Content uploaded by Eamon Fulcher
Author content
All content in this area was uploaded by Eamon Fulcher on Mar 15, 2018
Content may be subject to copyright.
Fulcher, E., Mathews, A. and Hammerl, M. (2008) Rapid acquisition
of emotional information and attentional bias in anxious children.
Journal of Behavior Therapy and Experimental Psychiatry, 39 (3).
pp. 321-339. ISSN 0005-7916 Available from: http://eprints.uwe.ac.uk/13372
We recommend you cite the published version.
The publisher’s URL is:
http://dx.doi.org/10.1016/j.jbtep.2007.08.003
Refereed: Yes
(no note)
Disclaimer
UWE has obtained warranties from all depositors as to their title in the material
deposited and as to their right to deposit such material.
UWE makes no representation or warranties of commercial utility, title, or fit-
ness for a particular purpose or any other warranty, express or implied in respect
of any material deposited.
UWE makes no representation that the use of the materials will not infringe
any patent, copyright, trademark or other property or proprietary rights.
UWE accepts no liability for any infringement of intellectual property rights
in any material deposited but will remove such material from public view pend-
ing investigation in the event of an allegation of any such infringement.
PLEASE SCROLL DOWN FOR TEXT.
Rapid Acquisition of Emotional Information and Attentional Bias in Anxious Children
Eamon P . Fulcher
University of the West of England
Andrew Mathews
University of California at Davis, USA
Marianne Hammerl
University of Regensburg, Germany
Abstract
This study reports on the relationship between evaluative learning (EL) and attentional
preference in children with varying degrees of anxiety, as measured by the
Multidimensional Anxiety Scale for Children, and varying degrees of parental anxiety, as
measured by scores on the Beck Anxiety Inventory. In the first experiment, 3 age groups
(7-8, 10-11, and adults with mean age 26.8 years) were compared on a novel EL method,
in which neutral images “morphed” over 1 s into either smiling or angry adult faces.
There were no differences in EL between the age groups – each showing a strong EL
effect. In 2 subsequent experiments, we examined learning and attention to stimuli
following EL trials in 7- to 8-year olds. In Experiment 2, Panic/Separation Anxiety (PSA)
and the mothers' BAI predicted the overall magnitude of EL. In addition, high PSA
children were more likely to attend to a neutral stimulus previously paired with a negative
stimulus than were low PSA children. In Experiment 3, only PSA was positively
associated with the magnitude of EL. In the attention trials, high PSA children had longer
fixation times on frowning faces than did low PSA children. These results indicate that
associations between learning, attention, and emotional information are influenced by
separation anxiety and maternal anxiety.
Keywords: Information processing; Perception of threat; Anxiety; Children; Conditioning
1. Introduction
A normal response to a situation that involves some form of extreme threat is to focus
one's attention on the source of the threat and to monitor its development. However,
when a threat is only mild or moderate, individuals with an anxiety disorder and often
people who score highly on an anxiety measure demonstrate preferential processing of
the threat (MacLeod & Mathews, 1988; Mathews & MacLeod, 1994). Cognitive bias,
which is the tendency to preferentially attend to threatening stimuli or to perceive
ambiguous stimuli in threatening ways, has been studied and observed using several
different methods.
Firstly, a threatening item is more likely to capture attention when presented alongside a
neutral or innocuous item (MacLeod, Mathews, & Tata, 1986). Two items are presented
at two different locations on a monitor, and one of them is replaced by a probe or target.
The task is to press a button as soon as the target appears. The logic is that the response
time to the target should be speeded when the target appears in the same location as the
one being attended to, and slowed when the target appears in the other location. The
findings of a number of studies suggest that attentional bias occurs without awareness or
intent (Fox, 1996; MacLeod & Rutherford, 1992).
Secondly, a threatening item can interfere with an ongoing task. In a version of the
Stroop task, emotion words are printed in different colours with the task to name each
colour. Words related to threat (e.g., collapse, failure, and so on) tend to slow anxious
individuals on this task, and this may be because for them threatening material is
unavoidable (Williams, Mathews, & MacLeod, 1996).
Thirdly, a threatening item once noticed can hold attention for longer than a non-
threatening item. When a target is presented immediately after the presentation of an item
in another region of a visual display, anxious individuals are slower to disengage from a
threatening item than from a non-threatening item (Yiend & Mathews, 2001).
Fourthly, a sentence or word that has more than one meaning is more likely to be
interpreted negatively by anxious than non-anxious individuals. For example, the
homophone stroke can have two interpretations: disease or caress . Anxious individuals
are likely to access the more threatening meaning of ambiguous words than non-anxious
individuals (Mathews, Richards, & Eysenck, 1989; Richards & French, 1992). It has been
suggested that the cognitive processes that promote a negative interpretation of
ambiguous material are the same that drive attention towards threat (Mathews &
Mackintosh, 1998).
The association between cognitive bias and anxiety may arise because moderate to high
states of anxiety can promote a particular cognitive style. For example, anxiety levels can
predict the presence of an attentional bias: it is commonly present in clinically anxious
populations but disappears after successful treatment (e.g., Mathews, Mogg, Kentish, &
Eysenck, 1995). Therapy can have effects even on unconscious attentional bias (Lavy,
Van Den Hout, & Arntz, 1993). However, in addition to the possibility that anxiety
induces bias, cognitive bias may be one of many causal agents in the development of
chronic anxiety. If threatening information receives processing priority it may induce
anxiety, and a habitual cognitive style of attending to threat over prolonged periods may
exacerbate, maintain, or initiate anxiety states.
A possible causal hypothesis is that in early development cognitive bias is acquired either
through some form of learning, either as a result of traumatic or mild but repeated fearful
experiences. Bias might also be acquired indirectly by observing others, especially
mothers and other care givers, attending to threat or interpreting innocuous stimuli as
threatening. Another route through which it may be acquired is via verbal transmission of
information about potential dangers and sources of threat. Certain situations, people,
events, and other stimuli may thus become labelled as potentially threatening, even in the
absence of confirmatory experience. The range of stimuli labelled in this way may
increase and the threshold for the detection of threat may become lowered. Furthermore,
with repeated practice selective attention to threat may become automatised, such that it
is done effortlessly, rapidly, without voluntary intention or control. If so, a likely effect is
an increase in the regularity of the experience of fear and anxiety, more intense
experiences of anxiety, the adoption of a number of coping strategies, such as avoidance
of potential sources of threat, and an increase in the range of the behavioural symptoms
of anxiety ( Mathews & MacLeod, 1994) .
Although there is considerable evidence of a relationship between anxiety and cognitive
bias, only a handful of studies can be considered to have addressed the causal hypothesis.
In the absence of a longitudinal study in which anxiety and cognitive bias are measured
in both children and parents from childhood into adulthood, we have only indirect
methods at our disposal. However, there is at least some evidence for several aspects of
the version of the causal hypothesis outlined above.
Firstly, there is convincing evidence that parental fears can be transmitted to their
children. Parental reaction to a traumatic event, for example, can influence the
development of post traumatic stress disorder (PTSD) in their children (Foa, 2000).
Furthermore, Muris, Steerneman, Merckelbach, and Meesters (1996) found that children
of mothers who tend to verbalise their fears regularly had an elevated number of fears of
specific situations and stimuli, and that these fears were related to the mother's fears.
Other research suggests that the responses of a caregiver can influence a child's fear
reaction (Klinnert, Campos, Sorce, Emde, & Sredja, 1983). In a study by Field, Argyris,
and Knowles (2001) it was found that vicarious learning in children, in the form of
negative information given by adults about a fictitious character, can lead to increased
fear beliefs.
Although these studies are not evidence that attentional bias is transmitted from parent to
child, they are consistent with the notion that fears and anxiety can be acquired
vicariously in children. Other studies have shown that vicarious fear conditioning (by
observing others experiencing a mild shock or by being informed about shock
contingencies) can be as effective as Pavlovian conditioning in which shocks are actually
delivered (Olsson & Phelps, 2004) and that a mother's fearful facial-expression can act as
strong unconditioned stimuli for their young children (Gerull & Rapee, 2002). Taken
together, the studies reported here are consistent with one aspect of the causal hypothesis,
which is that anxiety can be transmitted from parent to child through verbalised negative
information.
A second important line of research is to examine cognitive bias in children with anxious
parents. Moradi, Neshat-Doost, Teghavi, Yule, and Dalgleish (1999) examined
attentional bias in children of mothers with PTSD. They found that in comparison to a
control group, children of mothers with PTSD showed greater attentional interference
from threat-related words than from emotionally neutral words. Relating this finding to
the causal hypothesis, the study is consistent with the view that attentional bias is one
consequence of the emotional background in the family and that the bias may predate
manifest anxiety. However, since anxiety levels were not measured in the children, the
study does not rule out the possibility that high levels of anxiety were already present in
the children of mothers with PTSD. In a recent study of our own, we found an association
between attentional bias in children with anxiety in their mothers (Fulcher, Mathews,
Emler, Catherwood, & Hammerl, 2006). This association was independent of the child's
level of anxiety. This could be interpreted as evidence that attentional bias (vigilance) is
present before the child acquires a measurable level of anxiety, which is entirely
consistent with the causal hypothesis.
A third line of research is to test the most important element of the causal hypothesis,
which is that attentional bias promotes a direct increase in anxiety. Mathews and
Mackintosh (2000) and Mathews & MacLeod (2002) demonstrated that interpretations of
emotionally ambiguous words can be biased in either a positive or negative direction by
brief training methods. Furthermore, congruent changes in anxiety do occur when the
induced bias is subsequently deployed in processing mildly stressful events. Hence
emotional processing styles may play a causal role in producing emotional changes.
These studies support the notion of a causal relationship between attentional bias and
anxiety, but clearly more evidence is needed.
In terms of how the perception of threat may spread to new stimuli, we recently identified
a potential mechanism for this, namely evaluative learning. In this form of learning, a
neutral stimulus when paired with a stimulus of positive valence comes to be valued more
positively than before, and a neutral stimulus that is paired with a stimulus of negative
valence comes to be valued more negatively (Baeyens, Eelen, & Van den Berg, 1990;
Fulcher & Cocks, 1997; Hammerl & Grabitz, 1996; Levey & Martin, 1975). It is unclear
whether this change in the affective significance of the stimulus is due to the fact that it
predicts the occurrence of another affective stimulus or whether the stimulus has
somehow itself become imbued with valence (Fulcher, 2001). A large number of studies
have shown that the change in valence brought about through evaluative learning occurs
without awareness of the stimulus contingencies, that is the effect is present when
participants are unable to recall, recognise, or even perceive the affective stimulus that
was paired with a particular neutral stimulus (De Houwer, Thomas, & Baeyens, 2001;
Fulcher & Hammerl, 2001; Hammerl & Fulcher, 2005). In a recent study of ours, stimuli
that acquired negative valence (and hence some threat value) through evaluative learning,
were more likely to capture attention in anxious but not less anxious individuals (Fulcher,
Mathews, Mackintosh, & Law, 2001; Mackintosh and Mathews, 2003). These studies
demonstrate that threat value can be spread onto a previously neutral stimulus by a
contagion it has become associated with. Furthermore, they show that stimuli with
acquired threat value can attract attention in anxious individuals. Evaluative learning may
therefore be an important element of the causal hypothesis since it shows how perceived
threat value can spread and thereby extend the range of stimuli that are able to capture
attention. If so, the ease with which evaluative learning occurs and/or the extent to which
such learning leads to attentional bias effects may be characteristic of those who are
vulnerable to develop anxiety states (Fulcher, et al., 2001).
In the present study, we therefore attempted to assess the ease of evaluative learning and
any consequent attentional bias to newly emotionalised stimuli, in anxious children. We
are unaware of any previous study that had done this. The first aim was to compare the
magnitude of evaluative learning in three age groups to assess any developmental
differences. We then observed attentional preference following evaluative learning to
ascertain whether stimuli imbued with valence would indeed capture attention in children
with anxiety or with anxious parents. Next, we attempted to find out whether anxious
children or those from anxious parents showed any difficulty in disengaging from
negative images and from images with imbued valence.
The anxiety measure used was the Multidimensional Anxiety Scale for Children (MASC;
March, Parker, Sullivan, Stallings, & Conners, 1997). This was chosen because it
measures a number of dimensions, such as physical symptoms, harm avoidance, social
anxiety, and panic/separation anxiety (SA). The latter may be particular important since a
number of theories have postulated a key role for separation anxiety in adult
psychopathology. For example, SA in children has been thought to (1) predispose to a
variety of anxiety disorders (Tyrer, 1985), (2) be a precursor to dependent personality
disorder in adults, and (3) be a predictor of an adult form of SA (Silove &
Manicavasagar, 2001).
The causal hypothesis predicts an association between parental (and especially maternal)
anxiety and attentional bias. It also predicts that this association will exist in the absence
of child anxiety, although its presence does not falsify the hypothesis since anxiety may
already be present. The causal hypothesis also predicts that child anxiety will be
associated with an evaluative learning bias, manifested as a rapid learning of negative
associations (since negative emotional stimuli may be perceived as more salient than
positive emotional stimuli in anxious children).
2. EXPERIMENT 1
The evaluative learning paradigm has been almost exclusively used on adult samples and
these authors are aware of only a single study, Field (2006), that has adapted the method
for children. In that study, neutral cartoon characters were paired with liked food (e.g.,
ice-cream) and disliked food (e.g., Brussel sprouts). Characters paired with liked food
became more liked than characters paired with disliked food. In this experiment we
attempted to compare different age groups on an evaluative learning task. We also wished
to pilot a new method for presenting the pictures, which involves morphing a neutral
image into an emotional image. A “morph” is an animation that transforms one picture to
another and in a smooth manner. The question is whether a neutral picture that is
transformed into a smiling face will be more liked than a neutral picture that is
transformed into a frowning face.
2.1 Method
2.1.1 Participants
Participants were 15 children in Year 3 Primary of ages 7-8 years, 15 children in Year 5
primary of ages 10-11, and 15 adults (mean age 26.8 years) recruited from a university
campus. The children were recruited from a local urban school in the UK . Recruitment
was via a random sampling procedure, where each child had an equal chance of being
selected once they had opted in. Signed permission was obtained from the caregivers as
well as from the participating children, after a brief outline of the procedure had been
given. The experimental procedure was approved by the Ethics Committee at Worcester
University and The Worcestershire NHS Ethics Committee.
2.1.2 Materials
The stimuli consisted of 8 separate morphs, which were developed as MPEG files,
created from 16 pictures (8 neutral pictures, 4 of smiling faces and 4 of frowning faces).
Each morph consisted of a Japanese ideograph (representing a neutral picture)
transformed over 1 s into either a smiling or frowning human face. For the purposes of
counter balancing, two morphs were created for each ideograph: one that transformed
into a smiling face and one that transformed into a frowning face. Two sets of morphs
were created, with each picture appearing only once in each set. Thus, half of the neutral
stimuli morphed into one of the smiling faces in set 1 and into one of the frowning faces
in set 2. Similarly, the other half of neutral stimuli morphed into one of the frowning
faces in set 1 but morphed into one of the smiling faces in set 2. Participants were then
divided into two groups on a random basis with half being shown set 1 and the other half
shown set 2.
The eight ideographs were selected from a set that was hand drawn by a Japanese-
speaking colleague. The affective pictures consisted of eight photographs of faces taken
from four adults (each model was photographed twice: once when asked to pose with an
angry expression and once when asked to pose with a smile). These photographs were
used in a previous unpublished study and were selected from a large sample on the basis
of the threat rating of the angry expression and the pleasantness rating of the smiling
expression by a panel of raters. Those pairs that provided the highest ratings for both
expressions were selected. The morphing transformations were created by identifying a
set of 12 dots on each ideograph and 12 corresponding dots on the target image. The
purpose of the dots is to match up key features in each picture and to instruct the program
how the first picture should morph into the second picture, so that the transformation is
smooth. The morphing program then rendered the images to create a movie file that was
subsequently converted into an MPEG file. Each movie file was presented on a computer
monitor controlled by a script written for the Inquisit software. Pictures measured 12 cm
x 9 cm . Each participant was exposed to eight morphs over two trials.
2.1.3 Procedure
The experiment consisted of three phases: pre-trial rating, learning trials, post-trial rating,
and recall. In the pre- and post-trial rating phases, which were identical, the (static)
ideographs were presented separately and the task was to rate each one on a scale ranging
from 0 to 100 with intervals of 10 (11 anchor points in all). The participants were told
that the more they liked the picture, the higher the number they should assign, and the
less they liked a picture the lower the number they should assign. In the learning trials,
the order of the morphs was randomised. A trial began with a fixation point (a yellow
cross in the centre of the screen) for 3 s that was replaced by a blank screen for 2 s. The
morph then appeared for 1 s and there was an 8 s pause between trials. Thus, each trial
lasted 14 s. The morphs were shown twice (in two blocks) and hence there were 16 trials
in all. This is quite an unusually low number of trials for an evaluative learning
experiment (8 pairs each shown twice), and we did this for two reasons: (1) The planned
Experiments 2 and 3 involved additional phases and we were limited in how long we
could test each child, and (2) the morphs were quite dramatic (in the sense that the
movement enhanced the emotional tone of the appearing face); and hence we suspected
that evaluative learning would occur rapidly.
In the recall phase, the ideographs were presented individually and the participant had to
state whether the ideograph morphed into a happy/smiling face or an angry/frowning
face. Including time for providing instructions, testing lasted about 15 minutes for each
participant.
2.2 Results
For each participant, the mean rating of the ideographs was calculated. This was done
separately for the pre-trial and the post-trial rating phases and according to the valence of
the paired image. The data was then entered into a Valence (paired with smiling or
frowning) x Time (pre- or post-learning trial) x Age group (7-8, 10-11, or adult) mixed
ANOVA. The predicted interaction of valence and time was statistically significant, F (1,
42) = 9.849, p < .005. There was no interaction between valence, time, and age group, F
(2, 42) = 2.296, p > .05. The means were as predicted for the participants as a whole and
indicate that ideographs paired with smiling faces showed an increase in liking after the
learning phase (M pre-learning = 43.7 and M post-learning = 47.5), whereas ideographs
paired with frowning faces showed a decrease in liking (M pre-learning = 45.6 and M
post-learning = 42.5). There were no other main effects or interactions.
In terms of the ability to recall the ideograph-face pairs the mean recall for each age
group was calculated and entered into an age group one-way ANOVA, which was not
statistically significant. Each age group recalled about 50% of the pairs. In order to
examine the role of contingency awareness further, a correlation was carried out between
an overall evaluative learning index and recall score for each participant. The Evaluative
Learning Index (ELI) was calculated by:
Evaluative Learning Index = (V + 2 – V + 1 ) – (V - 2 – V - 1 )
where V refers to the mean rating of an ideograph, + or – refers to whether the ideograph
morphed into smiling or frowning face, respectively; 1 refers to pre-trial rating, and 2 to
post-trial rating. The higher the ELI, the stronger the learning effect. An ELI of 0 implies
no learning effect at all, and a negative ELI implies an effect orthogonal to that predicted.
The overall ELI was calculated to be +6.9. The correlation between ELI and recall score
was not statistically significant, r (45) = –.067, p > .05. Further, recall of positive
associations did not correlate with a positive ELI (calculated by V + 2 – V + 1 ; and
calculated to be +3.8 overall), r (45) = 0.028, p > .05, and recall of negative associations
did not correlate with a negative ELI (calculated by V - 2 – V - 1 ; and calculated to be
+3.1 overall), r (45) = .061, p > .05. A further analysis consisted of splitting participants
into two groups by a median split of their recall scores and then comparing the overall
ELI of the two groups. This analysis revealed no statistically significant difference
between the two recall groups, t (43) = 0.288, p > .05.
2.3 Discussion
The most important finding is that the morphing of a neutral picture into a valenced one
is a valid and novel method to induce evaluative learning. What is remarkable about this
method is that only two presentations of each morph were sufficient for learning to occur
and this may be due to the high potency of the morphed images. The second finding is
that all age groups show evaluative learning effects. This is an important finding for the
subsequent experiments. A third finding is that evaluative learning appears to be
independent of an ability to recall the pairs, and this adds to the already weighty body of
evidence that evaluative learning is not dependent on contingency awareness.
3. EXPERIMENT 2
One aim of the next experiment was to attempt to test the prediction that anxious children
and/or children with anxious parents would show a bias in evaluative learning. In other
words, we asked whether (1) the mother's or father's anxiety would predict the extent of
evaluative learning in their children, (2) whether any of the measures of child anxiety
would predict the strength of evaluative learning, or (3) whether the child's and parent's
anxiety levels would interact with the strength of evaluative learning. A second aim was
to examine the role of attention to previously neutral stimuli, imbued with valence
following evaluative learning, and to determine whether vulnerable children show an
attentional bias not only to the unpleasant images, but also to the images imbued with
valence. The same morphing procedure was used as in Experiment 1 (with the same
stimuli) and the ideographs were also evaluated before and after learning trials. One
difference in the evaluative learning procedure between this experiment and the previous
one was that we noticed that some of the lower aged children had difficulty in using the
rating scale and required regular reminding of how to use the scale. For this reason, we
employed a ranking technique used previously in Field, Argyris, and Knowles, (2001), so
participants had merely to express preference rather than identify some point on a scale.
3.2 Method
3.2.1 Participants
Participants were 44 children (22 boys and 22 girls) in Year 3 Primary of ages 7-8 years
recruited from 10 local schools, including 1 independent school (urban), 4 urban state
schools, and 5 rural state schools. Signed permission to participate was obtained from the
main caregiver as well as from the child. A brief outline of the main task and the type of
materials that would be shown was presented before requesting permission. The
experimental procedure was approved by the Ethics Committee at University of
Worcester and The Worcestershire NHS Ethics Committee. Recruitment from each
school was via a random sampling procedure, where each child who had opted in had an
equal chance of being selected (with the constraint that we required equal numbers of
boys and girls). All children showed willingness to participate. Donations were made to
each participating school, and the amount was based on the number of children tested.
Each child was given a small toy for their participation.
Materials
3.2.2 Beck Anxiety Inventory (BAI). The BAI (Beck, Epstein, Brown, & Steer, 1988)
questionnaire was distributed to all parents of selected children at participating schools.
Parents or guardians (both male and female carers where possible) responded to 21 items.
Each item is descriptive of subjective, somatic or pain related symptoms of anxiety. Each
question asks how much each of the symptoms had bothered them in the past week.
3.2.3 Multidimensional Anxiety Scale for Children (MASC). This scale has been
developed for and tested on children between the ages 8 to 16 (March, et al. 1997). It is a
39-item questionnaire and factor analysis reveals that it has four sub-scales that measure
Physical Symptoms ( tension, restlessness, and somatic/autonomic symptoms) , Harm
Avoidance ( perfectionism and anxious coping) , Social Anxiety ( humiliation/rejection
and public performance fears) , and Panic/Separation Anxiety (PSA), and also provides a
total anxiety score (sum of the sub-scales). It has been shown that girls score significantly
higher on all factors than boys; that parent-child agreement is moderate, but highest for
easily observable symptoms (March et al., 1997), and that parent-teacher agreement is
good (Fulcher et al., 2006). This questionnaire was completed by parents of the children
selected to participate.
3.2.4 Stimuli. In this experiment the same stimuli as in Experiment 1 were used.
3.2.5 Eye-tracker. An ASL eye-tracker system was used in this experiment, consisting of
an infra-red camera, which was mounted next to the TFT monitor. The system was
controlled by “e-5Win” software, which manipulated the camera to capture the
participant's pupil and corneal reflex. A monitoring screen displayed the participant's eye.
The camera operates remotely from the participant and no headgear is required. The
software records the x-y coordinates of the eye fixation once every millisecond and stores
this in a fixation file. Additional software written by the first author controlled the
presentation of all stimuli and this software also signalled unique codes to the fixation file
for all stimulus onset and offset times, making it possible to marry eye fixations with the
particular stimulus being presented. Further software, also written by the first author, was
used to compute first fixations on each of the trials. This software was tested and shown
to be accurate when compared with a sample of computations carried out manually. The
system was configured for each participant at the start of the experiment, although only
eye movements during the attentional trials were recorded.
3.2.6 Procedure
The same procedure was used as in Experiment 1, with the following exceptions: (1) a
preference procedure was used to measure liking/disliking, (2) there was no test of
contingency awareness, and (3) there was an additional final phase consisting of a
number of attentional trials.
3.2.7 Ranking Phase. The eight neutral images were presented in their static forms and in
pairs (and in random order for each participant), and the participant merely had to say
which of the two images they preferred or neither image by pressing a corresponding key.
Every combination of pairs was presented (28 trials in all). The computer logged a 1-
point rank value for each picture that was preferred. The maximum rank value for a
picture could therefore be 7 and the minimum 0. The ranking phase occurred before and
after learning trials.
3.2.8 Attentional Trials Phase. In this phase, two pictures were presented at the same
time, one on the left of the monitor and one on the right. The pairs consisted of either one
smiling and one frowning face, or one ideograph previously paired with a smiling face
and one ideograph previously paired with a frowning face. All pictures were those used
in the learning phase, but this time pictures were shown in their static form. Participants
were required to fixate on a central cross on the screen for 1000 ms before the pictures
appeared. The pictures remained visible until a key was pressed or for 2000 ms if no key
was pressed. Before this and 500 ms after the pictures appeared, a small image of
ladybird appeared on one of the pictures and the participant's task was to press the a key
if the target appeared on the left picture, and the 5 key on the numeric keypad of the
keyboard if the target appeared on the right picture. There was a 1500 ms inter-trial
interval before the fixation point appeared. Since there were four different smiling faces
and four different frowning faces, we paired up every combination, including reversing
their left/right positions, which is a total of 12 trials. We did the same for the pairs of
ideographs, which added another 12 trials. Each pair was then shown twice, making a
total of 48 trials. Reaction times to respond to the target were recorded, as well as the eye
fixations throughout the attention trials.
3.3 Results
3.3.1 BAI and MASC
Overall, 43 BAI forms were returned by mothers and 32 by fathers, and 41 MASCs were
completed. The mean scores of the BAI showed that mothers had significantly higher
scores than fathers, t (30) = 2.813, p < .01. The BAI of mothers correlated with many
other measures, such as the BAI of the father, r (32) = .388, p < .05; the MASC total
score, r (41) = .335, p < .05; and Physical Symptoms, r (41) = .546, p < .001. Apart from
correlating with the mother's BAI score, the father's BAI scores correlated only with
Physical Symptoms, r (32) = .46, p < .001.
3.3.2 Evaluative Learning
The mean rank values for pictures paired with smiling faces and for those paired with
frowning faces were computed, and for both pre-learning and post-learning phases. A
repeated measures ANOVA of time (pre-rank or post-rank scores) and valence
(ideographs paired with smiling faces or ideographs paired with frowning faces), revealed
an interaction between valence and time, F (1, 43) = 4.62, p < .05, such that ideographs
paired with smiling faces showed an increase in liking after learning (M pre-learning =
3.67 and M post-learning = 4.18), and ideographs paired with frowning faces showed
decrease in liking after learning (M pre-learning = 2.99 and M post-learning = 2.53).
We calculated an Evaluative Learning Index (ELI) in the same way we did in Experiment
1 in order to examine any relationship with the anxiety measures. We conducted a
stepwise multiple-regression analysis on the ELI with parental BAI and MASC subscales
as the main predictor variables. There was a highly statistically significant association
between Panic/Separation Anxiety and BAI of the mother on the ELI [Multiple R =
0.359; F (2, 28) = 7.841, p < .002. Together, Panic/Separation Anxiety and Mother's BAI
accounted for 35.9% of the variance. Both variables were positively associated with the
ELI. The regression coefficient beta for Panic/Separation Anxiety was 0.131 ( t = 2.832,
p < .01); and for Mother's BAI beta was 0.125 ( t = 2.078, p = .05). In sum,
Panic/Separation Anxiety and Mother's BAI predict the overall magnitude of evaluative
learning. In order to determine whether more vulnerable children showed a bias to learn
the negative associations better than the positive associations, we conducted separate
regression analyses for a positive ELI and a negative ELI. These revealed no significant
associations.
To be sure that vulnerable children were not biased to learn negative associations more
quickly than positive associations, we split participants into two groups based on a
median split of their Panic/Separation Anxiety (PSA) scores, split mothers according to a
median split of their score on the BAI, and computed a multivariate ANOVA of PSA
group x Mother's Anxiety group on the three dependent variables: overall ELI, positive
ELI, and negative ELI. This analysis revealed a main effect of PSA group on negative
ELI, F (1, 36) = 5.789, p < .025, with the low PSA group having a mean negative ELI of
–0.636, and the high group with a mean of 0.895 (the higher the value of the negative
ELI, the greater is the amount of evaluative learning). There were no other main effects
or interactions.
Reaction Time to Human Faces
The mean reaction times were calculated for each participant according to the type of
trial. A mixed ANOVA on reaction time to emotional faces, with valence of the face as
the within-subjects variable and PSA group and Mother's BAI group as the between-
subjects variables revealed no main effects or interactions. Thus, there was no
preferential looking for human smiling or frowning faces. Since using the individual
measures dichotomously does not exploit the full range of scores, we carried out a paired
t -test on valence alone. This did not reveal a significant difference in reaction time. We
next computed a reaction time index (RTI) by: reaction time when the target was in the
location of a smiling face minus the reaction time when the target was in the location of a
frowning face. A positive RTI implies that the participants were mostly viewing the
frowning faces by the time the targets appeared, and a negative RTI implies that the
participants were mostly viewing the smiling faces by the time the target appeared. We
then carried out a stepwise multiple regression on RTI with Mother's BAI and PSA as the
first block of predictor variables, and the remaining anxiety measures as the second block
of predictor variables. No variable entered at .05.
3.3.3 Reaction Time to Ideographs Imbued with Valence
A mixed ANOVA on reaction time to the ideographs, with valence of the associated
facial expression as the within-subjects variable and PSA group and Mother's BAI group
as the between-subject variables, revealed a highly significant interaction between PSA
and valence, F (1, 36) = 8.407, p < .01. The high PSA group were quicker to respond to
the target when it appeared in the location of an ideograph paired with a frowning face
(M = 751 ms) than when it appeared in the location of an ideograph with a smiling face
(M = 882 ms). The opposite effect occurred with the low PSA group who were quicker to
respond to the target when it appeared in the location of an ideograph paired with a
smiling face (M = 721 ms) than when it appeared in the location of an ideograph paired
with a frowning face (M = 825 ms). To corroborate this finding, we computed a reaction
time index (RTI): reaction time when the target was in the location of an ideograph
paired with a smiling face less the reaction time when the target was in the location of an
ideograph paired with a frowning face. A positive RTI implies that the participants were
mostly viewing the frowning faces by the time the targets appeared, and a negative RTI
implies that the participants were mostly viewing the smiling faces by the time the target
appeared. We then conducted a stepwise linear regression with Mother's BAI and the
child's PSA as the main predictor variables, and the other anxiety measures as secondary
predictor variables. There was a highly statistically significant positive association
between PSA on the RTI, Multiple R = 0.402; F (1, 38) = 7.309, p < .01. PSA accounted
for 40% of the variance and was negatively associated with the reaction time index. The
regression coefficient beta for PSA was 0.402 ( t = –2.704, p < .01). Thus, PSA predicts
reaction time and shows that as PSA increases, there is an increased attentional
preference for ideographs that were previously associated with frowning faces.
Eye Movements
We recorded the first eye fixations on each trial and calculated the mean fixations for
each participant based on the trial type. For trials with pictures of human faces, we then
conducted an ANOVA on mean number of first fixations with valence as a within-
subjects measure and PSA group and Mother's BAI group as between-subjects factors.
This time we found a main effect of valence, F (1, 36) = 6,489, p < .025, such that
participants tended to fixate first on frowning human faces ( M = 1.09) rather than on
smiling faces ( M = 0.69). Therefore, unlike the reaction time data, the eye movement
data implies that on the whole participants tended to fixate on the frowning faces earlier
than the smiling faces. To determine whether any individual measures of anxiety
predicted orientation, we computed a fixation index (number of first fixations to frowning
faces less the number of first fixations to the smiling faces) and ran a regression analysis.
No variable entered at .05. Thus, in line with the reaction time data, there does not appear
to be any preferential attending to emotional faces that is predicted by any measure of
anxiety.
Next we conducted an ANOVA on mean number of first fixations on the ideographs,
with valence of the ideograph's pair as a within-subjects measure and PSA group and
Mother's BAI group as between-subject variables. There was a significant interaction
between PSA group and valence, F (1, 36) = 6.318, p < .025, such that the high PSA
group tended to fixate first on ideographs paired with frowning faces (M frowning = 3.64,
M smiling = 3.0), and the low group tended to fixate first on ideographs paired with
smiling faces (M smiling = 3.73, M frowning = 2.53). There were no other main effects
or interactions. We conducted a regression analysis on a fixation index (computed by
number of first fixations on ideographs paired with frowning faces less the number of
first fixations on ideographs paired with smiling faces) with the anxiety measures as
predictor variables, but no measure entered at .05.
4. Discussion
The results obtained can be summarised in the following way: The participants as a
whole demonstrated evaluative learning, however, Panic/Separation Anxiety and the
mother's anxiety predicted the overall magnitude of evaluative learning. Furthermore,
children high on Panic/Separation Anxiety showed stronger evaluative learning with
frowning faces than did children low on this measure. In terms of attention, there was a
tendency for the group as a whole to attend to frowning faces before smiling faces;
however, Panic/Separation Anxiety and mother's anxiety predicted attentional vigilance
for frowning faces, and children high on Panic/Separation Anxiety tended to fixate more
on ideographs that predicted frowning faces than did children low on this measure. This
experiment has therefore revealed important differences between children high and low
on PSA in learning and attention. We have therefore identified learning and attentional
biases in children prone to separation anxiety. Given the importance of this result we
carried out a replication with a number of modifications.
5. EXPERIMENT 3
In this experiment, we attempted to examine whether stimuli imbued with valence would
acquire the capacity to hold greater attention in children more vulnerable to anxiety than
with less vulnerable children. The evaluative learning trials differed from the previous
experiment because we wished to exploit the Yiend and Mathews (2001) paradigm,
which is a location based task. The idea is to present a target image in a different location
from the emotional stimulus and then measure the time it takes to respond to the target.
Slowed responses imply slower disengagement from the emotional stimulus. Hence the
method warranted evaluative learning trials with the neutral and affective stimuli
appearing at different locations. For this reason, it was not feasible to use the morphing
procedure of the previous two experiments. Instead, we used the standard paradigm used
in visual evaluative learning, which is to pair static images. Although we were unsure as
to whether we would observe evaluative learning under these conditions, the method did
allow us to compare evaluative learning when the affective stimulus and the neutral
stimulus are presented at the same or different locations, which is something that has not
been examined in previous studies. Using this method, we were able to assess the amount
of disengagement from affective stimuli as well as stimuli that had been previously
associated with other affective stimuli between children with various anxiety-related
dispositions.
5.1 Method
5.1.1 Participants
Participants were 34 children (19 boys and 15 girls) in the 7-8 year range. The children
were sampled from six local schools in the UK (four urban and two rural). As with the
previous experiments, signed permission to participate was obtained from the main
caregiver as well as from the child. A brief outline of the main task and the type of
materials that would be shown was presented before requesting permission. The
experimental procedure was approved by the Ethics Committee at University of
Worcester and The Worcestershire NHS Ethics Committee. Recruitment from each
school was via a random sampling procedure, where each child who had opted in had an
equal chance of being selected (with the constraint that we required equal numbers of
boys and girls). All children showed willingness to participate. Donations were made to
each participating school, and the amount was based on the number of children tested.
Each child was given a small toy for their participation.
5.1.2 Materials
Twelve abstract images created with imaging and drawing software were used for the
neutral stimuli. These consisted of different shapes, colours, and textures. Two
constraints governed their creation: They should have an abstract appearance, and they
should be highly distinctive from each other. Eight of the 12 neutral images were used as
neutral stimuli for evaluative learning trials, and four were used as novel stimuli in the
attention trials. Four photographs of smiling faces and four photographs of angry faces
were used, and these were the same as those in the previous experiments. BAI and
MASC measures were also taken as in the previous experiment. We also used the same
eye tracking equipment as in Experiment 2. However, in 80% of trials, the participant
made a response without making a fixation on (or even towards) the target. Since so
many made use of parafoveal vision, we collected too little data to be able to conduct any
useful analysis on eye movements. One possible explanation for this might be that the
stimuli used were less visually demanding or interesting than those used in the previous
experiment where the images were constantly moving. The results section is based only
on ranking and reaction time data, therefore.
5.1.3 Procedure
There were four phases in this experiment: pre-ranking, learning, attention trials, and
post-ranking.
5.1.4 Pre- and Post-Ranking. The ranking procedure used for the pre- and post-ranking
phases was the same as that used in Experiment 2.
5.1.5 Learning Phase. Each of the eight neutral pictures was consistently paired with one
of the pictures of a human face. Furthermore, this pairing either occurred in the same
location or at a different location on the monitor. On each trial, the neutral picture
appeared on the left side of the monitor, and the face appeared either on the left side or
the right side of the monitor, replacing the first picture. In the previous experiments, we
obtained evaluative learning in only two trials using the morphing procedure, and we
expected this owing to the visual potency of the images. However, with static images we
suspected that more trials would be required to achieve evaluative learning (based on
previous reported experiments) and hence there were 4 learning trials for each stimulus
pair. These were separated four learning blocks so that the same pair did not appear
consecutively. Pairings were counterbalanced across participants in the same way as the
previous experiment. The location of the picture of the face was also counterbalanced
across participants.
5.1.6 Attention Trials . The attentional trials began with the presentation of one of the
images for 2 s, after which a small arrow appeared that pointed up or down in either the
same location or on the right of the picture. When the arrow appeared at the same
location, the picture was removed. When the arrow appeared at the different location, the
picture remained on for a further 2 s or until a key was pressed. There was a 1500 ms
inter-trial interval. The participant's task was to detect whether the arrow was pointing up
or down by pressing a corresponding key on the keyboard. On valid trials, the probe
appears at the same location as the picture, and hence responses should be quick
regardless of valence. On invalid trials, the probe appeared at a different location, and
hence the time taken to respond to the probe is indicative of the time to disengage form
the picture. Long response times to detect the probe implies greater attentional
engagement (and hence slowed disengagement) to the picture.
5.2 Results
5.2.1 BAI and MASC
Overall, 33 BAI forms were returned by mothers and 26 by fathers, and 34 MASCs were
completed. The BAI of the mother correlated with three of the other measures: with
Physical Symptoms, r (33) = .548, p < .001; with Social Anxiety, r (33) = .361, p < .05;
and with Panic/Separation Anxiety, r (33) = .388, p < .05. The father's BAI correlated
only with Physical symptoms, r (26) = .423, p < .05. All MASC subscales correlated with
each other at .05, except Harm Avoidance, which did not correlate with any other
measure.
5.2.2 Evaluative Learning
We computed the overall rank values for each neutral picture, noting the valence of the
face it was paired with, whether the paired face appeared in the same or different
location, and whether the ranking was done pre- or post-learning. We computed a
repeated measures ANOVA of location of the face (same or different), valence of the
face (smiling or frowning), and time (pre- or post-learning). There was a highly
significant interaction between valence and time, F (1, 33) = 53.82, p < .001. Ideographs
paired with smiling faces showed an increase in liking after learning (M pre-learning =
2.97 and M post-learning = 4.97), and ideographs paired with frowning faces showed
decrease in liking after learning (M pre-learning = 3.72 and M post-learning = 1.95).
There were no other main effects or interactions.
We computed correlations between PSA and the ELI (computed in the same way as in
the previous experiments), as well as between BAI of the Mother with the ELI, to
determine whether those relationships found in Experiment 2 were present with this
sample. PSA correlated significantly and positively with the ELI, r (34) = .356, p < .05,
and while BAI scores of mothers was positively correlated with ELI, it was not
statistically significant, r (33) = .187, p > .05. Moreover, a regression analysis on ELI
with the anxiety measures as predictor variables failed to reveal a significant predictor
variable at .05. Likewise, no predictor variables entered at .05 on positive ELI or negative
ELI.
5.2.3 Reaction Time to Emotional Faces
We calculated a Fixation Index, FI, by reaction time to respond to the target when a
frowning face was present less reaction time to respond to the target when a smiling face
was present. A positive FI implies greater fixation time on frowning faces than on
smiling faces, and a negative FI implies the reverse. We computed a univariate ANOVA
on FI with PSA group and Mother's BAI group as independent variables. The only
significant effect was a main effect of PSA group, F (1, 30) = 11.399, p < .002. The low
PSA group had a mean FI of –264.9, implying longer fixations on smiling faces than on
frowning faces, while the high PSA group had a mean FI of 133.2, implying longer
fixations on frowning faces than on smiling faces. This was confirmed by a multiple
regression analysis on FI, with Mother's BAI and PSA as primary predictor variables and
the remaining anxiety measures as secondary predictor variables, with only PSA
positively associated with FI, F (1, 23) = 9.414, p < .005; beta 20.3, and PSA accounted
for 26% of the variance. Furthermore, PSA correlated highly significantly and positively
with the FI, r (34) = .567, p < .001, and although the BAI of the mother was positively
correlated with FI, it was not statistically significant, r (33) = .176, p > .05. Thus, in these
analyses, higher PSA scores are associated with longer fixation times on frowning faces.
5.2.4 Reaction Time to Ideographs Imbued With Valence
Next, we computed the FI for invalid trials with ideographs in the same way as above,
and carried out a univariate ANOVA on FI with PSA group and Mother's BAI group as
independent variables. The only near-to-significant effect was that of PSA group, F (1,
30) = 3.599, p = .067. To test this further, we carried out a similar multiple regression
analysis as above, but no variable entered at .05. This was confirmed by correlational
analyses, which showed that there was no significant correlation between FI and PSA, r
(34) = –.138, p > .05, nor was there a significant correlation between BAI of the mother
and FI, r (33) = –.108, p > .05. In order to determine whether the participants as a whole
showed any preferential attention before responding to the target, we conducted a one-
sample t -test on the FI, but it did not differ significantly from 0, t (33) = –1.204, p > .05.
Thus, there was no consequential effect of evaluative learning on attention in this
experiment.
5.3 Discussion
Similar to the findings of Experiment 2, a strong evaluative learning effect was
demonstrated by the sample as a whole but with Panic/Separation Anxiety associated
with the magnitude of evaluative learning, although the mother's anxiety did not interact
significantly in this experiment. Panic/Separation Anxiety was also associated with
longer fixation times on frowning faces, and children high on this measure had a slower
disengagement from them than children low on the measure. Unlike Experiment 2, where
there were effects on attention in anxiety of stimuli imbued with valence, there were no
such effects here.
6. GENERAL DISCUSSION
All three experiments revealed that children as young as 7-8 years can learn to like and
dislike through evaluative learning, and Experiment 1 shows that children appear to learn
associations without any awareness of the stimulus contingencies. This is consistent with
the many previous findings and supports the contention that evaluative learning is an
implicit form of learning. Experiments 2 showed that although most of the children
changed their preferences as a result of being exposed to the associations, there is an
evaluative learning bias in anxious children (high on Panic/Separation Anxiety, PSA),
which interacted with the mother's anxiety. Furthermore, it appears that this bias takes the
form of inflated learning of negative associations in high PSA children. These results are
consistent with the causal hypothesis because they show firstly that the mother's anxiety
level can predict the strength of evaluative learning and secondly that anxious children
have a evaluative learning bias for negative information. Experiments 2 also
demonstrated that attentional bias is associated with the child's anxiety (PSA). This
finding mirrors that found in adults, but places the focus on separation anxiety. The third
experiment showed that children with high PSA were slow to disengage from threatening
material, a finding that also mirrors the effect found in adults, but again places the focus
on separation anxiety.
Of all the subscales of the MASC, only separation anxiety was associated with the
dependent measures. Separation anxiety was associated with the strength of evaluative
learning, attentional bias, and disengagement failure. These findings support those
perspectives that postulate a major role for separation anxiety in the development of adult
anxiety. One way of testing this further, other than through a longitudinal study, is to use
a retrospective measure of separation anxiety on adults (e.g., the Separation Anxiety
Symptom Inventory, Silove et al., 1993) to determine whether those who score higher on
this measure show the same learning and attention biases as did the high PSA children in
this study.
What may be of particular interest in the development of anxiety, is the finding in
Experiment 2 that greater attention is paid to the predictors of threat (stimuli imbued with
valence) by children high on PSA, while all children tended to direct attention to the
threat itself (angry faces). Thus, this could imply that children with separation anxiety,
borrowing a term from the classical conditioning literature, show an enhanced
anticipatory effect in learning. One explanation of this effect is that the inflated
evaluative learning for negative associations with high PSA promotes attentional bias to
those stimuli. However, equally possible is that negative stimuli attract greater attentional
resources and this leads to increased evaluative learning for those stimuli. Further
research could determine whether evaluative learning bias promotes an attentional bias or
the other way around by comparing fixation times on neutral stimuli in negative
associations and neutral stimuli in positive associations.
Experiment 3 shows that high PSA children show slowed disengagement from with
threatening pictures, and again this mirrors the effect found in adult anxiety. There was
no slowed disengagement to stimuli acquiring negative valence when compared with
stimuli acquiring positive valence. Combining this result with the attentional bias towards
stimuli imbued with negative valence, stimuli with imbued valence may have less
potency than an overtly threatening stimulus; so while they capture attention, they do not
hold it. An alternative explanation is that a stimulus imbued with negative valence (via
evaluative learning) is also one that predicts the occurrence of a negative stimulus. The
result is that they may capture less sustained attention than an overt threat. One way of
examining this issue further would be to record eye movements during learning and to
interleave learning and attentional trials. This way, it may be possible to construct a more
detailed picture of the effects of evaluative learning on attention as imbued valence is
developed over time. This may uncover any relationship between rates of learning and
vulnerability to anxiety that may exist.
Our study implies a cognitive style in anxious (high PSA) children that appears to take
the following pattern: (1) They learn rapidly about negative information, (2) they
selectively attend to predictors of threat, and (3) they are slow to disengage from
threatening material. Evaluative learning is therefore an important learning mechanism
through which an attentional bias for threat may be spread or transferred to other
previously innocuous stimuli. Taken together, the results from the experiments provide
support for a learning mechanism in the development of attentional bias. The study also
demonstrates associations between maternal anxiety, child separation anxiety, evaluative
learning, and attentional bias in children. A longitudinal study on the present sample may
help to determine which of these come to be good predictors of adult anxiety.
As mentioned in the introduction, many studies demonstrate that evaluative learning can
operate without awareness of the stimulus contingencies. Our study shows that such
learning in children may also operate outside of awareness. In the case of an evaluative
learning bias, this implies that it may be difficult to repair. The question also arises as to
whether there are differences in contingency awareness between high and low anxious
individuals. It may be, for example, that anxiety and contingency awareness are
positively associated. Furthermore, previous studies show that when participants are
informed of the relationship between neutral stimuli and paired affective stimuli, learning
is inhibited (Fulcher & Hammerl, 2001). One question is whether such inhibition applies
equally to high and low anxious individuals.
Finally, Mathews and colleagues (e.g., Yiend & Mathews, 2002) have addressed the
question of whether cognitive biases have a causal role in anxiety by attempting to induce
cognitive bias in non-anxious individuals. Since such cognitive training increases anxiety,
it is reasonable to suppose that cognitive biases are a contributory factor in the
development of anxiety disorders. The question the present study raises is whether a
similar induced evaluative learning bias can promote attentional bias and levels of
anxiety. If so, then there exists the prospect of using evaluative learning techniques in
therapy.
In conclusion, we have identified that children with a degree of separation anxiety show a
learning bias that mainly concerns the acquisition of negative emotional information and
an attentional preference for learned signals of threat. Therefore, in order to understand
the development and maintenance of anxiety further, it seems important to understand the
role of evaluative learning in anxiety.
7. REFERENCES
Baeyens, F., Eelen, P., & Van den Bergh, O. (1990). Contingency awareness in
evaluative conditioning: A case for unaware affective-evaluative learning. Cognition and
Emotion, 4, 2-18.
Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for measuring
clinical anxiety: Psychometric properties. Journal of Consulting and Clinical Psychology,
56 , 893-897.
De Houwer, J., Thomas, S., & Baeyens, F. (2001). Associative learning of likes and
dislikes: A review of 25 years of research on human evaluative conditioning.
Psychological Bulletin, 127 , 853-869.
Field, A., Argyris, N. G., & Knowles, K. A. (2001). Who's afraid of the big bad wolf: a
prospective paradigm to test Rachman's indirect pathways in children. Behaviour
Research and Therapy, 39, 1259–1276.
Fox, E. (1996). Selective processing of threatening words in anxiety: The role of
awareness. Cognition and Emotion, 10 , 449-480.
Fulcher, E. P. (2001). Neurons with attitude: A connectionist account of human
evaluative learning. In S. Moore & M. Oaksford (Eds.), Emotional cognition: From brain
to behaviour (pp. 75-109) . Amsterdam : John Benjamins.
Fulcher, E. P., & Cocks, R. (1997). Dissociative storage systems in human evaluative
conditioning. Behaviour Research and Therapy, 35, 1-10.
Fulcher, E. P., & Hammerl, M. (2001). When all is revealed: A dissociation between
evaluative learning and contingency awareness. Consciousness and Cognition, 10 , 524-
549.
Fulcher, E. P., Mathews, A., Emler, A., Catherwood, D., & Hammerl, M. (Under review).
Parental anxiety, child anxiety, and attention to threat. Manuscript submitted for
publication.
Fulcher, E. P., Mathews, A., Mackintosh, B., & Law, S. (2001). Evaluative learning and
the allocation of attention in anxiety. Cognitive Therapy and Research, 25, 261-280.
Hammerl, M., & Fulcher, E. P. (2005). Reactance in affective-evaluative learning:
Outside of conscious control? Cognition and Emotion, 19, 197-216.
Hammerl, M., & Grabitz, H.-J. (1996). Human evaluative conditioning without
experiencing a valued event. Learning and Motivation, 27, 278-293.
Kindt, M., Bögels, S., & Morren, M. (2003). Processing bias in children with separation
anxiety disorder, social phobia, and generalised anxiety disorder. Behaviour Change, 20,
143-150.
Kindt, M., & Van den Hout, M. (2001). Selective attention and anxiety: A perspective on
developmental issues and the causal status. Journal of Psychopathology and Behavioral
Assessment, 23 , 193-202.
Levey, A. B., & Martin, I. (1975). Classical conditioning of human “evaluative”
responses. Behaviour Research and Therapy, 13, 221-226.
Mackintosh, B., & Mathews, A. (2003). Don't look now: Attentional avoidance of
emotionally-valenced cues. Cognition and Emotion, 17, 623-646.
MacLeod, C., & Mathews, A. (1988). Anxiety and the allocation of attention to threat.
Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 40A,
653-670.
MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional bias in emotional disorders.
Journal of Abnormal Psychology, 95, 15-20.
MacLeod, C., & Rutherford, E. (1992). Anxiety and the selective processing of emotional
information: Mediating roles of awareness, trait and state variables, and personal
relevance of stimulus materials. Behaviour Research and Therapy, 30, 479-491.
March, J. S., Parker, J. D. A., Sullivan, K., Stallings, P. K., & Conners, C. K. (1997). The
multidimensional anxiety scale for children (MASC): Factor structure, reliability, and
validity. Journal of the American Academy of Child and Adolescent Psychiatry, 63 , 554-
565.
Mathews, A., & Mackintosh, B. (1998). A cognitive model of selective processing in
anxiety. Cognitive Therapy and Research, 22, 539-560.
Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional
disorders. Annual Review of Psychology, 45 , 25-50.
Mathews, A., Richards, A., & Eysenck, M. (1989). Interpretation of homophones related
to threat in anxiety states. Journal of Abnormal Psychology, 98, 31-34.
Mathews, A., & Sebastian, S. (1993). Suppression of emotional Stroop effects by fear-
arousal. Cognition and Emotion, 7, 517-530.
Richards, A., & French, C. C. (1992). An anxiety-related bias in semantic activation
when processing threat/neutral homographs. Quarterly Journal of Experimental
Psychology: Human Experimental Psychology, 45A, 503-525.
Silove, D., & Manicavasagar, V. (2001). Early separation anxiety and its relationship to
adult anxiety disorders. In M. R. Dadds & M. W. Vasey (Eds.), The developmental
psychopathology of anxiety (pp. 459-480). London : Oxford University Press.
Silove, D., Manicavasagar, V., O'Connell, D., Blaszcynski, A., Wagner, R., & Henry, J.
(1993). The development of the separation anxiety symptom inventory (SASI).
Australian and New Zealand Journal of Psychiatry, 27 , 477-488.
Tyrer, P. (1985). Classification of anxiety. British Journal of Psychiatry, 144, 78-83.
Williams, J. M. G., Mathews, A., & MacLeod, C. (1996). The emotional Stroop task and
psychopathology. Psychological Bulletin, 120, 2-24.
Yiend, J., & Mathews, A. (2001). Anxiety and attention to threatening pictures. Quarterly
Journal of Experimental Psychology: Human Experimental Psychology, 54A, 665-681.
Yiend, J., & Mathews, A. (2002). Induced biases in the processing of emotional
information. Advances in Psychology Research, 13, 43-68.
Acknowledgements
This research was supported by a Wellcome Trust grant (# 065667) awarded to the first
author.
