781

British Journal of Health Psychology (2016), 21, 781–795

© 2016 The Authors. British Journal of Health Psychology published by
John Wiley & Sons Ltd on behalf of British Psychological Society
www.wileyonlinelibrary.com

Children’s implicit understanding of the stress–

illness link: Testing development of health cognitions
Tara J. Cheetham, Julie M. Turner-Cobb* and Tim Gamble
Department of Psychology, University of Bath, UK

Objective. Innate knowledge and developmental stage theory have been used to
explain children’s understanding of concepts relating to health, illness, and stress. The aim
of this study was to investigate the degree to which children demonstrate unconscious
cognitive associations between the concepts of stress and illness.
Design. The study employed an experimental design using an age appropriate implicit
association task.
Methods. Thirty-two children (5–11 years of age) completed the Preschool Implicit
Association Test (PSIAT), a computer-based measure of reaction time to consistent
(stress and illness) and inconsistent (stress and health) concept pairings.
Results. Whilst age group had a significant effect on reaction times (older children
generally displaying faster reaction times than younger children), those as young as 5–6 years
of age were able to demonstrate implicit associations between stress and illness using the
PSIAT. There was also some indication that this association peaks at around 7–8 years of age.
Conclusions. Findings support a combination of developmental stage theory and the innate
theory of children’s understanding. Whilst sample size is small, this study is the first to apply the
PSIAT to the context of implicit cognitive associations between stress and illness in children.
Findingshavepotentialimplicationsforthedeliveryofinterventionstofacilitatehealthpromotion
anddevelopmentofpositivehealthbehavioursinchildrenandindicatethatevenchildrenasyoung
as 5–6 years have some ability to relate to the concept that stress may influence illness.

Statement of contribution
What is already known on this subject?
The way in which children understand health and illness is commensurate with their developmental
stage and experience of illness. Children also appear to have a degree of innate understanding of health
and illness and their causes. Furthermore, recent work suggests children have some innate
understanding and knowledge of the concept of stress.
What does this study add?
 This is the first study to use an implicit association task to assess children’s understanding of stress
and illness.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which
permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no
modifications or adaptations are made.
The copyright line for this article was changed on 25th February 2016 after original online publication.

*Correspondence should be addressed to Julie M. Turner-Cobb, Department of Psychology, University of Bath, Claverton Down,
Bath BA2 7AY, UK (email: j.m.t.cobb@bath.ac.uk).

DOI:10.1111/bjhp.12181
782 Tara J. Cheetham et al.

 Implicit stress–illness associations were seen in children as young as 5–6 years of age.
 Whilst sample size was small, this proof-of-concept study bridges disciplines to further
understanding of health.

The understanding of stress as a contributory factor in physical health and well-being in

adults has received increasing recognition in health research and practice. Yet
understanding of this link in healthy children has received relatively little attention for
over a quarter of a century. The manner in which children conceptualize health and illness
was largely advanced by Bibace and Walsh in the 1980s; building on Piaget (1930) four
stages of cognitive development and applying it to the concept of illness. They identified
six developmental stages: Phenomenism and contagion (3–7 years); contamination and
internalization (8–10 years); physiological (11–13 years); and psychophysiological
(14 years and older), with an additional level of ‘incomprehension’ for the very youngest
3-year-olds (Bibace & Walsh, 1980). As with earlier criticism of Piagetian staging of
cognitive development, which pointed to the underestimation of children’s abilities
(Hughes & Donaldson, 1983), staging theory in relation to illness cognitions has received
similar criticisms.
A renewed interest in children’s understanding of health and illness is emerging. For
example, Koopman, Baars, Chaplin, and Zwinderman (2004) compared illness under-
standing in a sample of 8- to 12-year-old children with diabetes and a healthy sample.
Findings support the developmental staging process whilst acknowledging the role of
illness specific experience. Children with diabetes were found to have a more developed
understanding of diabetes compared to healthy controls but this understanding did not
transfer to understanding of other illnesses such as the common cold (Koopman et al.,
2004). An alternative approach to developmental staging is the theory of innate
understanding (Hergenrather & Rabinowitz, 1991) in which children are able to
understand more about health and illness than the staging theory would accept and they
do so from an earlier age (Bir & Podmore, 1990). This suggests an intuitive knowledge of
biological processes that could explain the ability of very young children to give
explanations and definitions of illness (Kalish, 1996). Goldman, Whitney-Saltiel, Granger,
and Rodin’s (1991) research with 4- to 6-year-olds also provides support for innate
understanding. They used three tasks to determine the level of illness comprehension of
27 children and found that most of the children showed some level of understanding of
illness causation, demonstrating a comprehension of health and illness from as young as
4 years.
Whilst the purest test of an innate response or association could be argued to be that
closest to birth, thus avoiding experiential input, in referring to innate theory in these
studies children of ages four and upwards are frequently the sampled population as such
paradigms require a degree of cognitive and verbal ability. We would concur with this
view and argue that innate ability can be assessed in young children of school age by
comparing across older school-age children. Furthermore, although termed ‘innate’, the
argument is not for ability necessarily demonstrable at birth but one that naturally occurs
or is allowed expression following development of necessary communication abilities, in
other words that children have a natural ability to make associations between stress and
illness and that this is not something that is evident only at the later psychophysiological
stages of development but occurs alongside and may influence the earlier stages of
development.
 Implicit understanding of stress and illness 783

Several studies have provided support for both staging theory and an innate
understanding of health in children. Normandeau, Kalnins, Jutras, and Hanigan (1998)
conducted structured interviews with 1,674 children aged 5–12 years and found age
differences across criteria, perception, consequences and threats to health. Yet results
also showed that level of understanding in each age group was more multidimensional,
suggesting that young children may also have an innate understanding of health and
illness. Similar results were found by Myant and Williams (2005) who interviewed 83
children aged 4–12 years about various contagious and non-contagious illnesses (cold,
chicken pox, asthma, toothache) and physical injuries (bruise, broken leg). They found
that definitions became more precise with age and that children as young as 4 years could
discuss illness to a certain extent, supporting the assertion that even young children have
some comprehension of illness.
To date, only one study has focused on children’s understanding of stress. Valentine,
Buchanan, and Knibb (2010) used semi-structured interviews to gather qualitative data
from 50 children aged 4–11 years about their knowledge of stress, and found that the
majority of the participants had some experience of stress, whether personally (direct) or
through a friend or relative (indirect). Older children provided more complex definitions
of stress than younger children, and the 10- to 11-year-olds give more multidimensional
definitions including emotional and physical/biological factors. Yet an innate under-
standing was demonstrated by even the youngest children (4 years) having some level of
understanding and knowledge of stress, mostly giving definitions including emotional
elements.
In this study, we examined children’s understanding of the link between stress and
illness, assessing both cognitive developmental stage of stress–illness understanding and
the level of intuitive understanding of psychosocial influences. The majority of previous
studies used self-report methods such as interviews, yet researchers have highlighted that
when investigating stress in children it is better to use a variety of assessments including
more objective alternatives to self-report measures where appropriate (Jessop & Turner-
Cobb, 2008; Turner-Cobb, Rixon, & Jessop (2008); Turner-Cobb, Rixon, & Jessop, 2011).
The present study builds on the findings by Valentine et al. (2010) and extends these
through the use of the Preschool Implicit Association Test (PSIAT) experimental paradigm
(Cvencek, Greenwald, & Meltzoff, 2011). The Implicit Association Test (IAT) is an
established cognitive assessment tool using reaction times (RTs) to investigate attitudes
across a range of topics including stereotyping or attitudes towards age, gender, race, and
age and of associations between self-esteem and self-concept (Greenwald & Farnham,
2000; Greenwald, McGhee, & Schwarz, 1998). It has only been applied to one area of child
health psychology; attitudes towards smoking (Andrews, Hampson, Greenwald, Gordon,
& Widdop, 2010) and has not previously been applied in the stress–illness context or to
measure the level of understanding of a health concept in children. As such, this is a proof-
of-concept study that focuses on conceptual design in bridging across disciplines of health
and cognitive/developmental psychology. It is acknowledged that the sample size is
representative of concept development, it is underpowered to fully test a null hypothesis
and as such findings are likely to be suggestive rather than conclusive.
We hypothesized that children of all ages, even the youngest participants, would
demonstrate implicit understanding of an association between stress and illness, showing
faster RTs and greater accuracy on the PSIAT when consistent stress and illness pairings
were presented, compared to inconsistent pairings of stress and health. We reasoned that
if the youngest participants were able to significantly distinguish between consistent and
inconsistent pairing as effectively as the older participants then this would provide
784 Tara J. Cheetham et al.

support for the innate theory of understanding. We also hypothesized that there would be
a gradual increase in RT (speed) on PSIAT performance, with older children demonstrat-
ing significantly faster RTs than the younger children, indicative of a developmental
pattern consistent with staging theory. Our expectation was for evidence of develop-
mental staging complemented by an underlying innate understanding. The latter was
given greater consideration than afforded in some of the original work in the area.

Method
Participants
Participants (n = 32; 13 girls and 19 boys) were aged 5–11 years and of a similar
educational level and socio-economic background to one another. All were white
Caucasian, of standard educational level, attending school in a semi-rural location of
relative affluence. Further sociodemographic data were not collected as parents were not
asked to contribute data for the study. Participants were drawn from a local school using
an opt-in procedure. Invitation letters were sent to the parents of all children in the school,
to which there was a 32% response rate for their child to opt in for participation in the
study.

Materials
Explicit understanding
A test of explicit understanding of the stress–illness link was initially assessed. The explicit
measure was a story, based on one used by Valentine et al. (2010) to assess children’s
understanding of stress. The version in this study was adapted to incorporate health and
illness. The story described two characters, one who encountered various stressors and
one who did not encounter stressors but carried out health promoting behaviours. After
the story, participants were asked about why they thought the character who
encountered the stressors in the story became ill. Answers were coded as 0 or 1 as to
whether they mentioned stress as a possible cause for the character’s illness.

The Preschool Implicit Association Test

The implicit association between stress and illness was measured using a modified version
of the PSIAT (Cvencek et al., 2011) adapted from the original adult version of the IAT
(Greenwald et al., 1998). The IAT is a computer-based task that measures RTs to stimulus
items. Participants have to sort the stimulus items into one of four categories using two
response buttons (where each response button corresponds to two categories). This
allows researchers to measure associations between a target concept and an attribute
concept by measuring whether people are quicker to respond when consistent/
associated or inconsistent/unassociated pairs are paired on the same response button). A
faster reaction time indicates a stronger link between concept and attribute. A meta-
analysis by Greenwald, Poehlman, Uhlmann, and Banaji (2009) of the predictive validity of
the IAT found it to be a better predictor of attitudes than explicit measures (r = .25 and
r = .13, respectively).
The IAT has been adapted very effectively for use with children to include colour-
coded response buttons, auditory presentation of any word stimuli by a neutral female
voice rather than written word, and pictures of children’s faces rather than adults’ faces
 Implicit understanding of stress and illness 785

(Baron & Banaji, 2006; Dunham, Baron, & Banaji, 2006; Rutland, Cameron, Milne, &
McGeorge, 2005; Steffens, Jelenec, & Noack, 2010). Although IATs have been used
successfully with children as young as 6 years, it was not until a few years ago that they
were adapted by Cvencek et al. (2011) for use in children under 6 years of age. Cvencek
et al. (2011) created the PSIAT, an effective tool for measuring 4-year-olds’ implicit
attitudes towards objects and gender adapted the standard IAT to present both auditory
and visual stimuli simultaneously. Reading level was not relied upon and the PSIAT used
fewer presentations of stimulus items (trials) than standard adult IATs to ensure
concentration throughout the experiment and use of both words and images, image
category reminders, and coloured stripes (green and orange) corresponding to response
buttons (Cvencek et al., 2011). IATs typically involve 180 trials (Greenwald, Nosek, &
Banaji, 2003) but Rutland et al. (2005) suggested a reduction in the number of trials is
appropriate for young children to avoid fatigue and boredom effects. Work by Cvencek
et al. (2011) reduced the number of trials in young children by 20% to 144. The current
study employed 112 trials in total (64 of which were critical trials) with an intertrial
interval of 400 ms.
The stimuli presented in the PSIAT were a combination of words and images, relating
to the four ‘target concept’ categories of ‘health’, ‘illness’, ‘stress’, and ‘no stress’, and
were presented in the centre of the laptop screen. Stimulus images were selected from a
combination of copyright-free image websites, the Karolinska Directed Emotional Faces
scale, and the International Affective Picture System. Stimulus words were selected from
the Affected Norms for English Words (Bradley & Lang, 1999), some used by Cvencek
et al. (2011). During material preparation and selection two additional researchers
independently coded, rated and verified the choice of pictures and words prior to testing.
A total of 36 words and 32 images were collected and rated for how strongly they were
associated with the four target concepts. The four images rated as best representative of
their target category were selected and used as the category images. These category
images remained on the screen throughout the critical blocks of the task (one in each of
the four corners of the screen) and can be seen in the bottom pictures in Figure 1. The
next highest rated 16 images and 16 words were used as stimulus items (four words and
four images per category). For example, the words in the stress category included
nervous, scared, worried, and busy, and the pictures showed children looking fearful and
upset. Similarly words in the illness category included fever, hospital, medicine and sick
and pictures showed medical situations and doctors.
A laptop running ePrime was used to display the stimuli and participants responded
using a response button box (only two response keys were needed). Stimulus items were
presented in a random order generated by ePrime and each item was presented once.
Headphones were used to play the audio files that accompanied the stimulus items (beeps
for images, verbalizations for words). The 32 stimulus items in the PSIAT were presented
in five blocks: Block 1 contained the 16 stress and no stress words/images, blocks 2 and 4
contained the 16 health and illness words/images, and blocks 3 and 5 included all 32
items. Only two categories were shown in blocks 1, 2, and 4 (see Figure 1 upper images);
therefore, each response button corresponded to one category. Conversely, in blocks 3
and 5 four categories were shown, two categories per response button. The purpose of
these two blocks is to test implicit understanding of the stress–illness relationship when
the two concepts stress and illness are paired together (associated/consistent pairs) or
conversely paired (unassociated/inconsistent pairs). In the consistent block the category
images for stress and illness were shown on the same side of the screen and shared a
response button, and the categories of no stress and health were shown on the other side
786 Tara J. Cheetham et al.

Happy

Hospital

Figure 1. Examples of Preschool Implicit Association Test (PSIAT) blocks presented. The two upper
images are examples of two practice blocks; (left) block 1 (no stress and stress) and (right) block 2 (health
and illness). The two lower images are examples of two test blocks; (left) block 3 displays associated pairs.

of the screen and also shared a response button. In the inconsistent block stress and health
were paired together and no stress and illness were paired together on the screen and the
same response button (see Figure 1 lower images).
During all five blocks the orange and the green stripes with the two/four category
images remained on the screen as a reminder to participants as to which button
corresponds to which category. The two response buttons each corresponded to a
category: The left/orange button to the category on the left and the right/green button to
the category on the right. The participants used these response buttons to indicate which
category the stimulus word/image belonged to. Counterbalancing ensured that half the
participants were randomly allocated the associated pairs first (order A) and half were
presented with the unassociated pairs first (order B) as in Cvencek et al. (2011). The
health and illness categories alternated between the right and left position on the screen
(blocks 2 and 4) whereas the stress and illness categories remained unchanged
throughout the task, to cause minimal confusion to participants.

Procedure
Written consent was obtained from parents and verbal assent was provided by child
participants in advance of testing. Participant experience of illness was indicated on the
consent form by their parents. Children were tested in a quiet room in the school by a
female experimenter, who remained present throughout testing (approximately 15 min
per participant). A story and question were used to assess participants’ explicit
 Implicit understanding of stress and illness 787

knowledge of the stress–health relationship and to check their comprehension of the

words that were to be presented in the implicit task. The researcher wrote down
participants’ answers verbatim. Participants then sat in front of the laptop to complete the
PSIAT using the response buttons. Participants were given verbal instructions and shown
example screens, after which they completed the PSIAT. The researcher gave positive
feedback and prompted participants to ‘guess’ which group a word or image belonged to
if they were unsure. This ensured that participants responded as quickly as possible to the
stimulus. At the end of the experiment participants were debriefed.

Statistical analysis
For the explicit task, responses were coded dichotomously as either mentioning/not
mentioning stress as a cause of illness (coded 0/1 respectively). In the implicit task,
performance was measured by RT to each stimulus item. RTs and accuracy of responses
(correct/incorrect) were recorded by ePrime. A higher RT to the associated pairs than to
the unassociated pairs would indicate a higher level of understanding of the link between
stress and illness (the associated concepts). The RT data were analysed according to the
scoring algorithm created by Greenwald et al. (1998). A split-plot ANOVA was used to
compare RT for consistent and inconsistent pair responses across age groups, using a 2
(RT: Consistent, inconsistent) 9 4 (Age: 5–6, 7, 8–9, 10–11 years) design. In addition, we
have also presented the data using a more recent scoring algorithm which calculates D
scores rather than RT scores (Greenwald et al., 2003) and analysed using a univariate
ANOVA.
The additional D score analysis, whilst informative, is treated with caution given the
small sample size but presented for consistency; therefore, findings are based on the RT
scoring as was deemed appropriate for this cohort. The rationale for using RT as opposed
to D scores is mostly due to the small sample size. The calculation of D scores is complex
and controls for numerous factors such as age and cognitive level. It therefore requires
more power and with a small sample such as this one we could not achieve the power
necessary for this; a probable explanation for why we found no statistically significant
results. As the hypothesis for this study focuses on whether children are faster to respond
to consistent or inconsistent concepts, it is appropriate to use the RT data for consistent
and inconsistent pairs. Although we have found statically significant results using RT data,
they should be treated with caution due to the small sample size.

Results
Age group and concept pairings: Reaction time
Descriptives for the demographic information and dependent variables can be found in
Table 1. A paired-samples t-test showed that RTs were faster for consistent than
inconsistent pairs for all age groups, demonstrating that participants respond more
quickly when stress and illness were paired on the same response button, t(31) = 5.32,
p < .001, 95% CI ( 853.58, 380.20). There was a roughly equal gender split in the
sample (59% boys), only 21% of the sample had experience of illness, and the majority of
participants (56%) explicitly stated a link between stress and illness.
A split-plot ANOVA revealed that age group did exert a significant effect on mean RT,
F(1, 28) = 28.37, p < .001, with participants demonstrating faster RTs for consistent
concept pairings (M = 2068.59, SD = 787.191) than for inconsistent concept pairings
788 Tara J. Cheetham et al.

Table 1. Number and percentages for the demographic information and means and standard deviations
(SD) of the dependent variables by child age (n = 32)

Child age (years)

5–6 (n = 8) 7 (n = 7) 8–9 (n = 10) 10–11 (n = 7)

Gender, N (%)
Male 2 (25) 3 (43) 8 (80) 6 (86)
Female 6 (75) 4 (57) 2 (20) 1 (14)
Experience of illness, N (%)
Yes 2 (25) 0 (0) 4 (40) 1 (14)
No 6 (75) 7 (100) 6 (60) 6 (86)
Explicit answer included ‘stress’, N (%)
Yes 2 (25) 4 (57) 6 (60) 6 (86)
No 6 (75) 3 (43) 4 (40) 1 (14)
Reaction time (RT) (ms) and D scores, mean (SD)
RT for consistent pairs 2,851 (947) 2,066 (466) 1,969 (438) 1,319 (403)
RT for inconsistent pairs 3,114 (655) 2,940 (1,091) 2,723 (812) 1,889 (476)
D scores 0.24 (0.50) 0.61 (0.60) 0.6 (0.30) 0.53 (0.75)

(M = 2685.48, SD = 876.56). Post-hoc tests showed a significant difference in mean RTs

between 5–6 and 10–11 years (p = .01) with 95% CIs (459.77, 2296.65). These results show
that older participants (10–11 years) had faster RTs than younger participants (5–6 years).
See Figure 2a for a comparison of RT to consistent and inconsistent pairings for each group.

Age group and concept pairings: Response accuracy

A mean score for accuracy of responses for consistent and inconsistent pairs was
calculated for each participant. Accuracy of responses refers to how many trials
participants responded to correctly; the closer the mean score is to 1 the more
correct responses there were, the closer it is to 0 the more incorrect responses there
were. A paired-samples t-test showed that participant responses were more accurate
when responding to consistent pairs than to inconsistent pairs, t(31) = 2.88,
p = .001, 95% CI (0.02, 0.09) with a medium to large effect size (r = .46). A series
of paired-samples t-tests demonstrated significant differences in accuracy to consistent
and inconsistent pairs for 8–9 years, t(9) = 8.34, p = .001, 95% CI (0.06, 0.10), and
10–11 years, t(6) = 2.83, p = .03, 95% CI (0.01, 0.11), demonstrating the two older
age groups responded more accurately to consistent than inconsistent concept pairs,
suggesting that they were better able to accurately distinguish between associated
and unassociated images and words.

Explicit understanding
An independent t-test showed that for consistent concept pairs participants who
explicitly identified stress as a cause of illness had significantly faster RTs than the
participants who did not explicitly link stress and illness, t(29) = 2.12, p = .03, 95% CI
(46.65, 1146.46) with a medium effect size (r = .37). No significant difference was found
for inconsistent pairs between those who had an explicit understanding of stress and
those who did not.
 Implicit understanding of stress and illness 789

(a)

(b)

Figure 2. Preschool Implicit Association Test (PSIAT) mean reaction times (RTs) in milliseconds
(ms) for consistent and inconsistent concept pairings by age group (a) and mean D scores (positive
scores indicate a preference for associated pairs and negative scores suggest a preference
for unassociated pairs) (b) across ages 5–11 years. Error bars represent 95% confidence
intervals.

Effect of illness experience

An independent t-test revealed no significant differences in mean RTs to consistent pairs
(p = .76) and inconsistent pairs (p = .78) of participants who had experience of illness
and those who did not have experience of illness, nor was there a significant difference
between accuracy of responses to consistent pairs (p = .77) and inconsistent pairs
(p = .59) for those with experience of illness and those without experience of illness.
790 Tara J. Cheetham et al.

Gender differences
An independent t-test showed there to be a significant effect of gender, with male
participants displaying significantly faster RTs to both the consistent, t(30) = 2.28,
p = .03, 95% CI ( 1138.11, 62.00), and inconsistent concept pairings, t(30) = 2.22,
p = .03, 95% CI ( 1269.72, 53.49), with a medium effect size (r = .38). However,
when gender was added into the ANOVA as a variable there was no significant
interaction. An independent t-test showed no significant difference in accuracy of
responses between males and females for consistent (p = .33) or inconsistent concept
pairs (p = .37).

IAT assumptions: Order effect, reliability and validity

To ensure that the PSIAT met the assumption that the order in which the blocks were
presented to participants had no significant effect on RTs an independent t-test was
administered. This showed no significant effect of order on RTs to consistent (p = .12) or
inconsistent concept pairs (p = .65). This study demonstrates the effectiveness of the
PSIAT in terms of reliability and validity in research on children’s understanding of the
relationship between stress and illness. As discussed in the method, the IAT is a reliable
tool for measuring implicit beliefs (Nosek, Greenwald, & Banaji, 2007) and has been
shown to have high predictive validity (Greenwald et al., 2009). The version of the PSIAT
used in the present study also meets the criteria for internal validity as discussed by Nosek
et al. (2007) who stated that internal validity can be ensured as long as it is unambiguous
which stimulus items belong to each category, the presentation of stimulus items is
random, and the consistent and inconsistent concept blocks are counterbalanced. Whilst
IATs are commonly used to measure attitudes, implicit beliefs and in some cases concepts,
they have rarely been used previously to examine health concepts and this is the first use
of the IAT to assess stress–illness associations or understanding. The categories of health,
illness, stress and no stress were clear and unambiguous.

Alternative analysis: D scores

As previously discussed, IAT data can be analysed using RT or D scores. IAT performance is
measured by response latency (the speed of the response or RT) and response accuracy
(whether the responses were correct or incorrect) to each stimulus item. The response
latency and accuracy data for each participant were transformed into D scores using the
improved IAT scoring algorithm from Greenwald et al. (2003) which has been used
effectively in IATs adapted for children including the PSIAT (Cvencek et al., 2011). D
scores represent the difference between the RT scores for associated/consistent and
unassociated/inconsistent concept pairs. A positive D score suggests a preference for the
consistent pairs (equivalent to a faster RT to those pairs) and a negative score indicates a
preference for the inconsistent pairs.
The new algorithm differs from the previous algorithm in relation to which extreme
values were deleted; in the original algorithm, RT scores were eliminated or recoded if
they were outside the range of 300–3,000 ms; however, the new algorithm eliminates
scores outside the range of 300–10,000 ms and participants who have more than 10% of
trials with a latency of <3,000 ms, making this method more inclusive and the RT method
more stringent. Cvencek et al. (2011) excluded participants based on the above criteria
and also those who had an error rate or 35% or above. Only one participant in the present
study had over this level of errors (50%) and so was excluded from the analysis. The data
 Implicit understanding of stress and illness 791

were screened for outliers, normality, linearity and multicollinearity and no issues arose. D
scores were analysed using a univariate ANOVA with D scores entered as the DV, age and
gender entered as the IVs and order of blocks, explicit understanding and experience of
illness entered as covariates. Order was the only variable that was found to be significant,
F(1, 17) = 9.83, p = .006, with a medium effect size (r = .37), whereas age (p = .821),
gender (p = .857), explicit understanding (p = .207), and experience of illness (p = .934)
were not significant. It appears that people who were presented with order A (M = 0.81,
SE = 0.10) which showed the consistent concept pairs first showed a stronger preference
for the consistent pairs than those who were presented with order B (M = 0.19,
SE = 0.12) and saw the inconsistent pairs first.
Although age and gender were not significant, there is an interesting pattern found
when the D scores for each age group are visually assessed. Figure 2b shows that 5-year-
olds have a preference towards the inconsistent pairs suggesting they lack an
understanding of the link between stress and illness. However, all of the other age
groups show a preference towards consistent pairs. As D scores represent the difference
between RTs to associated and unassociated pairs, a similar pattern was expected with the
RT and D score analyses. Reasons for the discrepancy are considered in the discussion.

Discussion
In this sample of 32 children aged 5–11 years, we found that children as young as 6 years
of age demonstrated explicit understanding of an association between stress and illness
using the PSIAT. There was some indication from D score evaluation that this association
peaks at around age 7–8 years of age with no effect found for the very youngest age group
of 5- to 6-year-old children. As expected, older children generally displayed faster RTs than
younger children indicative of both cognitive and developmental maturity.
The results of the present study give support for both innate and developmental
theories of children’s understanding of the relationship between stress and illness. The
majority of the younger participants completed the PSIAT with reasonable RTs and these
accelerated across age groups, demonstrating a developmental pattern of understanding.
Yet evidence for an innate understanding of the link between stress and illness was only
partially supported and the relative weight of the findings was in support of a
developmental rather than an innate understanding. Children of all ages responded
faster to consistent concept pairings (e.g., stress and illness) than inconsistent concept
pairings (e.g., stress and health) as expected. This pattern of results suggests that it was
easier for the participants to categorize stimuli when associated concepts shared a
response button. This could be because the concepts of stress and illness are strongly
linked in cognition and are therefore more easily and quickly accessed than when stress
and health are paired together. An alternative explanation of these results might suggest
that it is children’s understanding of the task that develops with age, as opposed to the
development of their understanding of the stress–health relationship. However, this
cannot explain how even the youngest children were able to complete the task with only a
few errors; therefore, the first explanation appears to be substantiated by the findings.
Although significant differences were not observed between all of the age groups, a clear
age-related pattern emerged.
Findings of the present study support research by Normandeau et al. (1998) and
Myant and Williams (2005) who found evidence of early intuitive understanding, as well as
support for children’s understanding of illness following a developmental pattern.
Valentine et al. (2010) found a similar pattern in children’s understanding of stress. The
792 Tara J. Cheetham et al.

present study demonstrates that even very young children have an understanding of stress
although this develops and becomes more comprehensive with age. The present study
builds upon qualitative work by Valentine et al. (2010) using an experimental paradigm to
address understanding of stress and illness. This link appears to follow the developmental
pattern of cognitive abilities in other domains. Yet there is some support for the youngest
children having some concept of the stress–illness link which has not been documented
previously and which supports the notion of stage theories underestimating young
children’s intellectual ability (e.g., Siegler, 1998). Innate theories suggest underlying
capability for implicit understanding of stress and illness associations which if overlooked
may limit the potential for building more sophisticated understanding of the nature of
illness and staying healthy. Such understandings could be harnessed and appropriately
encouraged in children to promote health and resilience, avoiding over protective or
disabling approaches that may generate more harm than good.
Although the results using RT data show clear support for the hypotheses, D score
analysis did not reveal significant effects. When analysed using ANOVA, neither age nor
gender had an effect on D scores, but in the RT data there were significant age and gender
effects. As the RT data and D scores represent the same data, it is unexpected for them to
show differing effects. We suggest this discrepancy is due to small sample size causing the
D score analysis to be underpowered. Nonetheless, the pattern is conceptually of interest
as it suggests that understanding may not follow a linear trajectory. D scores indicate a
peak in understanding at ages 7–8 years.

Limitations of the present study

The present study aimed simply to make an initial attempt to apply a well-established
paradigm to a new concept. We acknowledge a major limitation in sample size. A larger
sample would be required to improve the overall power of the study and enable more
meaningful D score interpretations. This is particularly true for the younger age groups
(<7 years old) where a larger sample would enable more in depth breakdown of ages for
closer scrutiny of innate understanding. Similarly for the 7- to 8-year-old age group, who
were suggested by the data as being at a potential pivotal age for understanding the stress–
illness link but effects were not sufficiently powered to determine significance. As
acknowledged in the introduction, the sample size used in this study is representative of
concept development and as such suggestive rather than conclusive. Nonetheless, it
represents some important ideas and methodological advances which further under-
standing of health. Issues relating to application of the PSIAT to a stress concept include
the lack of uniformity of image stimulus items, being of a variety of colours and styles. A
prevalence of male images over female images occurred due to practical issues of image
accessibility and may have explained why boys had faster RTs than girls. Furthermore, the
categories of ‘no stress’ and ‘health’ were selected as counterparts to ‘stress’ and ‘illness’
although they are not exact opposites, as ‘no stress’ is not a precise description and
‘health’ is not merely the absence of illness. Despite this, the concept descriptors serve
their purpose as opposing categories to the main concepts of interest.

Implications for future research

The application of psychosocial concepts of stress and health to an established implicit
association paradigm requires further validation, testing with a greater sample size and
range of ages. Yet the use of the PSIAT paradigm as a tool for assessing children’s
 Implicit understanding of stress and illness 793

understanding of stress and illness has proved extremely valuable. A measure of implicit
understanding was used to develop a more complete picture of children’s understanding,
not always accessible using explicit self-report measures. The importance of this area of
research was highlighted by Burbach and Peterson (1986) who suggested that
understanding of children’s knowledge can improve the way in which health
professionals communicate health messages to children, beneficial for both prevention
of illness and treatment interventions. Yet previous research has suggested that in
practice, health care professionals do not adapt health messages for children based on
patient age, developmental level or experience of illness, instead aiming their information
at ‘mid-school age level’ (Perrin & Perrin, 1983, p. 877). The findings of the current study
provide initial support for the inclusion of a child’s age or level of understanding of the
stress–illness link in health messages given to children. Issues concerned with the
prevention of illness are particularly important in relation to the health education given by
parents and schools. More broadly, the implicit associations tested could also be explored
in adolescent and adult populations to gain a fuller understanding of the differences
between child and adult understandings. This work assumes an implicit understanding of
stress and illness in adults, but this has not yet been systematically examined using the
methods outlined here. This raises important question for future research that could
potentially inform health interventions across the lifespan.
In conclusion, the present study is the first to use the experimental PSIAT paradigm to
address children’s understanding within the health psychology field. It is also the first
study to address children’s comprehension of associations between stress and illness.
Using this novel experimental method, the findings illustrate that children as young as 5–
6 years of age demonstrate some knowledge, at the level of implicit understanding, of a
relationship between stress and illness, and there is some indication that this unconscious
association is strongest in children aged 7–8 years. This study highlights the importance of
using psychosocial information from health psychology research informed by cross-
disciplinary work, to appropriately tailor health messages to children of different ages.
The findings have potential implications for the delivery of interventions to facilitate
health promotion and development of positive health behaviours in children. Whilst such
messages requires age appropriate delivery, unconscious understanding or capacity for
implicit associations between stress and illness should not be underestimated, even in the
youngest of school-age children.

Acknowledgements
The authors would like to thank the participants and schools who assisted in this study and
Emma Ashwin for providing the audio recordings which accompanied the PSIAT.

Conflict of interest
All authors declare no conflict of interest.

References
Andrews, J. A., Hampson, S. E., Greenwald, A. G., Gordon, J., & Widdop, C. (2010). Using the Implicit
Association Test to assess children’s implicit attitudes toward smoking. Journal of Applied
Social Psychology, 40, 2387–2406. doi:10.1111/j.1559-1816.2010.00663.x
794 Tara J. Cheetham et al.

Baron, A. S., & Banaji, M. R. (2006). The development of implicit attitudes: Evidence of race
evaluations from ages 6 and 10 and adulthood. Psychological Science, 17, 53–58. doi:10.1111/
j.1467-9280.2005.01664.x
Bibace, R., & Walsh, M. E. (1980). Development of children’s concepts of illness. Pediatrics, 66,
912–917. doi:10.1016/S0002-7138(09)61657-5
Bir, J. E., & Podmore, V. N. (1990). Children’s understanding of health and illness. Psychology &
Health, 4, 175–185. doi:10.1080/08870449008408151
Bradley, M. M., & Lang, P. J. (1999). Affective norms for English words (ANEW): Instruction
manual and affective ratings (Technical Report C-1). Gainsville, FL: The Centre for Research in
Psychophysiology, University of Florida.
Burbach, D. J., & Peterson, L. (1986). Children’s concepts of physical illness: A review and critique of
the cognitive-developmental literature. Health Psychology, 5, 307–325. doi:10.1037/0278-
6133.5.3.307
Cvencek, D., Greenwald, A. G., & Meltzoff, A. N. (2011). Measuring implicit attitudes of 4-year-olds:
The Preschool Implicit Association Test. Journal of Experimental Child Psychology, 109, 187–
200. doi:10.1016/j.jecp.2010.11.002
Dunham, Y., Baron, A. S., & Banaji, M. R. (2006). From American city to Japanese village: A cross-
cultural investigation of implicit race attitudes. Child Development, 77, 1268–1281.
doi:10.1111/j.1467-8624.2006.00933.x
Goldman, S. L., Whitney-Saltiel, D., Granger, J., & Rodin, J. (1991). Children’s representations of
“everyday” aspects of health and illness. Journal of Pediatric Psychology, 16, 747–766.
doi:10.1093/jpepsy/16.6.747
Greenwald, A. G., & Farnham, S. D. (2000). Using the Implicit Association Test to measure self-
esteem and self-concept. Journal of Personality and Social Psychology, 79, 1022–1038.
doi:10.1037/0022-3514.79.6.1022
Greenwald, A. G., McGhee, D. E., & Schwarz, J. L. K. (1998). Measuring individual differences in
implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology,
74, 1464–1480. doi:10.1037/0022-3514.74.6.1464
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the Implicit
Association Test: I. An improved scoring algorithm. Journal of Personality and Social
Psychology, 85, 197–216. doi:10.1037/0022-3514.85.2.197
Greenwald, A. G., Poehlman, T. A., Uhlmann, E. L., & Banaji, M. R. (2009). Understanding and using
the Implicit Association Test: III. Meta-analysis of predictive validity. Journal of Personality and
Social Psychology, 97, 17–41. doi:10.1037/a0015575
Hergenrather, J. R., & Rabinowitz, M. (1991). Age-related differences in the organization of
children’s knowledge of illness. Developmental Psychology, 27, 952–959. doi:10.1037/0012-
1649.27.6.952
Hughes, M., & Donaldson, M. (1983). The use of hiding games for studying coordination of
viewpoints. In M. Donaldson, R. Grieve & C. Pratt (Eds.), Early childhood development and
education (pp. 245–253). Oxford, UK: Basil Blackwell.
Jessop, D. S., & Turner-Cobb, J. M. (2008). Measurement and meaning of salivary cortisol: a focus on
health and disease in children. Stress, 11, 1–14. doi:10.1080/10253890701365527
Kalish, C. (1996). Causes and symptoms in preschoolers’ conceptions of illness. Child
Development, 67, 1647–1670. doi:10.2307/1131723
Koopman, H. M., Baars, R. M., Chaplin, J., & Zwinderman, K. H. (2004). Illness through the eyes of
the child: The development of children’s understanding of the causes of illness. Patient
Education and Counseling, 55, 363–370. doi:10.1016/j.pec.2004.02.020
Myant, K. A., & Williams, J. M. (2005). Children’s concepts of health and illness: Understanding of
contagious illnesses, non-contagious illnesses and injuries. Journal of Health Psychology, 10,
805–819. doi:10.1177/1359105305057315
Normandeau, S., Kalnins, I., Jutras, S., & Hanigan, D. (1998). A description of 5- to 12-year old
children’s conception of health within the context of their daily life. Psychology & Health, 13,
883–896. doi:10.1080/08870449808407438
 Implicit understanding of stress and illness 795

Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2007). The Implicit Association Test at age 7: A
methodological and conceptual review. In J. A. Bargh (Ed.), Automatic processes in social
thinking and behavior (pp. 265–292). New York, NY: Psychology Press.
Perrin, E. C., & Perrin, J. M. (1983). Clinicians’ assessments of children’s understanding of illness.
American Journal of Diseases of Children, 137, 874–878. doi:10.1001/
archpedi.1983.02140350048012
Piaget, J. (1930). The child’s conception of physical causality. London: Routledge & Kegan Paul
PLC.
Rutland, A., Cameron, L., Milne, A., & McGeorge, P. (2005). Social norms and self-presentation:
Children’s implicit and explicit intergroup attitudes. Child Development, 76, 451–466.
doi:10.1111/j.1467-8624.2005.00856.x
Siegler, R. S. (1998). Children’s thinking (3rd ed.). London: Prentice Hall.
Steffens, M. C., Jelenec, P., & Noack, P. (2010). On the leaky math pipeline: Comparing implicit
math-gender stereotypes and math withdrawal in female and male children and adolescents.
Journal of Educational Psychology, 102, 947–963. doi:10.1037/a0019920
Turner-Cobb, J. M., Rixon, L., & Jessop, D. S. (2008). A prospective study of diurnal cortisol
responses to the social experience of school transition in four-year-old children: Anticipation,
exposure, and adaptation. Developmental Psychobiology, 50, 377–389. doi:10.1002/
dev.20298
Turner-Cobb, J. M., Rixon, L., & Jessop, D. S. (2011). Hypothalamic-pituitary-adrenal axis activity and
upper respiratory tract infection in young children transitioning to primary school.
Psychopharmacology (Berl), 214, 309–317. doi:10.1007/s00213-010-1965-x
Valentine, A., Buchanan, H., & Knibb, R. (2010). A preliminary investigation of 4 to 11-year-old
children’s knowledge and understanding of stress. Patient Education and Counseling, 79,
255–257. doi:10.1016/j.pec.2009.08.011

Received 7 November 2014; revised version received 10 November 2015