Attention as longitudinal predictor of early outcomes 1
Attention across modalities as a longitudinal predictor of early outcomes:
The case of fragile X syndrome
Gaia Scerif1, Elena Longhi1, Victoria Cole1, Annette Karmiloff-Smith2 and Kim
1Attention, Brain and Cognitive Development Group, Department of Experimental
Psychology, University of Oxford, UK
2Birkbeck Centre for Brain & Cognitive Development, University of London, UK
3Centre for Developmental Psychiatry & Psychology, Faculty of Medicine,
Monash University, Melbourne, Australia
Total word count: 6,000 (excluding key points and acknowledgment page)
Abbreviated title: Attention as longitudinal predictor of early outcomes
First version of manuscript submitted on 24th July 2011
Revised and re-submitted on 22nd September 2011
Attention as longitudinal predictor of early outcomes 2
Background: Fragile X syndrome (FXS) is an early diagnosed monogenic
disorder, associated with a striking pattern of cognitive/attentional difficulties and
a high risk of poor behavioural outcomes. FXS therefore represents an ideal
model disorder to study prospectively the impact of early attention deficits on
Methods: 37 boys with FXS aged 4 to 10 years and 74 typically developing (TD)
boys took part. Study 1 was designed to assess visual and auditory attention at
two time-points, one year apart. Study 2 investigated attention to multimodal
information. Both tested attention markers as longitudinal predictors of risk for
poor behaviour in FXS.
Results: Children with FXS attended less well than mental-age matched TD
boys and experienced greater difficulties with auditory compared to visual stimuli.
In addition, unlike TD children, they did not benefit from multimodal information.
Attention markers were significant predictors of later behavioural difficulties in
boys with FXS.
Conclusions: Findings demonstrate, for the first time, greater difficulties with
auditory attention and atypical processing of multimodal information, in addition
to pervasive global attentional difficulties in boys with FXS. Attention predicted
outcomes longitudinally, underscoring the need to dissect what drives differing
developmental trajectories for individual children within a seemingly
Attention as longitudinal predictor of early outcomes 3
Keywords: Fragile X syndrome, Attention deficits, Longitudinal predictors of
Neurodevelopmental disorders diagnosed early in life with a clear genetic
aetiology and a recognised phenotype can provide important clues to
understanding the neurodevelopmental origins of other disorders that are
currently defined only by their childhood phenotype (e.g., attention
deficit/hyperactivity disorder, ADHD). Fragile X syndrome (FXS) represents a
model disorder in this respect because it is a well-recognized cause of hereditary
developmental delay, with an estimated incidence of 1 in 2500 world-wide
(Hagerman, 2008), associated with the silencing of a single X-linked gene
(Garber, Visootsak & Warren, 2008) and cases are identified as early as infancy,
with an average diagnosis age of 37.9 months (Bailey et al., 2009). At the
neurocognitive level, FXS is characterised by syndrome-specific proficiencies
and deficiencies that distinguish it from other neurodevelopmental disorders
(Bertone et al., 2010). Highly notable are the striking visual attention difficulties
(e.g., Cornish, et al., 2007; Hooper et al. 2008; Munir, Comish, & Wilding, 2000;
Scerif et al., 2004, 2005, 2007) that persist across development. At the
behavioural level and by mid childhood, FXS is associated with increased risk of
clinically-relevant difficulties (e.g., Rogers, Wehner& Hagerman, 2001; Sullivan et
Attention as longitudinal predictor of early outcomes 4
As such, FXS affords the opportunity to study the links between a single
genetic aetiology, cognitive attention difficulties and subsequent behavioural
outcomes prospectively from early childhood. Critically, however, both theoretical
discussions and emerging evidence warn against construing these gene-brain-
cognition-behaviour links as linear and unidirectional (Karmiloff-Smith, 2009;
Scerif & Karmiloff-Smith, 2005): basic cognitive attentional difficulties
characterise FXS from infancy and early childhood (Cornish et al., 2007; Scerif et
al., 2004, 2005, 2007), and this in turn is likely to dynamically interact over
development with atypical everyday experiences, leading to variable outcomes.
Indeed, within this seemingly homogeneous genetic disorder, individual
differences are noted as early as in infancy (Roberts et al., 2009) and outcome
differences for individuals with FXS are clinically as crucial as known group
differences from the normal population (Chonchaiya, Schneider, & Hagerman,
2009). However, the extent to which early neurocognitive attentional difficulties
drive these poor and variable behavioural outcomes (e.g., classroom
hyperactivity) has not been tested.
A further limitation is that existing work has focused primarily on visual
difficulties, but growing evidence, albeit scarcer, points to more general deficits
including auditory processing and attention difficulties. Mouse models (e.g., Chen
& Toth, 2001), as well as adults and school-aged boys with FXS (e.g., Castren et
al., 2003) display atypical responses to auditory stimuli. A single study has thus
far pitted visual and auditory attention measures directly against each other in
children with FXS (Sullivan et al., 2007). Sullivan and colleagues used auditory
Attention as longitudinal predictor of early outcomes 5
and visual continuous performance tasks (CPT) with children with FXS (8- to 13-
year-olds), a large proportion of whom were on stimulant medication. Only 61%
of 56 boys tested on visual CPT and 54% of 52 boys tested on the auditory task
were able to complete the two tasks respectively. Furthermore, children with FXS
performed poorly on both, and relatively more so for the auditory version.
Although pioneering in the collection of data on attention across modalities, a
number of concerns limit these conclusions. The tasks were neither matched in
terms of stimulus complexity nor duration, and were not counterbalanced, with
the auditory task always following the visual one. Yet, understanding whether
attentional difficulties generalise across vision and audition is clinically relevant: if
stimuli in a certain modality were relatively easier for young children with FXS to
attend to, this might open opportunities for optimisation of learning materials
through the development of targeted educational and clinical interventions that
could begin prior to formal schooling.
Moreover, individuals with FXS show atypical sensitivity to complex or
multimodal environments (e.g., Baranek et al., 2008). Yet, in typical
development, multimodal information provides clear redundancies that aid
perception, attention and learning (Spector & Maurer, 2009). For example, visual
perceptual judgments (e.g., deciding whether a ball is small) are easier for young
children with congruent auditory information (e.g., concurrent presentation of a
high pitch sound), even when this is not directly relevant to the visual task. These
cross-modal binding processes accrue invaluable benefits in a complex
multimodal environment, but, to our knowledge, no study has assessed how
Attention as longitudinal predictor of early outcomes 6
attentional difficulties impact on processing stimuli in multimodal settings that
more closely resemble everyday environments (e.g., classrooms) for individuals
The current study therefore had three principal aims. The first was to
investigate trajectories of attention across two modalities (visual and auditory)
using directly comparable tasks to understand whether attentional difficulties in
young boys with FXS are similar across modalities. The second aim was to
investigate whether, and if so, how, children with FXS have difficulties in coping
with attention in a multimodal environment. As there are striking developmental
changes in attentional abilities over childhood, these first two aims required
mapping typical developmental trajectories as well as those in children with FXS,
starting from as early as possible. Thirdly, through our prospective longitudinal
design, we aimed to investigate attentional predictors of risk for subsequent poor
classroom outcomes, such as inattention, hyperactivity, strengths and difficulties
in the classroom a year later. The latter aim required larger samples than have
normally been studied in the context of FXS, and to our knowledge, this is the
first and largest experimental study of attention across modalities in medication-
naïve boys with FXS as young as 4 years of age.
Study 1 – Attention across modalities
In Study 1, we predicted that boys with FXS would find sustaining
attention more difficult than TD children, regardless of modality, but that they
would in addition experience greater difficulties with auditory compared to visual
Attention as longitudinal predictor of early outcomes 7
attention. Secondly, attentional markers would predict individual differences in
outcomes a year later.
Thirty-seven boys with a confirmed diagnosis of FXS (mean age at Time
1: 8 years;6 months; range: 4-10 years) were recruited through the national
support group for children and families with FXS as part of a larger study of
attentional difficulties. To obtain as representative a sample as possible of
abilities in all boys with FXS, we attempted to administer the current measures to
all boys in the full sample with FXS, but desisted if boys failed to complete an
experimental block, or clearly did not comprehend/comply with task instructions.
So, differences across boys with FXS whose data are reported here or not but
are still part of the broader sample define completion rates [see Table 1 for their
characteristics and Supplemental Materials for full sample details].
A group of 74 typically developing (“TD” henceforth) boys with no reported
family history of FXS (M=6:10; range: 3-10 years) was recruited locally. From this
larger TD sample, a comparison group of 33 TD boys (“controls”) was obtained,
whose mean and range of non-verbal ability (Leiter International Performance
Scale-Revised, “Leiter”, Roid & Miller, 1997) matched that of the boys with FXS.
All children were followed and re-assessed 12 months later (Time 2). Signed
informed consent was obtained from parents following ethical procedures
approved by the appropriate institutional review board.
[Insert Table 1]
Attention as longitudinal predictor of early outcomes 8
The current sample (N=37) excluded children who were taking medication
for the treatment of inattention and hyperactivity symptoms, because work on
ADHD highlights how stimulants modify neurocognitive function and structure.
However, as part of our broader study and to maintain a clear picture of the
sample as a whole, all children who had volunteered to take part in the study
(N=59) were followed at all time-points. At both time-points this broader FXS
study sample (N=59) contained a relatively small proportion of children on
stimulant medication (max 11.8%, i.e., 7 out of 59 children) compared to rates
reported for other samples of boys with FXS (e.g., 33.4%, Sullivan et al., 2006)
[see Supplemental Materials details and for a discussion]. Our TD control sample
also excluded any child who scored at or above standardised clinical thresholds
on the ADHD Index, and we excluded children with poor hearing or visual acuity
in both groups.
Non-verbal IQ. The Leiter International Performance Scale-Revised (Roid &
Miller, 1997) is a standardized assessment for individuals aged 2-20 designed to
be administered entirely non-verbally. The Brief IQ score is a composite of 4
different scales: Figure Ground segregation, Form Completion, Sequential
Ordering and detecting Repeated Patterns. It was chosen because of the
communication challenges experienced by boys with FXS.
Behavioural Outcomes. The Conners Teacher Rating Scale (“CTRS”,
Conners, 1997) is a commonly used standardized screening instrument that
targets ADHD symptomatology and consists of 28 items, measuring indices of
Attention as longitudinal predictor of early outcomes 9
oppositional behaviour problems, hyperactive behaviour and cognitive/inattention
problems across the school setting in 3-17 year olds. The Strengths and
Difficulties Questionnaire, Teacher version (“SDQ”, Goodman, 1997) is a 25-item
behavioural screening questionnaire about 3-16 year olds, asking more broadly
about emotional symptoms, conduct problems, hyperactivity/inattention, peer
relationship problems and prosocial behaviours. Both were completed at both
time-points by teacher(s) best acquainted with individual children.
Visual Attention. This task was an analogue to a standard continuous
performance task, providing a baseline measure of attention to centrally
presented targets of higher contrast than other more frequent but low contrast
non-targets (see Figure 1). These controlled stimuli allowed us to ensure that all
children could discriminate targets similarly, as indeed demonstrated by
performance in a separate task in which we established similar visual contrast
thresholds for children with FXS and controls, t(60)=.254, p=.801, and selected
stimuli for the attentional task that were clearly supra-threshold for both groups.
In contrast, using stimuli solely designed for TD individuals in standardised
assessment tools could yield differences for perceptual (Bertone et al., 2010),
rather than cognitive reasons.
Auditory Attention. Temporal parameters of stimulus presentation and overall
duration were identical to the Visual task, but the task presented pure tone
targets of higher intensity amongst lower intensity non-targets. Again, children
with FXS and controls did not differ in their auditory intensity thresholds assessed
Attention as longitudinal predictor of early outcomes 10
through a separate but similar task, t(59)=-1.112,p=.272, and the intensity
discrimination chosen for the attention task was supra-threshold for both groups.
[Insert Figure 1]
Children were seen at school, in a quiet space close to their classroom.
For the attention tasks, they sat at a small table at ~30cm from the monitor and
two speakers, facing the button box. Individual short blocks for each task and
standardised assessment scales were alternated in presentation across
participants to limit differential effects of fatigue and practice. The visual attention
task was presented as a fishing game. Children were asked to watch the
“moving water” (low contrast gabor patches), and to look out for the “big waves”
(high contrast targets), as this meant “there was a fish swimming past.” When the
target wave was detected, children pressed a target key with their dominant hand
(to “catch the fish.”). If they responded correctly, a cartoon fish image appeared
during practice, alongside auditory feedback (“yippee!”) throughout the
experiment. For auditory attention, children were asked to help a hungry mouse
grab some cheese delivered behind a closed door, and in order to do so they
needed to listen to the knock at the door (the “cheese knock”) amongst a
continuous run of quieter knocks. Pressing the target key opened the door and
“caught the cheese”. If children responded correctly, an animated cartoon
character grabbed the cheese. For both attention tasks, slow practice trials were
followed by real-time practice and by test blocks. Test trials were divided into 3
blocks, each lasting 1 minute and including a total of 15 targets presented at
Attention as longitudinal predictor of early outcomes 11
pseudo-random intervals. Children completed at least one block, providing a
maximum total of 45 targets across the 3 blocks.
The primary dependent measures were accuracy of target detection
(percentage hits) and reaction time to hits (time taken to press the target button).
We also measured false alarms to the non-target stimuli, and calculated an
unbiased measure of discrimination, d-prime, appropriately adjusted for the low
frequency of targets. The extent to which attention measures predicted outcomes
a year later for boys with FXS was assessed through preliminary correlations,
followed by multiple regression models for the significantly correlated variables.
Typical developmental attention across modalities
We first aimed to establish the robustness of developmental changes for
our experimental measures in a larger TD sample. As chronological age (“CA”)
was continuously distributed, we assessed the effects of CA through mixed
design linear regression models (Thomas et al., 2009). There were significant
main effects of Time and CA on accuracy (percentage hits)
[F(1,72)=32.845,p<.001,η2=.313 and F(1,72)=72.382,p<.001,η2=.501] and
reaction time (RT in ms) [F(1,72)=95.954,p<.001,η2=.568 and
F(1,72)=91.186,p<.001,η2=.559], driven by an improvement in accuracy and
faster responses from Time 1 to Time 2 and by lower accuracy and slower RT for
younger compared to older children. For condition means, SEM, and other
statistically significant effects, please see Supplemental Materials.
Attention as longitudinal predictor of early outcomes 12
Critically, we wanted to assess whether these longitudinal improvements
indexed developmental change or more simply depended on practice effects.
Performance at Time 2 by 20 control children (here labelled as “Time 2 sample”)
was compared to that of a different group of 20 TD children who had performed
the tasks at Time 1, and therefore for the first time (“Time 1 sample”). The two
groups were matched in terms of CA and MA (ps>.05).There was no significant
main effect of Sample on accuracy [F(1,38)=1.513,p=.226,η2=.038] or RT
[F(1,38)=.349,p=.558,η2=.009], and no significant interaction effect of Sample
and Modality on either accuracy [F(1,38)=.917,p=.344, η2=.024] or RT
[F(1,38)=.501,p=.483,η2=.013], suggesting that improvements observed at Time
2 are unlikely to be accounted for by practice, because children presented with
the tasks at Time 2 performed no better than children at Time 1 who had
encountered them for the first time. Furthermore, task measures related
specifically to everyday inattention and hyperactivity observed in the classroom
(see Supplemental Materials).
Attention across modalities in boys with FXS: Trajectories and predictors
[Insert Figure 2]
Figure 2 presents accuracy and reaction times in Auditory and Visual
Attention for boys with FXS and controls at Time 1 and Time 2. There was a
main effect of Time[F(1,68)=47.572,p<.001,η2=.412] with higher accuracy at Time
2 (67.9%) compared to Time 1 (57.2%), and a main effect of Group
[F(1,68)=36.818,p<.001,η2=.351] with higher accuracy for TD (76.5%) than FXS
Attention as longitudinal predictor of early outcomes 13
(48.6%), but no significant interaction effect between Time and Group, p=.806.
There was a significant interaction of Modality and Group
[F(1,68)=11.709,p=.001,η2=.147], driven by greater accuracy for auditory (78.9%)
compared to visual stimuli (74.2%, F(1,68)=6.403,p=.014,η2=.086) in controls, but
the opposite pattern for FXS, i.e., greater accuracy for visual (50.6%) compared
to auditory stimuli (46.6%, F(1,68)=5.310, p=.024,η2=.072). None of the other
main effects or interactions reached statistical significance.
In terms of RT, there were main effects of Group [F(1,68)=39.960,
p<.001,η2=.370], with boys with FXS responding more slowly (1278.79ms) than
controls (877.73ms), and Time [F(1,68)=50.672,p<.001,η2=.427], with slower
responses at Time 1 (1171.11ms) compared to Time 2 (985.41ms). These were
moderated by an interaction of Time and Group
[F(1,68)=15.699,p<.001,η2=.188], driven by smaller but significant improvements
in speed from Time 1 to Time 2 for FXS [1319.96ms vs 1237.62ms,
F(1,68)=5.282,p=.025,η2=.072] compared to controls [1022.25ms vs 733.19ms,
F(1,68)=58.072,p<.001,η2=.461]. There was also a main effect of Modality
[F(1,68)=8.206,p=.006,η2=.108], with slower responses for auditory (1107.55ms)
compared to visual stimuli (1048.96ms). None of the other main effects reached
Group differences in accuracy and reaction time indicated poorer attention
in FXS overall, improvements with time for both groups, but relatively poorer
auditory vs visual attention in FXS. However, boys with FXS also produced
significantly more false alarms to non-targets overall compared to controls across
Attention as longitudinal predictor of early outcomes 14
conditions (Mann-Whitney U, lowest Z=-2.646, p=.008), and more false alarms to
visual than auditory stimuli, at both Time 1 (Wilcoxon Signed Ranks, Z=-
5.062,p<.001) and Time 2 (Z=-3.600,p<.001), suggesting they may have been
responding more impulsively in general, and more frequently to visual compared
to auditory stimuli. We therefore computed d-prime, an unbiased estimate of
children’s ability to discriminate visual and auditory targets amongst non-targets.
All main effects and interaction effects reported above were consistent with the
analysis on percentage hits, including the critical interaction of Group and
Table 2 reports Pearson’s correlations between accuracy and reaction
time on Auditory and Visual Attention at Time 1, non-verbal ability and scores on
subscales of the CTRS (indexing particularly ADHD-relevant behaviours) and
SDQ (related to broader strengths and weaknesses) at Time 2 for FXS. It also
reports regression model statistics for significant Time 1 predictors of Time 2
outcomes, controlling for non-verbal IQ. Greater Visual Attention accuracy and
faster Visual Attention RTs significantly predicted lower Time 2 hyperactivity
ratings in the classroom on the CTRS and on the SDQ, lower Total Difficulties on
the SDQ and better Pro-social Behaviour on the SDQ. Greater Visual Attention
accuracy also strongly predicted lower Time 2 ADHD Index scores on the CTRS.
None of the other relationships reached significance.
[Insert Table 2]
Attention as longitudinal predictor of early outcomes 15
To recapitulate, Study 1 embodied a prospective longitudinal approach to
attention difficulties across modalities in FXS. We first traced trajectories in a
large sample of TD children, to validate our experimental measures and their
longitudinal changes. Against this backdrop, young boys with FXS showed clear
difficulties in attending to visual stimuli, consistent with attentional difficulties
being core impairments as reported later in the lifespan, but they also responded
more poorly to auditory stimuli, for which control children showed an advantage.
This difference across modalities is consistent with reports of auditory processing
difficulties (e.g., Castren et al., 2003) and extends these to attention, ruling out
gross differences in auditory vs. visual paradigms (cf. Sullivan et al., 2007). Of
note, clear attentional difficulties across modalities and the relative modality
difference for boys with FXS are to be placed in the context of overall
improvements over time, a pattern also found elsewhere (Cornish et al.,
submitted) and one that therefore argues against a view of static developmental
freeze in FXS.
Finally, within-group differences in attention markers at Time 1 predicted
ADHD-relevant symptoms across multiple classroom measures as well as pro-
social behaviours, highlighting the importance of understanding early predictors
of individual differences for specific dimensions of everyday outcomes in the
classroom, even within a group of relatively homogeneous genetic aetiology such
as boys with FXS. Interestingly, visual attention measures were a stronger
predictor of later differences than auditory attention measures in FXS (see
Roberts et al., 2009, for parallel findings for non-attentional markers). This
Attention as longitudinal predictor of early outcomes 16
asymmetry across modalities lends further support to our emphasis on the need
to study both general and specific attentional difficulties in FXS.
Study 2 – Attention in multimodal environments
Having investigated similarities and differences in attention difficulties for
boys with FXS when dealing with simple stimuli, we aimed to test empirically
commonly reported clinical observations of atypicalities in dealing with
multimodal environments (Hagerman & Hagerman, 2002). We predicted that TD
children would benefit from information presented in multiple modalities, whereas
children with FXS would not.
One boy with FXS who participated in Study 1 did not complete Study 2,
whereas one TD child completed Study 2 and not Study 1. Group characteristics
for Study 2 did not change significantly compared to those reported in Table 1.
Apparatus, Measure and Procedure
The standardised and outcome measures were identical to Study 1.
Crossmodal Attention. The task was analogous to the Visual Attention task in
Study 1, in that children were told to pay attention to the ‘big waves’ infrequently
presented amongst lower contrast ‘waves’ to help a hungry boy catch fish.
However, this time they were told that there would be noises in the background
(the pure tones used in the Auditory Attention task), but that the noises would not
help catch the fish. On the majority of non-target trials, low contrast Gabors and
quiet tones were presented simultaneously. Infrequent visual targets were
presented under two conditions: in the Visual-only condition, in which target high
Attention as longitudinal predictor of early outcomes 17
contrast Gabors appeared but were not accompanied by a louder irrelevant tone;
in the Bimodal condition (illustrated in Figure 1), in which the visual change was
accompanied by a concurrent irrelevant louder tone. In Auditory-catch trials, a
loud tone was presented alone, to measure the extent to which children’s
attention was nonetheless captured by the irrelevant tone, leading them to an
incorrect button press. There were a maximum of 15 Visual-only trials, 15
Bimodal trials and 15 Auditory-catch trials. Infrequent irrelevant tones were also
presented alone (Auditory-Catch trials). As for the previous tasks, three blocks of
trials (1 minute each) were intermixed with the other tests to limit fatigue.
Typical developmental trajectories of attention in multimodal environments
For our larger TD sample there were main effects of Time and CA on
accuracy [F(1,73)=20.052,p<.001,η2=.215, F(1,73)=55.807,p<.001,η2=.433] and
RT [F(1,72)=13.041,p=.001,η2=.153, F(1,72)=54.480,p<.001,η2=.431], with
greater accuracy and faster responses at Time 2 compared to Time 1, and more
accurate and faster responses by older children. Auditory-only catch trials did not
yield significant changes in accuracy over time [F(1,73)=.824,p=.367,η2=.011],
but younger children committed more errors than older children
(F(1,73)=42.66,p<.001,η2=.369]. See Supplemental Materials for condition
means, SEM and other effects. Did improvements depend on simple practice?
Again, there were no significant differences in accuracy
[F(1,38)=1.356,p=.252,η2=.034] or RT [F(1,38)=1.697, p=.201,η2=.043] between
the older Sample assessed at Time 1 and the longitudinal Sample (Time 2), nor
Attention as longitudinal predictor of early outcomes 18
significant interaction effects of Sample and Condition on either accuracy
[F(1,38)=2.814, p=.102,η2=.069] or RT [F(1,38)=.001,p=.981,η2=.000], ruling out
simple practice effects. Again, correlations supported the validity of our
experimental attention measures as predictors of classroom behaviours (See
Atypical multimodal attention: Trajectories and predictors of outcomes
[Insert Figure 3]
Figure 3 represents accuracy and reaction times for Visual-Only and
Bimodal targets for FXS and controls at Time 1 and Time 2. In terms of accuracy,
there was a main effect of Time [F(1,68)=23.202,p<.001,η2=.254], with better
performance at Time 2 (66.3%) than Time 1 (57.3%), and a main effect of Group
[F(1,68)=31.444,p<.001, η2=.316], with boys with FXS performing significantly
more poorly (48%) than controls (75.6%). In addition, there was an interaction
effect of Time and Condition [F(1,68)=5.719,p=.020,η2=.078], and an interaction
effect of Condition and Group [F(1,69)=7.159,p=.009,η2=.095]. The latter was
driven by significantly better accuracy for controls on the Bimodal condition
(79%) compared to the Visual-Only condition [72.3%,
F(1,68)=8.137,p=.006,η2=.107], and no difference between the two conditions for
boys with FXS [F(1,68)=.817,p=.369,η2=.012], who performed more poorly than
controls on both the Visual-Only [49%, F(1,68)=19.929,p<.001,η2=.227] and the
Bimodal condition [47%, F(1,68)=38.316,p<.001,η2=.360]. The TimeXCondition
interaction was driven by greater differences between Conditions at Time 2
[F(1,68)=8.420,p=.005,η2=.110] and by larger improvements in Time for Bimodal
Attention as longitudinal predictor of early outcomes 19
[F(1,68)=28.888,p<.001,η2=.298] than for Visual-Only trials
In terms of RTs, there was a main effect of Time
[F(1,62)=7.858,p=.007,η2=.112], driven by faster responses at Time 2 (959.48ms)
than Time 1 (1009.04ms); one of Group [F(1,62)=30.340,p<.001,η2=.329], with
boys with FXS being slower (1219.59ms) than controls (838.92ms), but no
interactions between Time and Group [F(1,62)=1.006,p=.320,η2=.016],
suggesting similar improvements across groups. On Auditory-catch trials, there
were no significant group differences [F(1,68)=1.439,p=.235,η2=.021], with, if
anything, boys with FXS tending to produce fewer errors on Auditory-catch trials
(43.2%) than controls (49%), nor interactions between Group and
Table 2 reports correlations between accuracy and reaction time for
Visual-Only and Bimodal targets at Time 1, non-verbal ability, CTRS and SDQ
scores at Time 2 for boys with FXS. It also reports regression model statistics for
significant Time 1 predictors of outcomes at Time 2, controlling for non-verbal IQ.
Greater accuracy and faster responses in Bimodal trials significantly predicted
lower Hyperactivity ratings on the CTRS and lower ADHD Index scores. Greater
accuracy on Bimodal trials also predicted lower Hyperactivity and Total
Difficulties on the SDQ, and lower speed was related to greater Pro-social
Attention as longitudinal predictor of early outcomes 20
Study 2 is the first to explore multimodal attention in FXS, mimicking the
benefits and distractions encountered from congruent or incongruent multimodal
information in everyday environments. These questions were driven by atypical
responses reported for everyday environments both by clinicians and parents
(Hagerman & Hagerman, 2002). Younger TD children detected an infrequent
visual target better when presented with irrelevant but helpful auditory
information. They were also more likely to be incorrectly led to respond by
irrelevant auditory information than older children, consistent with the crossmodal
developmental literature (e.g., Spector & Maurer, 2009). Unlike TD children, boys
with FXS did not benefit from congruent bimodal information, suggesting that
either they did not bind auditory and visual information as controls do, or simply
did not process auditory information as well. Relatedly, children with FXS did not
produce more errors than controls on auditory-catch trials, despite their generally
poorer performance on all other attentional parameters. This further suggests
that they were not as strongly captured by auditory stimuli as TD children.
Critically, when faced with multimodal stimuli, within-group differences in
attention predicted boys with FXS’ later teacher-rated hyperactivity, ADHD-like
behaviours and pro-social behaviours as measured by two of the most commonly
used assessments of behavioural strengths and weaknesses in the classroom.
Intriguingly, within-group differences for the bimodal condition (i.e., the condition
that most clearly differentiated boys with FXS attention from that of TD children)
were the strongest predictors of Time 2 outcomes. All together, these findings
pinpoint more atypical multimodal attention by boys with FXS than expected
Attention as longitudinal predictor of early outcomes 21
given their developmental level, and an absence of the benefits that multimodal
information (albeit irrelevant) brings to TD children. In turn, this atypical attention
to multimodal stimuli predicted subsequent difficulties related to specific
behavioural outcomes in the classroom.
Conclusions and Implications
Pervasive attention difficulties in young boys with FXS were more
pronounced for auditory than for visual attention, consistent with clinical and
empirical reports of auditory processing differences in individuals with FXS
(Baranek et al., 2008; Castren et al., 2003; Sullivan et al., 2007). These relative
weaknesses reinforce practical suggestions to present ‘to-be-attended’ materials
in visual rather than auditory format. Furthermore, multimodal information did not
benefit children with FXS, mirroring reports of atypical responses to multimodal
environments (Hagerman & Hagerman, 2002) and suggesting that the common
educational practice of providing redundant multisensory information may not
necessarily help boys with FXS specifically, although this would also need to be
assessed when relevant redundant stimuli are presented. Of note, despite poorer
performance than ability-matched younger controls, boys with FXS improved
significantly over time when followed longitudinally, counter to the prevailing view
of static or plateauing profile that can emerge from cross-sectional data alone
(Cornish et al., in press). Finally, better visual attention and more typical
responses to multimodal stimuli predicted improved classroom outcomes a year
later, stressing the need to move away from group findings to individual children
and predictors of better outcomes (Conchaiya et al., 2009; Roberts et al., 2009)
Attention as longitudinal predictor of early outcomes 22
even in a seemingly homogeneous group as boys with FXS. In turn, they
therefore highlight how studying developmental dynamics over time is part and
parcel of understanding how complex behavioural outcomes unravel over early
childhood, and what predicts their trajectories (Karmiloff-Smith, 2009; Scerif &
Karmiloff-Smith, 2005). Isolating these predictors will guide more appropriate
and effective intervention tailored to the needs of individual children with FXS.
Attention as longitudinal predictor of early outcomes 23
This research was supported by a project grant from The Wellcome Trust to GS,
AK-S and KC (WT079326AIA). We express our deepest thanks to all families
who participated in the research and to the Fragile X Society for their continued
support to our efforts. We are indebted to Hannah Broadbent, Nela Cicmil,
Victoria Leggett and Katy Theobald for their invaluable input to data collection
and analysis. We also acknowledge Ben Harvey’s programming input and Justin
Cowan’s contributions to task design and the initial stages of data collection.
Correspondence address: Dr Gaia Scerif, Department of Experimental
Psychology, University of Oxford, South Parks Road, Oxford OX1 3UD, United
Kingdom, telephone: +44-1865-271403
Attention as longitudinal predictor of early outcomes 24
Bailey DB Jr, Raspa M, Bishop E, Holiday D. (2009).No change in the age of
diagnosis for fragile x syndrome: findings from a national parent survey.
Baranek GT, Roberts JE, David FJ, Sideris J, Mirrett PL, Hatton DD, Bailey DB.
(2008) Developmental trajectories and correlates of sensory processing
functions in young children with fragile X syndrome. Physical and
Occupational Therapy in Pediatrics, 28, 79-98
Bertone A, Hanck J, Kogan C, Chaudhuri A, Cornish, KM (2010). Using
perceptual signatures to define and dissociate condition-specific neural
etiology: Autism Spectrum Disorder and fragile x syndrome as model
conditions. Journal of Autism and Developmental Disorders, 40, 1531-40.
Castren M, Paakkonen A, Tarkka IM, Ryynanen M, Partanen J. (2003).
Augmentation of auditory N1 in children with fragile X syndrome. Brain
Topography, 15, 165-71.
Chen L, & Toth M. (2001). Fragile X mice develop sensory hyper-reactivity to
auditory stimuli. Neuroscience, 103, 1043-50.
Chonchaiya W, Schneider A, Hagerman RJ. (2009). Fragile X: a family of
disorders. Advances in Pediatrics, 56, 165-86.
Cornish K, Scerif G, Karmiloff-Smith A (2007). Tracing syndrome-specific
trajectories of attention across the lifespan. Cortex, 43, 672-685.
Cornish KM, Cole V, Longhi E, Karmiloff-Smith A, Scerif G. (in press). Mapping
developmental trajectories of attention and working memory in fragile X
Attention as longitudinal predictor of early outcomes 25
syndrome: developmental freeze or developmental change? Development
Garber KB, Visootsak J, Warren ST (2008). Fragile X syndrome. European
Journal of Human Genetics, 16, 666-672.
Hagerman PJ (2008). The fragile X prevalence paradox. Journal of Medical
Genetics, 45, 498-499.
Hagerman RJ & Hagerman PJ. (2002). Fragile X syndrome: diagnosis,
treatment, and research.3rd Edition. Baltimore, Maryland: Johns Hopkins
Hooper SR, Hatton D, Sideris J, Sullivan K, Hammer J, Schaaf J, et al. (2008).
Executive functions in young males with fragile X syndrome in comparison
to mental age-matched controls: baseline findings from a longitudinal
study. Neuropsychology, 22, 36-47.
Karmiloff-Smith A. (2009). Nativism Versus Neuroconstructivism: Rethinking the
Study of Developmental Disorders. Developmental Psychology, 45, 56-63.
Munir F, Cornish KM & Wilding J. (2000). A neuropsychological profile of
attention deficits in young males with fragile X syndrome.
Neuropsychologia, 38, 1261-1270.
Roberts, Mankowski, Sideris, Davis Goldman, Hatton, Mirret, Baranek, Reznick,
Long, Bailey (2009). Trajectories & Predictors of the development of very
young boys with FXS. J. Pediatr.Psychol., 34, 827-836.
Rogers SJ, Wehner DE, Hagerman R. (2001).The behavioral phenotype in fragile
X: symptoms of autism in very young children with fragile X syndrome,
Attention as longitudinal predictor of early outcomes 26
idiopathic autism, and other developmental disorders. Journal of
Developmental and Behavioural Pediatrics, 22, 409-17.
Scerif G, Cornish K, Wilding J, Driver J, Karmiloff-Smith A. (2004). Visual search
in typically developing toddlers and toddlers with Fragile X or Williams
syndrome. Developmental Science, 7, 116-130.
Scerif G, Cornish K, Wilding J, Driver J, Karmiloff-Smith A. (2007). Delineation of
early attentional control difficulties in fragile X syndrome: focus on
neurocomputational changes. Neuropsychologia, 45, 1889-1898.
Scerif G & Karmiloff-Smith A. (2005). The dawn of cognitive genetics? Crucial
developmental caveats. Trends in Cognitive Sciences, 3, 126-135.
Scerif G, Karmiloff-Smith A, Campos R, Elsabbagh M, Driver J, Cornish K.
(2005).To look or not to look? Typical and atypical development of
oculomotor control. Journal of Cognitive Neuroscience, 17, 591-604.
Spector F & Maurer D. (2009). Synesthesia: a new approach to understanding
the development of perception. Developmental Psychology, 45, 175-89.
Sullivan K, Hatton DD, Hammer J, Sideris J, Hooper S, Ornstein PA, et al.
(2007). Sustained attention and response inhibition in boys with fragile X
syndrome: Measures of continuous performance. American Journal of
Medical Genetics Part B-Neuropsychiatric Genetics, 144B, 517-532.
Sullivan K, Hatton D, Hammer J, Sideris J, Hooper S, Ornstein P, et al. (2006).
ADHD symptoms in children with FXS. American Journal of Medical
Genetics Part A, 140A, 2275-2288.
Attention as longitudinal predictor of early outcomes 27
Thomas MS, Annaz D, Ansari D, Scerif G, Jarrold C, Karmiloff-Smith A. (2009).
Using developmental trajectories to understand developmental disorders.
Journal of Speech Language and Hearing Research, 52, 336-358.
Attention as longitudinal predictor of early outcomes 28
Table 1. Sample Characteristics. Demographic and clinical characteristics for
children with FXS and typically developing (TD) children at Time 1 and Time 2.
Values in brackets represent the range of scores. Abbreviations: Leiter = Leiter
International Performance Scale-Revised (Roid & Miller, 1997), for the IQ
measure, population mean of 100, standard deviation of 15; CTRS = Conner’s
Teacher Rating Scale – Revised: Short Form (Conners, 1997), standardised t-
scores are available based on age and gender, with a mean of 50, and SD of 10,
with t-scores greater than 70 reported as ‘severely abnormal’. Children with FXS
differ from control children: * p<.05, ** p<.005, ***p<.001
Table 2. Correlations across Measures. Pearson correlation coefficients
between experimental attentional measures at Time 1 (T1) and classroom
outcome measures 12 months later (T2), for boys with FXS. For significant
correlations, residual predicted variance (after controlling for non-verbal IQ) is
also reported, with the significance level of attentional markers as unique
predictors for that step. +p<.10, *p<.05, ** p<.005, ***p<.001.
Attention as longitudinal predictor of early outcomes 29
Figure 1. Trial Schematics. Trial sequence for the Visual Attention, Auditory
Attention (Study 1) and Crossmodal Attention Task (Study 2). 1A. For Visual
Attention, targets were high contrast Gabor patches presented infrequently in a
stream of lower contrast Gabor patches. 1B. For Auditory Attention, pure tones
were presented, and the targets had higher intensity than the frequent standard
tones. 1C. For Crossmodal Attention, on the majority of non-target trials, low
contrast Gabors and quiet tones were presented simultaneously. Infrequent
visual targets were presented in the Visual-only condition or in the Bimodal
condition (illustrated in the figure). Auditory-catch trials presented louder intensity
Figure 2. Study 1. A) Accuracy scores (% hits) for visual and auditory attention
for boys with FXS and TD controls at Time 1 and Time 2. B) Mean reaction time
(ms) for visual and auditory attention for boys with FXS and TD controls at Time
1 and Time 2. Error bars indicate standard errors of the mean.
Figure 3. Study 2. A) Accuracy scores (% hits) for the Visual-only and Bimodal
conditions for boys with FXS and TD control children at Time 1 and Time 2. B)
Mean reaction time (ms) for the Visual-only and Bimodal conditions for boys with
FXS and TD control children at Time 1 and Time 2. Error bars indicate standard
errors of the mean.
Attention as longitudinal predictor of early outcomes 30
Table 1. Sample Characteristics.
Time 1Time 2
Boys with FXS TD controlsTD Group Boys with FXS TD controlsTD Group
Mean(range)Mean (range)Mean (range) Mean(range)Mean
Age at Test (mths)102.91 (52-
Leiter Mental AgeEquiv
60.54 (38-77) 63.65 (43-79)89.59 (43-
Leiter IQ64.08 (44-109)*** 106.48 (79-
CTRS Oppositional59.11 (44-89)*52.21 (44-89)51.15 (44-89) 55.21 (45-89)50.18 (44-
CTRS Cognitive/Inat67.44 (41-88)*58.93 (41-88) 52.76 (41-88)66.62 (41-88)** 56.97 (41-
CTRS Hyperactivity62.85 (47-77)*** 50.06 (43-68)48.45 (43-68)62.35 (47-52.0 (43-71)50.11 (43-71)
Attention as longitudinal predictor of early outcomes 31
CTRS ADHD Index 64.58 (48-79)*** 51.54 (43-67)48.89 (41-67) 63.62 (47-
Attention as longitudinal predictor of early outcomes 32
Table 2. Correlations across Measures.
Study 1 Attention Measures (T1) Study 2 Attention Measures (T1)
T2 Leiter IQ .051.241 -.057-.013 -.306+
-.034 -.058-.040 .173 .058 -.275.004 .183
-.258 -.010.028 -.209-.013 -.119 .173 .315
-.181 -.150 .222 .198-.174-.089.048.088
-.087-.148 .115 .137 -.032-.258 -.034 .224
.040-.176 .044 .243 .023-.034 .094 .132
Attention as longitudinal predictor of early outcomes 33
Figure 1. Trial Schematics.
“Catch the cheese when you hear a loud knock on the
“Catch the fish when you see a big wave in the water”
“Catch the fish when you see a big wave in the water.
Remember: the sounds don’t help!”
Attention as longitudinal predictor of early outcomes 34
Figure 2. Study 1.
Attention as longitudinal predictor of early outcomes 35
Figure 3. Study 2.
Attention as longitudinal predictor of early outcomes 36
What is known:
Neurodevelopmental disorders of known genetic origin can
offer unique insights into the early pathways and
mechanisms leading to childhood behavioural difficulties
What is new:
Tracking attention trajectories in young children with fragile
X syndrome revealed impaired attention, differentially
greater difficulties when dealing with auditory stimuli and
smaller benefits from multimodal information than control
Individual differences in attention also predicted
longitudinal differences amongst children with FXS in
behaviours relevant to classroom outcomes
What is clinically relevant:
Our findings highlight the importance of capturing dynamic
trajectories of attention over developmental time, as these
predict differing longitudinal outcomes even for young
children with a well understood monogenic disorder