and Precht (1993) found that the design requiring the user to interact with the label in
order to use the heater was more effective in attracting the user's attention that a traditional label or a
color-coded "ski-pass" label attached near the male end of the electric cord. However, none of them were
effective in mediating safe user behavior.
An experiment on the noticeability of warnings on alcoholic beverage containers indicated that warnings on the front
label were found more quickly than warnings appearing in any other location. Also, warnings printed horizontally were
found more quickly than warnings printed vertically. The signal phrase ("Government Warning"), as well as
the amount of "noise" or clutter on the surrounding label, significantly influenced warning detection
times. The authors suggested that proper manipulation of these design features could make the mandated warning more
noticeable (Godfrey et al. 1993).
Godfrey and colleagues (1993c) also looked at the effect of the physical size and location of the warnings. Warnings
were posted on a copy machine, a public telephone, a water fountain, and two sets of doors. Most people did not use
the copy machine and the telephone when they had the warnings on them. The warning on the water fountain was not
effective when a single, small warning was used. When that plus a larger, more forceful warning was used, most people
did not drink from the water fountain. The warnings on the doors were not obeyed unless there was a convenient
alternative exit available. Warnings with a low cost of compliance are apparently more effective. Additionally, the
warnings cannot be too small and must be placed strategically.
A paradigm was developed to examine the effectiveness of warnings in a laboratory task
(Wogalter, Fontenelle, and Laughery, 1993). Compliance (use of mask and gloves) was affected by the inclusion of the warning as well as by its
location. Greatest compliance occurred when the warning was placed prior to the instructions. The addition of a
printed statement placed before the instructions (with warning at the end) to read through the instructions before
beginning produced intermediate compliance that was not significantly different from the warning beginning and end
conditions. Observation revealed that when the warning message was at the end of the instructions subjects complied
only when they saw the warning message before starting the task. These results indicate that if warnings are placed
in front of instructions the consumer is more likely to read and comply.
Another chemistry lab format was used by Wogalter, Kalsher, and Racicot (1992) to examine the efficacy of two
warning-related factors to produce cautionary behavior. Experiment 1 compared the effects of a posted-sign warning
and a within instruction warning on behavioral compliance. The results showed that a warning embedded in a set of
task instructions produced significantly greater compliance (the wearing of protective gear) than a similar, larger
warning posted as a sign nearby. Experiment 2 reexamined the effect of location and also examined the influence of
the presence versus absence of pictorials. The results of Experiment 2 confirmed the location effect of Experiment 1.
No influence of pictorials was noted, although there was a nonsignificant increase in compliance when pictorials were
added to the within-instruction warning. The results indicated that warning placement is important for eliciting
behavioral compliance to safety messages. Explanation such as difference in field of view and perceived relevance are
discussed.
FMC Corporation (1985), as part of their safety program, emphasized the appropriate placement of safety signs and
labels to reduce the occurrence of accidents. Frantz (1992) found that contrary to current, recommended practice,
substantially more subjects read and complied with warnings and instructions that appeared in the directions for use
rather than the "precautions" section. On average, moving an instruction from the precautions into the
directions increased the reading rate from 37% to 89% and increased the compliance rate from 48% to 83%.
3.1.6 AUDIENCE VARIABLES
3.1.6.1 General considerations
McCarthy and colleagues (1984) examined approximately 400 articles on the effectiveness of on-product warning labels
and were able to categorize the literature based on the manner in which the warning issue was addressed:
Qualitative - editorial. These appear to be based on the expectation that providing
safety information is related to an increase in safety behavior.
Recognition - recall. These are the most represented quantitative works in the
literature addressing recognition of the warning message and ability to recall that message at a later time.
Standards and label systems. Articles on consensus, governmental, and industrial
standard systems.
Quantification of effectiveness. These are among the least represented.
The authors concluded for nonquantitative studies that not one of the design variables discussed in label creation
"is related in any known way to the actual effectiveness of warnings in changing behavior and reducing
injuries." Further, they felt that all available quantitative studies showed the warnings under investigation to
have no impact on safety. Horst (1993) reviewed the literature of experimental psychology and communication theory
and pointed out that people learn to filter out most of the information that bombards our senses and to attend to
things that they have learned to view as important or interesting. If safety information is perceived as familiar or
redundant, it may simply be dropped from short term memory, and have no further effect on behavior. Alternatively, if
the information is perceived as inconsistent with existing knowledge and beliefs, it may be rejected as not credible.
Even if the information is previously unknown and is accepted as true, the recipient may be insufficiently motivated
to alter behavior.
Similarly, Miller and Lehto (1986) looked at 388 reference sources on the subject of warnings and found only 10
sources that they considered true analyses of warnings effectiveness, and of these, only 6 actually cite any
experimental results. Purswell, Krenek, and Dorris (1993) do not find this paucity surprising because of the
difficulties associated with performing these studies. There is the general problem of collecting experimental
observations and the specific problem of inadvertently introducing an experimental bias because the subject is
influenced to read the warning and behave safely if he or she perceives that the purpose of the study is to evaluate
warning effectiveness.
Horst (1993) also reported that there are widely encountered guidelines - based primarily on preference studies –
that warnings should comprise four elements: a standard signal word, statements of the hazard, the potential
consequences, ad how to avoid the hazard. Horst et al. claim it is erroneous to suggest that all of this information
must be provided to affect behavior or that inclusion of all of this information will guarantee effectiveness. The
authors gave the examples of stop signs and restroom signs – neither of which carry all four elements but are
effective warnings. Stop signs are effective because the negative consequences of failing to respond appropriately
are universally familiar and highly probable. Similarly, the high probability of social embarrassment is clearly a
motivator in the effectiveness of restroom signs.
3.1.6.2 Gender
Laughery and Brelsford (1993) reviewed the literature and reported "reasonably clear trends" that indicate
women report a greater likelihood than men to look for and read warnings. They are also more likely to comply with
warnings.
3.1.6.3 Age
Kotwal and Lerner (1995) concluded from their comprehensive and excellent review of literature on warning labels that
the age of the user appears to have an effect on the perceived level of hazard conveyed by common signal words.
Leonard, Hill, and Karnes (1989) found significant effects between younger college students and older college
students (over age 25). Kotwal and Lerner (1995) summarized their findings with the following statement:
"Older subjects generally used signal words that implied greater hazards to represent the amount of risk
involved in a given situation. Since older users have indicated that a given signal word implies a lower level of
hazard than the same word implies for younger users, it may be necessary to apply relatively strong signal words for
older users in order to connote a given level of hazard."
Desaulniers (1991) reported that older people, 40 and above, are more likely to take precautions in response to
warnings as reflected in their behavioral intentions. However, Collins & Lerner, (1982) found that older subjects
exhibited lower levels of comprehension for safety signs involving pictorials. For Laughery and Brelsford (1993)
these studies taken together suggest that older receivers are more influenced by warnings, but greater attention to
issues of comprehension may be necessary.
3.1.6.4 Hazard and risk perception
A consistent finding in warning research is that people's perception of the hazardousness associated with a product
or situation can determine the effectiveness of a warning or label (Laughery and
Brelsford, 1993). Many researchers
have acknowledged that the actual design of the warning label may not be as important as the expectations that the
user brings to the situation (Kotwal and Lerner, 1995). Research has indicated that perceived severity of
consequences has a strong influence on behavioral intentions. Numerous studies have shown that the greater the a
priori perception of hazardousness, the more likely people will look for and read a warning and the more likely they
will comply by taking precautions (Donner and Brelsford, 1988; Friedman, 1988;
Otsubo, 1988). Additionally, the more
explicit the warning is about the severity of injury consequences, the greater the perceived hazardousness and recall
of warning information.
Purswell, Schlegel, Kejriwal, and Sashi (1993) reported that there have been surprisingly few attempts to develop a
useful measure of risk-taking behavior, given the importance of this factor in safe behavior. They developed a model
to predict whether a consumer would use a product safely as a function of sixteen different individual variable.
Subjects were presented with four consumer products to use in an experimental setting where the true purpose of the
study was concealed. Discriminant analysis was used to develop a prediction model to classify subjects into
categories of safe or unsafe behavior. Prediction accuracy ranged from 68 - 86 percent for different types of
behavior.
Other researchers looked at people's willingness to read warnings on household pest-control products and their
likelihood of purchasing these products (Silver and Wogalter, 1991b). Seventy college-aged and 20 older participants
examined a set of pest-control products and responded to a questionnaire assessing perceptions of the products, the
packaging, and the warnings. Results show that product hazardousness, warning understandability, and warning
attractiveness were strongly related to people's willingness to read the warnings. A different set of variables was
related to purchasing intentions. Participants reported greater willingness to purchase products that were more
familiar and had more attractive packaging. Participants were more willing to read warnings that contained more
statements and had readability scores at higher grade levels, a result that appeared to be due to their common
relationship with perceived hazard. The findings suggested that manufacturers can place appropriate and effective
warnings on pest-control products without necessarily reducing buying intentions.
Bogett and Rodriguez (1987) explored the influence of a perception of danger particularly regarding product-borne
warnings and safety instructional programs. The results of their study, together with supporting data from the
literature implied that a perception of danger (i.e., an unacceptable risk of loss or injury) must exist in order to
elevate a person's safety behavior.
Following Pages: Table from Kotwal and Lerner (1995) Summary of Research Findings Regarding User
Age, Gender, Perceived Hazardousness, and Familiarity With Product
| AUTHOR |
TITLE |
METHODOLOGY |
FINDINGS/
CONCLUSIONS |
| Godfrey, S.S., Allender, L., Laughery, K.R., and Smith, V.L.
(1983) |
Warning messages: Will the consumer bother to look? Proceedings
of the Human Factors Society 27th Annual Meeting, 950-954. |
EXP 1
- Subjects: 32 undergraduate college students
- Subjects were asked to imagine themselves in various purchasing situations and were presented 8 generic common
consumer products (plant food, oven cleaner, pesticides) in a booklet format
- Subjects were asked to rate products on four hazard scales (inhalation, swallowing, skin irritation, and
overall), three likelihood scales (likelihood to look for warning message, to read the message, and to comply with
the message), and on familiarity with the product |
- Found that subjects were more likely to look for warnings on
less familiar and more hazardous products
- Found that females were more likely to look for a warning than males
- However, for products that were most hazardous (e.g., pesticides), subjects' degree of familiarity did not
matter; subjects still reported that they would look for and read warnings regardless of their degree of
familiarity with the product |
| |
|
EXP 2
- Subjects: 88 undergraduate college students
- Study was conducted in a manner similar to a lab practical, using 12 different "stations"
- 40 household products were used in the experiment; each subject saw 12 of the 40 products
- Subjects were asked to respond to questions regarding familiarity with the product, similarity to other
products, frequency of use, recency of use, and perceived hazardousness |
- Findings indicate that subjects have lower hazard perceptions
for products that are familiar |
| Godfrey, S.S. and Laughery, K.R. (1984) |
The biasing effects of product familiarity on consumers'
awareness of hazard. Proceedings of the Human Factors Society 28th Annual Meeting, 483-486. |
- Subjects: 110 female students (most were undergraduate and
graduate college students; 17 were high school students)
- Surveyed women on their awareness of the hazards of tampon use, knowledge of the symptoms of toxic shock
syndrome, and awareness of warnings |
- Found that women who were more familiar with tampon products
were less likely to notice warnings when they switched brands
- Data suggest that being familiar with the product or a similar product leads to lower perception of hazard |
| Leonard, S.D., Matthews, D., and Karnes, E.W. (1986) |
How does the population interpret warning signals? Proceedings
of the Human Factors Society 30th Annual Meeting, 116-120. |
Subjects: 368 college students
- Studied whether different signal words produce different perceptions of risk
- Subjects viewed warning signs which varied according to signal word (DANGER, WARNING, CAUTION), signal word
color (RED or BLACK), signal word size, and information pertaining to the consequences of disregarding the warning
(consequence vs. no consequence)
- Subjects were told where the signs might be located and were provided scenarios in which the warning sign might
have relevance; subjects rated the perceived risk associated with each sign and the likelihood that they would
obey the sign displayed in the scenarios provided |
- Found that, given a description of the hazard being warned
against, subjects did not rate risk differentially as a function of the signal words
- Neither the size of the signal word nor the color significantly affected the perception of risk
- The perception of risk seemed to be predicated on the information content of the sign
- Concluded that information about consequences could raise the average perception of risk
- Note that one possible explanation of the results is that prior experience has generated a certain level of
perception of risk, based on personal experiences and possibly on vicarious experiences through the communications
media |
| Leonard, D.C., Ponsi, K.A., Silver, N.C., and Wogalter, M.S.
(1989) |
Pest-control products: Reading warnings and purchasing
intentions, Proceedings of the Human Factors Society 33rd Annual Meeting, 436-440. |
- Subjects: 70 undergraduate college students (17-19 yrs) and
20 older adult undergraduate students (mean age 37 yrs, std dev 7.7 yrs)
- Warnings on 22 pest-control products were evaluated
- Subjects were given a questionnaire to assess perceptions of the product's packaging, labeling, and warnings;
responses were recorded using a 9-point Likert scale (0 to 8)
- Examined whether several objective measures of the warning readability (statements, words, grade level) would be
related to the willingness-to-read variable |
- Warnings with more information and having more difficult
material may cause perceptions of hazardousness
- Found that perceived hazardousness is an important determinant of willingness to read warnings
- Hazard perception is more important than perceptions of familiarity with regard to willingness to read warnings
on dangerous products |
| Wogalter, M.S., Desaulniers, D.R., and Brelsford, J.W., Jr.
(1986) |
Consumer products: How are the hazards perceived? Proceedings
of the Human Factors Society 31st Annual Meeting, 615-619. |
EXP 1
- Subjects: 28 undergraduate college students
- Subjects were presented 72 generic products in a list format
- Subjects rated each product on such attributes as perceived hazardousness, likelihood of injury, frequency of
use, familiarity, level of knowledge of hazards using a 9-point Likert scale |
- Results indicate that severity of injury is the best single
predictor of hazard perception
- Findings suggest that products that are less frequently used and less familiar are perceived to be more
hazardous |
| Wogalter, M.S., Desaulniers, D.R., and Brelsford, J.W., Jr.
(1986) |
Consumer products: How are the hazards perceived?
Proceedings of the Human Factors Society 31st Annual Meeting, 615-619. |
EXP 2
- Subjects: 70 undergraduate college students
- Subjects were presented a list of 18 products and were asked to rate perceived hazardousness, to indicate
possible accident scenarios associated with each product including type of injury, severity of injury, and
likelihood of injury |
- Results indicated that perceived hazardousness was the most
important determinant of willingness to read warnings
-Familiarity with the product reduced the likelihood of reading the warning, but familiarity may not be as
important as perceived hazardousness |
3.1.6.5 Stress
Along with the perception of hazard, another important external factor with respect to warning effectiveness is
stress. Magurno and Wogalter (1994) used a chemistry task format in which the stressor was a combination of two kinds
of stress: time pressure and social evaluation by another person. There were four conditions evaluated: stress (lower
versus higher) and warning location (posted sign vs. within task instructions). The results showed that higher stress
produced significantly lower compliance. Warning placement was also found to produce a main effect on behavioral
compliance. More participants complied with the within-instruction warning than the posted-sign warning, even though
the sign was over 40 times larger in terms of area.
3.1.6.6 Product familiarity
A study of one hundred students using hammers in a laboratory setting resulted in none even noticing the warning
labels on the hammers (Dorris and Purswell, 1977). The authors concluded that a warning label on a familiar product
is not necessarily a salient item and may be filtered out by the user's attentional mechanisms and never processed at
a conscious level.
Godfrey and Laughery (1993c) found that women were not so likely to notice warnings about toxic shock syndrome on or
in tampon packages when they switched products. The authors found a real biasing effect from product familiarity.
DeJoy (1989) reviewed the literature on warnings and found seven studies where familiarity decreased the likelihood
of noticing, reading, or complying with warnings. Dejoy noted, however, that these factors do not appear to be as
important as the user's product-related expectations. Similarly, interview research into the effectiveness of warning
signs on scaffolding showed that both inexperienced and experienced subjects would react to a new warning on
scaffolding they were unfamiliar with but not to the same new warning on equipment that was familiar to them
(Johnson, 1993). This study, however, was based on measuring behavioral intention through interviews, not behaviors.
Many other studies have looked at the effects of familiarity on whether or not subjects will look for, notice, and
read a warning. Godfrey, et al. (1989), Godfrey & Laughery (1993c), LaRue & Cohen (1987), and Otsubo (1988)
have all reported that people are less likely to look for, notice and read warnings on consumer products with which
they are familiar than they are with unfamiliar products. Laughery and Brelsford (1993) considered the familiarity
effect consistent and robust and suggested a possible reason: the more people use a product without experiencing a
safety problem, the less hazardous they perceive the product to be. In the authors' words: "products that are
used repetitively pose special warning problems."
Ayres, et al. (1994) suggested, after an extensive literature review, that warnings are unlikely to be effective
unless a series of conditions are met. The failure of many intended warnings, including most on-product warning
labels, to reduce accidents reflects the difficulty of overcoming problems inherent in their use.
3.1.6.7 Color
Color affects behavior. Braun, Sansing, and Silver (1994) pointed to an extensive literature on the effects of color
but note that vast majority of product warning research has not considered color beyond the application of standards
and guidelines. Dunlap, Granda, and Kustas (1986) surveyed 1169 subjects across several different language groups
including English, German, Scandinavian, and Spanish. Subjects rate the color words red, orange, yellow, blue, green,
and white according to the level of perceived hazard. The results demonstrated that the hazard information
communicated by different colors followed a consistent pattern across language groups. Red resulted in the highest
hazard ratings followed by orange, yellow, blue, green, and white, respectively.
Researchers have examined a variety of attributes that influence a warning's ability to communicate important product
hazards but Kline, Braun, Peterson, and Silver (1993) consider the attribute of color to have been neglected in the
literature on warnings. The authors looked at the appropriateness of achromatic stimuli in product labels and found
that color labels were perceived as more readable and hazardous than achromatic labels.
Braun and Silver (1994) examined the effect of color on compliance with printed warnings. The participants in the
study indicated a higher likelihood of injury associated with products printed in red than green or black. Barbera
and Gill (1987) attempted to assess the ability of two warning label designs on a consumer product to: (1) attract
the user's attention, and (2) alter the user's behavior. For one condition, the warning label consisted of a one by
three inch black and white tag affixed to the back of a portable electric heater. The alternative condition involved
the same product but with a three by five inch red and black color coded warning label with pictographs. This second
warning label was mounted at the end of the heater's electrical cord in the form of a "ski-pass" tag.
Results revealed neither warning label to be effective. Even the "proper" warning label could not induce
the subjects to read it nor alter their behavior.
Building on their previous work, Gill, Barbera, and Precht (1993) moved away from solely passive label designs. Three
different warning label designs for a portable electric heater were tested: (1) a traditional non-human factored
label; (2) a color-coded "ski-pass" label attached near the male end of the electric cord; and (3) a
color-coded "interactive" label that required the user to interact with the label in order to use the
heater. The results showed that the interactive design was most effective in attracting the user's attention, but
none of them were effective in mediating safe user behavior, the ultimate goal.
Braun, Sansing, and Silver (1994b) examined the interaction of signal words and colors. A sample of 30 undergraduates
rated the perceived hazard of 105 signal word/color combinations printed in specific hazard colors. Of the colors
used, red conveyed the highest level of perceived hazard followed by orange, black, green, and blue. More
importantly, however, it was noted that a signal word such as DEADLY connotated less hazard when printed in green
than red ink.
Additional studies in the area of color and warnings support the earlier findings that the level of hazard
communicated by signal words varies as a function of the color in which they are printed (Braun, Kennedy, and Silver,
1994c). Importantly, these findings suggest that signal word and color combinations create a continuum of perceived
hazard. This model of a continuum may prove helpful to designers of warnings.
3.1.6.8 Cost of Compliance
Cost of compliance refers to the amount of effort a user must exert in order to comply with a warning
(Kotwal and
Lerner, 1995). The cost to the user of complying with a warning can be reduced, for example, by including protective
wear (e.g., gloves, mask, ear plugs, etc.) with the hazardous product. Kotwal and Lerner point out that the issue of
cost of compliance as a formal research topic is relatively new, as most of the literature which addresses this issue
has been published within the past six years. Therefore, it is not surprising that, although several recent articles
have demonstrated how strongly cost correlates with compliance, cost of compliance is not even mentioned in major
industry standards (ANSI, 1991; Westinghouse Printing Division, 1981; ISO, 1984). Kotwal and Lerner conclude that the
amount of effort required by the user to comply with the warning has not been a factor during the process of
designing a warning, but rather, an afterthought to the design. However, the changes in compliance rates which result
from lower cost to the user provide strong evidence for the inclusion of cost of compliance as a design factor in new
labeling guidelines.
Intuitively, the lower the effort required by the user to comply with the warning, the more impressive the increase
in compliance rates. In a field study (Dingus et al.,1991) university students agreed to evaluate the marketing
potential of a "new" cleaning product. Compliance involved wearing rubber gloves while using the product.
Subjects in the low cost condition had gloves provided with the product, while those in the high cost condition had
no gloves provided. The results showed a 25% compliance rate for the high cost condition, increasing to 87-88%
compliance for the low cost condition. Hunn et al. (1992) found that "in addition to reducing cost of
compliance, the inclusion of gloves as part of product packaging increased the perception of danger associated with
the product use." Therefore, increased compliance, in some cases, is probably a result of lower cost, as well as
an increased perception of danger.
Similar increases in compliance rates were reported by Kotwal and Lerner during their literature review (1995).
Increases in compliance rates range anywhere from 23% to almost 94%, depending on the given experimental situation.
Wogalter, McKenna, and Allison, (1988) used a chemistry laboratory setting to determine high and low cost compliance
rates (17% and 73%, respectively) for wearing protective gloves and masks. Studies on a racquetball court found high
costs resulted in no compliance with wearing protective eyewear, while low cost gave 60% compliance (Hathaway and
Dingus, 1992; Dingus, Hathaway, and Hunn, 1991).
Relating the cost of compliance to other warning features, Hunn et al. (1992) found that low cost of compliance had a
greater effect on compliance than either warning content or warning interactivity. Additionally, Hathaway et al.
(1992) found that the benefits of a low cost of compliance could be increased by adding specific consequence
information to the warning. The researchers concluded that providing the user with information regarding frequency
and severity of injuries associated with the hazard (in this case, eye injuries while playing racquetball), as well
as providing the tools required to exhibit safe behavior (e.g., protective eyewear) can markedly improve warning
effectiveness.
Table from Kotwal and Lerner (1995)
Summary of Research Findings Regarding Cost of Compliance
| AUTHOR |
TITLE |
METHODOLOGY |
| Dingus, T., Hathaway, J., and Hunn, B. (1991) |
A most critical warning variable: Two demonstrations of the
powerful effects of cost on warning compliance. Proceedings of the Human Factors Society 35th
Annual Meeting,1034-1038. |
EXP 1
- Subjects: 920 racquetball players at large centers in 2 universities
- Manipulated cost by varying the amount of effort required to obtain eye wear - Manipulated the level of warning
- Measured compliance as a function of whether or not subjects wore protective eye wear
- Low cost condition = eyewear provided just outside door
Middle cost condition = subjects were required to walk 60 feet to checkout booth to get eyewear
High cost condition = no eyewear provided on-site |
| |
|
EXP 2
- Subjects: 318 university students
- Experiment was presented under the guise of a marketing study; subjects thought they were evaluating the
marketing potential of a "new" cleaning product
- Measured compliance as a function of whether or not subjects wore rubber gloves
- Low cost condition = gloves provided
High cost condition = gloves not provided |
Dingus, T.A., Wreggit, S.S., and Hathaway, J.A. (1993) |
Warning variables affecting personal protective equipment use. Safety Science, 16, 655-673. |
EXP 2
- Subjects: 224 adult volunteers
- Subjects were given a package containing a spray bottle with cleaning solution
- Manipulated three levels of interactivity (no physical interaction, "billboard", and "trigger
guard" - see Hunn and Dingus, 1992), label content (generic product instructions, ANSI warning, and ANSI plus
specific consequence warning), and cost of compliance (low cost: gloves and mask provided, high cost: no gloves
and mask provided)
- Subjects were given a questionnaire after using the product for one week; were asked about product satisfaction,
perception of risk, and behavior in dealing with the product
- Subjects were also asked what they thought the purpose of the study was; if the subject answered this question
correctly, that subject's data was included in the analysis |
Hathaway, J. and Dingus, T. (1992) |
The effects of compliance cost and specific consequence information on the use of safety equipment. Accident
Analysis & Prevention, 24(6), 577-584. |
- Subjects: 420 racquetball players from a large university (339 males, 81 females); player experience ranged from
1 session to 13 years; each subject was used in only one condition
- 2 x 3 complete factorial between subjects design
- Manipulated cost
Low cost = eye protection provided
High cost = eye protection not provided
- Manipulated warning information
Baseline = warning not provided
ANSI standard warning
ANSI standard warning plus specific consequence information
- Baseline condition: strictly observation (i.e., eye protection not provided, no warnings)
- Experimental conditions: all manipulations of the two independent variables (i.e., cost of compliance and
warning information) |
3.1.7 MEASUREMENT AND PROTOCOLS
Over time the focus of research on warnings has shifted from a debate on whether warnings work to systematic
investigation of the factors that do or could influence the behavior of users of safety product. Ayers and his
colleagues ( 1992) pointed out that the logical test of a warning must be reduction of the frequency and/or severity
of accidents and injuries. This should be the key measurement.
Delegates from eight societies including: the American Society for Testing and Materials, American Association of
Textile Chemists and Colorists, American Pharmaceutical Association, American Psychological Association,
Illuminating Engineering Society, Optical Society of America, Textile Color Card Association of the United States,
and U.S. Pharmacopoeia Convention collectively assembled into a group known as the Inter-Society Color Council, or
ISCC (Billmeyer, 1991). The ISCC originally assembled to standardize color terms and color names used to describe the
colors of pharmaceuticals. Even after the color standardization problem was solved, the ISCC continued to exist with
the intent to advance the knowledge and use of color in art, science, and industry. This organization may be able to
provide important assistance with the international harmonization efforts.
The Commission Internationale de l'Eclairage (CIE) sanctions the internationally accepted method of color
specification and measurement that is designed to mimic human color perception (Hale, 1991). ANSI Z535 is a set of
standards to guide technical communicators in the development of effective hazard alert messages. ANSI Z535.3-1991 is
the ANSI standard entitled "Criteria for Safety Symbols" which contains a test method for determining the
effectiveness of a pictogram and a criterion for success of 85% correct responses with no more than 5% critical
confusion. The latter term refers to when the message conveyed is the opposite of the intended message, which for
safety warnings could result in death. A score below this ANSI level does not mean the pictogram cannot be used but
that it cannot be used alone and must be coupled with some other medium, such as a written warning. ANSI Z535.4 is
the ANSI standard for Product Safety Signs and Labels.
The International Standards Organization issued a standard, ISO 9186, Procedures for the
Development and Testing of Public Information Symbols, that recommends testing methodologies to evaluate
symbols intended to be used internationally. This standard sets a lower level of acceptability compared to the ANSI
standard. "For critical referents (e.g. safety symbols), the 66% criterion should be rigorously adhered
to." These standards represent a major starting point for international harmonization but Kemnitz (1991)
includes a reminder of the considerable time and effort required to understand and use any of these standards.
3.1.8 STANDARD PHRASES
The literature contains much investigation into the role of standardized phrases for warnings, particularly in the
use of "signal words" such as DANGER or WARNING. Research previously cited by Godfrey and colleagues (1993)
indicated that the use of a signal phrase such as "Government Warning" significantly influenced warning
detection times. Consequently, the repetition of words or phrases through government mandate may be an important way
to avoid confusion and increase recognition. Other studies on standard phrases are included in the MSDS portion of
this report (Section 3.3).
3.1.8.1 Signal words
ANSI Z535.4-1991 entitled "Product Safety Signs and Labels" indicates when the following signal words
should be used:
Danger indicates an imminently hazardous situation which, if not avoided, will result in
death or serious injury. This signal word is to be limited to the most extreme situations.
Warning indicates a potentially hazardous situation which, if not avoided, could result
in death or serious injury.
Caution indicates a potentially hazardous situation which, if not avoided, may result in
minor or moderate injury. It may also be used to alert against unsafe practices.
The importance that ANSI and the U.S. military attaches to signal words may not be shared by the population at large.
Given the lack of training, it is questionable how the public interprets different signal words. Leonard and Matthews
(1993) used questionnaires with college student test subjects to determine how the population in general
differentiates levels of warnings. No differences were found in ratings of perception of risk to different signal
words. Further, size of the signal word and color of the signal word had no effect on perception of risk. Statements
of consequences of disregarding the warnings and type of risk situation did affect rated perception of risk. Also,
circumstances in which the subjects might be placed affected ratings of likelihood of disregarding warnings.
Silver and Wogalter (1993) tested the arousal effects of signal words on college students and found that DANGER
connoted greater strength than WARNING and CAUTION. The results failed to show a difference between WARNING and
CAUTION. Among other words tested, DEADLY was seen as having the strongest arousal connotation, and NOTE the least.
From a long list of 84 terms, a "short" list of 20 signal words was developed based on understandability,
low variability. shortness of word, and frequency of use. The authors suggested that an expanded list of signal words
might alleviate potential problems of habituation from overuse of the currently recommended terms. This research was
built upon similar, earlier work (Wogalter and Silver, 1990).
Although many organizations have guidelines for the determination of what signal words are to be used with specific
hazards, these are usually unknown to the public. Leonard, Hill, and Karens (1989) developed information about how
the general public perceives the degree of danger represented by signal words in warnings. For 15 items that had been
rated for the seriousness of risk, 288 subjects were asked to indicate which signal word they would use to inform
others of the hazard. Signal words that had been found to rate high in seriousness by Leonard,
Karens, and Schneider
(1988) tended to be used more with items rated as higher risks. Differences were found among age groups with older
subjects using signal words that carried more serious connotations.
There are two systems for categorizing flammable liquids, one promulgated by the NFPA in 1986
(NFPA 321) and the
other found in the Federal Hazardous Substances Act, 16 CFR 1500. Implicit in both systems is that consumers will
recognize the degree of hazard associated with the terms "flammable" and "combustible." Main,
Rhoades, and Frantz (1994) showed 46 drivers of step vans two photos of lighter fluid that were identical except one
contained the warning "flammable" and the other "combustible." The subjects were asked which
presented the greatest fire hazard. Only 23.9% of the subjects correctly identified the fluid marked
"flammable" as the greater fire hazard while 47.8% incorrectly identified "combustible" as
presenting the greater A second evaluation in the same study looked at how well subjects could use the generic flame
symbol to choose correct specific actions to take with an extremely flammable adhesive. Less than 40% of the subjects
reported that vapor flammability was a reason to ventilate a work area and only 4 out of 100 specifically mentioned
that they would avoid using the product near pilot lights, the most frequently documented cause of accidents with
these adhesives and the cause of a ban by the CPSC. This research indicates that the terminology and symbols of two
current systems do not convey critical information to consumers, according to Main, Rhoades, and Frantz. The CPSC ban
was initiated due to a study by Nelson (1976) that found the required labeling was not effective. The CPSC concluded
that there was not a labeling scheme which would achieve the desired outcome and moved to a ban.
Wogalter et al. (1992) investigated the influence of warning signal words and a signal icon on perceptions of hazard
for consumer products. Under the pretext of a marketing research study, 90 high school and college students rated
product labels on variables such as product familiarity, frequency of use, and perceived hazard. Results showed that
the presence of a signal word increased perceived hazard compared to its absence. Between extreme terms (e.g. NOTE
and DANGER), significant differences were noted, but not between terms usually recommended in warning design
guidelines. The presence of the signal icon had no significant effect on hazard perception.
The standard wisdom of whether four components are needed for warnings was evaluated by
Wogalter, Desaulniers, and
Godfrey (1993). Four-statement signs contained a signal word, a hazard statement, a consequence statement, and an
instruction statement, as generally recommended. Four additional three-statement signs, each with a different
component systematically removed from the four-statement sign, were used, for a total of 5 signs for each hazard
situation. The results of Experiments 1 and 2 indicated that removing content statements reduced perceived
effectiveness. Hazard and instruction statements were the most important, showing the greatest decrease in
effectiveness when deleted. Signs for the most hazardous situations were perceived as the most effective warnings.
Other results suggested that the deletion of redundant statements, particularly signal words, had less influence on
effectiveness. The hazard statement showed the lowest redundancy consistent with it producing the greatest
effectiveness decrement when deleted.
Ley (1995) looked at effectiveness of label statements for the Australian government. This work was provided to the
IOMC as part of their call on chemical hazard communication. The study indicated that the following signal words were
not well understood: combustible, flammable, hazardous, irritant, lethal, and toxic.
3.1.8.2 Standardization of label format
The Environmental Protection Agency has wrestled with whether standardizing pesticide labels is an effective course
(EPA, 1986). After meeting with numerous stakeholder groups and establishing an advisory committee, three arguments
were considered against standardization:
- The variety of shapes and sizes of containers and the varying audiences reduced the practicality;
- standardization would eliminate any opportunity for marketing individuality by manufacturers; and
- standardized labels would allow (or even promote) users to simply skip over the information that they thought
was not important.
The arguments for standardization outweighed the concerns above. Standard formats were
considered useful to the user to readily find information regarding precautions, practical treatment, registration
number, etc. Standardized formats also allow the user to compare products, uses, risks, etc. and to determine whether
for their situation the benefits outweigh the risks. Standardized labels were seen as greatly facilitating training
efforts. The EPA reviewers agreed that standardization should be done within categories of pesticide products, rather
than one standard label. For instance, there should be one format for agricultural pesticides and another for home
garden products. Finally, the elements in a standardized label should be in the logical sequence so that if a person
followed each step as they read that step, they would be using the pesticide safely and correctly. This is what the
EPA's consultant called the "hierarchy of information" with the Primary Visual Panel at the top of the
hierarchy (U.S. EPA, 1987).
The EPA (1986) received an overwhelming response from stakeholders and experts in their evaluation of pesticide
labeling that the required phrase, "Keep out of the reach of children" had no impact because of overuse. It
was considered the least read section of the labels. A follow up report the next year (U.S. EPA, 1987) recommended
discontinuing the child warning and replacing it with fresh statements that might be rotated every two to three years
to attract attention.
3.1.9 ADDITIONAL RESEARCH NEEDS
Dorris and Purswell (1978) presented a list of needed research, including:
- Optimum amount of information to be presented;
- Symbolic versus verbal warning effectiveness;
- Need for an appropriate methodology for studying behavior; and
- Need to understand the factors that influence responses to warning.
Purswell, Krenek, and Dorris (1993) built on the previous list and included the observation
that the conceptual model of the warning process developed by Miller and Lehto (1986) also needs to be refined to
highlight all of the additional variables that need to be studied. Purswell and colleagues highlighted the following
areas for more research:
Stimulus: Energy Level and Contact
Most users of a product will not read the warning information because of a well documented "filtering"
process.
The following variables affect filtering:
- Information overload
- Warning lists frequently include too many items;
- Contents of a single warning may be too extensive;
- There may be too many individual warnings placed in the field of view; and
- There may be other non-warning stimuli.
- Faulty risk assessment - The amount of risk information provided does not significantly influence the subjective
rating of hazard perceived.
- Benign experience versus a warning - Karnes et al. (1986) hypothesize that being regularly exposed to a warning
about a hazard while at the same time, the hazard does not cause an injury, filtering will take place.
Comprehension of Warnings
- The meaningfulness of "signal" words
- The reading comprehension level required to understand the warning - particularly, the trade-off between the use
of a smaller number of words with a more exact meaning versus a larger number of smaller words with less exact
meanings.
- The meaningfulness of symbols
- The meaningfulness of warning as a function of the task being performed – this is the effectiveness of
warnings when presented in the context of instructions as compared to presenting the warnings in a separate list
without the true context of performing some operation.
Warning and Memory
Few warnings are stored in long-term memory if they are more than six or seven lines long or
address more than this number of hazards in using a product. More research is needed to better understand the reasons
for the lack of long-term recall.
The Decision Making Process
Purswell, Krenek, and Dorris conclude their list of needed research with this item, which they identify as the most
important. They hypothesize a threshold of perceived probability of an injury which must be reached before a person's
behavior will be influenced by risk information. This threshold may be relatively high, i.e. 1/100 before most
persons will respond to knowledge about hazards as presented in a warning. Since few products carry such risks, most
warnings go unheeded. Purswell et al. cite Godfrey et al. (1985) concept of "cost of compliance." There may
be benefits of noncompliance that are of significant importance to the individual. Consequently, individuals appear
to construct some type of utility function for making the trade-off. This is another important area for research.
3.2 TRAINING
3.2.1 SUMMARY OF FINDINGS AND RECOMMENDATIONS
The following information represents a summary of the general areas of consensus in the literature on training. The
summary also includes what the authors of this report.
- Reduced worker injury rates were found after the introduction of worksite first aid training programs in several
studies, suggesting that this kind of instruction increases consciousness to job safety concerns.
- There appears to be a clear link between training and the establishment of healthy and safe working conditions.
The studies reviewed in one meta analysis were "near unanimous" in the benefits that training can achieve
such as increased hazard awareness and safe workplace practices.
- Evaluation of the U.S. Hazard Communication Standard in selected manufacturing plants showed that the most
important variable in determining better compliance was explicit support for the program by plant-level management.
- In programs preparing health and safety professionals (at both undergraduate and graduate levels), very little
course work prepares the students for their roles as communicators or trainers.
- Literacy and language issues have not been substantially explored regarding hazard communication and therefore
should be evaluated.
3.2.2 GENERAL CONSIDERATIONS
Hazard communication is accomplished through a process that includes several tools for transferring information and
influencing behavior. Laugher and Brelsford (1993) note that "numerous studies have shown the greater the a
priori perception of hazardousness, the more likely people will look for and read a warning, and the more likely they
will comply by taking safety precautions (Donner and Brelsford, 1988; Godfrey et al, 1983;
Friedmann, 1988; LaRue and
Cohen, 1987; Leonard e | |
