Do puzzle pieces and autism puzzle piece logos evoke negative associations?

Abstract

Puzzle pieces have become ubiquitous symbols for autism. However, puzzle-piece imagery stirs debate between those who support and those who object to its use because they believe puzzle-piece imagery evokes negative associations. Our study empirically investigated whether puzzle pieces evoke negative associations in the general public. Participants' ( N = 400) implicit negative associations were measured with an Implicit Association Task, which is a speeded categorization task, and participants' explicit associations were measured with an Explicit Association Task, which is a standard task for assessing consumers' explicit associations with brands (and images of those brands). Puzzle pieces, both those used as autism logos and those used more generically, evoked negative implicit associations ( t(399) = -5.357, p < 0.001) and negative explicit associations ( z = 4.693, p < 0.001, d = 0.491). Participants explicitly associated puzzle pieces, even generic puzzle pieces, with incompleteness, imperfection, and oddity. Our results bear public policy implications. If an organization's intention for using puzzle-piece imagery is to evoke negative associations, our results suggest the organization's use of puzzle-piece imagery is apt. However, if the organization's intention is to evoke positive associations, our results suggest that puzzle-piece imagery should probably be avoided.

Full text

Do puzzle pieces and autism puzzle piece
logos evoke negative associations?
Morton Ann Gernsbacher1, Adam R Raimond1,
Jennifer L Stevenson2, Jilana S Boston1 and Bev Harp3
Abstract
Puzzle pieces have become ubiquitous symbols for autism. However, puzzle-piece imagery stirs debate between those
who support and those who object to its use because they believe puzzle-piece imagery evokes negative associations.
Our study empirically investigated whether puzzle pieces evoke negative associations in the general public. Participants’
(N = 400) implicit negative associations were measured with an Implicit Association Task, which is a speeded categorization
task, and participants’ explicit associations were measured with an Explicit Association Task, which is a standard task
for assessing consumers’ explicit associations with brands (and images of those brands). Puzzle pieces, both those
used as autism logos and those used more generically, evoked negative implicit associations (t(399) = –5.357, p < 0.001)
and negative explicit associations (z = 4.693, p < 0.001, d = 0.491). Participants explicitly associated puzzle pieces, even
generic puzzle pieces, with incompleteness, imperfection, and oddity. Our results bear public policy implications. If
an organization’s intention for using puzzle-piece imagery is to evoke negative associations, our results suggest the
organization’s use of puzzle-piece imagery is apt. However, if the organization’s intention is to evoke positive associations,
our results suggest that puzzle-piece imagery should probably be avoided.
Keywords
free associations, IAT, Implicit Association Test, puzzle
!
Puzzle pieces have become pervasive symbols for autism
(McGuire, 2012; McGuire and Michalko, 2011; Sarrett,
2011; Smukler, 2005). Puzzle pieces that represent autism
appear on numerous products. From t-shirts to designer jewelry,
car magnets to credit cards, and beverage holders to
smartphone cases, “the puzzle piece as a symbol for autism
… is ubiquitous” (Grinker and Mandell, 2015: 643).
Numerous autism organizations use puzzle pieces in their
logos. For example, the US organization Autism Speaks has
officially trademarked a blue puzzle piece as its logo. The
Autism Society of America has common-law trademarked a
ribbon comprising puzzle pieces, and puzzle pieces feature
prominently in the logos of many of the Autism Society of
America’s state chapters (Stevenson et al., 2011).
The depiction of autism as a puzzle piece has permeated
scientific vernacular. Research institutions have titled
their press releases “Finding the Missing Puzzle Piece of
Autism” (O’Connor, 2009) and “Missing Piece Surfaces in
the Puzzle of Autism” (CNRS, 2015). Researchers have
titled their journal articles “The Puzzling Life of Autistic
Toddlers” (Van Rijn and Stappers, 2008), “Solving the
Autism Puzzle a Few Pieces at a Time” (Schaaf and
Zoghbi, 2011), “The Neurobiology of Autism: New Pieces
of the Puzzle” (Acosta and Pearl, 2003), and “Another
Piece of the Autism Puzzle” (State, 2010). The International
Society for Autism Research has used puzzle imagery on
the cover of its annual International Meeting for Autism
Research (IMFAR) program. The conception of autism as
a puzzle piece is pervasive.
The debate over puzzle piece imagery
for autism
Common lore traces the use of puzzle pieces to represent
autism to the United Kingdom. In 1963, the logo of a weeping
child superimposed on a puzzle piece was created by a
(non-autistic) parent-member of the National Autistic
Society. Another (non-autistic) parent-member later related
_____________________________________________________
1University of Wisconsin–Madison, USA
2Ursinus College, USA
3University of Kentucky, USA
Corresponding author:
Morton Ann Gernsbacher, Department of Psychology, University of
Wisconsin–Madison, Madison, WI 53706, USA.
Email: MAGernsb@wisc.edu
https://doi.org/10.1177/1362361317727125
Autism
2018, Vol. 22(2) 118-125
© The Author(s) 2017

119!Autism [author copy]
that “the puzzle piece is so effective because it tells us
something about autism: Our children are handicapped by a
puzzling condition; this isolates them from normal human
contact and therefore they do not ‘fit in’” (National Autistic
Society, 2007).
In current day, the US state of Pennsylvania’s Bureau of
Autism Services claims that the “puzzle piece [remains]
the international symbol for autism” because the puzzle
piece conveys “the mystery still surrounding this condition”
(Healy, 2008: 141; emphasis ours). The US state of
Alaska’s Special Education Service Agency claims that
puzzle pieces continue to represent autism because of the
“many years [that] parents and professionals alike” have
been “baffled by this disorder” (Crawford, 2004: 2).
Images of puzzle pieces representing autism flourish on
the Internet and in other visual media. Surfing, scouting,
running, wrestling, making music, and many activities in
which autistic children engage have been represented by
puzzle pieces. Sometimes autistic children themselves are
caricatured as puzzle pieces (e.g. with the puzzle pieces’
top tab representing the children’s heads and the puzzle
pieces’ side tabs representing the children’s arms). Other
images depict autistic children with a puzzle piece missing
from their brain (as on IMFAR program covers). Such
imagery is recapitulated by typically developing children’s
conception of autistic children. When asked why some
children have neurodevelopmental disabilities, such as
autism, typically developing 9- to 12-year old children
report “there is a puzzle piece missing in their brain”
(Nowicki et al., 2014: 77).
Numerous autistic persons have objected to the use of
puzzle pieces to represent autism and autistic people. For
example, Yergeau writes in the journal College English:
Puzzle pieces … symbolize so much of what is wrong with
popular autism discourse—representing autistic people as
puzzling, mysterious, less-than-human entities who are “short
a few cognitive pieces,” who are utterly self-contained,
disconnected, and [who] need to “fit in” (Heilker and Yergeau,
2011: 494).
Brook (2016) begins a letter to the editor of the journal
Autism, which uses puzzle-piece imagery both on its cover
and in its masthead, by writing, “I will not attempt to detail
why the puzzle piece as a symbol for autism is offensive;
it is easy to search online and find many and varied reasons
from different authors” (p. 251) (see, for example, Endow,
2014; Hillary, 2013; Ised, 2012; Nakhla-Thometz, 2013;
Regan, 2015; Tucker, 2008; Wyatt, 2008).
Thus, while many autism organizations continue to use
puzzle-piece imagery, many autistic people object. Ergo,
puzzle-piece imagery stirs debate (Edmunds, 2013; Goin-
Kochel, 2016; Grinker and Mandell, 2015; Muzikar, 2015).
To our knowledge, however, no research has empirically
examined whether puzzle-piece images evoke negative
associations in the general public. That was the purpose of
our study. We measured participants’ implicit and explicit
associations to puzzle pieces, both those used as autism
logos and those used more generically.
Members of the general public completed an Implicit
Association Test (Greenwald et al., 1998), which is a
speeded categorization task often used to measure implicit
attitudes (including attitudes toward images of brands,
Maison et al., 2004). Participants categorized images of
puzzle pieces and matching images of non-puzzle piece
shapes. During test trials, the images of puzzle pieces or
shapes were paired with either negative or positive concepts.
Shorter latencies to categorize puzzle pieces versus
shapes when they were paired with negative versus positive
concepts indicated negative implicit biases. Participants also
completed an Explicit Association Task, which is a standard task
for assessing consumers’ explicit associations with brands
(using images of those brands; Koll et al., 2010; Krishnan,
1996; Spears et al., 2006).
Method
Participants
We pre-specified our sample size according to Nosek’s
(2005) extensive analysis of Implicit Association Tasks. In
each of Nosek’s analyses, an average of 200 participants
were assigned to each material set. Because our study had
two material sets, our sample size was 400 adults who
were recruited through Amazon Mechanical Turk. Data
from 30 other participants were excluded: 19 participants
responded too rapidly (more than 10% of their response
times were faster than 300 ms, which requires removal
from the data set according to Greenwald et al., 2003); 4
participants had complications from either their Internet
connection or their own computers; and 7 participants
failed to provide a Mechanical Turk authorization code.
Of the 400 participants, 53.00% identified as men;
46.00% identified as women; 0.75% identified outside the
gender binary; and 0.25% did not wish to disclose their
gender. The participants’ average age was 33.76 years
(standard deviation (SD) = 10.37), 97.99% reported that
English was their native language, and 98.80% reported
living in the United States for most of the past five years.
Fewer than 1% (0.50%) identified as having an autism
spectrum diagnosis; 3.00% identified as working with
autistic persons; 12.00% as being a friend of an autistic person;
and 15.00% as having an autistic relative.
Participants were kept naive about the research hypothesis.
The experiment was titled “Concepts and Images,”
and, when recruited, participants were told that “In this
study, you will be asked to classify written concepts into
two categories and classify visual images into two categories.”
Participants were compensated US$1 for their participation
in the 10-minute study.

Gernsbacher et al. 120
Stimuli
For the Implicit Association Task, the image stimuli comprised
10 images of generic PUZZLE pieces with 10
matching geometric SHAPEs and 10 images of autism
PUZZLE piece logos with 10 matching geometric SHAPE
logos. The generic puzzle piece images were royalty-free
clip art, and the autism puzzle piece logos were taken from
autism organizations’ logos and products. The geometric
shapes, which were rectangles, ovals, stars, triangles,
rounded rectangles, and the like, matched their respective
puzzle pieces in color, texture, border, shading, and size.
All stimuli are available at the Open Science Framework,
https://osf.io/mv2zt/.
The concept stimuli comprised 10 nouns representing
NEGATIVE concepts (e.g. grief, agony, cancer) and 10
nouns representing POSITIVE concepts (e.g. cheer, paradise,
laughter), which had been used in previous studies
(e.g. Greenwald et al., 1998). The 10 NEGATIVE
and 10 POSITIVE concept words did not differ in length,
as measured in letters (M = 6.500, SD = 1.434 and
M = 6.100, SD = 1.197; t(18) = 0.677, p = 0.507, d = 0.303)
or syllables (M = 2.000, SD = 0.817 and M = 2.100,
SD = 0.738; t(18) = –0.287, p = 0.777, d = –0.129). One
NEGATIVE and one POSITIVE concept, as well as one
generic PUZZLE piece and its matching generic SHAPE
and one autism PUZZLE piece logo and its matching
autism SHAPE logo, were used to illustrate the task and
allow participants to practice during the trials that preceded
the test blocks.
Material sets
Four material sets were constructed, Generic-Material-Set
A (illustrated in Figure 1), Logo-Material-Set A (illustrated
in Figure 2), Generic-Material-Set B, and Logo-Material-
Set B, by counterbalancing two between-subjects variables:
(1) whether PUZZLE pieces were associated with
NEGATIVE concepts during the first test block and
POSITIVE concepts during the last test block (Generic-
Material-Set A and Logo-Material-Set A) or PUZZLE
pieces were associated with NEGATIVE concepts during
the last test block and POSITIVE concepts during the first
test block (Generic-Material-Set B and Logo-Material-Set
B) and (2) whether the images were Generic PUZZLE
pieces and SHAPEs (Generic-Material-Set A and Generic-
Material-Set B) or Autism Logos PUZZLE pieces and
SHAPEs (Logo-Material-Set A and Logo-Material-Set B).
The order of image and concept stimuli was identical
across the four material sets and was pseudo-randomly
arranged so that each image or concept occurred only once
within each of the two test blocks, and no more than three
trials containing concepts or images occurred consecutively.
One hundred participants were randomly assigned to
each material set and did not differ in age (F(3,394) = 1.241,
Figure 1. Example Implicit Association trials for Generic-
Material-Set A.
p = 0.294, η2p = 0.009), gender (χ2(3) = 6.028, p = 0.110,
Cramer’s V = 0.123), having an autism spectrum diagnosis
(χ2(3) = 2.022, p = 0.568, V = 0.073), being friends with an
autistic person (χ2(3) = 1.170, p = 0.760, V = 0.051), working
with autistic persons (χ2(3) = 2.103, p = 0.551, V = 0.074), or
having an autistic relative (χ2(3) = 5.018, p = 0.171,
V = 0.051).
Procedure
Implicit Association Task. The Implicit Association Task
comprised 206 trials, divided into seven blocks, as illustrated
in Figures 1 and 2. During Block 1, participants
practiced categorizing images that were either PUZZLE
pieces or SHAPEs. For participants assigned to Generic-
Material-Set A (Figure 1) or Logo-Material-Set A (Figure
2), the category label SHAPE appeared on the left and
PUZZLE appeared on the right side of their computer
screen; therefore, participants pressed the ‘E’ key (on the
left side of their computer keyboards) if the image was a
SHAPE or the ‘I’ key (on the right side of their keyboards)
if the image was a PUZZLE. For participants assigned to
Generic-Material-Set B or Logo-Material-Set B, the category
label PUZZLE appeared on the left and SHAPE
appeared on the right; therefore, participants pressed ‘E’ if
the image was a PUZZLE or ‘I’ if the image was a SHAPE.

121!Autism [author copy]
During Block 2, participants practiced categorizing
lexical concepts that had POSITIVE (e.g. cheer ) or
NEGATIVE connotations (e.g. grief ). For all participants,
the category label POSITIVE appeared on the left and
NEGATIVE appeared on the right; therefore, participants
pressed ‘E’ if the concept was POSITIVE or ‘I’ if the concept
was NEGATIVE. During Block 3, participants practiced
categorizing images (PUZZLE or SHAPE) while
also categorizing concepts (POSITIVE or NEGATIVE).
During Block 4, participants continued to categorize
images (PUZZLE or SHAPE) while also categorizing concepts
(POSITIVE or NEGATIVE), and the 72 trials in
Block 4 served as the first set of test trials.
During Block 5, participants again practiced categorizing
only images because the response keys (and the
location of the image category labels) were reversed. For
participants assigned to Generic-Material-Set A (Figure
1) or Logo-Material-Set A (Figure 2), the category label
PUZZLE appeared on the left and SHAPE appeared on
the right; therefore, participants pressed ‘E’ if the image
was a PUZZLE or ‘I’ if the image was a SHAPE. The
opposite was true for participants assigned to Generic-
Material-Set B or Logo-Material-Set B. During Block 6,
participants practiced categorizing images while also
categorizing concepts, with the response key mapping
Figure 2. Example Implicit Association trials for Logo-
Material-Set A.
established during Block 5. During Block 7, participants
continued to categorize images while also categorizing
concepts, and the 72 trials in Block 7 served as the last
set of test trials.
Participants were encouraged to rest their index fingers
on the ‘E’ and ‘I’ keys for the duration of the task. They
were encouraged to make their categorizations as rapidly
as possible, without making too many errors. If participants
made an error, a red X appeared on the screen, and
they were required to press the correct key to continue.
Explicit Association Task. After completing the Implicit
Association Task, participants completed a short questionnaire,
which also assayed demographic information (e.g. age, gender).
Two items in the short questionnaire provided the Explicit
Association Task: “When I see a shape, the first few thoughts
that come to my mind are ____” and “When I see a puzzle
piece, the first few thoughts that come to my mind are ____.”
The participants were given five lines for typing five free
associations for each of the two statements.
Data analysis
The Implicit Association Task data were analyzed via traditional
analyses of variance (ANOVAs) and planned comparison
t-tests on participants’ mean correct latencies and
error rates for categorizing images in each of the two test
blocks. The Implicit Association Task data were also analyzed
via the Implicit Association Test scoring algorithm
(Greenwald et al., 2003). For both the ANOVAs and the
Implicit Association Test scoring algorithm, categorization
latencies longer than 10,000 ms were removed. The
Explicit Association data were analyzed with chi-square
and z-tests. For all analyses, an alpha level of 0.001 was
adopted to protect against type I error.
Mean correct latencies and error rates. The primary depend-
ent variables comprised participants’ mean correct latencies
and error rates to categorize images in the two test
blocks (i.e. Block 4 and Block 7 in Figures 1 and 2). One
block of test trials provided participants’ mean correct
latencies and error rates for categorizing PUZZLE pieces
paired with NEGATIVE concepts and SHAPEs paired
with POSITIVE concepts (Block 4 for Generic-Material-
Set A and Logo-Material-Set A; Block 7 for Generic-
Material-Set B and Logo-Material-Set B). The other block
of test trials provided participants’ mean correct latencies
and error rates for categorizing PUZZLE pieces paired
with POSITIVE concepts and shapes paired with NEGATIVE
concepts (Block 4 for Generic-Material-Set B and
Logo-Material-Set B; Block 7 Generic-Material-Set A and
Logo-Material-Set A).
Implicit Association Scores. The third dependent variable com-
prised participants’ Implicit Association Scores, computed

Gernsbacher et al. 122
via the algorithm specified by Greenwald et al. (2003). A
mean correct latency was computed for each participant for
each of the two test blocks, averaging across all trials in each
test block. For each participant, a pooled SD (of each participant’s
correct latencies) was computed across the two test
blocks. Next, each participant’s error latencies were replaced
with that participant’s mean correct latency (for that test
block) plus 600 ms. Then, a difference score was computed
for each participant, subtracting one test block from the other
test block so that a negative Implicit Association Score indicated
negative implicit associations to puzzle pieces. Finally,
each participant’s difference score was divided by that participant’s
pooled SD, producing an Implicit Association Score.
Explicit Associations. The fourth dependent variable comprised
the participants’ Explicit Associations, which were
coded independently by two of the authors (JLS and JSB),
naive to which material set the participants had been randomly
assigned (the coders were unable to remain naive to
the prompt to which participants had responded because of
the nature of participants’ responses, for example, responding
“where does it fit?” or “jigsaw” in response to “When
I see a puzzle piece, the first few thoughts that come to my
mind are ____” and responding “oval” and “square” in
response to “When I see a shape, the first few thoughts that
come to my mind are ____”).
Participants’ Explicit Associations were classified into
four mutually exclusive categories: Negative Associations
(e.g. “problem,” “bad,” “ugly,” or “sad”), Positive
Associations (e.g. “fun,” “happy,” “good,” or “pleasant”),
Neutral Associations (e.g. “where does it fit?”, “jigsaw,”
“oval,” or “square”), and No Associations (i.e. the participant
left the line blank). The coders agreed on 96.13% of the cals-
sifications, and disagreements were resolved by consensus.
Results
Mean correct latencies and error rates
Participants’ mean correct latencies and error rates are illustrated
in Figure 3. As shown in the left panel, the shortest
latencies were observed when participants responded with
the key that paired PUZZLE pieces with NEGATIVE concepts
(M = 846 ms, SD = 194 ms). These PUZZLE-NEGATIVE
pairings were categorized with significantly
shorter latencies than SHAPE-NEGATIVE pairings
(M = 916 ms, SD = 214 ms, t(399) = –7.661, p < 0.001),1
PUZZLE-POSITIVE pairings (M = 889 ms, SD = 203 ms,
t(399) = –4.761, p < 0.001), and SHAPE-POSITIVE pairings
(M = 885 ms, SD = 204 ms, t(399) = –6.749, p < 0.001),
resulting in a significant interaction between image category
(PUZZLE vs SHAPE) and concept connotation (NEGATIVE
vs POSITIVE, F(1,398) = 19.16, p < 0.001, η2p = 0.046).
As shown in the right panel of Figure 3, PUZZLE-NEGATIVE
pairings were also categorized at a significantly
Figure 3. Participants’ mean correct latencies (left side of
figure) and error rates (right side of figure) for categorizing
images as PUZZLE pieces or SHAPEs when paired with
NEGATIVE or POSITIVE concepts. Error bars are 99.9%
confidence intervals (CIs) of the mean.
lower error rate (M = 6.431%, SD = 7.218%) than SHAPE-
NEGATIVE pairings (M = 8.188%, SD = 8.538%,
t(399) = –4.207, p < 0.001), although error rates to PUZZLE-
NEGATIVE pairings did not differ significantly from error
rates to PUZZLE-POSITIVE pairings (M = 6.945%,
SD = 7.705%, t(399) = 1.318, p = 0.188) or SHAPE-POSITIVE
pairings (M = 5.667%, SD = 7.632%, t(399) = 2.550, p = 0.011).
The significantly lower error rates for PUZZLE-NEGATIVE pairings
than SHAPE-NEGATIVE pairings resulted in a significant
interaction between image category (PUZZLE vs SHAPE) and
concept connotation (NEGATIVE vs POSITIVE; F(1,398) = 16.95,
p < 0.001, η2p = 0.041).
Neither participants’ mean correct latencies nor their
error rates led to a significant three-way interaction among
image type (Generic Images vs Autism Logos), image category
(PUZZLE vs SHAPE), and concept connotation
(NEGATIVE vs POSITIVE; F(1,398) = 3.511, p = 0.062,
η2p = 0.009 for mean correct latencies; F(1,398) = 2.132,
p = 0.145, η2p = 0.005 for error rates). The lack of a significant
three-way interaction suggests that participants’ negative
implicit associations to puzzle pieces were unaffected
by whether the puzzle pieces were generic images or
autism logos.
Indeed, Generic Image PUZZLE-NEGATIVE pairings
(M = 796 ms, SD = 142 ms) were categorized significantly
faster than Generic Image SHAPE-NEGATIVE pairings
(M = 885 ms, SD = 206 ms, t(199) = –7.041, p < 0.001),
Generic Image PUZZLE-POSITIVE pairings (M = 851 ms,
SD = 184 ms, t(199) = –4.708, p < 0.001), and Generic
Image SHAPE-POSITIVE pairings (M = 836 ms,
SD = 168 ms, t(199) = –5.374, p < 0.001). Similarly, Autism
Logo PUZZLE-NEGATIVE pairings (M = 896 ms,
SD = 224 ms) were categorized significantly faster than
Autism Logo SHAPE-NEGATIVE pairings (M = 946 ms,
SD = 218 ms, t(199) = –3.872, p < 0.001) and Autism Logo
SHAPE-POSITIVE pairings (M = 934 ms, SD = 224 ms,

123!Autism [author copy]
t(199) = –4.292, p < 0.001), although not significantly
faster than Autism Logo PUZZLE-POSITIVE pairings
(M = 926 ms, SD = 215 ms, t(199) = –2.226, p = 0.027).
To summarize, participants’ mean correct latencies and
error rates demonstrated that members of the general population
have negative implicit biases against puzzle pieces.
Participants’ negative implicit biases against puzzle pieces
encompass generic puzzle pieces and autism logo puzzle
pieces.
Implicit Association Scores
Figure 4 presents the distribution of participants’ Implicit
Association Scores. The dashed line at zero indicates zero
negative or positive implicit bias. Participants’ mean Implicit
Association Score, –0.137 (SD = 0.511), was significantly
more negative than zero (t(399) = –5.357, p < 0.001), indicating
a significant negative bias against puzzle pieces. The
mean Implicit Association Score of participants who categorized
Generic Images (M = –0.199, SD = 0.529) did not differ
significantly from the mean Implicit Association Score of
participants who categorized Autism Logos (M = –0.075,
SD = 0.486; F(1,398) = 5.887, p = 0.016, η2p = 0.015).
To summarize, participants’ Implicit Association Task
Scores, like their mean correct latencies and error rates,
indicate that members of the general population have a
negative implicit bias against puzzle pieces. Participants’
negative implicit bias against puzzle pieces encompasses
autism logo puzzle pieces and generic puzzle pieces.
Explicit Associations
Participants’ negative versus positive Explicit Associations
to “seeing a puzzle piece” differed significantly from their
negative versus positive Explicit Associations to “seeing a
shape” (χ2(1) = 22.00, p < 0.001, d = 0.324). Participants
produced significantly more negative Explicit Associations
in response to “seeing a puzzle piece” (50% ± 6.76%, 99.9%
CI) than in response to “seeing a shape” (33% ± 9.44%;
z = 4.693, p < 0.001, d = 0.491). Participants also produced
significantly fewer positive Explicit Associations in
Figure 4. Distribution of participants’ Implicit Association
Scores. The dashed line at zero indicates zero negative or
positive implicit bias.
response to “seeing a puzzle piece” (50% ± 6.76%) than in
response to “seeing a shape” (67% ± 9.44%; z = –4.693,
p < 0.001, d = 0.441).
Participants produced significantly more negative
Explicit Associations in response to “seeing a puzzle piece”
than “seeing a shape” regardless of whether they had performed
the Implicit Association Task on Generic Images or
Autism Logos (χ2(1) = 0.320, p = 0.572, d = 0.058).
Similarly, participants produced significantly fewer positive
Explicit Associations in response to “seeing a puzzle
piece” than “seeing a shape” regardless of whether they had
performed the Implicit Association Task on Generic Images
or Autism Logos (χ2(1) = 0.368, p = 0.544, d = 0.056).
To summarize, participants’ Explicit Associations, like
their Implicit Associations, demonstrate that members of
the general population have negative explicit associations
with puzzle pieces.
Discussion
This study examined the general public’s associations to
puzzle piece images, both those used as autism logos and
those used more generically. Using the Implicit Association
Task, this study demonstrated that the general public has a
negative implicit bias against puzzle pieces. Participants
categorized images of puzzle pieces significantly more
rapidly and accurately when the response key was paired
with identifying negative connotations. Participants’ average
Implicit Association Score was also significantly negative.
Although of moderate magnitude, participants’ mean
Implicit Association Score resembles the magnitude
reported in studies examining the general population’s
implicit bias against images of ethnic minorities (Devos
and Ma, 2008; Foroni and Bel-Bahar, 2010; Nosek, 2005).
Using an Explicit Association Task, this study also
demonstrated that the general public has negative explicit
associations to puzzle pieces. Half the participants’ negative
versus positive explicit associations to puzzle pieces
were negative, whereas only a third of their explicit associations
to non-puzzle piece shapes were negative.
Although for some phenomena, people’s explicit associations
differ from their implicit associations (Nosek, 2005);
the general public’s negative explicit associations to puzzle
pieces corroborate their negative implicit associations.
Stronger corroboration between people’s explicit and
implicit associations indicates a more valid assessment of
each type of association (Greenwald et al., 2009).
This study also demonstrated that the general public’s
negative bias against puzzle pieces applies to generic puzzle
pieces, as well as to puzzle pieces used in autism logos.
Our participants’ explicit associations inform us as to why.
Participants associated puzzle pieces with imperfection,
incompletion, uncertainty, difficulty, the state of being

Gernsbacher et al. 124
unsolved, and, most poignantly, being missing. Some of
the participants’ explicit associations even referenced negative
aspects of puzzle pieces’ physical shape with associations
such as odd shaped, jagged, and uneven.
Apart from the Wikipedia logo, the puzzle-piece imagery
used in non-autism logos usually differs from the puzzle-piece
imagery used in autism logos (see https://osf.io/mv2zt/
for the most frequent results of a Google image search).
Puzzle-piece imagery used in non-autism logos usually
comprises interlocking, completed, and integral puzzle
pieces; in contrast, puzzle-piece imagery used in autism-related
logos usually comprises single puzzle pieces in isolation
or incomplete puzzles with a piece missing, reifying
our participants’ associations of puzzles pieces as reflecting
incompletion, imperfection, and the state of being unsolved.
This study’s results not only support many autistic
persons’ claims about puzzle piece imagery but also bear
public policy implications. If an autism organization’s
intention for using puzzle-piece imagery is to evoke negative
associations, then the use of puzzle-piece imagery
is apt. If, instead, an autism organization’s intention for
using puzzle-piece imagery is, as Grinker and Mandell
(2015) report, “not meant to devalue” autistic persons
“but, quite the opposite, to celebrate them,” (p. 644) our
results suggest that puzzle-piece imagery should probably
be avoided.
Availability of data and materials
The data set and stimuli supporting the conclusions of this article
are available in the Open Science Framework repository,
https://osf.io/mv2zt/.
Ethical approval and consent to participate
Participants provided informed consent, and the experiment was
approved by the Institutional Review Board at the University of
Wisconsin–Madison (protocol SE-2013-0965).
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) received no financial support for the research,
authorship, and/or publication of this article.
Note
1. As Lakens (2013) notes, for matched-pair t-tests, the t-value
and degrees of freedom provide the appropriate estimate of
effect size.
References
Acosta MT and Pearl PL (2003) The neurobiology of autism: new
!pieces of the puzzle. Current Neurology and Neuroscience
!Reports 3: 149–156.
Brook K (2016) A response to ‘notes on a puzzle piece.’ Autism
!20: 251–252.
CNRS (2015) Missing piece surfaces in the puzzle of autism.
!ScienceDaily. Available at: https://www.sciencedaily.com/releases/
2015/08/150804074037.htm
Crawford C (2004) Solving the autism spectrum puzzle. Alaska
!Special Education Service Area Newsletter 62(2): 2–4.
Devos T and Ma DS (2008) Is Kate Winslet more American than
!Lucy Liu? The impact of construal processes on the implicit
!ascription of a national identity. British Journal of Social
!Psychology 47: 191–215.
Edmunds DL (2013) A puzzle piece? The need for a new symbol
!for autistics. Psychology Today. Available at: https://
www.psychologytoday.com/blog/extreme-states-mind/201305/puzzle-
piece-the-need-new-symbol-autistics
Endow J (2014) Goodnight autism puzzle pieces. Ollibean.
!Available at: http://ollibean.com/goodnight-autism-puzzle-pieces/
Foroni F and Bel-Bahar T (2010) Picture-IAT versus Word-
!IAT: level of stimulus representation influences on the IAT.
!European Journal of Social Psychology 40: 321–337.
Goin-Kochel RP (2016) Musings on the puzzle piece. Autism 20:
!250.
Greenwald AG, McGhee DE and Schwartz JLK (1998)
!Measuring individual differences in implicit cognition: the
!Implicit Association Test. Journal of Personality and Social
!Psychology 74: 1464–1480.
Greenwald AG, Nosek BA and Banaji MR (2003) Understanding
!and using the Implicit Association Test, I: an improved scoring
!algorithm. Journal of Personality and Social Psychology
!85: 197–216.
Greenwald AG, Poehlman TA, Uhlmann EL, et al. (2009)
!Understanding and using the Implicit Association Test, III:
!meta-analysis of predictive validity. Journal of Personality
!and Social Psychology 97: 17–41.
Grinker RR and Mandell D (2015) Notes on a puzzle piece.
!Autism 19: 643–645.
Healy EM (2008) Improving autism policies beyond the enigma:
!current challenges and future considerations for improving
!policies to support individuals living with autism. Speaker’s
!Journal 8: 141–156.
Heilker P and Yergeau M (2011) Autism and rhetoric. College
!English 73: 485–497.
Hillary A (2013) Red flags. Yes, That Too. Available at: http://
yesthattoo.blogspot.com/2013/01/red-flags.html
Ised E (2012) Puzzling people. The Standard Review. Available
!at: http://diversityrules.typepad.com/my_weblog/2012/04/puzzling-
people.html
Koll O, von Wallpach S and Kreuzer M (2010) Multi-method
!research on consumer–brand associations: comparing free
!associations, storytelling, and collages. Psychology &
!Marketing 27: 584–602.
Krishnan HS (1996) Characteristics of memory associations: a
!consumer-based brand equity perspective. International
!Journal of Research in Marketing 13: 389–405.
Lakens D (2013) Calculating and reporting effect sizes to facilitate
!cumulative science: a practical primer for. Frontiers in
!Psychology 4: 863.
McGuire AE (2012) On the “puzzle of autism” and the incompleteness
!of autism awareness. Journal on Developmental
!Disabilities 18: 96–100.

125!Autism [author copy]
McGuire AE and Michalko R (2011) Minds between us: autism,
!mindblindness and the uncertainty of communication.
!Educational Philosophy and Theory 43: 162–177.
Maison D, Greenwald AG and Bruin RH (2004) Predictive validity
!of the Implicit Association Test in studies of brands,
!consumer attitudes, and behavior. Journal of Consumer
!Psychology 14: 405–415.
Muzikar D (2015) The autism puzzle piece: a symbol that’s
!going to stay or go? The Art of Autism. Available at: http://the-art-
of-autism.com/the-autism-puzzle-piece-a-symbol-of-what/
Nakhla-Thometz A (2013) Today on “what really grinds my
!gears”: a puzzle piece as a symbol for autism. Angry Autie.
!Available at: https://angryautie.wordpress.com/2013/06/14/today-
on-what-really-grinds-my-gears-a-puzzle-piece-as-a-symbol-for-
autism/
National Autistic Society (2007) Perspectives on a puzzle piece.
!Available at: http://web.archive.org/web/20070714093137/http:/
www.nas.org.uk/nas/jsp/polopoly.jsp?d=364&a=2183
Nosek BA (2005) Moderators of the relationship between
!implicit and explicit evaluation. Journal of Experimental
!Psychology: General 134: 565–584.
Nowicki EA, Brown J and Stepien M (2014) Children’s structured
!conceptualizations of their beliefs on the causes of
!learning difficulties. Journal of Mixed Methods Research
!8: 69–82.
O’Connor J (2009) Finding the missing puzzle piece of autism.
!New Science 17: 14.
Regan R (2015) Why the puzzle piece is an offensive symbol.
!Social Skills for Autonomous People. Available at: http://
realsocialskills.org/post/115578535827/why-the-puzzlepiece-is-
an-offensive-symbol
Sarrett JC (2011) Trapped children: popular images of children
!with autism in the 1960s and 2000s. Journal of Medical
!Humanities 32: 141–153.
Schaaf CP and Zoghbi HY (2011) Solving the autism puzzle a
!few pieces at a time. Neuron 70: 806–808.
Smukler D (2005) Unauthorized minds: how “theory of mind”
!theory misrepresents autism. Mental Retardation 43:
!11–24.
Spears N, Brown TJ and Dacin PA (2006) Assessing the corporate
!brand: the unique corporate association valence (UCAV)
!approach. Journal of Brand Management 14: 5–19.
State MW (2010) Another piece of the autism puzzle. Nature
!Genetics 42: 478–479.
Stevenson JL, Harp B and Gernsbacher MA (2011) Infantilizing
!autism. Disability Studies Quarterly 31(3): 1675.
Tucker T (2008) Post-puzzle piece autism. I Am An Autism
!Parent. Available at: http://iamanautismparent.com/post-puzzle-
piece-autism/
Van Rijn H and Stappers PJ (2008) The puzzling life of autistic
!toddlers: design guidelines from the LINKX project.
!Advances in Human-Computer Interaction 2008: 639435.
Wyatt CS (2008) Logos, symbols, ribbons … The Autistic Me.
!Available at: http://theautisticme.blogspot.com.au/2008/04/logos-
symbols-ribbons.html