electrical brain responses in language-impaired children reveal grammar-specific deficits ·...

Electrical Brain Responses in Language-ImpairedChildren Reveal Grammar-Specific DeficitsElisabeth Fonteneau¤, Heather K. J. van der Lely*

UCL Centre for Developmental Language Disorders and Cognitive Neuroscience, University College London, London, United Kingdom

Abstract

Background: Scientific and public fascination with human language have included intensive scrutiny of language disordersas a new window onto the biological foundations of language and its evolutionary origins. Specific language impairment(SLI), which affects over 7% of children, is one such disorder. SLI has received robust scientific attention, in part because ofits recent linkage to a specific gene and loci on chromosomes and in part because of the prevailing question regarding thescope of its language impairment: Does the disorder impact the general ability to segment and process language or aspecific ability to compute grammar? Here we provide novel electrophysiological data showing a domain-specific deficitwithin the grammar of language that has been hitherto undetectable through behavioural data alone.

Methods and Findings: We presented participants with Grammatical(G)-SLI, age-matched controls, and younger child andadult controls, with questions containing syntactic violations and sentences containing semantic violations. Electrophys-iological brain responses revealed a selective impairment to only neural circuitry that is specific to grammatical processingin G-SLI. Furthermore, the participants with G-SLI appeared to be partially compensating for their syntactic deficit by usingneural circuitry associated with semantic processing and all non-grammar-specific and low-level auditory neural responseswere normal.

Conclusions: The findings indicate that grammatical neural circuitry underlying language is a developmentally uniquesystem in the functional architecture of the brain, and this complex higher cognitive system can be selectively impaired. Thefindings advance fundamental understanding about how cognitive systems develop and all human language is representedand processed in the brain.

Citation: Fonteneau E, van der Lely HKJ (2008) Electrical Brain Responses in Language-Impaired Children Reveal Grammar-Specific Deficits. PLoS ONE 3(3): e1832.doi:10.1371/journal.pone.0001832

Editor: Steven Pinker, Harvard University, United States of America

Received February 7, 2008; Accepted February 16, 2008; Published March 12, 2008

Copyright: � 2008 Fonteneau, van der Lely. This is an open-access article distributed under the terms of the Creative Commons Attribution License, whichpermits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This research and both authors were supported by a Wellcome Trust University Award (Grant No: 063713) to Heather van der Lely.

Competing Interests: The authors have declared that no competing interests exist.

* E-mail: [email protected]

¤ Current address: Department of Psychology, Goldsmiths College, University of London, London, United Kingdom

Introduction

Grammar is an exclusively human and complex ability[1,2], yet

by age 3 years, most children produce grammatically correct

sentences. We still have little understanding of the biological and

evolutionary changes that enable humans to do this or the changes

that prevent them doing this normally as is found in children with

SLI, who continue to make grammatical errors, sometimes into

adulthood[3].

SLI variably affects the acquisition of subsystems or ‘‘compo-

nents’’ of language[3]; that is both grammatical components such

as syntax (the structural rules combining words into sentences);

morphology (the rules combining words or parts of words into new

words, e.g., jump+ed); and phonology (the rules for combining

sounds into words); word-storage (vocabulary) and other aspects of

the conversational (discourse) and social use (pragmatics) of natural

language.

The discovery of the subgroup Grammatical(G)-SLI provides

rare insight into the neural systems in the human brain and thus its

nature and origins are hotly debated. The controversy surround-

ing G-SLI focuses on whether it results from a domain-general

deficit in auditory processing speed or capacity[4,5], or whether it

results from a developmentally specialised grammatical subsystem

in the brain that can be selectively impaired[3]. Preliminary

evidence from G-SLI reveals familiar clustering of language

impairment that is consistent with an autosomal dominant

inheritance[6]. However the nature of the language impairment

in family members varies, suggesting a more complex inheri-

tance[6]. The G-SLI impairment is life-long, and affects

grammatical rules underlying structures in syntax, morphology,

and phonology[3]. G-SLI teenagers make errors that normally-

developing children rarely make after 5 years of age. For example,

they make errors in knowing who him or himself refers to in the

sentence Mowgli said Baloo was tickling him/himself, or produce

errors when asking questions (Who __ Joe see someone?)[3]. In

contrast, individuals with G-SLI show good understanding of

social and world knowledge when they communicate[7], do not

show any consistent auditory deficits[8] (see supporting Data S1,

Table S1 and Figures S1 and S2) and are of average

intelligence[3,7]. However, behavioural data alone cannot tell us

whether the deficit is restricted to only grammar or impacts on

more general language–related processing.

PLoS ONE | www.plosone.org 1 March 2008 | Volume 3 | Issue 3 | e1832

Electrophysiological measurements provide direct assessment of

brain activity and have the necessary temporal resolution to

distinguish between the two hypotheses: generally slow auditory

and language mechanisms[5] versus a selective impairment in

grammatical mechanisms alongside normal functioning in other

language mechanisms[3]. Such residual normality is claimed not

to exist[9]. Specifically, electrophysiological, event-related mea-

surements can differentiate neural systems that appear to be

automatic, fast, and specific to only grammatical (syntactic)

processing (‘‘Early Left-Anterior Negative electrical brain response

around 100 ms (ELAN)[10], from systems associated with

language processing but which are not grammar-specific, such as

an anterior or central positive electrical brain response around

600 ms (P600), often associated with structural syntactic re-

analysis of sentences[10,11] and a posterior negative electrical

response around 400 ms (N400) associated with semantic

processing[12]. Importantly, whereas the P600 is elicited by a

range of different grammatical violations[13] as well as semantic

violations[14], the ELAN is only elicited by structural grammatical

violations[10]. These differences allow us to make clear predictions

for G-SLI. Whereas domain-general hypotheses predict that most

if not all ERP language-related components will be affected (e.g.,

delayed latency), the domain-specific hypothesis predicts that only

the grammar-specific component (ELAN), that reflects pure

syntactic structure[10], will be atypical or absent.

To investigate these alternative hypotheses we recorded

electrophysiological time-locked, event-related brain potentials

(ERPs) in 18 participants ages 10 to 21 years, age-matched

controls, and younger child and adult controls listening to

questions containing a syntactic violation (Experiment 1) (see

materials and methods). The particular syntactic violation we were

interested in concerns structural ‘‘syntactic dependencies’’ such as

those that occur between a question word (who, what) and the

word, that in declarative sentences follows the verb, but typically is

absent in questions (see supporting Methods S1). Such syntactic

dependencies make sentences such as ‘‘Who did Joe see someone?’’

ungrammatical, but ‘‘Who did Joe see ?’’ and ‘‘Joe saw someone’’

grammatical. We hypothesised that G-SLI children’s syntactic

impairment lies in the computational grammatical system

underlying such syntactic dependencies[3].

Results

First, we analysed ERPs in the time window 100–300 ms

(Figure 1) to assess participants’ automatic brain responses to the

structural syntactic violations (see methods, and supporting

Methods S1 and supporting data in Table S1, and Figure S4).

The ERPs for the G-SLI group were compared with those of the

age and language controls. The overall ANOVA revealed a

group6condition6region of interest (ROI) interaction (F16,

424 = 1.82, p,.027). The syntactic violation elicited an Early Left

Anterior Negativity (ELAN) in the age and language controls,

which was absent in the G-SLI group (Figure 1a). Individual

subject analysis revealed that whereas almost all the age control

subjects revealed an ELAN, the G-SLI subjects did not (Figure 1c).

A similar brain response, distributed equally on anterior sites was

found in our adults (Figure 1b) and, previously, has been elicited in

young children, some under 3 years old[15,16]. The ELAN is

considered to be a brain correlate of automatic syntactic structural

building and processing[10] and thus, core to the syntactic

system[2]. Moreover, the ELAN’s sensitivity appears domain-

specific to syntactic structure. It is insensitive to task demands or

violation frequency that incur other cognitive processes[17,18].

Thus, our pattern of results is exactly what would be expected if G-

SLI children were impaired in a specific mechanism underlying

grammar.

To test the hypothesis that our G-SLI children were ‘‘slow

processors’’[19], thereby producing an ELAN with a delayed

latency, we analysed ERPs from the following 300–500 ms time

window. We found a significant negativity with a posterior

distribution, rather than an anterior distribution, for the G-SLI

group, but not for the control groups, or the adult subjects

(Figure 2). Individual subject analysis reveals the consistency of this

negativity across the G-SLI children but not their age matched

controls (Figure 2c) (see also supporting Table S2 and Figure S5).

This electrical response resembles the component known as the

Figure 1. Syntactic dependency component (ELAN, 100–300 ms -grey area) elicited for the syntactic violation. a. ERPwaveforms for the groups from F5 (left frontal) electrode. b. Scalpdistribution of differences between the violation minus controlsentences for each group. The syntactic violation elicited a negativitydistributed on the left hemisphere for the age controls (Condition6He-misphere: F1,17 = 10.16, p,.005), and the language controls (Condi-tion6Hemisphere: F1,19 = 11.12, p,.003; Condition6Caudality6Hemi-sphere: F2,38 = 3.55, p,.05), with a maximum of difference on theanterior left sites for both groups (p,.006, p,.05 respectively). Thisnegativity is equally distributed on anterior sites for the adults(Condition6Caudality: F2,38 = 10.17, p,.001; anterior central p,.001).No effect was significant for the G-SLI children (F,1). c. Effect sizes forindividual G-SLI children and their age controls (numbers correspond tomatched individuals with increasing numbers corresponding toincreasing age). Effect size: mean amplitude difference (violation minuscontrol) in the Anterior Left ROI in the 100–300 ms time window. Weplot Negativity upward.doi:10.1371/journal.pone.0001832.g001

Brain Responses Reveal Deficit


N400, that is associated with semantic processing[12], but not

syntactic processing. Interestingly, syntactic violations have also

elicited an N400 in adults with an acquired grammatical disorder

(aphasia)[20]. Thus, it appears from this study that the G-SLI

children were not merely delayed in their response, but were

compensating for their impairment in structural syntactic

dependencies by using a different neural circuitry associated with

semantic mechanisms.

To assess whether the G-SLI children’s deficit extends to other

neural correlates that are elicited by the same syntactic violations

to the same words, we analysed the time window between 800–

1000 ms (Figure 3). These neural correlates are associated with

(secondary) re-analysis of the structure, rather than initial

structural syntactic processing[10,11,13]. Analysis revealed a

significant positive electrical response in all groups (overall

ANOVA: Condition6ROI (F8, 424 = 21.81, p,.0001, but no main

effect of Group, (F2,53 = 1.33, p..27), or interaction with Group

(p..68). This response, in this time window[15,16] is characteristic

of the P600 component, associated with such re-analysis or

syntactic integration. The brain maps (Figure 3b) show that it is

distributed on the anterior regions of the scalp for the age and

language controls, and is equally distributed on the anterior sites

for the adults. For the G-SLI group it is also significant on both

anterior sites but, interestingly, shows maximum amplitude on the

right. This time, individual analysis reveals in both individual G-

SLI and age control children a consistent positive electrical

response (Figure 3c) (see also supporting Figure S6). This frontal

distribution (cf. the centroparietally distributed P600[21]) is

commensurate with previous research in adults where, as in this

study, the sentence structure at the point of measurement is

unexpected, rather than ungrammatical, per se[11,13,21]. Further,

in contrast to the ELAN which appears to be domain-specific, this

frontal P600 is modulated by more general cognitive process-

es[22,23], and therefore is likely to reflect domain-general

processes. Our results, showing dissociation between the ELAN

(missing) and P600 (normal) when processing the same word in a

sentence in the G-SLI individuals strongly indicate that these two

components reflect different computations in syntactic processing.

Whereas fast, automatic grammatical structure processing is

missing, later sentence analysis is normal. Thus such dissociation

is found not only in the mature adult system[10] and patients with

lesions[20], but also in developmental disorders.

To investigate the possibility that brain responses to semantic

processing in G-SLI were impaired, in Experiment 2 (see materials

and methods) we investigated brain responses to sentences with

semantic violations (*Barbie bakes the people in the kitchen). We

analysed responses to these semantic violations in the time window

between 300–500 ms (Figure 4) (see also supporting Table S3).

Overall ANOVA revealed a significant effect of group

(F2,52 = 3.56, p = .035) but no significant interactions with this

factor. The electrical responses in the control groups and the G-

SLI participants (Figure 4a) were characteristic of an N400,

associated with the brain’s detection of semantic anomalies[12].

The group effect was accounted for by differences in the

distribution of the N400 in the younger language controls, where

we recorded the maximal negativity in the right hemisphere

(Figure 4b). In contrast, for the G-SLI children, like the age

controls, the N400 was distributed bilaterally in the posterior areas

(Figure 4b). Moreover, this N400 is strongly consistent across

individual G-SLI children and their age controls (Figure. 4c) (see

also supporting Figure S7). Our findings showing differences in the

distribution of the N400 according to age concur with previous

research[16].

To ensure that the N400 was not elicited later in the G-SLI

children due to slow processing, we analysed the peak latency of

this response over the posterior areas for the G-SLI children (mean

latency 379 6 20 ms) and the age controls (mean latency

345620 ms.). ANOVA did not reveal any significant differences

between groups (F1,34 = 1.39, p = .24) nor interactions with group

(F,1). Therefore, neural responses elicited by semantic violations

in the G-SLI children and age control children revealed a similar

distribution and occurred at a similar millisecond time-point after

hearing the beginning of the word.

Discussion

Overall, the G-SLI subgroup indicates normal semantic

processing of language and normal auditory processing speed.

Such evidence challenges the view that generally slow or impaired

auditory processing causes and maintains grammatical impair-

ment. Note this does not militate against different forms of SLI

possibly having different biological and neural instantiations and

Figure 2. Semantic component (N400, 300–500 ms –grey area)elicited for the syntactic violation for the G-SLI group. a. ERPsfrom three posterior electrodes (P3, left P4 right hemisphere and Pzmidline) for the G-SLI and Age Control groups. b. Scalp distribution ofthe differences between the violation minus control sentences. TheERPs from the G-SLI participants elicited a negativity distributed on theposterior area for the syntactic violation (Condition6Caudality:F2,34 = 3.08, p,.05). Note the raw data suggested a lateralisation ofthe N400 (Condition6Caudality6Laterality F2,34 = 3.75, p = .03) whereasthe normalised data indicated a non significant interaction (F2,34 = 1.93,p = .15). No other group showed this result for the 300–500 ms timewindow. c. Effect sizes for individual G-SLI children and their agecontrols. Effect size: mean amplitude difference (violation minuscontrol) in the Posterior Central ROI within the 300–500 ms temporalwindow. We plot Negativity upward.doi:10.1371/journal.pone.0001832.g002



different developmental outcomes. However, the G-SLI electro-

physiological signature reveals a selective developmental deficit in

neural circuitry. This neural circuitry is linked to particular aspects

of grammar, representing structural syntactic dependency rela-

tions, whose evolution is crucial, and possibly unique to the human

language faculty[1,2]. The results argue for grammar being a

highly specific, specialised subsystem in the human brain and a

particular developmental pathway to this exclusive neural system.

The findings indicate that developmental higher cognitive deficits

can be selective, which has significant implications for the

diagnosis and treatment of SLI. For G-SLI children and perhaps

other SLI subgroups too, a relative strength in semantic processing

could be targeted to help compensate for their syntactic

impairment. The findings provide basic knowledge about the

functional architecture of the brain and the development of

uniquely human and specialised higher cognitive systems.

Materials and Methods

MethodsSubjects. We recorded four groups: 18 G-SLI (mean age

14.3, 10–21 years old, 13 males, for selection criteria see[3]), 18

age controls (mean age 14.3, 10–22 years old, 13 males matched

with the G-SLI participants on age, sex, laterality and non verbal

IQ[24]), 20 language controls (mean age 8.1, 7–9 years old, 11

males, matched with the G-SLI participants on receptive

vocabulary[25]) and 20 students from UCL (mean age 23.5, 18–

38 years old, 8 males).

Experimental Design. In this study we manipulated the

animacy property of the first noun following a verb so that in

Experiment 1 we created a syntactic violation, and Experiment 2 a

Figure 3. Reanalysis component (P600, 800–1000 ms–greyarea) elicited for the syntactic violation. a. ERP waveforms, foreach group from Fz (frontal electrode). b. Scalp distribution of thedifferences between the violation minus control sentences for eachgroup. The syntactic violation elicited a positivity distributed on anteriorregions with a maximum on the right sites for the G-SLI participants(Condition6Caudality: F2,34 = 15.64, p,.0001, Condition6Caudality6He-misphere F2,34 = 6.39, p,.005); and on anterior regions for the agecontrols (Condition6Caudality: F2,34 = 8.93, p,.003), language controls(Condition6Caudality: F2,38 = 17.54, p,.001), and adults (Condition6Caudality: F2,38 = 9.61, p,.003). c. Effect sizes for individual G-SLIchildren and their age controls. Effect size: mean amplitude difference(violation minus control) in the Anterior Right ROI within the 800–1000 ms temporal window. We plot Negativity upward.doi:10.1371/journal.pone.0001832.g003

Figure 4. Semantic component (N400, 300–500 ms) elicited forthe semantic violation. a. ERP waveforms for each group from twoposterior electrodes (P3, left and P4 right hemisphere). b. Scalpdistribution of the differences between the violation minus the controlsentences for each group. The semantic violation elicited a posteriornegativity for the age controls (Condition6Caudality: F2,34 = 3.72,p,.05) and also the G-SLI group (Condition6Caudality: F2,34 = 7.15,p,.001). This negativity was maximal on the right hemisphere for thelanguage controls (Condition6Hemisphere: F1,18 = 6.92, p,.01), and onthe left posterior sites for the adults (Condition6Caudality: F2,38 = 6.07,p,.01, Condition6Hemisphere: F1,19 = 10.69, p,.001). Note that theN400 effect started as early as 100 ms for the G-SLI, age and languagecontrols. c. Effect sizes for individual G-SLI children and their agecontrols. Effect size: mean amplitude difference (violation minuscontrol) in the 3 Posterior ROIs within the 300–500 ms temporalwindow. We plot Negativity upward.doi:10.1371/journal.pone.0001832.g004



semantic violation. Crucially, the syntactic violation relied on a

structural syntactic dependency between two non-adjacent words

in the sentence, whereas the semantic violation relied on purely

lexical semantic restrictions of the preceding verb. Note,

technically the syntactic violation was an ‘‘unexpectancy’’ as the

following preposition rendered the sentence grammatical.

However, pre-testing of the sentences (see below) indicted, that

at the critical word to which our EEG recordings were time-

locked, the listener would perceive the word as a violation. What is

at issue here, is not whether the word is a violation or

unexpectancy, but identifying the different functional neural

circuitry that is used to detect such a violation/unexpectancy.

Experiment 1: Syntactic processing. Here we manipulated

the animacy properties of the wh-word (who [+animate] vs. what

[2animate]) in object questions in relation to those of the noun

(clown [+animate] vs. ball [2animate]) following the verb. We

constructed questions where the wh-word-noun pair either matched

(syntactic violation) or mismatched (control) (see materials). For

questions that contained the animacy match (syntactic violation), a

preposition and NP followed the critical noun, making the overall

question ungrammatical. For the mismatch pair (control) following

the critical noun we added only a preposition, making the overall

question grammatical. In doing so, we aimed to focus the

participant’s attention to the lexical animacy properties of the

wh-word-noun pair: e.g., Who did Barbie push the clown into the wall?

(animacy match- syntactic violation), Who did Barbie push the ballinto? (animacy mismatch- control questions). We computed and

analysed ERPs from the presentation of the nouns (clown/ball)in the object position. We aimed to identify which neural (and

language) systems are incurred when a subject encounters the

syntactic violation nouns, rather than the fact that they might later

consciously notice the animacy match-ungrammaticality

association.

Experiment 2: Semantic processing. Using declarative

sentences, we manipulated the animacy property of the noun

following the verb, in relation to the verb’s semantic selection-

restrictions; e.g., bread [-animate] is a possible noun following the

verb, bake, (Barbie bakes the bread in the kitchen–control sentence) but

people [+animate] is not (Barbie bakes the people in the kitchen–semantic

violation).

Electrophysiological recording and data analysisWe recorded ERPs using the EGI system (128 channels, 250 Hz

sampling rate, 0.1–100 Hz). ERPs were re-referenced according to

the average reference. Prior to off-line averaging, all single trial

waveforms with artefacts were rejected. For Experiment 1,

syntactic processing, behavioural responses were ignored because

we expected the G-SLI participants to make more errors

compared to controls. For Experiment 2, semantic processing,

ERPs were averaged from correct behavioural responses only. We

rejected one outlier subject from the language control group based

on his behavioural responses from Experiment 2. For the syntactic

experiment the number of averaged trials did not show any group

differences (33 = language control, 34 = G-SLI, 40 = age control,

F2,53 = 2.14, p.0.12). For the semantic experiment, a group effect

(F2,52 = 10.71, p,0.001) was due to fewer trials being available in

the average for the language control (25) compared to the G-SLI

(32) and age control group (38).

ERPs (1000 ms epochs) were quantified by mean amplitude

measures after the onset of the critical word (direct object noun)

for different time windows (TW): the ELAN from 100 to 300 ms,

the N400 from 300 to 500 ms and the P600 from 800 to 1000 ms

relative to the 100 ms prestimulus baseline. Note, we also analysed

the time window from 0 to 100 ms, but found no significant effect

for experimental condition or interactions with topographical

factors (but see Table S2–S3). Subsequent overall ANOVAs with

group (3: G-SLI, age and language controls), condition (2) and

ROI (9: the head was divided into nine Regions Of Interest, and

for each we computed a single mean amplitude from 6 to 11

electrodes, see supporting Figure S3). We then carried out further

ANOVA after rescaling the data[26] to assess differences in scalp

topography for each population. Thus, separate ANOVAs

(Condition (2): violation, control; Caudality (3): anterior, median,

posterior; Hemisphere (2): left, right) for each group as well as the

adults were carried out. We report significant effects only when the

raw data and the normalized data were both significant. The

Greenhouse-Geisser correction was applied to all analyses when

evaluating effects with more than one degree of freedom in the

numerator.

Ethical approval was granted from the UCL/UCLH ethics

committee (01/0150). Signed consent was obtained from partic-

ipants or their parents/guardians.

Supporting Information

Data S1 Electrical brain responses to auditory processing in

language impaired children

Found at: doi:10.1371/journal.pone.0001832.s001 (0.05 MB

DOC)

Table S1 Mean latency and amplitude for the N100, P200, and

P300 components for the G-SLI and age matched control groups.

Lat = Latency; Amp = Amplitude in mV; Mean SD = Mean

average Standard Deviation


DOC)

Table S2 Experiment 1: Syntactic processing: Mean amplitude

differences (violation minus control) for the syntactic task within

the different windows of interest (0–100 ms, 100–300 ms, 300–

500 ms and 800–1000 ms) for each region of interest (ROI), the

standard error is shown in italic. We performed a simple ANOVA

for each region of interest separately: *** p,.001; ** p,.01; *

p,.05. AC: Age Controls, LC: Language Controls.


DOC)

Table S3 Experiment 2 Semantic processing: Mean amplitude

differences (violation minus control) for the semantic task within

the different windows of interest (0–100 ms, 100–300 ms, 300–

500 ms and 800–1000 ms) for each region of interest (ROI), the

standard error is shown in italic. We performed a simple ANOVA

for each region of interest separately: *** p,.001; ** p,.01; *

p,.05. AC: Age Controls, LC: Language Controls.


DOC)

Figure S1 Superimposed plot of AEPs for the target and

standard tones for the G-SLI and Age control groups.

Found at: doi:10.1371/journal.pone.0001832.s005 (0.19 MB JPG)

Figure S2 Mean average map for the periods of interest for the

N100, P200 and P300 for the G-SLI and Age control groups.


Figure S3 9 Regions of Interest and the corresponding electrode

sites.


Figure S4 Syntactic processing: Effect sizes for individual

subjects for the adult and language control (LC) groups in the

100–300 ms temporal window (ELAN). Effect size: mean



amplitude differences (violation minus control) in the Anterior Left

ROI. Negativity is plotted upwards.


PDF)



300–500 ms temporal window for the syntactic task. Effect size:

mean amplitude differences (violation minus control) in the

Posterior Central ROI. Negativity is plotted upwards.


DOC)



800–1000 ms temporal window (P600). Effect size: mean ampli-

tude differences (violation minus control) in the Anterior Right

ROI. Negativity is plotted upwards.


PDF)

Figure S7 Semantic processing: Effect sizes for individual


300–500 ms temporal window (N400). Effect size: mean ampli-

tude differences (violation minus control) in the 3 Posterior ROIs.

Negativity is plotted upwards.


PDF)

Methods S1


DOC)

Acknowledgments

We thank the children and their families and the adults who participated in

this study, the schools (in particular, Dawn House, Moor House, Radlett

and Southgate) and the Speech and Language Therapists, and thank

Nichola Gallon for help preparing the linguistic material and Chloe

Marshall for discussions and comments.

Author Contributions

Conceived and designed the experiments: Hv. Performed the experiments:

EF. Analyzed the data: EF. Wrote the paper: Hv. Other: Designed the

auditory control experiment: EF. Prepared the linguistic materials with

help from Nichola Gallon, DLDCN Centre: EF. Conducted the pre-test

experiment for adults: EF. Wrote the first draft of the methods, summarised

the results and auditory processing experiment sections, figure captions and

constructed the Figures: EF. Commented on and discussed the drafts of the

manuscript: EF. Designed and conducted the pre-test experiment for

children and teenagers: Hv.

References

1. Pinker S, Jackendoff R (2005) The faculty of language: what’s special about it?Cognition 95: 201–236.

2. Hauser MD, Chomsky N, Fitch WT (2002) The faculty of language: What is it,who has it, and how did it evolve? Science 298: 1569–1579.

3. van der Lely HKJ (2005) Domain-specific cognitive systems: Insight from

Grammatical specific language impairment. Trends in Cognitive Sciences 9:53–59.

4. Leonard L (1998) Children with specific language impairment. Cambridge-Mass.: MIT Press.

5. Joanisse M, Seidenberg M (1998) Specific language impairment: a deficit in

grammar or processing? Trends in Cognitive Sciences 2: 240–247.6. van der Lely HKJ, Stollwerck L (1996) A grammatical-specific language

impairment in children: An autosomal dominant inheritance? Brain andLanguage 52: 484–504.

7. van der Lely HKJ, Rosen S, McClelland A (1998) Evidence for a grammar-specific deficit in children. Current Biology 8: 1253–1258.

8. van der Lely HKJ, Rosen S, Adlard A (2004) Grammatical language impairment

and the specificity of cognitive domains: Relations between auditory andlanguage abilities. Cognition 94: 167–183.

9. Thomas M, Karmiloff-Smith A (2002) Are developmental disorders like cases ofadult brain damage? Implications from connectionist modelling. Behav Brain Sci

25: 727–750; discussion 750–787.

10. Friederici AD (2002) Toward a neural basis of auditory sentence processing.Trends in Cognitive Sciences 6: 78–84.

11. Kaan E, Swaab TY (2003) Repair, revision, and complexity in syntactic analysis:An electrophysiological differentiation. Journal of Cognitive Neuroscience 15:

98–110.

12. Kutas M, Hillyard SA (1980) Reading senseless sentences: brain potentials reflectsemantic incongruity. Science 207: 203–205.

13. Hagoort P, Brown C, Osterhout L (1999) The neurocognition of syntacticprocessing. (Chap 9). In: Brown C, Hagoort P, eds (1999) The neurocognition of

language. Oxford: Oxford University Press. pp 273–316.14. Munte T, Heinze H, Matzke M, Wieringa B, Johannes S (1998) Brain potentials

and syntactic violations revisited: no evidence for specificity of the syntactic

positive shift. Neuropsychologia 36: 217–226.

15. Oberecker R, Friedrich M, Friederici AD (1995) Neural correlates of syntactic

processing in two-year-olds. Journal of Cognitive Neuroscience 17: 1667–1678.

16. Hahne A, Eckstein K, Friederici AD (2004) Brain signatures of syntactic and

semantic processes during children’s language development. Journal of

Cognitive Neuroscience 16: 1302–1318.

17. Hahne A, Friederici AD (1999) Electrophysiological evidence for two steps in

syntactic analysis. Early automatic and late controlled processes. J Cogn

Neurosci 11: 194–205.

18. Hahne A, Friederici AD (2002) Differential task effects on semantic and syntactic

processes as revealed by ERPs. Cognitive Brain Research 13: 339–356.

19. Montgomery J (2002) Examining the nature of lexical processing in children

with specific language impairment: Temporal processing or processing capacity

deficit? Applied Psycholinguistics 23: 447–470.

20. Hagoort P, Wassenaar M, Brown CM (2003) Real-time semantic compensation

in patients with agrammatic comprehension: Electrophysiological evidence for

multiple-route plasticity. Proceedings of the National Academy of Sciences 100:

4340–4345.

21. Friederici AD, Hahne A, Saddy D (2002) Distinct neurophysiological patterns

reflecting aspects of syntactic complexity and syntactic repair. J Psycholinguist

Res 31: 45–63.

22. Coulson S, King JW, Kutas M (1998) Expect the Unexpected: Event-related

Brain Response to Morphosyntactic Violations. Language and Cognitive

Processes 13: 21–58.

23. Ainsworth-Darnell K, Shulman H, Boland J (1998) Dissociating brain responses

to syntactic and semantic anomalies: evidence from event-related potentials.

Journal of Memory and Language 38: 112–130.

24. Raven JC, Court JH, Raven J, eds (1986) Manual for Raven’s Progressive

Matrices and Vocabulary Scales (Section 2)-Coloured Progressive Matrices.

London: Lewis.

25. Dunn L, Dunn L, Whetton C, Burley J (1997) The British Picture Vocabulary

Scale: Second Edition. Windsor, UK: NFER, Nelson.

26. McCarthy G, Wood CC (1985) Scalp distributions of event-related potentials: an

ambiguity associated with analysis of variance models. Electroencephalography

and Clinical Neurophysiology 62: 203–208.



Language Controls (LC)-5

-4

-3

-2

-1

0

1

2

3

41 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Individual subjects

Size

eff

ect (

μV

) for

the

Ant

erio

r Le

ft R

OI

Adults-5

-4

-3

-2

-1

0

1

2

3

41 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Individual subjects

Size

eff

ect (

μV

) for

the

Ant

erio

r Le

ft R

OI


-4

-3

-2

-1

0

1

2

3

4

5

6

71 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Individual subjects

Size

eff

ect (

mV

) for

the

Post

erio

r C

entr

al R

OI

Adults-5

-4

-3

-2

-1

0

1

2

3

4

5

6

71 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Individual subjects

Size

eff

ect (

μV

) for

the

Post

erio

r C

entr

al R

OI

Language Controls-3

-2

-1

0

1

2

3

4

5

6

7

8

9

101 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Individual subjects

Size

eff

ect (

μV

) for

the

Ant

erio

r R

ight

RO

I

Adults-3

-2

-1

0

1

2

3

4

5

6

7

8

9

101 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Individual subjects

Size

eff

ect (

μV

) for

the

Ant

erio

r R

ight

RO

I


-6

-5

-4

-3

-2

-1

0

1

2

31 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Individual subjects

Size

eff

ect (

μV

) for

the

3 Po

ster

ior

RO

Is

Adults-7

-6

-5

-4

-3

-2

-1

0

1

2

31 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Individual subjects

Size

eff

ect (

μV

) for

the

3 Po

ster

ior

RO

Is

Anterior Median PosteriorLeft Central Right Left Central Right Left Central Right

________________________ ________________________ ________________________0-100 msAdults -0.04 0.12 -0.07 0.11 -0.08 0.11 -0.01 0.10 0.23 0.10 0.10 0.11 -0.04 0.11 0.07 0.13 0.12 0.12

AC -0.33 0.27 -0.47 0.24 0.15 0.23 -0.34 0.17 -0.44 0.22 -0.01 0.21 -0.22 0.25 -0.23 0.24 0.20 0.22

LC -0.47 0.28 0.06 0.37 0.60 0.35 -0.18 0.29 0.32 0.34 0.97 0.22 ** -0.78 0.48 -1.00 0.38 -0.25 0.26

G-SLI 0.14 0.23 0.08 0.29 0.32 0.22 0.08 0.22 0.28 0.22 0.12 0.19 -0.19 0.29 -0.18 0.26 -0.04 0.19

100-300 msAdults -0.38 0.13 * -0.48 0.14 * -0.24 0.14 -0.03 0.11 0.23 0.14 0.34 0.13 0.27 0.13 0.56 0.14 ** 0.47 0.13 **

AC -0.96 0.37 *** -0.69 0.36 * 0.34 0.37 -0.36 0.20 0.26 0.27 0.65 0.25 0.00 0.27 0.29 0.45 0.88 0.35*

LC -0.73 0.45 * 0.64 0.52 0.81 0.39 * -0.71 0.31 * 0.22 0.35 1.00 0.31 ** -0.85 0.51 -1.53 0.49 * -0.69 0.41

G-SLI 0.13 0.29 0.12 0.32 0.37 0.33 0.35 0.27 -0.55 0.34 -0.09 0.28 -0.01 0.29 -0.75 0.29 * -0.51 0.34

300-500 msAdults -0.30 0.17 -0.37 0.20 -0.08 0.19 -0.10 0.13 0.08 0.16 0.25 0.13 0.15 0.16 0.34 0.20 0.30 0.17

AC -0.56 0.41 0.62 0.47 0.15 0.45 -0.27 0.29 0.08 0.24 0.59 0.23 * 0.23 0.38 0.65 0.50 1.22 0.43 *

LC -0.75 0.50 0.11 0.60 -0.18 0.53 -0.52 0.39 0.66 0.35 0.52 0.38 -0.12 0.52 -0.32 0.50 -0.81 0.46

G-SLI 0.02 0.44 0.24 0.48 0.74 0.38 * 0.43 0.35 -0.20 0.41 0.17 0.31 -0.22 0.31 -1.07 0.40 * -0.74 0.38 *

800-1000 msAdults 0.70 0.22 * 0.35 0.21 0.78 0.22 ** 0.11 0.19 -0.26 0.26 0.29 0.19 -1.02 0.23 *** -1.51 0.27 *** -0.82 0.25 **

AC 1.63 0.69 * 1.09 0.71 2.32 0.59 ** 0.91 0.38 * 0.53 0.41 1.60 0.45 * -1.48 0.48 * -2.34 0.66 ** -0.93 0.58 *

LC 2.93 0.65 ** 2.15 0.69 * 2.02 0.49 ** 0.77 0.37 0.04 0.41 0.39 0.43 -1.54 0.52 * -3.01 0.56 ** -3.02 0.56 **

G-SLI 1.33 0.58 1.14 0.78 2.47 0.41 *** 1.22 0.40 * -0.74 0.61 0.67 0.40 -1.25 0.44 * -2.92 0.67 ** -2.24 0.55 **

Anterior Median PosteriorLeft Central Right Left Central Right Left Central Right

________________________ ________________________ ________________________0-100 msAdults 0.22 0.20 0.29 0.23 0.32 0.13 0.08 0.09 0.12 0.13 0.12 0.09 -0.35 0.14 -0.43 0.15 -0.28 0.11

AC 0.10 0.31 0.05 0.29 0.04 0.21 -0.09 0.20 -0.32 0.21 -0.06 0.15 0.00 0.21 -0.24 0.34 -0.12 0.28

LC -0.04 0.47 -1.01 0.54 -1.03 0.42 0.50 0.25 -0.46 0.39 -0.70 0.33 * 0.84 0.52 0.28 0.53 -0.14 0.46

G-SLI 0.57 0.30 0.82 0.51 0.66 0.46 0.10 0.33 0.00 0.29 -0.26 0.18 -0.67 0.37 -0.91 0.54 -0.98 0.50

100-300 msAdults 0.28 0.18 0.28 0.17 0.21 0.14 0.02 0.11 0.00 0.13 0.10 0.08 -0.25 0.19 -0.38 0.23 -0.12 0.17

AC 0.15 0.41 0.86 0.30 * 0.93 0.23 ** -0.42 0.24 -0.01 0.26 0.35 0.24 * -1.39 0.25 *** -1.18 0.35 ** -0.93 0.35 *

LC 0.76 0.47 0.55 0.56 -0.55 0.47 0.64 0.32 0.17 0.42 -0.58 0.36 0.37 0.53 0.16 0.64 -0.82 0.43 *

G-SLI 0.45 0.39 0.74 0.48 0.81 0.39 * -0.18 0.38 -0.35 0.37 -0.16 0.20 -0.69 0.48 -0.70 0.47 -1.00 0.35 *

300-500 msAdults 0.16 0.20 0.69 0.23 ** 0.77 0.17 ** -0.24 0.14 0.09 0.16 0.35 0.14 * -0.72 0.22 ** -0.76 0.26 * -0.20 0.24

AC 0.07 0.42 0.38 0.43 0.63 0.34 -0.03 0.26 0.42 0.24 0.21 0.26 -1.01 0.34 * -1.10 0.39 * -1.19 0.43 *

LC 1.57 0.69 0.88 0.73 0.00 0.63 1.70 0.44 ** 0.31 0.54 -0.63 0.44 0.35 0.48 -0.92 0.58 -1.28 0.52 *

G-SLI 0.24 0.45 0.68 0.65 1.27 0.50 * -0.74 0.38 -0.88 0.38 0.04 0.35 -1.35 0.45 ** -0.85 0.32* -0.69 0.39

800-1000 msAdults -0.20 0.20 -0.20 0.32 -0.25 0.21 0.33 0.18 0.63 0.24*** 0.49 0.15 ** 0.77 0.23 ** 0.73 0.24 ** 0.91 0.25 **

AC -0.81 0.57 -0.63 0.72 0.59 0.64 -0.53 0.39 0.49 0.47 1.08 0.49 ** -0.40 0.56 0.15 0.74 0.93 0.51

LC -0.40 0.78 -0.19 0.84 -0.69 0.80 0.71 0.58 -0.36 0.58 -1.70 0.61 ** 0.41 0.67 -0.38 0.89 -0.55 0.63

G-SLI 0.35 0.61 0.40 0.87 0.23 0.72 -0.13 0.45 -0.03 0.51 -0.12 0.46 -0.52 0.59 -0.08 0.68 -0.58 0.51

1

Supplementary Methods S1

2.1 Experimental Materials:Experiment 1: Syntacticprocessing160 sentences were prepared. The nounphrases (NPs) in the subject position of thequestions consisted of 80characters/personalities that were highlyfamiliar to children and teenagers (e.g.Homer, Marge, and other cartoon charactersand personalities). Each of these subject NPswas heard twice, once in each condition, butwith a different verb and target object noun.80 transitive verbs were used which wererepeated once for each participant, butpresented with a different subject and objectNP in the two conditions (syntactic violation,control). The repetition order (violation,control) was controlled. The object NPsconsisted of a determiner (the, his, her) and160 nouns – 80 animate, and 80 inanimate.The nouns and verbs were controlled for age-of-acquisition (under 6 years), frequency,number of syllables and imageability betweenthe two conditions1-5. None of these objectNPs was repeated.

From this first list, we constructed a secondlist in which the wh-word was exchanged sothat who sentences became what questions,and the preposition phrase following thetarget noun was changed accordingly—producing a balanced quartet of questions:For example,List 1Who did Barbie push the clown into the wall?(animacy match- syntactic violation),Who did Alice in Wonderland push the ballinto?(animacy mismatch- control)List 2What did Barbie push the clown into?(animacy mismatch- control)What did Alice in Wonderland push the ballinto the wall? (animacy match- syntacticviolation)

Subjects heard either list 1 or 2.

2.1.1 Pre-testing the animacyproperty of direct object nouns inquestions.

Animacy properties of words enter into overtgrammatical processes in languages, such asRomance and American-Indian languages6.For example, two words (noun, verb) in asentence might have to agree in animacyproperties, in the same way as in Englishwords overtly agree in person (He jumps vs. Ijump) and number (A cat vs. Some cats).Thus, the grammaticalisation of properties ofwords such as, person and number, as well asanimacy, is generally found in languages.

We looked to see whether such animacyproperties also affect children and adults’language in English in the pre-tests of ourexperimental questions.

Two forms of questions in English are: a) wh-questions; and b) yes-no questions.

a) wh-question: Who did Barbiepush the car into______?b) yes-no question: Did Barbiepush the car into the clown?There is a direct structural relation betweenthese two types of questions. When the wh-word is present, it encodes one of the“arguments” (participants or things)associated with the verb or other predicate(e.g., preposition, in, with) in the question.Thus the wh-word has a structural syntacticrelation with the position that is normallyfilled by this noun. Specifically, the “who”wh-word encodes for an animate and the“what” wh-word for an inanimate propertyof the noun associated with a verb or otherpredicate. When the wh-word is not includedin the question, the noun phrase that the wh-word is standing for has to be present. Theexample a) above, shows that in wh-questions, the “who” stands for “the clown”which is present at the end of the question inthe yes-no question (b).

We hypothesised that when the participantheard the determine in our experimentalquestions, he/she would already “know” thatthe Noun Phrase (NP) was filled, andtherefore must posit that the wh-word wasstructurally related to another (missing) NP inthe question. In view of this, we checked to

2

see when participants completed questionscontaining a determiner (the his) but missingdirect object noun, whether the animacyproperty of this noun would match ormismatch those of the wh-word in the wh-questions, or the indirect object noun in theyes-no questions. Such a process wouldrequire checking of the animacy propertiesbetween words—specifically the wh-wordand the noun, or between the verb and thedirect and indirect object nouns. Suchsyntactic checking in wh-questions, perhapsprior to even hearing the noun, could result inan expectation that the object noun if presentwould have certain lexical-grammaticalproperties or features (e.g., it would be a nounrather than a verb, and be inanimate ratherthan animate).Based on these properties of question, we pre-tested the questions used in Experiment 1 intwo ways.

Pre-testing adultsIn the first pre-test 21 adults (universityundergraduates) were asked to complete 160written wh-questions used in Experiment 1.The questions were exact copies of thosequestions, but stopped after the determinerand before the critical noun. For example,c) Who did Barbie push the………d) What did Alice in Wonderland pushthe……Participants were asked to complete thequestions with the first words that came intotheir heads. Responses to the first noun in theobject position were coded according towhether they were animate (specifically, theycould be the answer to a “who” question), orwhether they were inanimate (specifically,they could be the answer to a “what”question). Analysis revealed that participantswere significantly more likely to continue thequestions that started with a who wh-wordwith an inanimate noun compared to when thewh-word was what, and vice versa (p<.0001).Therefore, adults do not expect that if the firstnoun phrase after the verb is filled in a wh-question, that it will have the same animacyproperties as that of the wh-word. Therefore,adults avoid using a noun phrase that could bea potential answer to the question.

Pre-testing children and teenagers

In the second pre-test of the questions, weused this structural syntactic relation betweena wh-word and the noun, and the contrastinganimacy properties of who and what, toconstruct incomplete yes-no questions for thechildren and teenagers to complete. Thus,using the wh-question from Experiment 1, weconstructed yes-no questions where the finalnoun would be either animate (matching theanimacy properties of the “who” wh-word) orinanimate (matching the animacy propertiesof a “what” wh-word). The direct object nounwas omitted as shown in the examples below.e) Did Barbie push the_____ into the clown?f) Did Alice in Wonderland push the _____into the car?Test-response forms were constructed with 6yes-no questions; 3 from the “who” questionlist and three from the “what” question lists.The questions were written in a random orderon the test-response forms. Writteninstructions asked the participant to completethe question. They were told that there was noright or wrong answer, but they were torespond with the first word they could thinkof.

All 320 questions to be used inExperiment 1 were used in the pre-testing.Thereby we provided exact pairs of wh-yes/no questions with all words matching thetest questions including the determiner (the,my, his) before the missing word.

776 school-aged children, evenlydistributed over the ages 7 to 19 years, fromschools in London participated. Participantswere given a form, each with 6 yes-noquestions. For each of the 80 verbs used in theexperiment we collected approximately 30who-question-related responses and 30 what-question-related responses (Total 4,650responses). The direct noun responses werecoded according to whether they wereanimate or inanimate, as in the adult pre-test.Analysis revealed a highly significant effectof animacy of the nouns according to whetherthe indirect object noun was animate or not.Significantly fewer nouns were animate whenthe final noun was animate than when thefinal noun was inanimate (t765=10.69,p<.0001). There was no significant effect ofage when the final noun was animate(r764=.019, p=.6). In contrast, a small effect ofage was found when the final noun wasinanimate, with more animate noun responses

3

with increasing age (r764=.100, p=.001).However, this small effect accounts for only1% of the variance in the data set.

These data indicate that children andadults process the animacy properties ofnouns and wh-words in questions.

2.2 Experimental Materials:Experiment 2 SemanticprocessingIn this experiment each participant listened to160 declarative sentences (80 per condition).The subject NPs consisted of the same 80characters/personalities that were used inExperiment 1. Each NP was repeated once inthe two conditions. 80 verbs were used thatwere different from those used in Experiment1. Each verb was repeated once but with adifferent subject and object NPs, balancedacross the two conditions (control andsemantic violation). 160 nouns were used inthe object NP position. None of the nounswere repeated, or appeared in Experiment 1.As in Experiment 1, the verbs and object NPswere controlled for age-of-acquisition,frequency, number of syllables, imageabilityand animacy across the conditions.

2.3 Acoustic recording of stimuliThe stimuli were digitally recorded in theUCL anechoic chamber by two native Englishfemale speakers at a normal-slow speakingrate. Following a behavioural pilot study, forExperiment 1 the wh-word at the beginning ofthe questions was slightly stressed to ensurethat the participants paid attention to the wh-word. For the sentences in the syntacticviolation condition, speakers spoke thesentence twice, once with the target wh-word(who/what) and once where the initial wh-word was replaced by where or when, makingthe sentence grammatical (When did Barbiepush the clown into the wall). The target wh-word from the syntactic violation conditionwas then spliced into the grammaticallycorrect sentence. This was done to avoid anyprosodic abnormalities which tended to occurwhen the sentences were un-grammatical. Thefinal sentence lists were balanced for thequartet wh-word, speaker, and 2 conditions.Analysis revealed no acoustic differencesbefore the critical word (intensity, pitch andlength) between the conditions for eitherexperiment.

2.4 ProcedurePresentation order of experiments wasbalanced across participants within groups.Each experiment was split into 3 blocks of 6minutes. The participants judged theappropriateness of the sentences (motordelayed response).

2.5 Supplementary Figure S3Regions of InterestSee Figure S3

Additional References

1. Masterson, J. & Druks, J. Descriptionof a set of 164 nouns and 102 verbsmatched for printed word frequency,familiarity and age-of-acquisition.Journal of Neurolinguistics 11, 331-354 (1998).

2. Fenson, L. et al. The MacArthurCommunicative DevelopmentInventories: User's Guide andTechnical Manual (SingularPublishing Group, San Diego, 1993).

3. Coltheart, M. The MRCPsycholinguistic Database. QuarterlyJournal of Experimental Psychology33A, 497-505 (1981b).

4. Bird, H., Franklin, S. & Howard, D.Age of Acquisition and imageabilityratings for a large set of words,including verbs and function words.Behavioral Research Methods,Instruments and Computers 33, 73-79(2001).

5. Druks, J. & Masterton, J. An Objectand Action Naming Battery(Psychology Press, London, 2000).

6. Baker, M. C. The PolysynthesisParameter. (Oxford University Press,New York, 1996).

1

Supplementary Data S1

Auditory Processing

To control for a more basic, lower-levelauditory processing deficit, we recordedAuditory Evoked Potentials (AEPs) ofG-SLI and age matched controlparticipants to pure tones in a classicalauditory oddball paradigm. Auditoryprocessing elicits earlyelectrophysiological responses known asthe N100/P200 (or N1/P2) complex. Thiscomplex is associated with perceptualdetection of a discrete change in theauditory environment. In addition, theyelicit a later P300 component thatreflects attentional control processes todetect and categorize a specific event.

We compared the G-SLI and age-matched control groups’ AEPs to puretones (50 ms long segments with 10 msrise and fall time) in an attended odd-ballparadigm. The targets were 2000 Hztones (high tones) presented 20 % of thetime, whereas the standards were 1000Hz tones (low tones) presented 80 % ofthe time. The tones were presentedthrough sound speaker at 60-70 dB. Theinterstimulus interval (ISI) was 800 msand tones were randomised so that nomore than two target stimuli occurredsuccessively. 60 target and 240 standardtones were presented over a sevenminute period.

Participants sat in a comfortablechair in an electrically shielded faradaycage approximately 100 cm from themonitor. The children looked at cartoonpictures of animals presented on themonitor (to prevent boredom). Wheneverthe target (high tone) was presented, theyhad to press a button on a button boxwith their dominant hand “as quickly aspossible”. An adult sat with eachparticipant during the task to monitorattention to the task and minimisemovement artefacts.

Electrophysiological recordings

The EEGs were recorded continuously ata sampling rate of 250 Hz, with a bandpass of 0.1 to 100 Hz from a montage of128 recording sites with the GeodesicSensor Net (manufactured by electricalgeodesic Inc). The Cz (vertex) electrodewas used as the recording reference. TheEEG was first divided into segmentsfrom 100 ms pre-stimulus to 800 mspost-stimulus. Segments containingvertical EOG (eye blink) or horizontalEOG (eye movement) activity greaterthan 150 V were excluded fromaveraging. This resulted in two sets(target and standard tones) of averagedevoked potentials for each participant.ERPs were re-referenced according tothe average reference. ERPs werequantified in relation to the 100 ms pre-stimulus baseline.

ERP analysis

Three dependent variables weremeasured: peak latencies (highest orlowest peak latency on the curve in ms),peak amplitudes (highest or lowest peakrelative to the prestimulus baseline inµV), mean amplitudes (mean averageamplitude within a time window ofinterest in µV). These were defined asfollows:Peak latencies and peak amplitudesfor three selected electrodes Fz, Cz, Pz

N100: largest negative-going deflection in the time window70-190 ms P200: largest positive-going deflection in the time window120-220 ms P300: largest positive-going deflection in the time window220-700 ms

Mean amplitudes for selectedelectrodes grouped into Regions OfInterestA description of the ROIs are provided

in the supplementary Methods S1 andFigure S-3.

N100: 80-140 ms

2

P200: 140-200 ms P300: 250-650 ms

The resulting AEPs from both groupsshowed an N100 after about 120 msfollowed by a P200 around 180 ms forboth tones, and a P300 component forthe target tones (Figure S-1). Note theGreenhouse-Geisser correction wasapplied to all analyses when evaluatingstatistical effects with more than onedegree of freedom in the numerator

Peak LatencyTo assess whether the G-SLI

children were slow auditory processors,we first analysed the peak latency ofeach component according to thedefinitions above (see supplementaryTable S-1, below).Analysis revealed a similar pattern ofresults for the N100, P200 and the P300for the two groups. There were nosignificant effects of group (N100, F1,

34=.38, p=.53; P200, F1, 34=1.27, p=.26;P300, F1, 34=.08, p=.76), but significanteffects of condition (N100, F1, 34=10.32,p=.002; P200, F1, 34=5.17, p=.02; P300,F1, 34 =5.6, p=.02) and electrodes for theN100 (F2, 68=9.67, p=.0002) and theP300, F2, 68=22.20, p<.0001). There wereno significant interactions with group forany of the components (F<1). The G-SLIchildren and Age controls showed ashorter latency for the target tones(N100, 117.7 ms; P200, 172.1 ms; P300,446.2 ms) than the standard tones (N100,125.4 ms; P200, 179.5 ms; P300, 491.6ms). Moreover, for the N100 and P300,for both conditions, the latency of theN100 is shortest at Pz (N100, 111.8 ms;P300, 405.4 ms) compared to Fz (N100,128.3 ms, p<.00006; P300, 548.7 ms,p<.00001) or Cz (N100, 124.5 ms,p=.0001; P300, (452.6 ms, p=.03). Thus,electrical brain responses associated withperceptual detection of a discrete changein the auditory environment(N100/P200), and attentional controlprocesses to detect and categorise a

specific auditory event (P300), were asfast for the G-SLI children as their agematched controls. These results revealedno evidence that G-SLI children wereslow in basic auditory processing whichcould have affected their processing ofsentences.

Peak Amplitude

The second set of analyses investigatedthe peak amplitudes of the threecomponents (supplementary Table S-1).Increased amplitude of components isoften associated with increased effort toprocess information or carry out a task.Therefore, such amplitude increases iffound in our study could reveal a subtledeficit in auditory processing.

Analysis for each of thecomponents revealed a similar pattern ofresults for both groups. The Peakamplitude show no significant effects ofgroup (N100, F1, 34=.86, p=.36; P200, F1,

34=2.72, p=.10; P300, F1, 34=0.27, p=.60),a significant effect of condition (N100,F1, 34=6.36, p=.01; P200, F1, 34=7.54,p=.009; P300, F1, 34=93.7, p< .0001) andelectrodes (N100, F2, 68=17.69,p<.00001; P200, F2, 68=53.21, p<.00001;P300, F2, 68=66.15, p<.0001), but nointeractions with group for any of theanalyses (F<1). Our results show that theamplitude for the target tones (N100, -2.82 µV; P200, 1.71 µV; P300, 5.91 µV)is larger compared to the standard tones(N100, -1.96 µV, P200, 0.93 µV; P300,2.02 µV). For the N100, for bothcondition, the amplitude is larger at Fz (-3.68 µV) compared to Cz (- 2.19 µV,p<.0004) or Pz (- 1.30 µV, p<.001);whereas for the P200 it is larger at Pz(3.10 µV) compared to Fz (- 1.18 µV,p<.0001) or Cz (2.05 µV, p<.01).

For the P300, analysis revealed asignificant interaction between conditionand electrodes (F2, 68=29.99, p<.0001),that reflected a larger amplitude for thetarget at Pz (9.9 µV) compared to Fz (1.1µV, p<.0001) and Cz (-6.6 µV, p<.0001;whereas for the standard tone the

3

amplitude was equally distributed on thePz and Cz electrodes (Fz: 0.86 µV, Cz:2.3µV, Pz: 2.90 µV).

Importantly, these results revealno evidence of any difference inamplitude for any component betweenthe G-SLI and the Age-Matched controlgroups.

Mean amplitudes

The third set of analyses was carried outon normalized data and took intoconsideration the whole montage of 128electrodes. This enabled us to assesswhether there were any topographicaldifferences between the groups in thescalp distribution of the electricalresponses. To do this we investigatedinteractions between the factor group andcondition or ROI. The brain maps (Fig.S-2) show that the distribution of thethree components is similar for the G-SLI and age control groups. Statisticalanalysis revealed no significantinteraction between groups and conditionor ROI for any of the components. Forboth groups the N100 component showsmaximum amplitude on the frontalelectrodes (Caudality, F2, 68=16.77,p<.0001), whereas for the P200 themaximum amplitude is on the centralregions (Caudality, F2, 68=112.1,p<.0001) and is larger for the target tonescompared to the standard tones(Condition x Caudality, F2, 68=7.7,p<.0001). For the P300, both groupsproduced a larger amplitude for thetarget tones compared to the standardtones (Condition, F1, 34=14.19, p<.0001),which was larger on the posterior lefthemisphere for the target compared tothe standard tones (Condition xCaudality x Hemisphere, F2, 68=8.25,p<.0001) (see Figure S2)

In sum, we discovered that children withG-SLI have age-appropriate waveformsfor the N100, P200 and the P300components (latency, amplitude,distribution on the scalp). Our results

reveal that G-SLI children have normalauditory processing during thediscrimination of pure tones.

N100 Age Control G-SLI______________________________________________________

Lat Amp Lat AmpMean SD Mean SD Mean SD Mean SD_______________________________________

Standard Fz 130.2 25.9 -2.7 1.5 135.5 17.2 -3.3 1.8Cz 122.8 16.4 -1.7 1.3 131.8 23.2 -2.0 1.5Pz 119.4 23.7 -1.0 1.0 112.9 21.4 -0.9 1.5

Target Fz 120.5 27.4 -3.8 2.2 127.2 20.1 -4.7 3.0Cz 120.0 21.9 -2.2 2.5 123.7 31.0 -2.7 2.1Pz 108.2 20.5 -1.7 3.3 107.1 22.3 -1.4 3.1

P200 Age Control G-SLI______________________________________________________


Standard Fz 166.1 24.9 -0.8 2.0 174.3 23.8 -1.2 1.8Cz 183.3 28.9 1.2 1.8 190.1 22.8 2.1 1.4Pz 183.4 29.3 1.7 1.2 179.7 40.4 2.6 1.4

Target Fz 162.1 32.1 -1.8 2.1 182.3 26.8 -0.7 2.5Cz 171.1 28.2 1.9 2.4 177.8 25.1 2.9 3.7Pz 168.1 29.3 3.4 2.4 170.6 33.3 4.4 4.2

P300 Age Control G-SLI______________________________________________________


Target Fz 495.2 151.8 1.1 2.4 532.2 155.6 1.1 2.8Cz 416.0 106.7 6.4 3.5 399.1 103.1 6.9 4.4Pz 401.7 95.8 9.5 4.2 433.1 124.4 10.3 4.6

electrical brain responses in language-impaired children reveal grammar-specific deficits ·...

Documents