Error-related processing dysfunction in children aged 9 to 12 years presenting putative antecedents of schizophrenia.
ABSTRACT Intervention aimed at preventing schizophrenia may be most effective if targeted at specific, but modifiable, functional impairments that present during childhood. We have developed a novel method of screening community samples aged 9 to 12 years to identify children who present a triad of putative antecedents of schizophrenia (ASz), defined as 1) speech and/or motor development lags/problems; 2) internalizing, externalizing, and/or peer-relationship problems in the clinical range; and 3) psychotic-like experiences. This study examined whether ASz children display brain function abnormalities during error processing that are similar to those exhibited by adults with schizophrenia.
Twenty-two ASz children and 26 typically developing (TD) children with no antecedents of schizophrenia completed an error-inducing Go/NoGo task during event-related potential recording. Group differences were examined in the amplitude and latency of four event-related potential components: the initial error-related negativity (ERN) and later error-positivity (Pe) elicited on false-alarm responses to NoGo trials, and the corresponding initial correct response negativity (CRN) and later correct response positivity (Pc) elicited during processing of correct responses to Go trials.
Relative to TD children, ASz children were characterized by reduced ERN amplitude but unaffected CRN, Pe, and Pc amplitudes. No group differences were observed in the latency of any component.
Children presenting a triad of putative antecedents of schizophrenia show error-processing dysfunction mimicking that observed in adults with schizophrenia using the same Go/NoGo paradigm. The ASz children displayed specific early error-processing deficits rather than a generalized deficit in self-monitoring.
Kristin R. Laurens; Page 1
In press: Biological Psychiatry
Error-related processing dysfunction in children aged 9-12 years presenting putative antecedents of
Kristin R. Laurens 1,2, Sheilagh Hodgins 1, Glenn L. Mould 1, Sophie A. West 1, Poppy L. A. Schoenberg 1,
Robin M. Murray 2, Eric A. Taylor 3
1 Department of Forensic Mental Health Science, Institute of Psychiatry, King’s College London, Box
P023, De Crespigny Park, London SE5 8AF, United Kingdom
2 Division of Psychological Medicine, Institute of Psychiatry, King’s College London, Box P063, De
Crespigny Park, London SE5 8AF, United Kingdom
3 Department of Child and Adolescent Psychiatry, Institute of Psychiatry, King’s College London, Box
P046, De Crespigny Park, London SE5 8AF, United Kingdom
Correspondence concerning this article should be addressed to:
Dr. Kristin R. Laurens, Department of Forensic Mental Health Science (Box P023), Institute of Psychiatry,
De Crespigny Park, London SE5 8AF, United Kingdom. Telephone: +44-20-7848-0964; Facsimile: +44-
20-7848-0921; E-mail: Kristin.Laurens@kcl.ac.uk
Key Words: error-related negativity; error positivity; event-related potentials (ERP); high-risk; child
Kristin R. Laurens; Page 2
Background: Intervention aimed at preventing schizophrenia may be most effective if targeted at specific,
but modifiable, functional impairments which present during childhood. We have developed a novel
method of screening community samples aged 9-12 years to identify children who present a triad of
putative antecedents of schizophrenia (ASz), defined as (i) speech and/or motor development
lags/problems; (ii) internalising, externalising, and/or peer-relationship problems in the clinical range; and
(iii) psychotic-like experiences. The present study examined whether ASz children display brain function
abnormalities during error processing that are similar to those exhibited by adults with schizophrenia.
Methods: Twenty-two ASz children and 26 typically-developing (TD) children with no antecedents of
schizophrenia completed an error-inducing Go/NoGo task during event-related potential (ERP) recording.
Group differences were examined in the amplitude and latency of four ERP components: the initial error-
related negativity (ERN) and later error-positivity (Pe) elicited on false-alarm responses to NoGo trials,
and the corresponding initial correct-response negativity (CRN) and later correct-response positivity (Pc)
elicited during processing of correct responses to Go trials.
Results: Relative to TD children, ASz children were characterised by reduced ERN amplitude, but
unaffected CRN, Pe and Pc amplitudes. No group differences were observed in the latency of any
Conclusions: Children presenting a triad of putative antecedents of schizophrenia show error processing
dysfunction that mimics that observed in adults with schizophrenia using the same Go/NoGo paradigm.
The ASz children displayed specific early error-processing deficits rather than a generalised deficit in self-
Kristin R. Laurens; Page 3
Despite ongoing developments in pharmacological and psychosocial therapies, schizophrenia remains a
persistent and severe disorder for a majority of affected individuals. The first episode presents a critical
period for influencing disease trajectory and outcome; however, significant structural and functional brain
abnormalities are present already by this stage of illness (1). Recognising and intervening with at-risk
individuals prior to illness onset offers the potential to prevent or minimise deviant development and the
resulting disability, but elucidating at-risk status is challenging. Although schizophrenia is highly heritable,
almost two-thirds (63%) of affected individuals have no first- or second-degree relatives with the disorder
(2), and the complex interplay of genes conferring susceptibility is undetermined (3). Several prospective
longitudinal investigations that assessed birth cohorts repeatedly during infancy, childhood, adolescence,
and adulthood [reviewed in (4)], have indicated that children who later develop schizophrenia display
multiple developmental indices that distinguish them from their peers. The study of such children has the
potential to identify biological and environmental factors that promote the development of schizophrenia.
Based on detailed review of prospective investigations that identified childhood antecedents of
schizophrenia (i.e., developmental impairments that precede illness), we constructed child- and caregiver-
questionnaires to screen community samples for these antecedents (5). We attempted to identify a group of
children at elevated and specific risk for schizophrenia by requiring the presence of three putative
antecedents of schizophrenia that previous studies indicated as consistently and most strongly associated
with the development of the disorder. We operationalised the definition of the triad of putative antecedents
of schizophrenia (ASz) to include (i) caregiver-reported speech and/or motor development lags or
problems; (ii) child-reported internalising problems, and/or caregiver-reported externalising and/or peer-
relationship problems in the clinical range; and (iii) child-reported psychotic-like experiences. Data
obtained to date from 1,172 children and caregivers indicate that 9.1% of children (12.4% boys, 6.2%
girls) present the triad of antecedents, whereas 25.3% (20.2% boys, 29.4% girls) display no triad
Kristin R. Laurens; Page 4
component. Triad prevalence is elevated among children of African-Caribbean ethnicity, mimicking the
elevated prevalence of schizophrenia in this group in the UK (6).
Only follow-up of the ASz children will determine the specificity and sensitivity with which the triad
of antecedents identifies individuals who subsequently develop schizophrenia. In the meantime, we are
undertaking a series of studies to characterize the ASz children and to determine whether they present
abnormalities similar to adults with schizophrenia. The present study employed event-related potential
(ERP) recordings to determine whether brain function abnormalities present in schizophrenia would
distinguish ASz children from typically-developing (TD) children without the antecedents.
A characteristic functional abnormality in schizophrenia, as revealed by ERP recordings, is a reduction
in amplitude of a frontocentrally distributed negative voltage component, the “error-related negativity”
(ERN) or “error negativity” (Ne), that peaks around 50-150 ms after commission of an erroneous response
(7, 8). The abnormality may contribute to patients’ difficulty in appropriately monitoring and adjusting on-
going behaviour. The ERN is believed to arise in anterior cingulate cortex [ACC;(9-11)], a region
characterised by functional and structural abnormality in clinically- or genetically-defined high-risk youth
in the years immediately preceding their transition to psychosis (12, 13). The component has been
proposed alternately to reflect an error detection mechanism (7, 14), an error signal to initiate remedial
behaviour (15), detection of conflict between competing response alternatives [i.e., the error and error-
correcting correct response (16)], or an affective response to errors (17, 18). A corresponding, though
smaller, “correct response negativity” (CRN) is elicited following correct responses, and is suggested to
manifest a response comparison process (18, 19), uncertainty regarding the correct response (17, 20), or an
affective reaction (21). Adults with schizophrenia are characterised by attenuated ERN amplitude during
error processing across a variety of experimental paradigms, including Go/NoGo (22, 23), Erikson-type
flanker (24, 25), Stroop colour-word naming (26), picture-word naming (27), and probabilistic learning
(28) tasks. By contrast, reports of CRN amplitude differences are variable, with relative attenuation (22),
Kristin R. Laurens; Page 5
no difference (23), and relative augmentation (24-27) of CRN amplitude reported in adults with
schizophrenia as compared to healthy adults.
On error trials, the ERN is followed by a positive deflection in the waveform peaking between 200-400
ms after the error response [the “error positivity” (Pe), (14, 19)], which has been suggested to index the
conscious evaluation, or motivational significance, of the error (29). Like the ERN, this component is
distributed maximally at midline electrodes, although more posteriorly on the scalp than the ERN (19, 30).
Pe amplitude appears unaffected in schizophrenia (23, 25-27), indicating a lack of generalised error-
processing deficit. Adults with and without schizophrenia display slower reaction times on trials
immediately following an error (24, 27, 31), although this post-error slowing is sometimes reduced in
schizophrenia (26, 32). Pe amplitude, but not ERN amplitude, is consistently correlated with post-error
slowing (33). Thus, the unattenuated Pe response in adults with schizophrenia may underlie their ability to
make some behavioural adjustment following errors.
Both ERN and Pe components are present by age 7 years, but the ERN, not the Pe (34, 35), increases in
amplitude from age 7-8 to age 9-11 years (36, 37), and further by age 13-14 years (35), plateauing by late
adolescence or early adulthood, in line with the ongoing physiological maturation of the ACC (38, 39).
The present study sought to examine whether ASz children aged 9-12 years displayed brain function
abnormalities in error processing during ERP recording of a Go/NoGo task. Based on the pattern of ERP
abnormalities observed in schizophrenia, we hypothesised that ASz children, relative to TD children,
would demonstrate relative reduction in amplitude of the ERN elicited on erroneous trials, but unaffected
Pe amplitude. Previous reports of relatively increased CRN amplitude in adults with schizophrenia on
correct response trials (24-27) have not been observed using the Go/NoGo task (22, 23), and thus, were not
anticipated among ASz children.
Methods and Materials
Kristin R. Laurens; Page 6
Eight hundred and sixty-nine children aged 9-12 years and their primary caregiver, recruited via
schools, independently completed antecedent screening questionnaires. The child questionnaire comprised:
(i) the self-report version of the Strengths and Difficulties Questionnaire [SDQ, (40)], which incorporates a
prosocial behaviour scale and four psychopathology scales assessing emotional symptoms, conduct
problems, hyperactivity-inattention, and peer relationship problems, and (ii) nine items assessing
psychotic-like experiences (PLEs), including five items adapted from the Diagnostic Interview Schedule
for Children (41). The caregiver questionnaires comprised four sections: (i) socio-demographic
information on the family, (ii) quantitative and qualitative indices of developmental delays or
abnormalities in speech and motor function, (iii) the parent-report version of the SDQ, and (iv) parent-
report versions of the PLEs. ASz children were those presenting: (i) a caregiver-reported delay or
abnormality in speech and/or motor development; (ii) an “abnormal” rating (i.e., top tenth percentile of UK
population norms) on at least one SDQ psychopathology scale (i.e., child-reported emotional symptoms, or
caregiver-reported conduct problems, hyperactivity-inattention, or peer relationship problems); and (iii)
child-reported “certain experience” of at least one PLE. TD children presented none of these putative
antecedents1. Among the 869 children and caregivers who completed questionnaires, 9.5% of children met
criteria for ASz and 24.6% for TD, and 57.8% (503) agreed to be contacted for further research. Of these
503 children, 9.8% met criteria for ASz and 23.0% for TD. An invitation to participate in the study was
extended to 40 ASz and 42 TD families, of whom 22 ASz and 29 TD children consented to participate in
laboratory assessments. Children provided written assent, and parents provided written informed consent,
Twenty-two ASz children and 26 TD children without the antecedents completed ERP recording
(additional data from three TD children were excluded due to computer problems arising during
recording). Groups were matched on demographic variables (age, sex, handedness, and ethnicity), IQ, and
scholastic achievement. Comparisons of the ASz and TD children are presented in Table 1. No child had
ever taken psychotropic medication.
Kristin R. Laurens; Page 7
Task parameters and procedure
Participants completed a visual Go/NoGo task programmed using Microsoft Visual Basic 6.0 software.
The task comprised 240 Go (80%) and 60 NoGo (20%) letter stimuli presented in two blocks of 150
stimuli each separated by a rest break. In each block, the order of stimulus presentation was pseudorandom
such that no two NoGo stimuli occurred in succession. Each stimulus was presented to participants for 250
ms, with a random inter-trial interval of between 1000-3000 ms (average ~2000). The letter stimuli were
presented within a fixation box on a computer monitor in black font against a pale grey background, and
subtended a visual angle of approximately 2.4 x 2.6 degrees.
Participants were instructed to respond as quickly and accurately as possible with a dominant index
finger button-press on a response-pad to each presentation of the letter “X” (Go stimulus). They were
instructed not to respond to the letter “K” (NoGo stimulus). Reaction times on correct hits to Go stimuli
(Go) were computed for trials in which participants responded within 150–1000 ms of stimulus onset.
Errors of commission (EoC) were responses that occurred with 1000 ms of the onset of a NoGo stimulus.
Correctly-rejected NoGo stimuli were defined by the absence of a motor response within 1000 ms of a
Prior to recording, participants completed a 30 sec practice session (repeated as necessary to ensure full
task comprehension before recording). Recordings were conducted in a quiet room, with participants
instructed to refrain from blinking and movement during stimulus presentation.
Scalp potentials were recorded from 30 Ag/AgCl sintered electrodes (Quik-Cap; Compumedics
Neuroscan) distributed according to the International 10-10 System of electrode placement. Only data from
the midline electrodes positioned at FCz (frontocentral), Cz (central), and CPz (centroparietal), where the
ERP components of interest are maximal in amplitude, are reported in this manuscript. To track eye
movements, horizontal and vertical electrooculograms were recorded from bipolar electrodes positioned
adjacent to the outer canthi of each eye, and from above and below the left eye, respectively. All electrodes
Kristin R. Laurens; Page 8
were referenced to an electrode positioned at the tip of the nose. Electrical impedance at each electrode
was maintained ≤ 10 kOhms. Electroencephalograhic (EEG) data were recorded using Neuroscan software
(SCAN 4.3) and hardware (Neuroscan Synamps). Data were sampled continuously online to computer
hard-disk at a digitisation rate of 500 Hz, amplified (gain 500) and bandpass filtered at 0.05 to 100 Hz.
ERP Data Processing
ERP data were analysed using Brain Vision Analyzer software (Brain Products). The EEG was filtered
with 2 Hz high-pass and 10 Hz low-pass zero-phase Butterworth filters with a 24db/octave slope and a
50Hz notch filter. Correction of blink and eye movement artefacts was made using the method of Gratton
et al. (1983) (42), and data then segmented into response-locked correct hits to Go stimuli and response-
locked errors to NoGo stimuli (segment length 600 ms, from 200 ms pre-response to 400 ms post-
response). Segments were baseline corrected to the 50 ms period between 200 and 150 ms pre-response.
Artefact rejection was applied to remove trials contaminated by muscular activity or amplifier blocking.
Single trials with voltages exceeding +/- 75 µV at any electrode were excluded. For each trial type,
segments were averaged to produce individual participant waveforms. Grand average waveforms for
correct hits and for errors were subsequently created from these individual averages.
Behavioural data. Task performance indices comprised accuracy and reaction time (RT) measures.
Groups were compared on Go and EoC trials using univariate ANOVAs. For RT analyses, Go trials were
differentiated into two types: (a) those that followed another correct hit to a Go trial (Go-Go), and (b) those
that followed an EoC to a NoGo trial (EoC-Go). RT indices were entered into a two group (ASz, TD) x
three condition (Go-Go, EoC-Go, EoC) repeated-measures ANOVA, employing Greenhouse-Geisser
correction for repeated measures. Follow-up simple main effects testing was conducted using bonferroni
correction for multiple comparisons.
Physiological data. Four ERP components were analysed at FCz, Cz, and CPz electrodes, including
the initial negativity and later positivity elicited in the response-locked error average waveform (ERN and
Kristin R. Laurens; Page 9
Pe, respectively), and the corresponding initial negative and later positive components measured in the
response-locked correct hits average waveform (CRN and Pc, respectively). Choice of latency windows for
peak detection was based on visual inspection of the grand average waveforms (ignoring group status) and
on previous studies (22, 25, 35). The ERN and CRN peak amplitudes were defined within their respective
waveforms as the largest negative peak elicited between 25 ms pre-response and 125 ms post-response.
The Pe and Pc peak amplitudes were identified as the largest positive peak occurring between 200 and 400
ms post-response in their respective waveforms.
Amplitude and latency of the ERN and CRN were examined using separate two group (ASz, TD) x
two condition (response-locked correct hits to Go trials, response-locked errors to NoGo trials) x three site
(FCz, Cz, and CPz) repeated measures ANOVAs. Pe and Pc amplitude and latency were likewise
examined in separate two group (ASz, TD) x two condition (response-locked correct hits to Go trials,
response-locked errors to NoGo trials) x three site (FCz, Cz, CPz) repeated measures ANOVAs.
Greenhouse-Geisser correction for repeated measures was employed, and follow-up simple main effects
testing conducted with bonferroni correction applied to maintain familywise Type I error rate below 0.05.
Reaction time and accuracy data for the TD and ASz groups, and a summary of statistical results, are
provided in Table 2. Both groups responded correctly on the majority of trials, with no significant
differences in group performance on the percentage of correct responses to Go trials or the percentage of
errors committed on NoGo trials.
The two group (ASz, TD) by three condition (Go-Go, EoC-Go, EoC) repeated-measures ANOVA
indicated no effect of group and a significant main effect of condition. Follow-up simple main effects
testing revealed that, across both groups, RT for errors were significantly shorter than for correct hits to Go
trials (i.e., both Go trials following another correct Go trial [Go-Go; p<0.001] and Go trials following an
error [EoC-Go; p<0.001]). Post-error slowing was observed, with RT on Go-Go trials significantly faster
Kristin R. Laurens; Page 10
than RT on EoC-Go trials (p<0.001). There was no significant interaction of group-by-condition,
indicating a lack of reaction time differences between ASz and TD children across all trial types.
No significant group differences were apparent in the percentage of trials excluded from individual
average waveforms due to contamination by muscular activity or amplifier blocking for either correct hits
to Go trials [F(1, 46)=3.0, p=0.09] or for errors of commission on NoGo trials [F(1, 46) = 0.02, p=0.9],
leaving an equivalent number of trials comprising the individual average waveforms across groups3.
Peak amplitude and latency values for the ERN, CRN, Pe, and Pc are provided in Table 3. ERN and
CRN amplitudes were maximal for both groups at FCz. Pe amplitude was maximal for ASz children at Cz
and CPz, and for TD children at FCz, with Pc amplitude maximal at CPz for ASz children and at Cz and
CPz for TD children. Grand average waveforms illustrating the ERN, CRN, Pe, and Pc at the three
electrode sites are illustrated in Figure 1 (waveforms for all 30 channels recorded are provided as
supplementary online material). A summary of significant statistical test results obtained for peak
amplitude and latency analyses is also provided in Table 3.
ERN and CRN amplitude. The two group (ASz, TD) x two condition (CRN, ERN) x three site (FCz,
Cz, and CPz) repeated measures ANOVA on peak negativity amplitude data indicated no group effect
[F(1, 46)=3.4, p=0.07], a significant main effect of condition, with ERN amplitude significantly greater
than CRN amplitude, and a significant main effect of site. Follow-up simple main effects testing
demonstrated that peak amplitude decreased from anterior to posterior sites, such that amplitude elicited at
FCz was significantly greater than at both Cz (p=0.02) and CPz (p=0.003), and amplitude at Cz was
significantly greater than at CPz (p=0.03). These effects were modified by a significant interaction of site-
by-condition, with simple main effects testing indicating that ERN amplitude was smaller at the most
posterior site (CPz) than at FCz (p=0.002) and at Cz (p=0.002), while CRN amplitude did not differ
significantly across sites. There was a significant group-by-condition interaction. Follow-up simple main
effects testing revealed that TD children were characterised by a significantly greater ERN amplitude than
Kristin R. Laurens; Page 11
ASz children [F(1, 46)=5.7, p=0.02], but that groups did not differ on CRN amplitude [F(1, 46)=0.005,
p=0.9]. Post-hoc examination of the distribution of ERN amplitude at FCz within the ASz and TD groups
indicated that the smallest ERN amplitude observed among the TD group was greater than that elicited in
nine (41%) of the 22 ASz children.
Pe and Pc amplitude. The two group (ASz, TD) x two condition (Pc, Pe) x three site (FCz, Cz, and
CPz) repeated measures ANOVA on peak positivity amplitude data demonstrated no group effect [F(1,
46)=2.3, p=0.1], and no effect of site, but a significant main effect of condition, with Pe amplitude
significantly greater than Pc amplitude. There were no interactions of group with condition or site.
ERN and CRN latency. The two group (ASz, TD) x two condition (CRN, ERN) x three site (FCz,
Cz, and CPz) repeated measures ANOVA on peak negativity latency data revealed no effect of group [F(1,
46)=0.2, p=0.7], but significant main effects of condition and of site, with ERN latency significantly
greater than CRN latency, and with latency increasing over more posterior sites (i.e., latency at FCz was
significantly shorter than at Cz and at CPz, and latency at Cz was significantly shorter than at CPz; all
p<0.001). These effects were modified by a significant site-by-condition interaction. Follow-up simple
main effects testing indicated that ERN latency was significantly longer than CRN latency at FCz only
[F(1, 46)=14.8, p<0.001]. There were no interactions of group with condition or site.
Pe and Pc latency. The two group (ASz, TD) x two condition (Pc, Pe) x three site (FCz, Cz, and CPz)
repeated measures ANOVA on peak positivity latency data revealed no effect of group [F(1, 46)=0.009,
p=0.9], but a significant main effect of condition, with Pe latency greater than Pc latency. This effect was
modified by a significant site-by-condition interaction. Follow-up simple main effects testing revealed
significantly greater Pe latency than Pc latency at Cz [F(1, 46)=11.2, p=0.002] and at CPz [F(1, 46)= 8.9,
p=0.005], but not at FCz. There were no interactions of group with condition or site.
The present study examined whether children aged 9-12 years presenting a triad of putative
antecedents of schizophrenia exhibited brain function abnormalities during error processing that are
Kristin R. Laurens; Page 12
characteristic of adults with schizophrenia. As hypothesised, ASz children were characterised by ERN
amplitude reduction relative to TD children, while Pe amplitude did not differ between groups. The
findings are similar to those reported previously in comparisons of adults with schizophrenia and healthy
individuals (23, 25-27). Based on current interpretations of the functional significance of early and late
error-related ERP components (43), the intact Pe response observed in ASz children implies that the
conscious evaluation of the error and/or its motivational significance is unaffected. However, earlier
processes associated with detection of the error and/or an affective response to the error, as indexed by
ERN amplitude, appear less efficient in ASz than TD children. The groups did not differ on CRN (or later
Pc) amplitude elicited on correct response trials, a finding divergent from the relative increase in CRN
amplitude previously reported to distinguish adults with schizophrenia from healthy individuals on some
error-inducing experimental tasks (24-27), although not for the Go/NoGo paradigm (22, 23). The absence
of a group difference in CRN amplitude indicates a lack of generalised impairment in response-related
function in ASz children. Instead, the functional brain abnormality in ASz children appears to be specific
to early error-related processes. No group differences were observed in the latency of any of the four
components examined (ERN, CRN, Pe, and Pc), implying that the timing of error-related processing did
not differ among ASz and TD children.
In the Go/NoGo paradigm employed, substantial task difficulty was conferred by perceptual similarity
of the “X” and “K” stimuli and by the strong pre-potent response tendency established using highly
probable Go trials. The necessity to maintain both speed and accuracy of responding was emphasised. In
spite of the physiological differences apparent between groups, ASz and TD children performed
comparably on measures of accuracy and reaction time, implying that ASz children were no more
impulsive in their responding than TD children. The mean percentage of errors elicited on NoGo trials
(i.e., response inhibition failures) was comparable to that obtained in previous studies of adults (23) and
children (37) employing Go/NoGo tasks, but the percentage of missed responses on Go trials in both
groups was relatively high. Children may have experienced difficulty sustaining attention consistently
Kristin R. Laurens; Page 13
throughout the task and/or maintaining speeded responses within the 1000 ms response window allowed. A
lack of group differences in the number of missed Go trials, and an absence of slowed response time in
ASz children, are inconsistent with results obtained in adults with schizophrenia relative to healthy
individuals on this task (22, 31). Such effects might emerge using a paradigm that titrates difficulty of
response inhibition and/or stimulus presentation speed according to the participant’s performance on
previous trials. Both ASz and TD children adjusted their behaviour in response to errors by slowing their
response on the subsequent trial, which is consistent with the lack of group difference in the Pe amplitude
that correlates with the magnitude of post-error slowing. Such adjustments are typically interpreted as
compensatory action taken to increase the likelihood of a correct response on the next trial (33).
Previously, we demonstrated that prevalence of the triad of putative antecedents was elevated among
children of African-Caribbean ancestry living in the UK, as is the prevalence of schizophrenia (6). The
present study provides further support for our hypothesis that children displaying the triad of putative
antecedents present elevated risk for schizophrenia (5). Only follow-up of ASz children will establish the
proportion that develop schizophrenia and the extent to which ERN amplitude reduction is a biomarker of
disease risk. Among ASz children there are likely three sub-groups: those who will develop schizophrenia,
those who will develop other spectrum disorders, and those who will not develop schizophrenia. For 41%
of ASz children, ERN amplitude was smaller than the lower limit of the amplitude distribution within the
TD group. Perhaps ERN amplitude reduction is so great in the sub-group of children developing
schizophrenia that it lowers the overall group mean sufficiently to distinguish the ASz children as a group
from the TD children. We also recognise that the developmental impairments we identified may not be
specific to schizophrenia. Follow-up could indicate that they are associated with a range of other
psychiatric or neurological conditions rather than schizophrenia only. Nevertheless, the antecedents were
identified from the literature on developmental impairments that precede schizophrenia, and the
associations we describe in this paper strengthen the possibility of a link to schizophrenia in particular.
Kristin R. Laurens; Page 14
Presently, however, we conceptualise the antecedents as risk markers for psychotic disorders rather than
necessarily the beginnings of the disorder itself.
The present study provides no evidence of a generalised deficit in self-monitoring of behaviour among
ASz children, but rather, suggests that these children are characterised by brain dysfunction specific to
early error processing. Previous research has demonstrated functional and structural abnormalities in the
purported generator of the ERN, the ACC (9-11), in clinically- or genetically-defined high-risk youth in
the years immediately preceding transition to psychosis (12, 13). The present results imply that the
functional ACC abnormalities might be present much earlier, by 9-12 years. Follow-up of these children
will afford an examination of the stability of the ERN amplitude reduction, and critically, will allow
determination of whether ERN amplitude reduction represents a general vulnerability to psychopathology
or confers specific risk for schizophrenia. Given that functional connectivity of the ACC develops well
into young adulthood (44), there is an exciting prospect of remediating function with early intervention in
childhood, and potentially preventing the development of psychosis in children who present the triad of
putative antecedents plus the specific ERN amplitude reduction.
Kristin R. Laurens; Page 15
The authors thank the children and caregivers who participated in the study, and Alexis Cullen, Hannah
Dickson, Christina Theodoridou, Helen Stockill, Petra Gronholm, Janine Friedlander, Freya Whitehead,
Shahir Uddin, Aynur Yalcin, Navin Mayani, Naomi Harris, Mariam Darwish, Shamsil Hazarika, Susan
Sadek, and Amie Doidge for their assistance in conducting the research assessments. We also thank Jeffrey
Dalton for programming the Go/NoGo task paradigm, and Dr. Dominic ffytche for use of the Neuroscan
Synamps equipment. This research was supported by funding awarded to KRL from a National Institute
for Health Research (NIHR) Career Development Fellowship, a Bial Foundation Research Grant, a
NARSAD Young Investigator Award, and the British Medical Association Margaret Temple Award for
schizophrenia research. SH holds a Royal Society Wolfson Merit Award. EAT gratefully acknowledges
the financial support of the UK Medical Research Council. All authors are affiliated with the NIHR
Specialist Biomedical Research Centre (BRC) for Mental Health at the South London and Maudsley NHS
Foundation Trust and Institute of Psychiatry, King’s College London, United Kingdom; GLM and PLAS
were supported financially by BRC funds.
This work is produced by under the terms of a Career Development research training Fellowship
awarded by the NIHR. The views expressed in this publication are those of the authors and not necessarily
those of the NHS, the NIHR, or the UK Department of Health.
The authors report no biomedical financial interests or potential conflicts of interest.