Product Safety

Product Safety

Safety Signs
& Labels Does compliance with ANSI Z535 increase
compliance with warnings?
By Stephen L. Young, J. Paul Frantz, Timothy P. Rhoades and Kristin R. Darnell

T THE AMERICAN NATIONAL STANDARDS This standard was revised and approved in 1998,
Institute (ANSI) accredits the National Electrical and the Z535 Committee is nearing completion of
Manufacturers Assn. (NEMA), which, in April 1992, another revision. The standard’s stated purpose is:
published ANSI Z535.4-1991, Product Safety Signs
and Labels. This standard is one of a series of stan-
1) to establish a uniform and consistent visual
layout for safety signs and labels applied to a
dards that includes: wide variety of products; 2) to minimize the
•Z535.1, Safety Color Code proliferation of designs for product safety
•Z535.2, Environmental and Facility Safety Signs signs and labels; and 3) to achieve application
•Z535.3, Criteria for Safety Symbols of a national uniform system for the recogni-
•Z535.5, Accident Prevention Tags (for Temp- tion of potential personal injury hazards for
orary Hazards) those persons using products [ANSI(b) 1].
Stephen L. Young, Ph.D., is director of research and development at Applied Following from this stated purpose, the standard
Safety and Ergonomics Inc. in Ann Arbor, MI. He received a Ph.D. from Rice focuses on the format and presentation of product
University in Engineering Psychology. Young is a member of the Human Factors safety signs and labels. For example, it includes speci-
and Ergonomics Society and program chair of its Safety Technical Group. He has fications for the use of signal words (e.g., DANGER,
also served as a member of ANSI’s Z535 Committee. In addition, he has served as a WARNING, CAUTION), the format of signal word
guest lecturer at Harvard University and the University of Michigan on topics panels (e.g., colors, use of the safety-alert symbol), var-
including human error in accident causation, and design of displays and controls. ious other items related to a warning’s format and, to
some extent, content (Figure 1). Because the standard
J. Paul Frantz, Ph.D., CPSM, CPE, is a principal research engineer and cofounder
applies to a wide range of products, it contains a lim-
of Applied Safety and Ergonomics Inc. He also serves as an adjunct assistant
ited number of requirements (i.e., “shall” statements)
professor at the University of Michigan, teaching product and occupational safety
and many recommendations (i.e., “should” state-
management. Frantz received a Ph.D. in Industrial and Operations Engineering
ments and advisory material in appendices that are
from the University of Michigan. Frantz is a member of ASSE’s Greater Detroit
not part of the standard).
Chapter and is the Society’s representative to the ANSI Z535 Committee.
ANSI Z535.4 can offer practical benefits for those
Timothy P. Rhoades, Ph.D., P.E., CPE, is a principal research engineer preparing product warnings or developing compa-
at Applied Safety and Ergonomics Inc. and an adjunct assistant professor at the ny-wide hazard communication programs. For
University of Michigan. He received B.S.E., M.S.E. and Ph.D. degrees in Industrial example, the standard can help streamline the warn-
and Operations Engineering from the University of Michigan. Rhoades is a former ing-development process by limiting the need to
member of ASSE’s Standards Development Committee and a professional member consider a wide variety of formatting issues that
of ASSE’s Greater Detroit Chapter. might otherwise consume considerable time and
Kristin R. Darnell, M.S.I., is a senior research associate and director effort; this, in turn, allows resources to be expended
of the Information Services Group at Applied Safety and Ergonomics Inc. on other potential HazCom challenges that may be
A professional librarian with a master’s degree from the University of Michigan, more relevant to the goal of promoting product safe-
she conducts and directs customized research database development and ty. In addition, from a product liability perspective,
various information services on subjects related to consumer and industrial compliance with the standard may be viewed as evi-
safety and health, ergonomics and product liability. Darnell also directs dence related to the “adequacy” of a warning in the
and manages the firm’s specialized libraries. She is a member of the event of “failure-to-warn” allegations.
Engineering Div. of the Special Libraries Assn. Aside from these benefits, another potential rea-
18 PROFESSIONAL SAFETY SEPTEMBER 2002 www.asse.org
Figure
Figure 1 1

Warning Features Specified in Z535.4

son for using this (or any other) partic-

ular style of warning would be the
expectation that it would reliably and
significantly increase safe behaviors
compared to another style. While this
is generally desired by the ANSI com-
mittee, it should be noted that the
tenets of the standard were not devel-
oped on theoretical or empirical bases
which would support such an expec-
tation (Martin and Deppa). For exam-
ple, as Dorris noted in 1991:
Warnings researchers have investi-
gated a number of independent
variables related to the design of
warning labels. For instance, the
signal word and the colors utilized
in a warning have been investigat-
ed with some thoroughness. In the
final analysis, there is no demon-
stration that either of these factors
are reliable predictors of the behav-
ior of those who are presented with
a warning sign (1075).
While many companies have em- 1The signal word panel and the exclamation point in the safety alert symbol would be printed in
braced the standard in whole or in orange when the signal word WARNING is used.
part, use and acceptance of the stan-
dard has been gradual and is limited in many study participants in a manner consistent with the
respects. While government safety agencies such as hierarchy defined in the standard. It is relatively
National Highway Traffic Safety Administration common to find that people fail to perceive consis-
(NHTSA) and Consumer Product Safety Commis- tent and/or meaningful differences between these
sion have considered the standard, they have not yet three signal words [e.g., Leonard, et al; Wogalter, et
adopted it completely or incorporated it, by refer- al(a), (b); Wogalter and Silver(a), (b); Young].
ence, into regulations. For example, during develop- •The three-tiered hierarchy of safety colors spec-
ment of sun-visor labeling regulations for air bags, ified by ANSI Z535.4 (i.e., red, orange, then yellow)
NHTSA conducted focus-group research related to are also inconsistently interpreted in the population,
people’s reactions to various colors and signal words with the possible exception of the color red [e.g.,
rather than simply adopting the ANSI Z535 scheme Wogalter and Silver(b); Young].
[NHTSA(b)]. Indeed, the agency explicitly rejected These rating studies focus on measures other than
the idea that consumers draw any distinction behavioral response to warnings. They also tend to
between the words WARNING and CAUTION evaluate individual elements of warnings rather than
when used in safety information labeling. In addi- warning signs or labels as a whole. In addition, they
tion, many consensus standards that specify warn- rarely evaluate label elements in the context of appli-
ing messages, such as those promulgated by cation to a product. This article evaluates a much
Underwriter’s Laboratory, have not incorporated, smaller body of research that addresses behavior in
by reference, the tenets of the ANSI standard. response to warnings. Specifically, it evaluates several
Given that the standard has been available for sev- studies that have examined people’s response(s) to
eral years, the authors endeavored to identify and ANSI-style warnings compared to warnings which
summarize studies which address the relative merits use non-ANSI formats. Given the limited number of
of ANSI-specified warning formats compared to non- such studies, this assessment includes some studies
ANSI styles. Most of the relevant literature measures that employed ANSI-formatted signs rather than just
perceptions of or subjective reactions to warnings. For those involving product labels. In these cases, one
example, many studies have assessed people’s per- might have referred to ANSI Z535.2, Environmental
ceptions of different signal words and colors. Example and Facility Safety Signs, which has much in common
measures of reactions to these elements include rat- with Z535.4 as it relates to the present study.
ings of perceived hazardousness and intent to comply
with a warning. While these studies are not the focal Behavioral Research Studies
point of this article, the reader may be interested to Following is a summary of several studies that pro-
know that across much of this research, one finds that: vide evidence regarding the behavioral effect(s) of an
•The three-tiered hierarchy of signal words spec- ANSI-style format compared to that of warnings
ified by ANSI Z535.4 (i.e., DANGER, WARNING, which have few, if any, ANSI elements. However, it
then CAUTION) has not been reliably interpreted by should be noted that none of these studies was
www.asse.org SEPTEMBER 2002 PROFESSIONAL SAFETY 19
Figure
Figure 2 2 Figure
Figure 3 3

Warning Signs from Warning Signs from

Shaver & Braun Smith-Jackson and Durak
Non-ANSI Non-ANSI
Style Style

ANSI Style ANSI Style

Source: Proceedings of the Human Factors and Ergonomics Society

The signal word panel and the exclamation point in the safety alert
44th Annual Meeting. Copyright 2000 HFES.
symbol for the ANSI-style sign were printed in orange.
designed specifically to evaluate the ANSI
standard. Therefore, it is sometimes true that the . . . a warning constructed in strict adherence
ANSI ANSI-style warnings in these studies do not comply to ANSI standards resulted in no greater com-
with the standard in every detail. In addition, many of pliance in this study than no warning at all.
Z535.4 the studies compare warnings that differ not only in Washington State University [where the study
terms of format, but also in terms of content. However, was conducted] has signs posted in areas near
can help in every case where content differences exist, the the racquetball courts, which read “EYE-
ANSI-style warning was arguably “stronger” than the GUARDS STRONGLY RECOMMENDED.”
streamline non-ANSI version. Thus, to the extent that differences Apparently, the addition of the ANSI warning
other than format play a role in people’s behavior in did not significantly alter subject attitudes
the warning- these studies, one would expect that they would favor toward the necessity of using eyewear to
the ANSI-style warnings over the non-ANSI formats. avoid injury (583).
development It should also be noted that most of these studies
evaluated issues and variables which are extraneous Avoidance of Construction Hazards
process by to the present discussion. For the sake of simplicity, In this study, Shaver and Braun observed the
descriptions of the experimental protocols and behavior of 4,620 individuals in response to a sce-
limiting research designs have been simplified to allow the nario where a scaffold was erected in front of three
direct examination of the effect of ANSI-formatted contiguous doors to a single building (290-293). A
the need signs in relation to other formats. Despite these issues, warning sign that stated, “Overhead Construction
effort has been made to present, as fairly as possible, Please Use Center Door,” was installed at eye level on
to consider those data that would allow a sufficient comparison. both sides of the left doorway, except for the control
condition in which no sign was present. As shown in
a wide Use of Protective Eyewear Figure 2, the sign was formatted in one of two ways:
In this study, Hathaway and Dingus unobtru- ANSI-style (a three-panel sign with a CAUTION sig-
variety of sively observed the behavior of 420 racquetball play- nal word, color surround, symbol and text message)
ers in response to signage indicating that protective and plain-text (text message only in black print on a
formatting eyewear should be worn when playing racquetball white background). The researchers found that when
(577-584). A non-ANSI formatted sign was printed a sign was present—regardless of format—there was
issues. in plain black text on a white background and a sec- a significant increase in the proportion of people
ond warning sign was developed in the ANSI style avoiding doors which presented a scaffold hazard.
(a three-panel sign with signal word, text message However, people’s behavior was not influenced by
and symbol). The researchers found that 10 percent the format of the sign, leading the authors to con-
of the players complied with the plain-text sign and clude that the “effectiveness of a warning was not
10 percent complied with the ANSI-style sign. The related to the color, symbol, explicitness or warning
authors concluded that: format” (292). They further stated that “it appeared
20 PROFESSIONAL SAFETY SEPTEMBER 2002 www.asse.org
Figure
Figure 4 4

Warning Signs from Frantz &

Rhoades, and Frantz, et al
that simply stating the hazard and pro-
viding mitigating directions was suffi- Non-ANSI
cient to significantly alter the proportion
of people passing through the designated Style
door. No significant benefit was realized
with the addition of the three-panel
warning and its components” (292-293).
Use of Personal Protective Equipment
In a study involving the use of per-
sonal protective equipment, Smith-
Jackson and Durak had 37 participants
perform a task that involved mixing ANSI Style
“chemicals” (which actually were harm-
less powders and liquids made to seem
potentially hazardous through context
and instructions) (115-118). One of two
signs was posted on a wall directly
above the workspace indicating that par-
ticipants should wear protective gloves
and safety goggles, which were located
on the table with the other materials. The
content of the two signs was identical,
but their format differed (Figure 3); one
The signal word panel and the exclamation point in the safety alert symbol for the
sign was designed with black letters on a
ANSI-style sign were printed in orange.
white background (text-only), the other
had a safety orange background and the
signal word “WARNING” at the top (ANSI-style). (74 percent) “complied” without having read any of
After completing the mixing task, participants the warning label (818-821).
were asked to provide a rating between 0 (“not at all In the Frantz and Rhoades study, a non-ANSI
likely”) and 100 (“absolutely would wear gloves and style warning was employed (Figure 4) (719-730).
goggles”) to indicate the likelihood that they would Looking at the two studies together, it is clear that
wear the provided protective gear. They reported that both had comparable rates of noticing the cabinet
they would be significantly more likely to wear label—93 percent in Frantz and Rhoades; 98 percent
gloves and goggles in response to a ANSI-style sign in Frantz, et al. Also, the number of participants who
(mean rating = 87.6 out of 100) than to one in the text- read at least some of the label was similar—67 per-
only format (mean rating = 69.1 out of 100). However, cent in Frantz and Rhoades; 57 percent in Frantz, et
during the actual task, no participant looked at or al. However, 40 percent of participants who were
read the warning sign, regardless of format, prior to exposed to the non-ANSI label in Frantz and
beginning the mixing task. Thus, although partici- Rhoades reported reading all of it, while no subjects
pants believed that such features would or should reported reading all of the ANSI-style label in
make a difference in how they behaved, no such effect Frantz, et al. Frantz and Rhoades found, at most, 53
on their actual behavior was observed. percent compliance while Frantz, et al found, at
most, only 17 percent compliance.
Use of Office Equipment Differences in compliance between these studies
Two different studies involving the use of file cab- were most likely due to a combination of factors,
inets and their warning labels (Frantz and Rhoades; including the warnings provided and the task being
Frantz, et al) allow for an assessment of the effects of performed. However, despite the fact that the ANSI-
warning styles. In both studies, participants set up an style label in the 2000 study interrupted the task to
office space under the guise that the study was exam- some extent and was noticed by nearly everyone,
ining how people might arrange office furniture and information within the label was not actually
supplies. Among the materials to be arranged was a processed at a more meaningful level than the non-
two-drawer file cabinet with a warning label ANSI warning used in the 1993 study.
attached to the front such that neither drawer could
be opened until the label was physically removed. Analysis of Lap-Belt Use for Ford Escorts: ‘91-’94
The Frantz, et al study employed a warning label Like many other auto manufacturers during the
formatted according to ANSI Z535.4 (Figure 4). In late 1980s and early 1990s (e.g., Chrysler, General
this study, 82 of the 84 participants (98 percent) Motors, Honda, Hyundai, Mazda, Mitsubishi,
noticed and interacted with the label, at least to Nissan, Subaru, Toyota), Ford Motor Co. sold cars
remove it and open the file drawers. However, only with a motorized shoulder belt and a manual lap
14 (16.7 percent) complied with the target statement belt. The motorized shoulder belt was provided to
indicating that the file cabinet should be loaded from meet government crash protection requirements and
the bottom drawer first. Of these 14 particpants, 10 the manual lap belt was voluntarily provided for
www.asse.org SEPTEMBER 2002 PROFESSIONAL SAFETY 21
Figure
Figure 5 5

Visor Label Styles

1991-94 Ford Escorts
additional occupant protection. For the
years 1991 through 1994, Ford Escorts Visor Alert
contained a warning on the driver- and Label Styles
passenger-side sun visors that instructed
occupants to wear both belts together
1991-92
(Figure 5, primary visor label). When the
visors were in their stowed position, a
This visor alert label (Figure 5) instructed
literature occupants to read the primary visor label 1993
on the opposite side. During this same
time period, the style of the warnings
review adds became increasingly more consistent
with the ANSI standard.
to the Data from the National Automotive
Sampling Survey Crashworthiness Data 1994
growing System (NASS CDS) were used to com-
pare lap-belt use rates for occupants of
body of these model-year vehicles for the period
1990-1999. The NASS CDS contains
research detailed data on thousands of minor,
serious and fatal crashes that involve Primary Visor
indicating passenger vehicles towed from the scene Label Styles
of an accident. Teams of trained crash
that factors investigators obtain data from crash 1991-
sites, vehicle inspections, victim inter- 93
other than views and review of medical records.
NHTSA uses this data to evaluate,
the format among other things, seatbelt use pro-
grams and the effectiveness of occupant
of a warning protection systems [NHTSA(a), (b)].
While NASS investigators assess safety
itself are belt use based on various factors, they
often rely primarily on self-reporting of
often greater belt use. As a result, the actual percent- 1994
age of belt use is likely lower than the
determinants reported values. Nonetheless, since the
objective of this study was to compare
of a person’s belt use between warning conditions
(i.e., model years for the Ford Escort), the
response to data are suitable for making such com-
parisons.
a warning. For the 1994 model year (m = 58 per-
The signal word panel and the exclamation point in the safety alert
symbol for the 1994 labels were printed in orange.
cent), belt-use rates for the driver and
front seat passenger were as low or
lower than for the 1991 (m = 58 percent), 1992 (m = ANSI-style warning increases compliance with a
60 percent) or 1993 model years (m = 65 percent). No warning compared to a different style or variation of
evidence suggests that any statistically reliable dif- the ANSI-style. On the whole, these studies indicate
ference exists between lap-belt use rates for those that elements of the ANSI-style can be, and perhaps
model years, ␹2 = 1.7, p = 0.63. Even a comparison of are, unimportant in terms of how people perceive
the two years with the highest and lowest belt-use and respond to safety information. Such conclusions
rates (1993 and 1994) showed no significant differ- are consistent with a larger body of literature related
ence in lap-belt use rates (p = 0.36). Given the lack of to various presentation features of warnings which
meaningful differences in belt-use rates between the fails to support the position that using a particular
four model years, it is clear that the addition of color or signal word reliably and positively affects
ANSI-style elements (e.g., signal word, color) to the attention to and compliance with product warnings.
labeling had no influence on the use of lap belts. Why do ANSI-style warnings not appear to yield
improvements compared to other styles? One reason
Discussion may be related to the way that the user population
Does compliance with ANSI Z535.4 increase perceives various warning features as specified by
compliance with warnings beyond that of other ANSI Z535.4. For example, the standard attaches
styles of warnings? Based on available research, the specific meanings to the signal words DANGER,
short answer is no. In general, this literature review WARNING and CAUTION, but there is little reason
would not support the proposition that using an to believe that people perceive consistent and mean-
22 PROFESSIONAL SAFETY SEPTEMBER 2002 www.asse.org
Figure
Figure 6 6

Meanings of Signal Words

ingful distinctions between these signal words in the warning. While benefits can be gained by conforming
way they are intended. Indeed, an annex to the stan- to the standard, research to date suggests that it is not
dard acknowledges this fact and calls for help in a necessity from a product safety perspective. 䡲
educating the public about signal words as they are
specified in the standard. References
American National Standards Institute (ANSI) a). “Product
However, even if users perceived the specifica- Safety Signs and Labels.” ANSI Z535.4-1991. New York: ANSI,
tions exactly as they were intended, it is possible that 1991.
the domain over which Z535.4 has control is too ANSI(b). “Product Safety Signs and Labels.” ANSI Z535.4-
inconsequential to have a significant influence on 1998. New York: ANSI, 1998.
how people actually behave. Put another way, the Dorris, A.L. “Product Warnings in Theory and Practice: Some
Questions Answered and Some Answers Questioned.” Proceedings
signal words, colors and other label items that the of the Human Factors Society 35th Annual Meeting. Santa Monica,
standard specifies may, in themselves, have little CA: Human Factors Society, 1991. 1073-1077.
influence in determining people’s actions, and these Frantz, J.P. and T.P. Rhoades. “A Task-Analytic Approach to
low-level manipulations may be insignificant in rela- the Temporal and Spatial Placement of Product Warnings.”
Human Factors. 35(1993): 719-730.
tion to other nonformat and nonwarning sources of Frantz, J.P., et al. “Assessing the Effects of Adding Messages
information about hazards. For example, Shaver and to Warning Labels.” Proceedings of the IEA 2000/HFES 2000
Braun concluded: Congress: Volume 4. Santa Monica, CA: Human Factors and
Ergonomics Society, 2000. 818-821.
The comparable performance of the [ANSI- Hathaway, J.A. and T.A. Dingus. “The Effects of Compliance
style] warning and control [non-ANSI] signs Cost and Specific Consequence Information on the Use of Safety
might have resulted from the qualities of the Equipment.” Accident Analysis & Prevention. 24(1992): 577-584.
situation rather than the signs themselves. In Leonard, S.D., et al. “How Does the Population Interpret
Warnings Signals?” Proceedings of the Human Factors Society 30th
particular, the scaffolding might have provid- Annual Meeting. Santa Monica, CA: Human Factors Society, 1986.
ed all the salient information needed to inter- 116-120.
pret the hazard (293). Martin, B.J. and S.W. Deppa. “Human Factors in the Revised
ANSI Z535.4 Standard for Safety Labels.” Proceedings of the Human
Yet another possibility is that the ANSI format Factors and Ergonomics Society 41st Annual Meeting. Santa Monica,
has the unintended effect of allowing users to actu- CA: Human Factors and Ergonomics Society, 1997. 821-825.
ally ignore or filter warning information (Frantz, et National Highway Traffic Safety Administration
al). Specifically, if a user is presented with safety (NHTSA)(a). “Effectiveness of Occupant Restraint Systems and
Their Use.” Report to Congress. Washington, DC: Government
information that is perceived to be of low utility (for Printing Office, 1993.
whatever reason), a format that clearly defines the NHTSA(b). National Automotive Sampling System/Crash-
warning and makes it stand apart from other infor- worthiness Data System/1994-1996 Annual Report. Washington, DC:
mation could allow users to bypass or ignore it alto- Government Printing Office, 1996.
Shaver, E.F. and C.C. Braun. “Effects of Warning Symbol
gether (also known as “pre-lexical” filtering). In such Explicitness and Warning Color on Behavioral Compliance.” Pro-
cases, it may be that the ANSI-style format actually ceedings of the IEA 2000/HFES 2000 Congress: Volume 4. Santa Moni-
helps users ignore safety information which they ca, CA: Human Factors and Ergonomics Society, 2000. 290-293.
have generally found not to be relevant, credible, Smith-Jackson, T. and T. Durak. “Posted Warnings,
Compliance and Behavior Intent.” Proceedings of the IEA 2000/
personally applicable, directly related to achieve- HFES 2000 Congress: Volume 4. Santa Monica, CA: Human Factors
ment of their goals, etc. and Ergonomics Society, 2000. 115-118.
In conclusion, this literature review adds to the Wogalter, M.S., et al(a). “Effects of Warning Signal Words on
growing body of research indicating that factors other Consumer-Product Hazard Perceptions.” Proceedings of the Human
than the format of a warning itself are often greater Factors Society 36th Annual Meeting. Santa Monica,
CA: Human Factors Society, 1992. 935-939. Your Feedback
determinants of a person’s response to a warning. Wogalter, M.S., et al(b). “Influence of Warning Did you find this article
However, it is not the purpose of this article to sug- Label Signal Words on Perceived Hazard Level.” interesting and useful?
gest that the recommendations provided in Z535.4 are Human Factors. 36(1994): 547-556.
Wogalter, M.S. and N.C. Silver(a). “Arousal Circle the corresponding
without merit or worth. For the reasons stated earlier,
Strength of Signal Words.” Forensic Reports. 3(1990): number on the reader
several benefits are associated with a consensus stan- 407-420. service card.
dard that addresses basic attributes of warnings Wogalter, M.S. and N.C. Silver(b). “Warning
which are likely to apply to a wide range of products Signal Words: Connoted Strength and Under-
standability by Children, Elders and Non-Native RSC# Feedback
and situations. However, the evidence presented here
English Speakers.” Ergonomics. 38(1995): 2188-2206. 30 Yes
calls into question the proposition that either compli- Young, S.L. “Connotation of Hazard for Signal 31 Somewhat
ance with or deviation from ANSI Z535.4 would reli- Words and Their Associated Panels.” Applied 32 No
ably or substantially influence people’s response to a Ergonomics. 29.2(1998): 101-110.
www.asse.org SEPTEMBER 2002 PROFESSIONAL SAFETY 23