- Email: [email protected]
Abnormal frequency discrimination in children with SLI as indexed by mismatch negativity (MMN)
Neuroscience Letters 413 (2007) 99–104
Abnormal frequency discrimination in children with SLI as indexed by mismatch negativity (MMN) Tanja Rinker a,∗ , Gregor Kohls a,b , Cathrin Richter a , Verena Maas a , Eberhard Schulz a , Michael Schecker a a
Neurolinguistic Laboratory, Department of Child and Adolescent Psychiatry and Psychotherapy, University of Freiburg, Freiburg, Germany b Child Neuropsychology Section, Department of Child and Adolescent Psychiatry and Psychotherapy, RWTH Aachen University, Aachen, Germany Received 28 August 2006; received in revised form 20 October 2006; accepted 15 November 2006
Abstract For several decades, the aetiology of specific language impairment (SLI) has been associated with a central auditory processing deficit disrupting the normal language development of affected children. One important aspect for language acquisition is the discrimination of different acoustic features, such as frequency information. Concerning SLI, studies to date that examined frequency discrimination abilities have been contradictory. We hypothesized that an auditory processing deficit in children with SLI depends on the frequency range and the difference between the tones used. Using a passive mismatch negativity (MMN)-design, 13 boys with SLI and 13 age- and IQ-matched controls (7–11 years) were tested with two sine tones of different frequency (700 Hz versus 750 Hz). Reversed hemispheric activity between groups indicated abnormal processing in SLI. In a second time window, MMN2 was absent for the children with SLI. It can therefore be assumed that a frequency discrimination deficit in children with SLI becomes particularly apparent for tones below 750 Hz and for a frequency difference of 50 Hz. This finding may have important implications for future research and integration of various research approaches. © 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Specific language impairment (SLI); Mismatch negativity (MMN); Frequency discrimination; Event-related potentials (ERP); Auditory processing deficit; Children
About 7% of all children are affected by specific language impairment (SLI), a developmental disorder that significantly compromises their speech and language abilities (for a review, see [12]). By definition, children with SLI have an IQ within normal range and their language problems cannot be explained by serious neurological, emotional, or sensory deficits. For several decades, an auditory processing deficit has been discussed to be associated with SLI. The basic idea is that acoustic speech perception must occur within milliseconds and that the ability to perceive rapidly changing temporal and spectral characteristics is essential for language development [1]. It has been found that children with SLI have difficulty discriminating high and low tones that are presented in rapid succession or that are of
∗ Corresponding author at: Transferzentrum f¨ ur Neurowissenschaften und Lernen (ZNL), Beim Alten Fritz 2, 89075 Ulm, Germany. Tel.: +49 731 500 62031; fax: +49 731 500 62049. E-mail address: [email protected] (T. Rinker).
0304-3940/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.neulet.2006.11.033
brief duration [2,13,23,24]. This deficit has been theorized to be responsible for their difficulty in perceiving and processing nonverbal as well as verbal material [22]. However, recently, this view has been challenged in various studies. These showed that not primarily the processing of the temporal characteristics of the input is problematic for children with SLI but the frequency discrimination that is required of them in these tasks (e.g. [15]). A recent study by Mengler et al. [17] showed that children with SLI performed consistently worse on a psychoacoustic frequency discrimination task whereby children had to match a middle tone to either the first or the last tone (AXBtask). The average frequency threshold of the controls was at 25 Hz, while the threshold of the children with SLI to discriminate tones was at 58 Hz. In a follow-up study, Hill et al. [7] focused more specifically on the difference between a temporal and a frequency discrimination deficit. They retested 10 children with SLI and two controls of the Mengler-study [17] 30 months after their first examination. Although in both groups the average frequency discrimination-threshold decreased from the first to
100
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
the second study (SLI: 37.5 Hz, controls: 18.7 Hz), the SLI group still displayed a significantly higher frequency discriminationthreshold. The majority of all children showed adult-like backward masking thresholds. The results suggest a significant contribution of a frequency discrimination deficit to SLI. Although the studies by Hill et al. and Mengler et al. [7,17] controlled for possible task-dependent effects of attention on the performance, further influences on the experimental results (e.g. rapid learning, motor response) cannot be entirely excluded. An attention-independent component of the auditory event-related brain potential called mismatch negativity [19] is ideally suited to investigate central auditory processing, particularly in child clinical populations. The MMN is elicited by a change in auditory stimulation after the onset of a deviant stimulus in the context of repetitive standard stimuli. This component is usually displayed as a difference wave between averaged standard and deviant stimuli, peaking around 100–250 ms after stimulus onset. It is thought that the MMN is the result of a comparison process between the incoming deviant stimulus and a neural trace set up by the standard stimuli [18]. The elicitation of the MMN depends on individual abilities of auditory discrimination. Thus it can be used to examine frequency discrimination in children with SLI in comparison to healthy controls, regardless of subject’s focused attention on the auditory stimulation. Tones are ideally suited to distinguishing clinical groups, such as language impaired children from controls because they test the basic discriminative ability of children independently of the effects on them of over-learned structures, such as words or syllabic material. The MMN has been used in a small number of studies for the investigation of children with SLI but results regarding frequency discrimination have been contradictory. Korpilahti and Lang as well as Holopainen et al. [9–11] employed the same sine tones (500 Hz versus 553 Hz). They showed that the MMN-amplitudes of children with SLI differ significantly from the control groups during frequency discrimination outside the focus of attention. Further, the studies by Korpilahti and Lang and Holopainen et al. [9–11] showed significant hemispheric differences which is in line with other electrophysiological findings [20]. However, the use of a larger frequency contrast of 200 Hz in a higher and more salient frequency range (1000 versus 1200 Hz) by Uwer et al. [25] did not reveal significant group differences between children with SLI and healthy controls. To conclude, the choice of frequency range as well the frequency differences between the stimuli have to be taken into account when discussing possible frequency discrimination impairments in children with SLI. In addition to the classical MMN, a subsequent component, a second negativity that peaks between 300–500 ms has been observed in a number of passive oddball studies (e.g. [4,8]). Previous studies reported significant differences between controls and clinical groups (e.g. children at-risk for developmental language disorders) for this second negativity [14]. The present study extends the current electrophysiological findings on SLI to date to a frequency difference of 700 Hz versus 750 Hz, placed between the controversial results of [9–11] and [25]. It is hypothesized that children with SLI show an attention-
independent frequency discrimination deficit in this frequency range. To this end, both ERPs to standard and deviant as well as MMN(s) were analyzed. Thirteen children with SLI and 13 control children (all male) were recruited from schools in Freiburg and two special education schools for the language impaired in the Freiburg region. Children were included if they were native monolingual speakers of German and had normal hearing (as tested by audiometer), normal or normal-to-corrected vision, and an IQ above 85 on a nonverbal IQ-test (Culture Fair Intelligence Test, CFT). Children with other known disorders (e.g. ADHD, autism) were excluded from the testing. Groups did not differ significantly regarding age (SLI: mean age: 108 ± 16 months, controls: mean age: 116 ± 11 months; p = .128) and IQ (SLI mean IQ: 105 ± 13; controls: mean IQ: 112 ± 13; p = .180). The language impaired group scored at least two standard deviations below the norm on either a receptive or expressive subtest of a German language development test (Heidelberger Sprachentwicklungstest, H-S-E-T; significant difference between children with SLI and controls: expressive subtest: p = 0.001 and receptive subtest: p < 0.001). Further, groups differed significantly on a German auditory discrimination test (Heidelberger Lautdifferenzierungstest, HLAD (p < .001) and on a German pseudo-word repetition test (Mottier-test; p < .001). The study was approved by the Ethics Committee of the Medical Department at the University of Freiburg and parents gave informed consent for their children. Stimuli were created using Cool Edit Pro (Cool Edit Pro 2002, version 2.0 Syntrillium Software Corporation, Phoenix, Arizona, USA). All stimuli were pseudo-randomized to ensure that no two deviants follow each other. The sine tones consisted of a 700 Hz tone (standard) and a 750 Hz tone (deviant). The deviant tone occurred with a probability of p = .15. The total number of tones was 1200. The duration of the tones was 150 ms each, including 20 ms rise and fall time for each tone. Interstimulus interval (offset-to-onset) was kept constant at 650 ms. Subjects were seated in a chair that was adjustable for height in an electrically shielded booth. Particularly younger boys were accompanied by a parent, sibling, or staff member in the cabin. Subjects watched a Disney cartoon of their choice on a computer monitor with the sound off. They were told to ignore the auditory input and to focus their attention on the movie as they were going to be asked detailed questions about it. The stimuli were presented binaurally through insert earphones (EAR Auditory Systems, Wheaton, IL, USA). The sound volume was kept at 70 dB sound pressure level (SPL). The EEG was digitally recorded with SynAmps amplifiers and Scan 4.0 software (NeuroScan, Inc., Sterling, Virginia, USA) from Ag/AgCl electrodes, using an extended 10–20-System electrode cap (EasyCap, Falk Minow Services, Herrsching-Breitbrunn, Germany), from midline sites Fz, Cz, and on each side F3/F4, F7/F8, FC1/FC2, FC5/FC6, FT7/FT8, T7/T8, C3/C4, CP1/CP2, CP5/CP6, TP9/TP10, P3/P4, O1/O2. The ground electrode was positioned on the midline at AFz, and Pz was used as on-line reference. The vertical electro-oculogram (VEOG) was recorded from above and below the right and left eye, and the horizontal electro-oculogram (HEOG) was recorded
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
101
Fig. 1. Difference wave (deviant minus standard) to sine tones at Fz. Standard (700 Hz) is the dotted line, and the deviant (750 Hz) the striped line. The bold line is the difference wave with two negative deflections. The response in control children is displayed in the left graph, the response of the children with SLI is displayed in the right graph. The second MMN is not significant in the SLI group at Fz and both, the first and the second MMN are not significant at central electrode Cz (not displayed here).
from the outer canthi of each eye. The on-line filter was at 0.1–70 Hz. The sampling rate of the electric signal was 500 Hz. Electrode impedance was kept below 10 k and below 20 k for VEOG and HEOG as these electrodes were placed directly on sensitive skin. The EEG raw data was stored on a hard disk and processed offline with BrainVision Analyzer 1.05 software (Brain Products GmbH, M¨unchen, Germany). Data was high-pass filtered at 0.2 Hz and low-pass filtered at 30 Hz. Filtered data were baseline corrected to −50 ms. The first five standards were excluded from the analysis. Standards directly following a deviant were also excluded. Eye movements and epochs with other artifacts were rejected if the activity was ±100 V. For the removal of eye blinks, the Gratton & Coles algorithm was applied. The difference wave was calculated as the difference between standard and deviant. Mastoid polarity reversal was inspected using the average reference. Then, linked mastoid reference was applied. In the sine tone paradigm, P1 and N250 were present, as can be seen in Fig. 1. Both groups demonstrated a clear early positivity and later negativity for standard and deviant tones, respectively. The grand-average wave forms showed a reduction in amplitude for the children with SLI. However, these differences were not statistically significant at Fz for standard or deviant ERPs (e.g. standard: P1 amplitude: t(24) = 1.347, p = .190; N250 amplitude: t(24) = −1.073, p = .294). As outlined in [5], age effects may be reflected in a reduction of the latency of the P1-component with increasing age. In the present data, a correlation of age and P1-latency confirmed
this finding, however, there was only a significant negative correlation for the control children (Fz: r = −.574, p = .020) but not for the children with SLI (Fz: r = −.174, p = .285). This finding may be an indication for a developmental delay in the children with SLI. Time windows for the analysis were determined if paired point-to-point t-test against zero was p < .05 in time intervals greater than 20 ms at Fz for the controls. Following this procedure, two time windows were determined for the sine tone paradigm from the data set of the controls: 140–280 ms (MMN1) and 420–620 ms (MMN2). Grand-average data is displayed in Fig. 1. Only area measurements (in V) are presented here (see Table 1). Especially in children, the inter-individual variability is very large. In the present study, MMN-amplitudes in controls varied at Fz from −1.0 to −7.2 V. Therefore, MMN-amplitudes were not employed for group comparisons in the present study. Area measurements revealed a significant MMN1 for controls at Fz (t(12) = −4.261, p = .001) and Cz (t(12) = −2.708, p = .019). For the children with SLI, a significant MMN1 could be observed at Fz (t(12) = −4.753, p < .001) but not at Cz (t(12) = −1.824, p = .093). Independent samples t-test revealed that this difference between groups was not statistically significant at Cz (t(24) = −0.793, p = 0.435). Other electrodes were also inspected and using independent samples t-test, significant group differences could be found for the more posterior electrodes CP5 (t(24) = −2.228, p = .035) and P3 (t(24) = −2.265, p = .033), and a trend towards significance for C3 (t(24) = −1.960, p = .062). These three electrodes, at which
Table 1 Mean activity and standard deviations (V) for MMN1 and MMN2 area for controls and SLI MMN1
MMN2
Fz Controls SLI
Cz
−1.98 V −1.30 V (±0.9)*
Level of significance: * p < 0.05.
(±1.7)*
Fz
−1.05 V −0.64 V (±1.3)
(±1.4)*
Cz
−1.57 V −0.72 V (±1.8)
(±1.5)*
−0.78 V (±1.0)* −0.31 V (±2.0)
102
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
Fig. 2. Topographical maps of the MMN-response to sine tones. Time windows: 140–280 ms (top) and 420–620 ms (bottom). The response of the controls is displayed on the left part of the graphs, the response of the children with SLI on the right part.
the greatest difference between the groups was observed, form a cluster of left-ward central-posterior electrodes. As this was seen as indication for hemispheric differences between groups, these group differences were further analyzed. A repeated measures ANOVA for left hemisphere electrode clusters (FC1/F3/C3/CP1) and right hemisphere electrode clusters (FC2/F4/C4/CP2) revealed a significant three-way-interaction of electrode × hemisphere × group (F(2165) = 3.273, p = .042) (Fig. 2). Area measurements revealed a significant MMN2 for controls at Fz (t(12) = −3.861, p = .002) and Cz (t(12) = −2.906, p = .013). For the children with SLI, a significant MMN2 could neither be observed at Fz (t(12) = −1.406, p = .185) nor at Cz (t(12) = −0.548, p = .593). However, independent samples t-test revealed that this difference between groups was neither
statistically significant at Fz (t(24) = −1.302, p = .205) nor at Cz (t(24) = −0.759, p = .455). Other electrodes were also inspected and, using independent samples t-test, significant group differences could be found for FC5 (t(24) = −2.243, p = .034), C3 (t(24) = −2.243, p = .014), and CP5 (t(24) = −2.269, p = .026). As for the early time window, a repeated measures ANOVA for left hemisphere electrode clusters (FC1/F3/C3/CP1) and right hemisphere electrode clusters (FC2/F4/C4/CP2) was conducted. A significant two-way-interaction of hemisphere × group was found (F(1000) = 7.109, p = .014), but not for electrode × hemisphere × group (F(1979) = 1.650, p = .203) (Fig. 2). In order to examine the possible effects of age on MMN1 and MMN2, Pearson’s Product Moment Correlation for MMN1 area and MMN2 area was carried out:
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
There was no effect of the group and age on the MMN1 area (Cz: r = −.83, p = .343; Fz: r = .027, p = .447). However, there was a significant negative correlation between age and MMN2 for the entire group (Fz: r = −.353, p = .039), and a trend at Cz (r = .−328, p = .051). This correlation was not significant for any of the groups (controls: Fz: r = −.409, p = .083; SLI: Fz: r = −.125, p = .342). In the present study, an oddball paradigm with sine tones was applied to investigate the frequency discrimination abilities of SLI and control children. A standard tone of 700 Hz and a 750 Hz deviant were used. ERP-responses (P1 and N250) did not differ significantly between groups. A significant age effect for the controls was found for P1-latency. The correlation was on the whole negative. This is in line with previous findings by Uwer et al. [25]. It has been argued that ERPs decrease with age due myelination processes increasing velocity [5]. However, this effect was not found for the children with SLI. This can be seen as an indication that different maturation processes are taking place (see also [3]). In two time windows (140–280 and 420–620), a significant fronto-central MMN1 and MMN2 were elicited for the controls with more left-localized activity for the second negativity. For the children with SLI, MMN1 was elicited only at Fz but not at Cz with a slightly right-localized focus. The area of MMN2 was not significant in children with SLI. This indicates a difference in the abilities of children with SLI to discriminate two tones of different frequency. The 700 Hz tone must be kept in sensory memory so that the incoming new tone of 750 Hz can be compared to it. The accuracy of this representation (and the role of auditory memory in this process) must therefore be questioned seeing as the discriminative response indexed by the MMN was quite weak in the children with SLI. Significant hemispheric differences were observed for both MMN1 and MMN2. This is in line with previous studies [9–11,20] where it has been argued that hemispheric specialization is different in children with SLI. A number of conclusions can be drawn from the studies by present as well as the studies by [9–11] and [25]: Even if the difference between standard and deviant is 50 ms as in [11], the duration of the stimulus is not as crucial as the frequency difference of the stimulus. The fact that Uwer et al. [25] also did not find group differences for a duration deviant lends support to this result. Therefore in the present study, the duration of the tone (150 ms) may have not have led to the observed group differences. Further, it may be assumed that the language difficulties in children with SLI are more strongly influenced by weak auditory discriminative abilities than by lower IQ. Even though the children in the present study had a higher IQ than the subjects of the Korpilahti-group, they nonetheless showed difficulties in the processing of a slight acoustic contrast. The independence of the influence of the IQ on the MMN has been reported before by Holopainen et al. [10]. In addition, the frequency discrimination deficit also appears to be less dependent on age. Even though the children in the present study were older than the children in the study by Uwer et al. [25], they showed a frequency discrimination deficit. This is supported by Korpilahti and Lang [11] and Holopainen et al. [9,10] who consistently showed a weaker or absent MMN despite investigating children of a wide age range.
103
The late component, MMN2 was not analyzed further in Uwer et al. [25] although group differences were observed for the latency of the second negativity. In the study by Korpilahti and colleagues [9–11], the time window was usually too short to analyze this component. In the present study especially the late component revealed group differences. Groups were in fact more discrepant in their responses in the later time window. MMN2 was absent in the children with SLI and overall activity was significantly reduced. The hemispheric differences between groups were quite striking – with the children with SLI showing a converse pattern of hemispheric activity than the controls. Previous studies with at-risk populations also demonstrated significant group differences in this second component [14]. These combined results suggest that the MMN2 may reflect similar comparison processes as the MMN1. It may be assumed that the characteristics of the MMN2 depend on the level of automatization of the comparison process as it was found to decrease with age [6]. For future studies, it may therefore be a worth-while endeavor to enlarge the window of analysis, especially in the investigation of group differences between clinical populations and controls. This study is only an exploratory study, therefore more studies are needed investigating other frequency ranges. We cannot draw any explicit conclusion, but the present study supports a frequency discrimination deficit in children with SLI as previously reported by a number of authors. This deficit may become apparent for small frequency changes (e.g. around 50 Hz or less) and, combining the results of the present study and [9–11], in a frequency range below 750 Hz. Previous studies that controlled speech stimuli for their spectral characteristics have found that differences between language impaired groups and controls arise from small acoustic transitions in the speech material [16,21]. This can be seen as evidence for the relevance of a possible frequency discrimination deficit for the language acquisition process and may have important implications for diagnosis and intervention. Acknowledgments We would like to thank all participating children and families. This research was supported by “Fonds f¨ur interdisziplin¨are Forschungsvorhaben”, University of Freiburg Rectorate. This study was a portion of the doctoral dissertation presented by the first author. References [1] A.A. Benasich, P. Tallal, Infant discrimination of rapid auditory cues predicts later language impairment, Behav. Brain Res. 136 (2002) 31–49. [2] A.A. Benasich, N. Choudhury, J.F. Friedman, T. Realpe-Bonilla, C. Chojnowska, Z. Gou, The infant as a prelinguistic model for language learning impairments: Predicting from event-related potentials to behavior, Neuropsychologia 44 (2006) 396–411. [3] D.V.M. Bishop, G.M. McArthur, Immature cortical responses to auditory stimuli in specific language impairment: evidence from ERPs to rapid tone sequences, Dev. Sci. 7 (2004) 11–18. ˇ [4] R. Ceponien˙ e, M. Cheour, R. N¨aa¨ t¨anen, Interstimulus interval and auditory event-related potentials in children: evidence for multiple generators, Electroen. Clin. Neuro.: Evoked Potentials 108 (1998) 345–354.
104
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
ˇ [5] R. Ceponien˙ e, T. Rinne, R. N¨aa¨ t¨anen, Maturation of cortical sound processing as indexed by event-related potentials, Clin. Neurophysiol. 113 (2002) 870–882. [6] M. Cheour, P. Korpilahti, O. Martynova, A.H. Lang, Mismatch negativity and late discriminative negativity in investigating speech perception and learning in children and infants: a review, Audiol. Neuro-Otol. (2001) 2–11. [7] P.R. Hill, J.H. Hogben, D.V.M. Bishop, Auditory frequency discrimination in children with specific language impairment: a longitudinal study, J. Speech Hear. Res. 48 (2005) 1136–1146. [8] P.R. Hill, G.M. McArthur, D.V.M. Bishop, Phonological categorization of vowels: a mismatch negativity study, Neuroreport 15 (2004) 2195–2199. [9] I. Holopainen, P. Korpilahti, K. Juottonen, A.H. Lang, M. Sillanp¨aa¨ , Attenuated auditory event-related potential (mismatch negativity) in children with developmental dysphasia, Neuropediatrics 28 (1997) 253–256. [10] I. Holopainen, P. Korpilahti, K. Juottonen, A.H. Lang, M. Sillanp¨aa¨ , Abnormal frequency mismatch negativity (MMN) in mentally retarded children and in children with developmental dysphasia, J. Child Neurol. 13 (1998) 178–183. [11] P. Korpilahti, A.H. Lang, Auditory ERP Components and MMN in dysphasic children, Electroen. Clin. Neuro. 91 (1994) 256–264. [12] L.B. Leonard, Children with Specific Language Impairment, MIT Press, Cambridge, MA, 1998. [13] A.D. Lowe, R.A. Campbell, Temporal discrimination in aphasoid and normal children, J. Speech Hear. Res. 8 (1965) 313–317. [14] U. Maurer, K. Bucher, S. Brem, D. Brandeis, Altered responses to tone and phoneme mismatch in kindergarteners at familial dyslexia risk, Neuroreport 14 (2003) 2245–2250. [15] G.M. McArthur, D.V.M. Bishop, Which people with specific language impairment have auditory processing deficits? Cogn. Neuropsych. 21 (2004) 79–94.
[16] G.M. McArthur, D.V.M. Bishop, Speech and non-speech processing in people with specific language impairment: a behavioural and electrophysiological study, Brain Lang. 94 (2005) 260–273. [17] E.D. Mengler, J.H. Hogben, P.T. Mitchie, D.V.M. Bishop, Poor frequency discrimination is related to oral language disorder in children: a psychoacoustic study, Dyslexia 11 (2005) 155–173. [18] R. N¨aa¨ t¨anen, Attention and Brain Function, Lawrence Erlbaum, Hillsdale, NJ, 1992. [19] R. N¨aa¨ t¨anen, A.W.K. Gaillard, S. M¨antysalo, Early selective-attention effect on evoked potential reinterpreted, Acta Psychol. 42 (1978) 313– 329. [20] V.L. Shafer, R.G. Schwartz, M.L. Morr, K.L. Kessler, D. Kurtzberg, Deviant neurophysiological asymmetry in children with language impairment, Neuroreport 11 (2000) 3715–3718. [21] J.E. Sussman, Vowel perception by adults and children with normal language and specific language impairment: based on steady states or transitions? J. Acoust. Soc. Am. 109 (2001) 1173–1180. [22] P. Tallal, Experimental studies of language learning impairments: from research to remediation, in: D.V.M. Bishop, L.B. Leonard (Eds.), Speech and Language Impairments in Children: Causes, Characteristics, Intervention and Outcome, Psychology Press, Hove, UK, 2000, pp. 131– 156. [23] P. Tallal, M. Piercy, Defects of non-verbal auditory perception in children with developmental aphasia, Nature 241 (1973) 468–469. [24] P. Tallal, M. Piercy, Developmental aphasia: impaired rate of nonverbal processing as a function of sensory modality, Neuropsychologia 11 (1973) 389–398. [25] R. Uwer, R. Albrecht, W. von Suchodoletz, Automatic processing of tones and speech stimuli in children with specific language impairment, Dev. Med. Child Neurol. 44 (2002) 527–532.
Abnormal frequency discrimination in children with SLI as indexed by mismatch negativity (MMN) Tanja Rinker a,∗ , Gregor Kohls a,b , Cathrin Richter a , Verena Maas a , Eberhard Schulz a , Michael Schecker a a
Neurolinguistic Laboratory, Department of Child and Adolescent Psychiatry and Psychotherapy, University of Freiburg, Freiburg, Germany b Child Neuropsychology Section, Department of Child and Adolescent Psychiatry and Psychotherapy, RWTH Aachen University, Aachen, Germany Received 28 August 2006; received in revised form 20 October 2006; accepted 15 November 2006
Abstract For several decades, the aetiology of specific language impairment (SLI) has been associated with a central auditory processing deficit disrupting the normal language development of affected children. One important aspect for language acquisition is the discrimination of different acoustic features, such as frequency information. Concerning SLI, studies to date that examined frequency discrimination abilities have been contradictory. We hypothesized that an auditory processing deficit in children with SLI depends on the frequency range and the difference between the tones used. Using a passive mismatch negativity (MMN)-design, 13 boys with SLI and 13 age- and IQ-matched controls (7–11 years) were tested with two sine tones of different frequency (700 Hz versus 750 Hz). Reversed hemispheric activity between groups indicated abnormal processing in SLI. In a second time window, MMN2 was absent for the children with SLI. It can therefore be assumed that a frequency discrimination deficit in children with SLI becomes particularly apparent for tones below 750 Hz and for a frequency difference of 50 Hz. This finding may have important implications for future research and integration of various research approaches. © 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Specific language impairment (SLI); Mismatch negativity (MMN); Frequency discrimination; Event-related potentials (ERP); Auditory processing deficit; Children
About 7% of all children are affected by specific language impairment (SLI), a developmental disorder that significantly compromises their speech and language abilities (for a review, see [12]). By definition, children with SLI have an IQ within normal range and their language problems cannot be explained by serious neurological, emotional, or sensory deficits. For several decades, an auditory processing deficit has been discussed to be associated with SLI. The basic idea is that acoustic speech perception must occur within milliseconds and that the ability to perceive rapidly changing temporal and spectral characteristics is essential for language development [1]. It has been found that children with SLI have difficulty discriminating high and low tones that are presented in rapid succession or that are of
∗ Corresponding author at: Transferzentrum f¨ ur Neurowissenschaften und Lernen (ZNL), Beim Alten Fritz 2, 89075 Ulm, Germany. Tel.: +49 731 500 62031; fax: +49 731 500 62049. E-mail address: [email protected] (T. Rinker).
0304-3940/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.neulet.2006.11.033
brief duration [2,13,23,24]. This deficit has been theorized to be responsible for their difficulty in perceiving and processing nonverbal as well as verbal material [22]. However, recently, this view has been challenged in various studies. These showed that not primarily the processing of the temporal characteristics of the input is problematic for children with SLI but the frequency discrimination that is required of them in these tasks (e.g. [15]). A recent study by Mengler et al. [17] showed that children with SLI performed consistently worse on a psychoacoustic frequency discrimination task whereby children had to match a middle tone to either the first or the last tone (AXBtask). The average frequency threshold of the controls was at 25 Hz, while the threshold of the children with SLI to discriminate tones was at 58 Hz. In a follow-up study, Hill et al. [7] focused more specifically on the difference between a temporal and a frequency discrimination deficit. They retested 10 children with SLI and two controls of the Mengler-study [17] 30 months after their first examination. Although in both groups the average frequency discrimination-threshold decreased from the first to
100
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
the second study (SLI: 37.5 Hz, controls: 18.7 Hz), the SLI group still displayed a significantly higher frequency discriminationthreshold. The majority of all children showed adult-like backward masking thresholds. The results suggest a significant contribution of a frequency discrimination deficit to SLI. Although the studies by Hill et al. and Mengler et al. [7,17] controlled for possible task-dependent effects of attention on the performance, further influences on the experimental results (e.g. rapid learning, motor response) cannot be entirely excluded. An attention-independent component of the auditory event-related brain potential called mismatch negativity [19] is ideally suited to investigate central auditory processing, particularly in child clinical populations. The MMN is elicited by a change in auditory stimulation after the onset of a deviant stimulus in the context of repetitive standard stimuli. This component is usually displayed as a difference wave between averaged standard and deviant stimuli, peaking around 100–250 ms after stimulus onset. It is thought that the MMN is the result of a comparison process between the incoming deviant stimulus and a neural trace set up by the standard stimuli [18]. The elicitation of the MMN depends on individual abilities of auditory discrimination. Thus it can be used to examine frequency discrimination in children with SLI in comparison to healthy controls, regardless of subject’s focused attention on the auditory stimulation. Tones are ideally suited to distinguishing clinical groups, such as language impaired children from controls because they test the basic discriminative ability of children independently of the effects on them of over-learned structures, such as words or syllabic material. The MMN has been used in a small number of studies for the investigation of children with SLI but results regarding frequency discrimination have been contradictory. Korpilahti and Lang as well as Holopainen et al. [9–11] employed the same sine tones (500 Hz versus 553 Hz). They showed that the MMN-amplitudes of children with SLI differ significantly from the control groups during frequency discrimination outside the focus of attention. Further, the studies by Korpilahti and Lang and Holopainen et al. [9–11] showed significant hemispheric differences which is in line with other electrophysiological findings [20]. However, the use of a larger frequency contrast of 200 Hz in a higher and more salient frequency range (1000 versus 1200 Hz) by Uwer et al. [25] did not reveal significant group differences between children with SLI and healthy controls. To conclude, the choice of frequency range as well the frequency differences between the stimuli have to be taken into account when discussing possible frequency discrimination impairments in children with SLI. In addition to the classical MMN, a subsequent component, a second negativity that peaks between 300–500 ms has been observed in a number of passive oddball studies (e.g. [4,8]). Previous studies reported significant differences between controls and clinical groups (e.g. children at-risk for developmental language disorders) for this second negativity [14]. The present study extends the current electrophysiological findings on SLI to date to a frequency difference of 700 Hz versus 750 Hz, placed between the controversial results of [9–11] and [25]. It is hypothesized that children with SLI show an attention-
independent frequency discrimination deficit in this frequency range. To this end, both ERPs to standard and deviant as well as MMN(s) were analyzed. Thirteen children with SLI and 13 control children (all male) were recruited from schools in Freiburg and two special education schools for the language impaired in the Freiburg region. Children were included if they were native monolingual speakers of German and had normal hearing (as tested by audiometer), normal or normal-to-corrected vision, and an IQ above 85 on a nonverbal IQ-test (Culture Fair Intelligence Test, CFT). Children with other known disorders (e.g. ADHD, autism) were excluded from the testing. Groups did not differ significantly regarding age (SLI: mean age: 108 ± 16 months, controls: mean age: 116 ± 11 months; p = .128) and IQ (SLI mean IQ: 105 ± 13; controls: mean IQ: 112 ± 13; p = .180). The language impaired group scored at least two standard deviations below the norm on either a receptive or expressive subtest of a German language development test (Heidelberger Sprachentwicklungstest, H-S-E-T; significant difference between children with SLI and controls: expressive subtest: p = 0.001 and receptive subtest: p < 0.001). Further, groups differed significantly on a German auditory discrimination test (Heidelberger Lautdifferenzierungstest, HLAD (p < .001) and on a German pseudo-word repetition test (Mottier-test; p < .001). The study was approved by the Ethics Committee of the Medical Department at the University of Freiburg and parents gave informed consent for their children. Stimuli were created using Cool Edit Pro (Cool Edit Pro 2002, version 2.0 Syntrillium Software Corporation, Phoenix, Arizona, USA). All stimuli were pseudo-randomized to ensure that no two deviants follow each other. The sine tones consisted of a 700 Hz tone (standard) and a 750 Hz tone (deviant). The deviant tone occurred with a probability of p = .15. The total number of tones was 1200. The duration of the tones was 150 ms each, including 20 ms rise and fall time for each tone. Interstimulus interval (offset-to-onset) was kept constant at 650 ms. Subjects were seated in a chair that was adjustable for height in an electrically shielded booth. Particularly younger boys were accompanied by a parent, sibling, or staff member in the cabin. Subjects watched a Disney cartoon of their choice on a computer monitor with the sound off. They were told to ignore the auditory input and to focus their attention on the movie as they were going to be asked detailed questions about it. The stimuli were presented binaurally through insert earphones (EAR Auditory Systems, Wheaton, IL, USA). The sound volume was kept at 70 dB sound pressure level (SPL). The EEG was digitally recorded with SynAmps amplifiers and Scan 4.0 software (NeuroScan, Inc., Sterling, Virginia, USA) from Ag/AgCl electrodes, using an extended 10–20-System electrode cap (EasyCap, Falk Minow Services, Herrsching-Breitbrunn, Germany), from midline sites Fz, Cz, and on each side F3/F4, F7/F8, FC1/FC2, FC5/FC6, FT7/FT8, T7/T8, C3/C4, CP1/CP2, CP5/CP6, TP9/TP10, P3/P4, O1/O2. The ground electrode was positioned on the midline at AFz, and Pz was used as on-line reference. The vertical electro-oculogram (VEOG) was recorded from above and below the right and left eye, and the horizontal electro-oculogram (HEOG) was recorded
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
101
Fig. 1. Difference wave (deviant minus standard) to sine tones at Fz. Standard (700 Hz) is the dotted line, and the deviant (750 Hz) the striped line. The bold line is the difference wave with two negative deflections. The response in control children is displayed in the left graph, the response of the children with SLI is displayed in the right graph. The second MMN is not significant in the SLI group at Fz and both, the first and the second MMN are not significant at central electrode Cz (not displayed here).
from the outer canthi of each eye. The on-line filter was at 0.1–70 Hz. The sampling rate of the electric signal was 500 Hz. Electrode impedance was kept below 10 k and below 20 k for VEOG and HEOG as these electrodes were placed directly on sensitive skin. The EEG raw data was stored on a hard disk and processed offline with BrainVision Analyzer 1.05 software (Brain Products GmbH, M¨unchen, Germany). Data was high-pass filtered at 0.2 Hz and low-pass filtered at 30 Hz. Filtered data were baseline corrected to −50 ms. The first five standards were excluded from the analysis. Standards directly following a deviant were also excluded. Eye movements and epochs with other artifacts were rejected if the activity was ±100 V. For the removal of eye blinks, the Gratton & Coles algorithm was applied. The difference wave was calculated as the difference between standard and deviant. Mastoid polarity reversal was inspected using the average reference. Then, linked mastoid reference was applied. In the sine tone paradigm, P1 and N250 were present, as can be seen in Fig. 1. Both groups demonstrated a clear early positivity and later negativity for standard and deviant tones, respectively. The grand-average wave forms showed a reduction in amplitude for the children with SLI. However, these differences were not statistically significant at Fz for standard or deviant ERPs (e.g. standard: P1 amplitude: t(24) = 1.347, p = .190; N250 amplitude: t(24) = −1.073, p = .294). As outlined in [5], age effects may be reflected in a reduction of the latency of the P1-component with increasing age. In the present data, a correlation of age and P1-latency confirmed
this finding, however, there was only a significant negative correlation for the control children (Fz: r = −.574, p = .020) but not for the children with SLI (Fz: r = −.174, p = .285). This finding may be an indication for a developmental delay in the children with SLI. Time windows for the analysis were determined if paired point-to-point t-test against zero was p < .05 in time intervals greater than 20 ms at Fz for the controls. Following this procedure, two time windows were determined for the sine tone paradigm from the data set of the controls: 140–280 ms (MMN1) and 420–620 ms (MMN2). Grand-average data is displayed in Fig. 1. Only area measurements (in V) are presented here (see Table 1). Especially in children, the inter-individual variability is very large. In the present study, MMN-amplitudes in controls varied at Fz from −1.0 to −7.2 V. Therefore, MMN-amplitudes were not employed for group comparisons in the present study. Area measurements revealed a significant MMN1 for controls at Fz (t(12) = −4.261, p = .001) and Cz (t(12) = −2.708, p = .019). For the children with SLI, a significant MMN1 could be observed at Fz (t(12) = −4.753, p < .001) but not at Cz (t(12) = −1.824, p = .093). Independent samples t-test revealed that this difference between groups was not statistically significant at Cz (t(24) = −0.793, p = 0.435). Other electrodes were also inspected and using independent samples t-test, significant group differences could be found for the more posterior electrodes CP5 (t(24) = −2.228, p = .035) and P3 (t(24) = −2.265, p = .033), and a trend towards significance for C3 (t(24) = −1.960, p = .062). These three electrodes, at which
Table 1 Mean activity and standard deviations (V) for MMN1 and MMN2 area for controls and SLI MMN1
MMN2
Fz Controls SLI
Cz
−1.98 V −1.30 V (±0.9)*
Level of significance: * p < 0.05.
(±1.7)*
Fz
−1.05 V −0.64 V (±1.3)
(±1.4)*
Cz
−1.57 V −0.72 V (±1.8)
(±1.5)*
−0.78 V (±1.0)* −0.31 V (±2.0)
102
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
Fig. 2. Topographical maps of the MMN-response to sine tones. Time windows: 140–280 ms (top) and 420–620 ms (bottom). The response of the controls is displayed on the left part of the graphs, the response of the children with SLI on the right part.
the greatest difference between the groups was observed, form a cluster of left-ward central-posterior electrodes. As this was seen as indication for hemispheric differences between groups, these group differences were further analyzed. A repeated measures ANOVA for left hemisphere electrode clusters (FC1/F3/C3/CP1) and right hemisphere electrode clusters (FC2/F4/C4/CP2) revealed a significant three-way-interaction of electrode × hemisphere × group (F(2165) = 3.273, p = .042) (Fig. 2). Area measurements revealed a significant MMN2 for controls at Fz (t(12) = −3.861, p = .002) and Cz (t(12) = −2.906, p = .013). For the children with SLI, a significant MMN2 could neither be observed at Fz (t(12) = −1.406, p = .185) nor at Cz (t(12) = −0.548, p = .593). However, independent samples t-test revealed that this difference between groups was neither
statistically significant at Fz (t(24) = −1.302, p = .205) nor at Cz (t(24) = −0.759, p = .455). Other electrodes were also inspected and, using independent samples t-test, significant group differences could be found for FC5 (t(24) = −2.243, p = .034), C3 (t(24) = −2.243, p = .014), and CP5 (t(24) = −2.269, p = .026). As for the early time window, a repeated measures ANOVA for left hemisphere electrode clusters (FC1/F3/C3/CP1) and right hemisphere electrode clusters (FC2/F4/C4/CP2) was conducted. A significant two-way-interaction of hemisphere × group was found (F(1000) = 7.109, p = .014), but not for electrode × hemisphere × group (F(1979) = 1.650, p = .203) (Fig. 2). In order to examine the possible effects of age on MMN1 and MMN2, Pearson’s Product Moment Correlation for MMN1 area and MMN2 area was carried out:
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
There was no effect of the group and age on the MMN1 area (Cz: r = −.83, p = .343; Fz: r = .027, p = .447). However, there was a significant negative correlation between age and MMN2 for the entire group (Fz: r = −.353, p = .039), and a trend at Cz (r = .−328, p = .051). This correlation was not significant for any of the groups (controls: Fz: r = −.409, p = .083; SLI: Fz: r = −.125, p = .342). In the present study, an oddball paradigm with sine tones was applied to investigate the frequency discrimination abilities of SLI and control children. A standard tone of 700 Hz and a 750 Hz deviant were used. ERP-responses (P1 and N250) did not differ significantly between groups. A significant age effect for the controls was found for P1-latency. The correlation was on the whole negative. This is in line with previous findings by Uwer et al. [25]. It has been argued that ERPs decrease with age due myelination processes increasing velocity [5]. However, this effect was not found for the children with SLI. This can be seen as an indication that different maturation processes are taking place (see also [3]). In two time windows (140–280 and 420–620), a significant fronto-central MMN1 and MMN2 were elicited for the controls with more left-localized activity for the second negativity. For the children with SLI, MMN1 was elicited only at Fz but not at Cz with a slightly right-localized focus. The area of MMN2 was not significant in children with SLI. This indicates a difference in the abilities of children with SLI to discriminate two tones of different frequency. The 700 Hz tone must be kept in sensory memory so that the incoming new tone of 750 Hz can be compared to it. The accuracy of this representation (and the role of auditory memory in this process) must therefore be questioned seeing as the discriminative response indexed by the MMN was quite weak in the children with SLI. Significant hemispheric differences were observed for both MMN1 and MMN2. This is in line with previous studies [9–11,20] where it has been argued that hemispheric specialization is different in children with SLI. A number of conclusions can be drawn from the studies by present as well as the studies by [9–11] and [25]: Even if the difference between standard and deviant is 50 ms as in [11], the duration of the stimulus is not as crucial as the frequency difference of the stimulus. The fact that Uwer et al. [25] also did not find group differences for a duration deviant lends support to this result. Therefore in the present study, the duration of the tone (150 ms) may have not have led to the observed group differences. Further, it may be assumed that the language difficulties in children with SLI are more strongly influenced by weak auditory discriminative abilities than by lower IQ. Even though the children in the present study had a higher IQ than the subjects of the Korpilahti-group, they nonetheless showed difficulties in the processing of a slight acoustic contrast. The independence of the influence of the IQ on the MMN has been reported before by Holopainen et al. [10]. In addition, the frequency discrimination deficit also appears to be less dependent on age. Even though the children in the present study were older than the children in the study by Uwer et al. [25], they showed a frequency discrimination deficit. This is supported by Korpilahti and Lang [11] and Holopainen et al. [9,10] who consistently showed a weaker or absent MMN despite investigating children of a wide age range.
103
The late component, MMN2 was not analyzed further in Uwer et al. [25] although group differences were observed for the latency of the second negativity. In the study by Korpilahti and colleagues [9–11], the time window was usually too short to analyze this component. In the present study especially the late component revealed group differences. Groups were in fact more discrepant in their responses in the later time window. MMN2 was absent in the children with SLI and overall activity was significantly reduced. The hemispheric differences between groups were quite striking – with the children with SLI showing a converse pattern of hemispheric activity than the controls. Previous studies with at-risk populations also demonstrated significant group differences in this second component [14]. These combined results suggest that the MMN2 may reflect similar comparison processes as the MMN1. It may be assumed that the characteristics of the MMN2 depend on the level of automatization of the comparison process as it was found to decrease with age [6]. For future studies, it may therefore be a worth-while endeavor to enlarge the window of analysis, especially in the investigation of group differences between clinical populations and controls. This study is only an exploratory study, therefore more studies are needed investigating other frequency ranges. We cannot draw any explicit conclusion, but the present study supports a frequency discrimination deficit in children with SLI as previously reported by a number of authors. This deficit may become apparent for small frequency changes (e.g. around 50 Hz or less) and, combining the results of the present study and [9–11], in a frequency range below 750 Hz. Previous studies that controlled speech stimuli for their spectral characteristics have found that differences between language impaired groups and controls arise from small acoustic transitions in the speech material [16,21]. This can be seen as evidence for the relevance of a possible frequency discrimination deficit for the language acquisition process and may have important implications for diagnosis and intervention. Acknowledgments We would like to thank all participating children and families. This research was supported by “Fonds f¨ur interdisziplin¨are Forschungsvorhaben”, University of Freiburg Rectorate. This study was a portion of the doctoral dissertation presented by the first author. References [1] A.A. Benasich, P. Tallal, Infant discrimination of rapid auditory cues predicts later language impairment, Behav. Brain Res. 136 (2002) 31–49. [2] A.A. Benasich, N. Choudhury, J.F. Friedman, T. Realpe-Bonilla, C. Chojnowska, Z. Gou, The infant as a prelinguistic model for language learning impairments: Predicting from event-related potentials to behavior, Neuropsychologia 44 (2006) 396–411. [3] D.V.M. Bishop, G.M. McArthur, Immature cortical responses to auditory stimuli in specific language impairment: evidence from ERPs to rapid tone sequences, Dev. Sci. 7 (2004) 11–18. ˇ [4] R. Ceponien˙ e, M. Cheour, R. N¨aa¨ t¨anen, Interstimulus interval and auditory event-related potentials in children: evidence for multiple generators, Electroen. Clin. Neuro.: Evoked Potentials 108 (1998) 345–354.
104
T. Rinker et al. / Neuroscience Letters 413 (2007) 99–104
ˇ [5] R. Ceponien˙ e, T. Rinne, R. N¨aa¨ t¨anen, Maturation of cortical sound processing as indexed by event-related potentials, Clin. Neurophysiol. 113 (2002) 870–882. [6] M. Cheour, P. Korpilahti, O. Martynova, A.H. Lang, Mismatch negativity and late discriminative negativity in investigating speech perception and learning in children and infants: a review, Audiol. Neuro-Otol. (2001) 2–11. [7] P.R. Hill, J.H. Hogben, D.V.M. Bishop, Auditory frequency discrimination in children with specific language impairment: a longitudinal study, J. Speech Hear. Res. 48 (2005) 1136–1146. [8] P.R. Hill, G.M. McArthur, D.V.M. Bishop, Phonological categorization of vowels: a mismatch negativity study, Neuroreport 15 (2004) 2195–2199. [9] I. Holopainen, P. Korpilahti, K. Juottonen, A.H. Lang, M. Sillanp¨aa¨ , Attenuated auditory event-related potential (mismatch negativity) in children with developmental dysphasia, Neuropediatrics 28 (1997) 253–256. [10] I. Holopainen, P. Korpilahti, K. Juottonen, A.H. Lang, M. Sillanp¨aa¨ , Abnormal frequency mismatch negativity (MMN) in mentally retarded children and in children with developmental dysphasia, J. Child Neurol. 13 (1998) 178–183. [11] P. Korpilahti, A.H. Lang, Auditory ERP Components and MMN in dysphasic children, Electroen. Clin. Neuro. 91 (1994) 256–264. [12] L.B. Leonard, Children with Specific Language Impairment, MIT Press, Cambridge, MA, 1998. [13] A.D. Lowe, R.A. Campbell, Temporal discrimination in aphasoid and normal children, J. Speech Hear. Res. 8 (1965) 313–317. [14] U. Maurer, K. Bucher, S. Brem, D. Brandeis, Altered responses to tone and phoneme mismatch in kindergarteners at familial dyslexia risk, Neuroreport 14 (2003) 2245–2250. [15] G.M. McArthur, D.V.M. Bishop, Which people with specific language impairment have auditory processing deficits? Cogn. Neuropsych. 21 (2004) 79–94.
[16] G.M. McArthur, D.V.M. Bishop, Speech and non-speech processing in people with specific language impairment: a behavioural and electrophysiological study, Brain Lang. 94 (2005) 260–273. [17] E.D. Mengler, J.H. Hogben, P.T. Mitchie, D.V.M. Bishop, Poor frequency discrimination is related to oral language disorder in children: a psychoacoustic study, Dyslexia 11 (2005) 155–173. [18] R. N¨aa¨ t¨anen, Attention and Brain Function, Lawrence Erlbaum, Hillsdale, NJ, 1992. [19] R. N¨aa¨ t¨anen, A.W.K. Gaillard, S. M¨antysalo, Early selective-attention effect on evoked potential reinterpreted, Acta Psychol. 42 (1978) 313– 329. [20] V.L. Shafer, R.G. Schwartz, M.L. Morr, K.L. Kessler, D. Kurtzberg, Deviant neurophysiological asymmetry in children with language impairment, Neuroreport 11 (2000) 3715–3718. [21] J.E. Sussman, Vowel perception by adults and children with normal language and specific language impairment: based on steady states or transitions? J. Acoust. Soc. Am. 109 (2001) 1173–1180. [22] P. Tallal, Experimental studies of language learning impairments: from research to remediation, in: D.V.M. Bishop, L.B. Leonard (Eds.), Speech and Language Impairments in Children: Causes, Characteristics, Intervention and Outcome, Psychology Press, Hove, UK, 2000, pp. 131– 156. [23] P. Tallal, M. Piercy, Defects of non-verbal auditory perception in children with developmental aphasia, Nature 241 (1973) 468–469. [24] P. Tallal, M. Piercy, Developmental aphasia: impaired rate of nonverbal processing as a function of sensory modality, Neuropsychologia 11 (1973) 389–398. [25] R. Uwer, R. Albrecht, W. von Suchodoletz, Automatic processing of tones and speech stimuli in children with specific language impairment, Dev. Med. Child Neurol. 44 (2002) 527–532.