Sensory bases of reading and language disorders

Submitted Abstracts

Perceptual centres in speech, phonological representation and dyslexia

Usha Goswami¹, Sophie Scott¹ and Johannes Ziegler²

¹Institute of Child Health, University College London, London, UK.

²CNRS, University of Provence, France.

Dyslexic children are thought to have difficulties in the accurate specification and neural representation of the sequential sounds of speech. As these sequential sounds are phonemes, the search for the basic neural mechanisms underlying the phonological deficit in dyslexia has focused at the phonemic level. Developmentally, however, children appear to represent syllables and onset-rimes prior to phonemes. Logically, therefore, a deficit in a neural mechanism that contributes to syllabic representation could be primary in explaining dyslexic children’s difficulties in phonological representation.

We propose and test a new candidate neural mechanism that may be impaired in developmental dyslexia, that which gives information about the 'perceptual centre' or actual moment of occurrence of sounds (both speech and non-speech). As 'P-centre' processing is based on the onsets of sounds around certain more sonorous frequencies, which in speech correspond to characteristics of the vowel onset, we propose that 'P-centre' processing is the perceptual correlate of the onset-rime division in any syllable (s-eat, sw-eet, str-eet). As phonological awareness of onsets and rimes is usually present prior to literacy, a deficit in P-centre processing would cause problems in the accurate specification of phonological information and would have consequent effects for literacy. Data from 2 studies is presented. In the first, we show significant differences in 'P-centre' processing between English dyslexic and normally-reading children, and we further show that these differences are related to reading development, phonological representation and verbal ability. In the second, we show that P-centre processing in normally reading French adults is related to performance in an auditory lexical decision task, such that adults with poorer P-centre processing show larger inhibitory neighbourhood density effects. Both results are consistent with our overall hypothesis that P-centre processing underpins the development of phonological representations for syllables and their specification in terms of onset-rime units.

Auditory perceptual processing in people with a specific language impairment: A comparison of behavioural and electrophysiological responses

G. M. McArthur and D. V. M. Bishop

Department of Experimental Psychology, University of Oxford, Oxford, UK.

The frequency discrimination and rapid auditory processing skills of people with a specific language impairment (SLI) were examined in two experiments. Experiment 1 collected behavioural responses to auditory backward recognition masking (ABRM) stimuli and unmasked stimuli from 16 people with an SLI and 16 people with normal spoken language skills. The results indicated that (1) there is a subgroup of people with SLI who have poor behavioural responses to masked and unmasked stimuli, (2) these people are characterised by poor nonword reading, and (3) their poor performance stems from an inability to discriminate between the frequencies of sounds rather than an inability to process rapid sounds. Experiment 2 collected electrophysiological responses (the mismatch negativity component of the event-related potential) to the same stimuli from the same group of people. These suggested that the poor behavioural scores of the SLI poor nonword readers to masked and unmasked stimuli in Experiment 1 did not stem from a low-level auditory processing deficit. The implications of this pattern of results will be discussed.

Deficits in processing efficiency in individuals with specific language impairment and dyslexia
D. E. H. Hartley and D. R. Moore
University Laboratory of Physiology, University of Oxford, Oxford, UK.

Claims have been made that children with SLI and dyslexia have difficulty separating rapidly presented or brief stimuli, which putatively degrades the ability to perceive the brief elements of speech. These claims, known as the 'temporal processing hypothesis', gained support through demonstrations that children with SLI (Wright, B.W., et al. Nature, 1997; 387:176-178) and adults with dyslexia (Wright, B.W., et al., unpublished data) had severe deficits for the perception of a brief tone presented before a masking noise (backward masking). Children with SLI also had deficits for perception of a tone presented after a noise (forward masking). However, when the tone was presented during the noise (simultaneous masking) participants with language and reading impairments performed nearly as well as controls. It was concluded from these studies that language and reading impairments are associated with poor temporal resolution. Temporal resolution is the minimum time interval within which different acoustic events can be distinguished. Here we report an analysis of data in terms of a model of temporal resolution (Moore, B.C.J., et al. JASA, 1988; 83:1102-1116). The model consists of four stages: i) a filtering device, ii) a non-linear device, which simulates the input-output function of the basilar membrane, iii) a sliding temporal integrator or ‘temporal window’ and, lastly, iv) a decision device. In terms of the model, poor temporal resolution suggests a widened temporal window. However, our results do not support the temporal processing hypothesis and, instead, suggest that poor processing efficiency can account for deficits in backward and forward masking amongst individuals with SLI and dyslexia. Processing efficiency encompasses all factors, aside from temporal resolution, that may affect detection, such as attention, cognition and motivational factors. In terms of the model, this suggests that, at threshold, subjects with SLI and dyslexia require a higher signal-to-noise ratio at the output of the temporal window, compared with controls, regardless of the specific masking task. Consistent with experimental data, the model also predicts that backward masking i) is worse in younger children, ii) is more prone to training effects, iii) has greater inter- and intra-subject variability and, iv) increases less with masker level, than other masking tasks.

Acknowledgements
Supported by the Wellcome Trust & Medical Research Council.

Why does colour help reading - and also reduce headaches and seizures?

A. J. Wilkins

Department of Psychology, University of Essex, Colchester, UK.

It is often assumed that magnocellular deficits (observed initially in dyslexia, but now also in migraine) provide an explanation for the improvements in reading fluency that result from the use of coloured filters (e.g. Lehmkuhle, 1993). Magnocellular deficits can not, however, explain the large individual differences that are observed with respect to the colour optimal for reading, or, indeed the specificity of this optimum. We will present evidence that for maximum reading fluency the colour has to be selected to within a precision of two jnd's. We will also present an explanation for these effects that can account for the benefits of coloured filters that we have recently demonstrated in patients with migraine and photosensitive epilepsy. The explanation is consistent with the co-morbidity of dyslexia and migraine, and with the family history of migraine often present in individuals whose reading improves with colour.

Lehmkuhle, S. (1993). Neurological basis of visual processes in reading. In D.M. Willows, R.S. Kruk and E. Corcos (eds.) Visual processes in reading and reading disabilities. Hove: Erlbaum

Only reading disabled individuals with additional cognitive difficulties have severe acoustic processing deficits

Karen Banai¹ and Merav Ahissar^2
1Interdisciplinary Center of Neural Computation, Hebrew University of Jerusalem, Israel.
²Department of Psychology, Hebrew University of Jerusalem, Israel.

Numerous studies documented the acoustic processing deficits of individuals with reading disabilities. Results are mixed. While some found substantial deficits, other studies found only marginal effects. In the present study we asked whether this variability could be attributed to different sampling of reading disabled participants. Thirty-four reading disabled (RD) and 36 controls (age and education matched) were tested on psychoacoustic tasks (frequency, duration and intensity discrimination), speech discrimination, phonological awareness, reading and cognitive tasks. As a group, RDs performed significantly poorer than controls on frequency (JND (SD) = 11(12)% and 3.6(4)% for RDs and controls respectively) and duration discrimination tasks, but not on intensity discrimination. This large difference stemmed from very poor performance of a subgroup of RDs (JND > 10%). The RDs with very poor psychoacoustic performance were not worse readers than other RDs, and were not poorer on phonological awareness. They were, however, significantly poorer than the other RDs (and controls) on speech discrimination and verbal memory tasks. Their performance on problem solving tasks was also significantly impaired. Particularly poor was their performance on Raven Matrices, a matrix completion task in which adequate performance requires adequate memory (typically verbal). Poor acoustic processing is thus clustered with difficulties in a range of cognitive tasks related to verbal memory. Consequently, categorizing reading disabled individuals according to their psychoacoustic performance aligns with the traditional, cognitive based, diagnosis of specific (“dyslexia”) and non-specific (“poor reading”) reading disability.

Temporal contrast sensitivity of reading disabled individuals is specifically impaired in retain-and-compare paradigms
G. Ben-Yehudah¹ and M. Ahissar^2
1Deptartment of Neurobiology, Hebrew University of Jerusalem, Israel.

²Department of Psychology, Hebrew University of Jerusalem, Israel.

Many studies, motivated by the hypothesis that the magnocellular system of a large portion of reading disabled (RD) is impaired, tested temporal contrast sensitivity in disabled readers. Different studies, using different stimuli and paradigms, found mixed results. While some found the predicted deficits in detecting low spatial frequencies presented at high temporal frequencies, others did not. In the present study, we assessed contrast sensitivity in a group of adult Israeli RDs and controls (matched for age and education), using both drifting and flickering gratings (0.5-16c/deg) under two different behavioral paradigms. We found both impaired and normal temporal contrast sensitivity within the same group of RDs. When we applied sequential presentation, using a temporal forced choice paradigm, RDs showed impaired sensitivity both to drifting and flickering gratings. However, under simultaneous presentation, using a spatial forced choice paradigm, RDs' sensitivity did not differ from controls'. Within each paradigm, RDs' sensitivity was poorer at higher temporal frequencies, consistent with the magnocellular hypothesis. To determine the relation between our finding that RDs' temporal contrast sensitivity is significantly impaired only in sequential presentations and the magnocellular hypothesis, we further assessed these individuals' sensitivity to whole screen flicker (effectively 0c/deg at 5-25Hz; temporal forced choice paradigm). Based on monkey lesion studies, a magnocellular deficit should be detrimental to the detection of this stimulus. Indeed, several RDs (8/38) were severely impaired in this task. However, within the RD group, there was no correlation between this measure and a measure of serial performance (ratio between sensitivity to flickering gratings under sequential and simultaneous presentations). These results suggest that RDs have a basic perceptual impairment in retaining-and-comparing perceptual traces, regardless of their magnocellular-related performance.

Magnocellular processing and reading disability

Sygal Amitay¹ and Merav Ahissar^1,2
1Interdisciplinary Center for Neural Computation, Hebrew University of Jerusalem, Israel.

²Department of Psychology, Hebrew University of Jerusalem, Israel.

Visual deficits often concur with reading disability. These deficits have been attributed to impaired magnocellular processing. Yet not all data are consistent with this interpretation, and, moreover, the relationship between the magnocellular impairment and reading is unclear. In this study we asked whether a magnocellular deficit is a good description of the visual deficits of any subgroup of reading disabled individuals, and whether any similar deficits are found in controls with similar age and education. Towards this goal we tested speed discrimination of drifting gratings, contrast sensitivity to flicker and drifting gratings, and coherent motion direction detection under several conditions. We found that there is indeed a subgroup of reading disabled subjects (8/30) with consistent difficulties in tasks probing mangocellular processing. However, the difficulties of this subgroup were not limited to mangocellular tasks, and were accompanied by difficulties in other, similar tasks not optimally probing magnocellular processing. Performance on magnocellular tasks did not account for the variability in reading and spelling after general cognitive abilities were partialled out, but was the major contributor in accounting for the variability in a task involving eye-hand coordination. These findings are consistent with dorsal stream dysfunction in a small subgroup of reading disabled individuals, but preclude a direct contribution of such a deficit to reading.

Could an auditory modulation processing deficit cause phonological impairment?
C. Witton, J. B. Talcott, C. J. Stoodley and J. F. Stein.
University Laboratory of Physiology, University of Oxford, Oxford, UK.

It is well known that deficits in phonological skills are central to the reading problems associated with developmental dyslexia. Reduced ability to attend to or utilise the sound structure of language has been suggested to result from impaired auditory processing of dynamic or temporal aspects of sound. In this presentation we address how processing of acoustic modulations in amplitude and frequency (AM and FM) might be related to phonological and orthographic skills in dyslexic and control readers. We then discuss whether the mild deficits in FM and AM processing found in developmental dyslexics could cause their phonological skill impairments.

Detection thresholds for AM and FM of a 1-kHz tone were obtained for a group of disabled readers and a group of age- and ability-matched controls, using a standard psychophysical technique. The reading-disabled group was less sensitive on average than the controls to 2 Hz FM and 20 Hz AM, but not to 2 Hz AM or 240 Hz FM. The significant group differences remained even after controlling for subjects whose high psychophysical thresholds could be caused by lapses in attention or short-term memory during the testing procedure. Thresholds for detecting 2 Hz FM and 20 Hz AM (but no other auditory stimulus) were independent statistical predictors of phonological skill. No auditory threshold predicted significant variance in orthographic skills. Could the correlations between auditory thresholds and phonological skills reflect a causal relationship?

The most likely route by which hearing could relate to phonological skill acquisition is via speech perception. These data will be discussed in terms of theories regarding the role of modulations in the early development of language skills and in speech perception.

Sensory processing variables and children’s reading skills: An examination of disability subtypes in a primary-school sample
Joel B. Talcott, Catherine J. Stoodley, Gill S. Hebb, Elizabeth A. Westwood, Susan J. France, Peter C. Hansen, Caroline Witton, and John F. Stein
University Laboratory of Physiology, University of Oxford, Oxford, UK.

Three hundred and fifty randomly selected primary school children completed a large psychometric and psychophysical test battery to ascertain relationships between the putative subtypes of reading (dis)ability as posited by the dual route model and sensitivity to dynamic visual and auditory stimuli. Contrasting children's performance on accuracy measures for exception words and nonwords showed that 19% of the sample had orthographic skills and/or phonological skills in the lowest 10th percentile for their age group. Although some of the children demonstrated a profile on these tasks characteristic of either developmental phonological dyslexia (7.4%) or developmental surface dyslexia (3.7%), the largest subtype group (8.9%) had impaired accuracy on both measures. The two groups with poorer orthographic skills were generally impaired, relative to both the controls and a group with relatively poorer phonological skills, on all of the dependent measures of processing time, suggesting convergence between the developmental surface dyslexic and the rate impaired subtype suggested by the double deficit model. On the psychophysical tasks, group differences on the visual motion task were significant only for the group with impairments in both orthographic and phonological processing. The group with poor phonological processing skills were impaired relative to controls and the group with poor orthographic skills on a measure of auditory frequency discrimination, suggesting the potentially important role of auditory processing skills to the development of phonological skills and deficits.

Auditory temporal resolution in ferrets following ear plugging
D. R. Moore, L. A. Hulvershorn, C. H. Parsons, O. Kacelnik and A. J. King

University Laboratory of Physiology, University of Oxford, Oxford, UK.

Auditory temporal resolution has received much recent attention because of evidence that deficits of temporal resolution contribute to a variety of language impairments in children. We have been examining the acoustic and neural bases of backward masking (BM), a common test of temporal resolution, in children and in ferrets. In children, we have have confirmed a correlation between BM and language impairment (see Hill and Moore, this meeting; Hartley et al., 2001, ARO, Brit. J. Audiol.). However, we have not found any relation between BM and a history of middle ear disease (MED; Hartley et al., 2001), despite various other hearing deficits seen in children who have had excess MED.

In this study we have examined a possible relationship between hearing loss and auditory temporal resolution further by measuring BM and simultaneous masking (SM) in ferrets reared with or without ear plugs. Ferrets were used because (i) we can control the duration and type of the hearing loss, (ii) we can impose the hearing loss from the time begins, and (iii) we can examine neural correlates of BM and SM. Ferrets were trained to distinguish ‘tone' trials (1 kHz) from ‘noise' trials. On ‘tone' trials, the tone was presented alone, or before or during a noise burst of fixed level (68 dB SPL) and duration (300 ms). Tone level was varied adaptively and tone duration was varied from 20 – 200 ms. On some trial blocks, the tone was presented without the noise, and the ferret discriminated between tone and silence, cued by a light. Ferrets were either (i) normally hearing, (ii) reared for 3-6 months from P26-28 with a unilateral plug that alternated monthly between the ears, (iii) received 3-6 months alternating monaural plugging in adulthood, or (iv) reared with bilateral ear plugs. Training on the task began 3 months after plugging started. The plug(s) were removed about 3 months later, before testing started.

We previously showed (Moore et al, ARO Abs, 2000, 2001) that ferrets have excess BM (by about 25 dB), compared with humans. In contrast, SM thresholds were more similar to those of humans. Unilaterally ear plugged ferrets performed as well as normally hearing ferrets on all tasks, after the ear plug was removed. However, bilaterally plugged (BP) ferrets were much poorer than normal on BM (20 ms tone, 0 ms delay). Mean BM tone thresholds in normal ferrets decreased from 75 to 69 dB (SPL) over three blocks of trials. BP ferret BM tone thresholds exceeded 90 dB in the first block of trials, and decreased to a threshold of 87 dB in the third trial block. SM thresholds using long (200 ms) tones were normal in the BP ferrets. We are currently studying whether other aspects of short tone perception (SM and pure tone thresholds) are impaired in the BP ferrets.

The processing of sound differs between infants with and without familial risk for dyslexia
Heikki Lyytinen^1,4, Tomi Guttorm¹ Paavo Leppänen² and Ulla Richardson^3
1Department of Psychology & Child Research Center, University of Jyväskylä, Finland
²Center for Molecular and Behavioral Neuroscience,Rutgers,The State University of New Jersey, USA
³Institute of Cognitive Neuroscience, University College London, UK
⁴Niilo Mäki Institute, Jyväskylä, Finland

In the Jyväskylä Longitudinal study of Dyslexia (JLD), 100 potentially dyslexic children (PD) born to a dyslexic parent (having not less than one close dyslexic relative) have been followed from birth. Comparisons with infants (100) born to families without such a risk have been made in the JLD using different technologies, including e.g. brain event-related potentials (ERP) and a head turn paradigm. The ERPs have been collected immediately following birth and at six months of age using a number of different stimulus set-ups and conditions. These include oddball paradigm (to measure mismatch negativity, MMN) and equiprobable presentation of synthetic (ba-da-ga) syllables of non-Finnish origin by content and natural (paa-taa-gaa) syllables of which first two occur in Finnish language usage. An other source of difference between the syllables was the voice onset time. The head-turn procedure was used at six months of age to assess categorical perception of duration of speech sound within a pseudo-word context (/ata/ - /atta/).

The results show that PD group differs from the control infants matched in terms of parental IQ and education, as well as the infant’s own IQ (as measured later using Bayley Scales) in a number of the contexts listed and on several variables. In the most early recordings at the age of a few days, the ERPs to syllable contrasts revealed (independent of the way of the analysis – traditional averages or in scores computed using principal component analysis of the ERP-waveform) a number of significant group x consonant x hemisphere interactions. The quantity discrimination was associated with differential MMN-type waves between groups with radically different MMN under certain conditions. In addition, at six months, the behavioral head-turn conditioning revealed a less sharp categorical perception of quantity associated with the /t/-consonant manipulated using segments of silence to lengthen its perceived duration. The PD-infant required a longer silence (similarly with their parents in comparison to controls) to perceive the sound as long. The ERPs and behavioral data reveal relatively strong but complicated correlations culminating in differential scores between hemisphere-specific ERP-amplitudes and perception. This same hemisphere-related difference also characterizes the PD group in the results of some other experimental conditions listed above.

Dissociating reading related from cognitive related perceptual processes
Merav Ahissar
Hebrew University of Jerusalem, Israel.

In the last 2 decades two largely independent lines of study characterized individual differences in perceptual processing. In one, studies of intelligence, it was found that thresholds in a broad range of simple perceptual tasks, including pure tone frequency discrimination, are higher in individuals with lower scores in cognitive tasks. In the second, studies of reading disability, it was found that many poor readers show poor performance in simple psychoacoustic tasks. In attempt to incorporate these sets of observations within a single framework, we characterized the performance of more than 100 adults. Half the participants were reading disabled and half were age and education matched controls. Each group was further divided according to individuals' scores in a problem-solving cognitive task (Raven Matrices). We compared perceptual performance across the resulting 4 groups. We reasoned that in a task which is only reading related, both groups of reading disabled would perform poorer than controls. In a task which is only cognitive related, the two groups with higher cognitive scores would perform better.

We found that in a range of tasks, including pure tone frequency discrimination, poor readers with poorer cognitive abilities perform worst; good readers with high scores in cognitive tests perform best; the other two groups have similar, intermediate, thresholds. Thus, compared with intelligence matched controls, reading disabled with good cognitive abilities are poorer performers. Yet, their scores are well within the range of other controls with fine reading abilities (though lower cognitive scores).

One testing condition yielded a clear reading related segregation. When pure tone frequency discrimination was measured with very brief inter-stimulus-intervals (100-200ms), both groups of reading disabled showed similarly high thresholds (10%). When intervals were increased to 0.5sec and longer (up to 2sec), the performance of reading disabled with high cognitive scores improved dramatically and reached the typically measured values (3%). This range of intervals (0.5-2sec) is typically used in psychophysical studies. Thresholds of the other 3 groups were not sensitive to inter stimulus intervals. We suggest that reading disabled with high cognitive abilities suffer from impaired ability in anticipating incoming stimuli at specific timings, rather than from a comprehensive perceptual deficit.

Binaural unmasking and temporal order discrimination in adult dyslexics

Y. M. Griffiths¹, N. I. Hill², P. J. Bailey¹ and M. J. Snowling¹

¹Department of Psychology, University of York, York, UK.

²Department of Psychology, University of Essex, Colchester, UK.

This presentation is concerned with two specific auditory deficits that have been reported for groups of dyslexic individuals relative to normal readers: (i) impaired ability to benefit from binaural cues when detecting a tone in noise, manifest as reduced binaural masking level differences (BMLD) [McAnally & Stein (1996), Proc. Roy. Soc. B, 263, 961-965], and (ii) impaired ability to judge the temporal order of rapidly-presented sounds [Tallal (1980), Brain & Lang., 9, 182-198].

Hill et al (1999) [J. Acoust. Soc. Am., 106, L53-L58] did not find systematic differences between BMLDs for dyslexic and normal-reading listeners, contrary to McAnally and Stein. The difference in outcome may have arisen because the two studies measured BMLDs at different frequencies: Hill et al sought to maximise the BMLD magnitude by using a relatively low frequency (200 Hz), whereas McAnally and Stein used a higher frequency (1 kHz). It is possible that differences between BMLDs for dyslexic and normal-reading listeners are more evident at higher frequencies, closer to the limits of availability of interaural time-difference cues. This hypothesis was tested by measuring BMLDs at 200 Hz and at 1 kHz in 18 adult dyslexic and 18 control listeners. The dyslexic listeners differed reliably from controls on measures of reading, spelling and phonological processing, but were matched on age. The mean BMLD at 200 Hz was approximately 17 dB, compared with 9 dB at 1 kHz. Consistent with our previous finding, there was no reliable group difference in the magnitude of the BMLD at either frequency. However, in contrast with our previous experiment and McAnally and Stein, mean detection thresholds for tones in noise were significantly higher for dyslexic listeners than control listeners.

In the second experiment reported here we assessed the ability of dyslexic listeners to extract temporal order information from successive tone pairs. Performance was assessed using a four-interval, two-alternative forced choice procedure. In each interval a pair of tones was presented, one of which was always at 1 kHz and the other at a higher frequency. The order of the tones in the pair was the same for three of the four intervals (low-high), but was reversed (high-low) in either the second or third interval at random. The listeners’ task was to select the target interval in which the tone pair order was reversed. The threshold frequency difference corresponding to 71% correct discrimination was determined using a 2-down, 1-up adaptive procedure. All tones had duration of 75ms, and the within-pair inter-stimulus interval (ISI) was either 20 ms or 200 ms. Data were obtained from twenty adult dyslexic listeners and twenty controls matched on age and IQ. Temporal order discrimination performance was better for both groups at the longer ISI. Although the dyslexic listeners differed reliably from control listeners on measures of reading, spelling and phonological processing, there was no reliable group difference in performance on the temporal order discrimination task at either ISI.

Acknowledgement
This work was supported by project grant 55671 from the Wellcome Trust.

Magnocellular processing and reading: Can negative results be reconciled?
Bill Lovegrove and Elizabeth Conlon
Griffith University, Australia

Three experiments are reported. In the first three measures of visual sensory processing were obtained on a sample of good and poor readers. These were spatial and temporal contrast sensitivity, visual coherent motion and spatial and temporal dot counting. Effect sizes for differences between groups on spatial and temporal CSF were negligible with significant differences found on the coherent motion and temporal dot counting tasks. These results were found using overall reading performance, measured by the WRAT. Poor phonological skill measured by non-word reading on its own did not discriminate between groups on the visual processing measures. The way different component sub-skills of reading were related to magnocellular processing was further investigated in Experiment 2 where an unselected sample of adult readers was used. Measures of visual attention, phonological and orthographic sub-skills of reading, word identification skills and coherent motion were obtained. A strong relationship was found between performance on the coherent motion task and general word identification skills (WRAT). Performance on a measure of visual orthographic processing accounted for an increased percentage of the variance but phonological awareness and visual attention did not. In the third experiment good and poor readers performed two tasks using random dot displays. In the first task global motion thresholds were obtained and in the second pattern thresholds obtained. The results demonstrate three things. First, that poor magnocellular processing may exist on a continuum in the population. Second, that performance may be related to the visual components of word processing not phonological processing skill. Third performance on the global motion task may be influenced by attentional components in addition to motion processing. The possible nature of this attentional component will be discussed.

Variability between and within studies: what does it mean?

D. V. M. Bishop

Department of Experimental Psychology, University of Oxford, Oxford, UK.

Theories that implicate perceptual deficits in causing language and literacy disorders are plausible and parsimonious, yet after decades of research on this topic, there is little agreement about empirical findings. Someone unfamiliar with the field who browsed through the abstracts for this meeting may come away with a sense of despair at the lack of solid, generalisable results. It seems that no sooner does one research group publish a finding demonstrating perceptual deficits than another publishes a failure to replicate. Rather than progressing forward, the field comes to resemble a ping-pong match, with each new claim being lobbed back over the net into the domain of scepticism. I aim to consider why this state of affairs exists, and to propose ways of remedying it. I shall start with a brief review of the contradictory findings, then move on to consider possible reasons for inconsistencies between studies, before discussing the related issue of variable findings of deficit within a clinical group. A tentative conclusion is that positive findings are not simply artefacts of poor methodology; the association between perceptual deficit and language/literacy problems is real but weak, with substantial overlap between control and impaired groups. If this is the case, we need to abandon the quest for theories that postulate a single underlying factor causing impairment, and to consider more complex etiological accounts. Three such accounts will be discussed: (a) etiological heterogeneity; (b) pleiotropy; (c) multiple risk factors.

Can auditory temporal order judgment be used as an effective measure of individual differences in dyslexics?
Steve Heath and John Hogben
Department of Psychology, The University of Western Australia

There has been considerable support for the contention that children with reading and oral language deficits have impaired perception of sequential sounds. However, it seems that the establishment of reliability and construct validity for measures of auditory temporal processing has been largely ignored. Fifty-two Dyslexics and Good Readers aged 8 to 11 years performed an adaptation of the Tallal (1980) Repetition Test, together with measures of phonological processing and oral language ability. A subgroup was retested on the perceptual task three months later, and after extended practice. A further subgroup completed a rate of auditory processing task.

Deficits in auditory temporal order judgment (ATOJ) were found in a subset of Dyslexics, but not in Good Readers. These were related to weakness in oral language and phonological processing. In spite of very significant practice effects, moderate test-retest reliability over time and after practice was observed. However, performance on the two perceptual tasks was not significantly correlated. These results call into question the use of ATOJ as an individual measure of auditory temporal processing ability.

Sensory deficits do not play a causal role in specific reading impairment
Franck Ramus¹, Stuart Rosen², Steven Dakin³ and Uta Frith^1
1Department of Psychology, University College London, UK.

²Department of Phonetics and Linguistics, University College London, UK.

³Institute of Ophthalmology, University College London, UK.

The phonological theory of dyslexia postulates that the cause of specific reading impairment is a deficit that is restricted to the representation of speech sounds. However, low-level sensory impairments, both in the auditory and in the visual modality, have also been shown in dyslexics, thereby questioning the specifity of the phonological deficit. But are these sensory deficits the cause of developmental dyslexia, or do they simply co-occur with it?

Under the assumption that high-achieving dyslexics are less likely to have several independent impairments, the present study focuses on university students. The dyslexic group was recruited on the basis of previous assessments by qualified psychologists and of their persisting reading and spelling difficulties.

Each subject was administered 1) standard psychometric tests (WRAT, WAIS), 2) standard phonological tests (non-word reading and repetition, spoonerisms, rapid automatic naming), 3) auditory tests (syllable categorisation and discrimination, backward/simultaneous masking, frequency modulation detection, temporal order judgement) and 4) visual tests (magno/parvo grating detection, magno/parvo speed discrimination, motion integration, Ternus display).

Results show that only a small fraction of our dyslexic subjects have a visual and/or auditory impairment, whereas all of them have a phonological impairment. This study therefore suggests that 1) sensory deficits can account neither for a phonological deficit, nor for pure varieties of developmental dyslexia (i.e., not associated with SLI, ADHD or other learning disabilities); and 2) a phonological deficit seems to be the best causal explanation for specific reading impairment.

Outliers in the association of auditory processing disorders with dyslexia

Stuart Rosen

Department of Phonetics and Linguistics, University College London, UK.

It has long been clear that people with dyslexia exhibit an auditory sensory deficit more often than would be expected from measurements on appropriate control groups. On the other hand, it is also clear that such sensory disorders affect only a minority of dyslexics. More telling is the observation that, at least in some studies, group differences, and significant correlations of reading ability with auditory processing abilities, appear only to arise because of a few very poorly performing subjects. Few if any researchers in this area explicitly employ statistical techniques for the identification of outliers. A number of previously published studies are systematically examined in order to determine the extent to which outliers are responsible for the associations reported.

Behavioral, neuroimaging and family/genetic studies help clarify the role of rapid auditory processing in language and reading

Paula Tallal¹, April Benasich¹, Judy Flax¹, Elise Temple², John Gabrieli², Russ Poldrack³ and Mike Merzenich⁴

¹Rutgers University, USA.
²Stanford University, USA.
³Harvard University, USA.
⁴UCSF, USA.

Considerable research has focused on the role of central auditory processing disorders in both developmental language impairments and dyslexia. Specifically, research has focused on an inability to process sensory input that enters the nervous system rapidly in succession, and the relationship between individual differences in processing rapid acoustic information and phonological processes important for both oral and written language. Results of prospective, longitudinal studies with infants, as well as behavioural and psychoacoustic studies with school age children and adults, have shown significant correlations between individual differences in various aspects of spectrotemporal auditory processing and various language learning abilities. Family/genetic studies have also shown a high degree of comorbidity between developmental language impairment, reading impairment and auditory rate processing impairments. Recently, fMRI was used to identify the brain basis of rapid acoustic processing for both non-linguistic and linguistic stimuli. Normal adult readers showed left prefrontal activity in response to both rapidly changing, relative to slowly changing, nonlinguistic as well as linguistic acoustic stimuli. Dyslexic readers showed no differential left frontal response to either the non-linguistic or linguistic stimuli, based on the rate of acoustic change. A subset of the dyslexic subjects participated in intense auditory training with both non-linguistic and linguistic stimuli, using the Fast ForWord training program. After training these subjects showed a normalization of response in the left prefrontal cortex.

Impaired processing of rapid stimulus sequences as a result of sluggish attention shifting in dyslexic adults: A new interpretation
Riitta Hari
Brain Research Unit, Low Temperature Laboratory, Helsinki University of Technology, Finland.

The problems of dyslexic subjects extend beyond the skills directly needed for reading. We have observed that adults dyslexics are impaired in processing of rapid sound sequences both in a directional hearing illusion task (Hari & Kiesilä, Neurosci Lett 1996, 205: 138-140) and in a pitch streaming task (Helenius et al., Brain 1999, 122: 907-913). Most likely this sluggishness cannot be explained by problems in neuronal phase locking (Hari et al., NeuroReport 1999, 10: 2347-2348) nor by slow conduction velocities, and thus some other core mechanism has to be searched for.

In an attentional blink task, the dwell time of visual attention was 30% longer in dyslexic than normal-reading adults (Hari et al., Neurosci Lett 1999, 271: 202-204). In another psychophysical task, adult dyslexics processed stimuli from the left visual hemifield more slowly than normal readers, a finding which we interpret as a sign of a left-sided "minineglect" (Hari et al, Brain 2001, in press). Furthermore, abrupt stimuli captured attention in both visual hemifields less effectively in dyslexics than in normal readers. These abnormalities could reflect sluggish attentional capture and shifting, resulting from hypofunction of the right parietal lobe. Although the observed minineglect is mild, its existence is in line with the magnocellular deficit hypothesis. We propose that the causal link from the magnocellular deficit to several sensory and cognitive problems involves impaired regulation of automatic attention which slows down processing of stimuli presented in rapid succession.

Is the magno system important for identifying embedded letters?
David Omtzigt¹, Angélique W. Hendriks² and Herman H.J. Kolk^1
1Nijmegen Institute for Cognition and Information, University of Nijmegen, Nijmegen, the Netherlands.

²Department of Special Education, University of Nijmegen, Nijmegen, the Netherlands.

According to the magnocellular-deficit hypothesis of developmental dyslexia, a defective magno system is causally related to the reading problems of developmental dyslexics. However, it is unclear what contribution a normally functioning magno system makes to the reading process. Noting that for 100-ms presentations of single letters dyslexics and controls make equal numbers of naming errors on letters that are isolated, whereas the former are poorer than the latter if the letters are flanked by x’s (Bouma & Legein, 1977, 1980), we hypothesized that the magno system might have a function in the identification of embedded letters.

Adult normal readers were presented with single letters that were or were not flanked by an x on each side. The letters appeared in colour or brightness contrast relative to the background. Because the magno system is sensitive to brightness contrast but relatively insensitive to colour contrast, the manipulation of contrast influenced the level of magno activity. Similar to Bouma and Legein’s experiments (1977, 1980), letters were presented centrally or one degree to the left or right of fixation. Stimulus duration was 105 ms. The time interval between fixation offset and letter onset was 315 ms for one group of subjects (N = 24) and 15 ms for a second group (N = 24). The latter condition was similar to the procedure used in Bouma and Legein’s experiments, where there was no gap between fixation offset and stimulus onset.

 
In line with our hypothesis, Contrast and Flanking were found to interact significantly, but only when the gap between fixation offset and letter onset was 315 ms. We repeated the experiment, but this time with a within-subject manipulation of the gap (300 ms vs. 0 ms; N = 24). Now, Contrast and Flanking interacted independently of gap duration.

In conclusion, our manipulation of contrast (which presumably brought about a manipulation of magno activity) had a similar interaction with Flanking as was found for the reader-group manipulation by Bouma and Legein (1977, 1980). This not only provides support for a magnocellular deficit in dyslexia, it also suggests how an intact magno system might be important to reading. As a further specification of this function, one possibility is that the magno system is involved in locating (target) information relative to other (distractor) information. This localisation may be important for optimising subsequent form processing by the parvocellular stream (see Vidyasagar, 1999).

References:
Bouma, H., & Legein, Ch. P. (1977). Foveal and parafoveal recognition of letters and words by dyslexics and by average readers. Neuropsychologia, 15, 69-80.

Bouma, H., & Legein, Ch. P. (1980). Dyslexia: a specific recoding deficit? An analysis of response latencies for letters and words in dyslectics and in average readers. Neuropsychologia, 18, 285-298.

Vidyasagar, T.R. (1999). A neuronal model of attentional spotlight: parietal guiding the temporal. Brain Research Reviews, 30, 66-76.

Abnormal auditory and visual masking associated with reading and language disorders
Beverly A. Wright, Steven G. Zecker, and Richard W. Bowen

Department of Communication Sciences and Disorders, Northwestern University, USA.

Individuals with reading and language disorders also often have impairments in nonlinguistic sensory perception. I will describe our investigations into the nature of one such perceptual problem: the impaired detection of nonlinguistic target stimuli in the presence of competing masker stimuli.

In one set of experiments, we tested the ability of adults with dyslexia and controls to detect a brief target that was presented either before, during, or after a longer masker. The target and masker were either a tone and noise, or a small and a large visual pattern. Compared to controls, adults with dyslexia were significantly impaired in their ability to detect the brief target when it was presented just before or at the beginning of the longer masker. The specific temporal condition that most clearly revealed the impairment depended on the properties of the auditory or visual pathway activated. However, the detection of all brief, masked, stimuli appeared to be affected, particularly when the target and masker had similar sound or spatial frequencies. We previously observed similar impairments in auditory perception (visual perception not yet tested) in children with either dyslexia, specific language impairment (SLI), or central auditory processing disorder (CAPD). These results point to a bisensory masking abnormality that is revealed when a brief target and longer masker are processed through the same frequency-selective sensory pathway.

To better understand this masking abnormality, in another experiment, we used the same conditions as in our previous experiments to evaluate both auditory frequency resolution and processing efficiency in children with either SLI or CAPD and controls. Frequency resolution refers to the ability to separate the individual frequencies of a complex sound. Processing efficiency encompasses all factors that affect detection aside from resolution. Compared to controls, children with SLI or CAPD had poorer processing efficiency overall. A subset of both groups with impairments also had extremely poor frequency resolution, but because this problem occurred only in ongoing noise, it appears to be a different manifestation of poor efficiency rather than the result of defective peripheral frequency resolution. Importantly, both deficits only affected the detection of brief target sounds. Thus, the bisensory masking abnormality we previously identified can be thought of as a deficit in processing efficiency. Tests of the ability to detect brief target stimuli in the presence of maskers could help refine the diagnosis and treatment of language and listening impairments in children and adults.

Acknowledgements

Supported by NIDCD and the McDonnell-Pew Program in Cognitive Neuroscience.

Sensory, perceptual, linguistic and cognitive factors in academic success and failure, in grades one and two of elementary school: the Benton-IU Project
Charles S. Watson and Betty U. Watson
The Department of Speech and Hearing Sciences, Indiana University, Bloomington, Indiana, USA.

This project was a collaborative effort of the Benton Community School Corporation and a multidisciplinary group of ten investigators from Indiana University. Its purpose was to determine the characteristics of children that predict academic achievement in the early grades, in reading but also in other academic subjects. Standardized sensory, perceptual processing, linguistic, educational and cognitive tests were administered to 472 children, approximately 98% of the students entering the first grade in the four elementary schools of Benton County, Indiana over a three-year period (1995-1997). The results of 36 tests and subtests administered to entering first graders were well-described by a four-factor solution. These factors and the tests that loaded most heavily on them were: Reading-Related Skills (phonological awareness, letter- and word-identification), Visual Cognition (visual perceptual abilities, spatial perception, visual memory), Verbal Cognition (language development, vocabulary, verbal concepts), and Speech Processing (the ability to understand speech under difficult listening conditions). A multiple-regression prediction of reading achievement in grades one and two, based on the four factor scores, yielded a correlation of 0.73 between predicted and obtained achievement. A cluster analysis identified nine groups of children, each with a different profile of scores on the four factors. Within these groups the proportion of students with unsatisfactory academic achievement in the first two years of elementary school (as reflected in teacher-assigned grades) varied from 3 to 40 per cent. The profiles of factor scores demonstrated a primary influence of the Reading-Related Skills factor on reading achievement and also on other areas of academic performance. The second strongest predictor of reading and mathematics grades was the Visual Cognition factor, while there was a weak influence of the Verbal Cognition factor. The Speech Processing factor accounted for no significant variance in achievement in any academic area. Selected measures of auditory processing, including backward masking and syllabic-rate thresholds also were related to neither general speech recognition skills nor to any aspect of academic achievement.

Acknowledgements

Research support was provided by the Benton County Community School Corporation, the U.S. National Institute for Deafness and other Communicative Disorders, and the Strategic Initiatives Fund of Indiana University.

Dyslexia in children and adults: who is affected and where is the deficit?
Maggie Snowling

Department of Psychology, University of York, York, UK.

It is now widely accepted that the core cognitive deficit in dyslexia is in phonological language processes. However, it is important to be clear that this conclusion has been reached from research on samples of dyslexic children who are defined by the traditional discrepancy definition, that reading skills are out of line with age and IQ. This paper reviews current research suggesting that the phonological deficit in dyslexia is in output rather than input systems. Broadening the definition of dyslexia to include children with oral language impairments or adults with well-compensated reading impairments should proceed cautiously if the research agenda is to identify the sensory impairments that are causally related to reading disability.

Controversy about the sensory basis of reading difficulties

John Stein

University Laboratory of Physiology, University of Oxford, Oxford, UK.

There is very little doubt among physiologists and psychophysicists that there is something different about the vision of many developmental dyslexics, although they might disagree about how restricted this difference is to the visual magnocellular system. Likewise there is increasing evidence that dyslexics’ hearing is different also. Yet there are few cognitive or linguistic psychologists who would accept that the mild low level vision and hearing impairments that have been demonstrated in dyslexics have much to do with their phonological or orthographic problems that seem to be much more marked. This is a most unsatisfactory state of affairs, and it confuses practitioners, teachers and parents who look to the research community for advice. I believe that it results from fundamental misunderstandings about the nature of the different kinds of evidence. The psychophysical tests usually reveal mild processing deficits in dyslexics which will only emerge as group differences if a sufficient number of subjects is studied. These low level weaknesses can then be amplified by additional, higher level, attentional and memory deficiencies to generate the large deficits in orthographic and phonological ability that are characteristic of dyslexia. I will briefly review evidence that shows that sensitivity to visual and auditory transients can account statistically for significant independent proportions of individual differences in orthographic and phonological reading skills in normal readers and dyslexics. This is true even when ‘outliers’ are excluded suggesting that these basic sensory capacities help to determine how well reading skills develop in everybody. But I will emphasise that the correlations between low level visual and auditory transient processing and reading skills can not account for all the phonological and orthographic deficiencies of dyslexics; there are many dyslexics who have normal sensory processing of transients. Their main defect must be attributed to higher level processes. However the existence of these dyslexics does not contradict the overall conclusion that visual and auditory transient processing abilities are important prerequisites of learning to read in everybody.

Neurobehavioral origins of variations in human performance abilities
M. M. Merzenich
Keck Center for Integrative Neurosciences, UCSF, San Francisco, USA.

Recent studies conducted largely in rodents and non-human primates have provided an increasingly clear understanding of how cortical plasticity processes contribute to the progressive refinement of representation of spectrotemporally complex inputs (e.g., aural speech, haptics, visual discrimination and recognition) and brain-controlled actions (e.g., vocalized speech, locomotion, postural control, oculomotor control, hand movements) through the course of child development. Maturing cortical systems "learn how to learn" on the basis of the statistics of behaviorally important inputs and outputs engaging the successive, complexly-interconnected levels of cortical processing machinery to progressively change. Variation in the functional outcomes of these experientially-driven processes in early life are hypothesized to account for much of the variance in perceptual and cognitive performance abilities expressed in different human children and adults.

Parallel behavioral, electrophysiological and brain imaging studies conducted in normal, and language-, reading- and motorically-impaired children and adults support this view. Evidence is growing that in humans, as in our animal models, abnormal, early, experientially-driven self-organization of the cortical processing machinery largely underlies the emergence of language, reading, and sensorimotor deficits in impaired individuals. They also show that in humans, as in other mammals, there is a life-long capacity for intensive-training-driven refinements of complex input processing and sensorimotor control abilities: Intensive training-driven refinements in complex signal and complex movement representations in impaired individuals commonly result in a rapid remediation of longstanding neurobehavioral deficits.

The more profound impairments in aural language abilities recorded in pervasively developmentally disabled (PDD) children appear to arise from representational catastrophe within an entry-level auditory cortical area on the superior temporal plane (LeWine, et al., 1999). We can easily generate this form of catastrophic de-differentiation of auditory processing within the primary auditory cortex in animal models, e.g., by a particular form of progressive learning. Studies are now focussed on understanding the nature of the inherited "weaknesses" that facilitate this catastrophic development process in PDD children. Strategies designed to re-differentiate the aural speech processing machinery of severely impaired PDD children are also being investigated.

Some of the important theoretical and practical implications of these findings will be discussed.

Acknowledgements

Research supported by NIH Grant NS-10414, HRI, the Coleman Fund, and the MIND Foundation.

Natural variation in visual processing and word recognition

Kristen Pammer and Piers Cornelissen

Department of Psychology, University of Newcastle, Newcastle-upon-Tyne, UK.

Awaiting abstract

Tone duration and training can affect frequency discrimination performance in 6- to 9-year-old children

P. A. Sutcliffe and D. V. M. Bishop

Department of Experimental Psychology, University of Oxford, Oxford, UK.

Frequency discrimination development was examined in 6- to 9-year-old children. The study investigated whether performance is affected by altering the target tone duration and through training. In a two-interval forced-choice paradigm the target tones, one of 400 Hz and the other of higher frequency, were 50 or 150 ms in duration. Each was followed after 320 ms by a 200 ms comparison tone of 400 Hz. The child had to select the interval containing the higher frequency target tone. All children participated in two sessions and completed both duration conditions in each. Recognition thresholds showed the 8- and 9-year-olds performed significantly better on both durations in both sessions compared to the 6- to 7-year-olds. However, there were significant improvements in the younger age group during the second session, especially in the 150 ms condition. Within each age group there was no significant relationship between language ability and frequency discrimination performance. These results show that auditory processing can be affected by tone duration and training, and age is a critical factor when examining auditory processing in young children.

Auditory backward recognition masking in dyslexic and normal reading adults

Y. M. Griffiths¹, N. I. Hill², M. J. Snowling¹ and P. J. Bailey¹

¹Department of Psychology, University of York, York, UK.

²Department of Psychology, University of Essex, Colchester, UK.

The ability of dyslexic listeners to extract information from successive tones was investigated using an auditory backward recognition masking (ABRM) task. Forty listeners participated in the study, twenty adult dyslexics and twenty controls matched on age and IQ. Individuals in the dyslexic group all had a known history of reading and spelling difficulties, and performed significantly below the level of controls on laboratory-administered measures of reading, spelling accuracy and phonological skill. Performance on a frequency discrimination task was assessed using a four-interval, two-alternative forced choice procedure, in which a target signal with frequency greater than 1 kHz was presented in either the second or third interval at random, and a 1 kHz standard signal in the other intervals. Four estimates of the frequency difference between target and standard corresponding to 71% correct performance were obtained adaptively in each of four conditions. In three ABRM conditions the signal tones were followed by masking tones, so the listeners’ task was to select the interval in which the frequency of the leading tone was incremented. The frequency of the masking tone in each interval was either fixed at 1 kHz, randomised about 1 kHz, or randomised about 2 kHz, the three conditions being tested separately. In a fourth condition the masker was absent. All tones had duration of 75 ms, and the inter-stimulus interval (ISI) between signal and masking tones in the ABRM conditions was either 20 or 200 ms. Mean thresholds for both groups were lower when the ISI was 200 ms, but there was no significant group difference for either ISI. In the ABRM conditions random maskers caused more mean threshold elevation than static maskers, particularly around 1 kHz. The only significant group difference occurred in the 2 kHz random masker condition when the target-masker separation was 200 ms. However, an analysis of the lowest of the four threshold estimates for each listener in each condition showed no significant group differences. The groups did not differ in the condition where the masker was absent, consistent with our previous findings [Hill et al. (1999) J. Acoust. Soc. Am. 106, L53-L58]. The results do not support the hypothesis that adult dyslexics have an auditory deficit that impairs their ability to process rapid tone sequences.

Acknowledgement
This work was supported by project grant 55671 from the Wellcome Trust.

Perceptual learning and attentional cues in nonsimultaneous masking

M. M. Rose, D. E. H. Hartley and D. R. Moore

University Laboratory of Physiology, University of Oxford, Oxford, UK.

This study investigated the training effects of three nonsimultaneous masking conditions, and the possible role of an attentional cue. Three groups of six participants trained on seven consecutive days on either i) backward masking (BM, 0-ms delay), ii) combined forward and backward masking with a short inter-masker delay (10-ms, FBS) or iii) combined forward and backward masking with a long inter-masker delay (110-ms, FBL). The signal was a 10-ms, 1-kHz tone pip, and the masker was a 300-ms, 1.6-kHz wide noise (40dBSPL/Hz), centered at 1 kHz. Training sessions had 15 runs (30 presentations/run), using a 2AFC, maximum-likelihood procedure.
The FBS condition showed the highest amount of masking, followed by the BM condition and the FBL conditions, respectively (p < 0.05). A small but significant improvement was found between training days when all three groups were combined (p < 0.001). Pre-training data showed, for most subjects, lower FBL than BM thresholds. This could indicate that the first noise burst in the FBL condition did not mask the signal, but rather acted as a temporal and attentional cue, helping to "listen at the right time".

Acknowledgements
Supported by the National Health Service, the Wellcome Trust and the Medical Research Council.

Auditory backward masking in emotionally and behaviourally disturbed adolescents

P. R. Hill and D. R. Moore

University Laboratory of Physiology, University of Oxford, Oxford, UK.

Studies of auditory masking in children with specific language impairment reveal deficits in backward, but not simultaneous masking. This suggests that their auditory temporal processing is abnormal. Children with emotional and behavioural disturbances (EBD) have also been identified as a group showing high incidences of speech and language difficulty. This project aimed to investigate whether such children were also impaired at masking tasks.

16 EBD boys (average age 13.6) and 10 controls (4 girls and 6 boys, average age 13.3) were tested. Pure tone and masked tone thresholds were estimated using a two alternative, forced choice paradigm, incorporated into a simple computer game. Tones of 1000Hz lasting 20ms were presented during (simultaneous masking) or before (backward masking) a 300ms narrow band (0.6-1.4 kHz) noise masker centred at 1000Hz. Presentation was (right ear) monaural. Children were also measured on three ability scales, measuring phonological skills (spoonerisms), verbal memory (digit recall), and non-verbal intelligence (matrices).

There was no difference between the EBD and control groups in absolute thresholds or in simultaneous masking. This suggests that all the subjects were equally able to perform this type of psychometric test. However, at the backward masking task, the mean threshold of the EBD group (mean 61.4 dB SPL; s.d. 9.6) was significantly higher than the mean threshold in the control group (53.7 dB SPL; s.d. 4.9; p= 0.03). Also, the difference between the absolute threshold and the masked threshold for each child showed that EBD children experienced more masking in the backward masking condition (mean 22.1 dB SPL; s.d.13.1) than did Controls (12.2 dB SPL; s.d. 5.7; p =0.002). However, about half the children in the EBD group had thresholds within the same range as the control group. This suggests that only some of the EBD children may have been impaired at backward masking.

The ability scores revealed that the Controls performed significantly better at the matrices and spoonerisms tests than the EBD children. The co-occurrence of poor phonological ability and elevated backward masking thresholds in the EBD group agrees with previous work (Wright et al, Nature, 1997, 387:176-8), strengthening the suggestion that impaired temporal processing may negatively affect phonological ability. However, a significant difference was also found between the two groups on the matrices test, indicating that additional factors, including non-verbal intelligence, may also contribute to differences in the backward masking thresholds between the groups.

Deficits in fast temporal processing - neither necessary nor sufficient for reading disability

Sygal Amitay¹, Merav Ahissar^1,2, and Israel Nelken^1,3
1Interdisciplinary Center for Neural Computation, Hebrew University of Jerusalem, Israel.

²Department of Psychology, Hebrew University of Jerusalem, Israel.

³Department of Physiology, Hebrew University - Hadassah Medical School, Jerusalem, Israel.

The finding that dyslexics are impaired on a two-tone frequency discrimination task when the inter-tone interval is very short formed the basis of the rapid temporal processing deficit hypothesis. Further studies showed impaired performance on tasks measuring temporal resolution, such as amplitude and frequency modulation detection. In this study we tested a population of disabled readers and their age and education matched controls on a large battery of psychoacoustic tasks that require processing of a wide range of stimulus time constants, from hundreds of microseconds to several seconds. About a third of the disabled readers had concurrent difficulties in most of the tasks tested: detection of frequency differences, detection of tones in narrowband noise, detection of amplitude modulation, detection of the direction of sound sources moving in virtual space, and perception of the lateralized position of tones based on their interaural phase differences. Nevertheless, across-channel integration was intact in these poor auditory processors since comodulation masking release was not reduced. Furthermore, phase locking was presumably intact since binaural masking level differences were normal. This subgroup of reading disabled subjects was not, moreover, specifically impaired on a two-tone frequency discrimination task with brief inter-tone intervals. On the other hand, specific deficits in frequency discrimination at brief inter-tone intervals were found in the complementary subgroup. The latter was not impaired on any standard psychoacoustic task measuring temporal resolution. There was no difference between the two reading disabled subgroups on any reading measure, but the subgroup of poor auditory processors scored lower on tests of general cognitive abilities. We conclude that the relationship between reading disabilities and auditory processing tasks are complex, and possibly different for different subjects.

Processing of transient stimuli in the visual and auditory domains, in children with Specific Language Impairment
M. Hayiou Thomas¹, D. V. M. Bishop¹, K. Plunkett¹, J. F. Stein² and P. C. Hansen^2
1Department of Experimental Psychology, University of Oxford, Oxford, UK.
²University Laboratory of Physiology, University of Oxford, Oxford, UK.

It has been shown that at least some dyslexics are less sensitive to transient stimuli than controls, in both the visual domain (detecting coherent motion; Cornelissen et al, 1995, 1998; Witton et al, 1998) and the auditory domain (detecting frequency modulation in tones; Witton et al, 1998). This study examines whether similar deficits can be demonstrated in children with Specific Language Impairment (SLI).

Visual Domain: Sensitivity to coherent motion was investigated using Random Dot Kinematograms in a two-interval forced-choice paradigm. The target RDK showed directional Left/Right motion, while the distracter showed Brownian motion, and thresholds were estimated according to a Kaernbach staircase procedure. A control condition measured sensitivity to coherent form, such that in the target image the orientation of stationary short lines created the percept of a circle. It was anticipated that if children with SLI have difficulty processing transient stimuli, they should show higher detection thresholds in the coherent motion but not the coherent form condition, when compared to a group of control children (matched on age and non-verbal ability). This was not the case, as there were no significant group differences in either condition; this suggests that unlike dyslexics, children with SLI do not have impaired processing of transient visual information.

Auditory Domain: Sensitivity to frequency modulated (FM) tones was investigated using a two-interval forced-choice paradigm. Thresholds estimates were obtained using an adaptive PEST procedure, for three rates of FM: 2 Hz, 20 Hz, and 240 Hz. In the slow FM conditions (2 Hz & 20 Hz), the modulation is perceived as a dynamic change in pitch; hence, impaired processing of transient auditory information should be reflected in high detection thresholds for these rates of FM. By contrast, 240 Hz frequency modulation is perceived by a spectral mechanism rather than by tracking the frequency changes; this provides a control condition, in which even subjects who have high thresholds in the slow FM conditions should perform as well as controls. However, children with SLI performed as well as control children in all three conditions, suggesting that their processing of transient auditory information, at least as indexed by FM, is not impaired.

These results do not support the hypothesis that children with SLI have impaired processing of transient stimuli, in either the visual or auditory domains, of the type found in dyslexia.

Acknowledgement

Many thanks to Caroline Witton, for her generous help.

Which acoustic cues are sufficient or necessary for speech comprehension?
Gary Green, Rebecca Millman, Michael Simpson and Adrian Rees
Department of Physiology, The Medical School, Newcastle upon Tyne, UK.

The current debate concerning the role of sensory processing in language disorders requires an understanding of which physical patterns in a signal are used in language processing. Since the early work of Dudley (1939) it has been known that the detailed spectral content of speech sounds is not necessary for speech comprehension. Dudley demonstrated that the envelope of speech signals could be used to modulate noise waveforms and that the resultant sound was highly intelligible. More recently, several groups (Drullman et al, 1995; Shannon et al, 1995; Shannon et al 1998) have confirmed that specific characteristics of the temporal structure of a sound are crucial for the comprehension of the contents of that sound. Intelligibility is high if the amplitude modulation (the envelope) content of a synthesised speech sound retains the original low modulation rates around 4Hz. These modulation rates are near the average syllabic rate of 3-4Hz. Drullman has also noted that there is no one-to-one relationship between the modulation transfer function and intelligibility. Manipulations of the peaks and troughs of the envelope demonstrate that peaks are more important than troughs.

Witton et al (this conference) have demonstrated that 2Hz FM sensitivity and 20Hz AM sensitivity can be used as predictors of phonological ability in both disabled readers and normal controls. Low rate AM sensitivity is not a predictor in either group. Thus there is a fundamental question concerning the role of high modulation rates in speech processing.

In this paper we demonstrate that bandpass filtering of the modulation waveform of speech signals around 28Hz, removing very low rates and very high rates, can result in highly intelligible speech. However, the detection and processing of these cues is not predicted by a modulation filter acting on mid-rate fluctuations in the envelope. It is demonstrated that although the synthesized speech only contains envelope cues around 16-40Hz, these cues are processed by mechanisms that are normally assumed to be concerned with low rates. Thus the processing of the envelope patterns in speech is highly non-linear and one cannot simply think of patterns processing in terms of slow and rapid events. The implications for the design of auditory tests for probing language processing will be discussed.

The effect of coloured filters on visual motion sensitivity and reading speed

G. S. Hebb¹, C. Clisby², S. F. Fowler² and J. F. Stein^1
1University Laboratory of Physiology, University of Oxford, Oxford, UK.
²Learning Difficulties Research Clinic, c/o Orthoptic Department, Royal Berkshire Hospital, Reading, UK.

The visual motion sensitivity with and without coloured filters of 49 unselected primary school children (mean age 8 years 1 month) was compared with their rate of reading. Their mean IQ was 111.84, prorated from the British Ability Scales (BAS) subtests: Matrices and Similarities. Their mean reading skill T score was 48.32 (St Dev 13.10) measured by the BAS single word Reading subtests. For this study rate of reading was measured with the Wilkins Rate of Reading Test under four conditions: without a filter, and with blue, yellow and neutral density filters. After a full orthoptic examination, visual functioning was assessed by means of a psychophysical test of motion sensitivity under the four filter conditions. The children's rate of reading over all filter conditions was found to correlate well with their coherent motion thresholds (R= -0.335, P=0.021). When the children wore yellow filters their motion detection and reading speed were most highly correlated (R= -0.477, P= 0.001); wearing blue filters the correlation was R= -0.361 (P=0.013); but reading speed and motion sensitivity were not significantly correlated when they were using neutral density or no filters. We conclude that the speed with which children read is related to their motion detection ability. Since the latter is related to visual magnocellular sensitivity and the magnocellular system is most sensitive to yellow, these results also imply that coloured filters affect the performance of the magnocellular system.

The perceptual weighting of rapidly changing acoustic speech cues by Dutch children
Ellen Gerrits
Utrecht Institute of Linguistics OTS, Utrecht University, Utrecht, The Netherlands.

There are numerous reports that demonstrate that children with Specific Language Impairment and children with dyslexia have impairments in expressive and receptive phonology (for an overview see Bishop, 1997; and also Leonard, 1998). Of special interest are the claims of Tallal and colleagues (e.g. Tallal & Stark, 1981; Tallal et al., 1996) that refer to the perception of formant-transition information. In Tallal’s hypothesis of a "temporal processing deficit" it is said that children with Specific Language Impairment have marked difficulty processing brief and rapidly changing acoustic information, notably formant transitions.

In Nittrouer (1999), it was tested whether children (aged 8 to 10) with poor phonological processing capacities could make phonemic decisions using formant-transition information. Nittrouer showed that, in making fricative labelling decisions ("Sue" or "shoe"), children with poor phonological processing abilities weighted formant transitions more heavily and fricative-noise properties less heavily than control children did. According to Nittrouer, this indicates a delay in cue weighting: children with poor phonological processing abilities are using the noise and formant-transition cues in a way shown by younger normally developing children (see Nittrouer, 1992, 1996; Nittrouer & Studdert-Kennedy, 1987), whereas the control children already weighted the noise and formant-transition cues in a way similar to adult listeners.

The results of Nittrouer (1999) contradict Tallal’s hypothesis: children with poor phonological processing abilities based their fricative classification more on formant transitions than the children with good phonological processing abilities did. No evidence was found in favour of a temporal processing deficit. However, the children tested in Nittrouer’s study were not language or reading impaired, as was the case in Tallal’s studies. It would be worthwhile to use the cue-weighting paradigm to test the perception of children with language disorders.

 
At present, a series of experiments is in preparation in which the weighting of formant- transition information for phoneme identification will be tested with 3-year-old Dutch children at risk of dyslexia, and with SLI children and controls. Yet, before using the cue-weighting paradigm to explain language and reading disorders, an experiment was conducted with normally developing children. The goals of this experiment were to replicate Nittrouer’s results with completely natural fricative stimuli (Nittrouer used synthetic noise portions concatenated with natural vowels) and to expand the fricative design with a stop consonant and vowel contrast. The findings of this experiment will be presented and discussed in relation to normal and disordered speech perception development.

References
Bishop, D.V.M. (1997). Uncommon understanding. Development and disorders of language comprehension in children. Hove: Psychology Press.

Leonard, Laurence B. (1998). Children with specific language impairment. Cambridge: MIT Press.

Nittrouer, S. (1999). Do temporal processing deficits cause phonological processing problems? Journal of Speech, Language and Hearing Research, 42, 925-942.

Nittrouer, S. (1992). Age-related differences in perceptual effects of formant transitions within syllables and across syllable boundaries. Journal of Phonetics, 20, 351-382.

Nittrouer, S. (1996). Discriminability and perceptual weighting of some acoustic cues to speech perception by 3-year-olds. Journal of Speech and Hearing Research, 39, 278-297.

Nittrouer, S. and Studdert-Kennedy, M. (1987). The role of coarticulatory effects in the perception of fricatives by children and adults. Journal of Speech and Hearing Research, 30, 319-329.

Tallal, P., Miller, S., Bedi, G., Byma, G., Wang, X., Nagarajan, S.S., Schreiner, C., Jenkins, W. and Merzenich, M.M. (1996). Language comprehension in language-learning impaired children improved with acoustically modified speech. Science, 271, 81-84.

Tallal, P. & Stark, R.E. (1981). Speech acoustic cue discrimination abilities of normally developing and language-impaired children. Journal of the Acoustical Society of America, 69(2), 568-574.

Diminished auditory mismatch fields in dyslexic adults
H. Renvall and R. Hari
Brain Research Unit, Low Temperature Laboratory, Helsinki University of Technology, Finland.

We have previously shown that dyslexic adults are slow in processing of binaural sounds that follow each other in rapid succession in an auditory illusion task (Hari and Kiesilä, 1996) as well as sounds of two different frequencies in a pitch streaming task (Helenius et al., 1999). One possibility is that dyslexic subjects would have a prolonged short-term sensory or cognitive buffer within which subsequent stimuli may interfere. We have suggested this prolongation to be secondary to sluggish shifting of automatic attention as a consequence of weak functioning of the magnocellular pathways. To study this issue further, we used an oddball paradigm to elicit magnetic mismatch fields (MMFs) that are assumed to reflect an automatic sensory comparison process.

Whole-scalp magnetoencephalographic (MEG) signals were recorded from 8 dyslexic and 11 normal-reading adults. The oddball sequence consisted of binaural 50-ms tones: the standards (86% of all) were 1000 Hz in frequency and the two deviants, each with probability of 7%, were of 920 Hz and of 1080 Hz. The subject was reading during the measurement.

The dyslexics were significantly (p < 0.001) slower in reading and word recognition speed. All stimuli elicited a N100m response, peaking at 105 ± 5 ms in controls and at 97 ± 6 ms in dyslexics (n.s.). The deviants also elicited a MMF, with maximum amplitude at 169 ± 6 ms in controls, and at 162 ± 6 ms in dyslexics (n.s). The N100m amplitudes did not differ between the groups. However, the MMFs were weaker in dyslexics than in controls; the difference was statistically significant (p < 0.002) in the left hemisphere. In dyslexics, MMFs were significantly smaller in the left than right hemisphere (p < 0.01) but did not differ in the control subjects between the hemispheres.

The diminished MMFs in a group of dyslexic adults suggest a general deficiency in the triggering mechanisms of automatic auditory attention. The present results are in line with electric scalp recordings that have shown diminished mismatch responses in children with learning problems (Kraus et al., 1996) and in dyslexic adults (Baldeweg et al., 1999).

Supported by the Academy of Finland.

The effect of a verbal labelling strategy on rapid auditory processing: Evidence from normal and dyslexic readers
Catherine M. Marshall, Margaret J. Snowling and Peter J. Bailey
Department of Psychology, University of York, York, UK.

The phonological difficulties of dyslexic children have been attributed to rapid auditory processing deficits, as measured by the Auditory Repetition Task (ART; Tallal, 1980). This paradigm requires children to reproduce sequences of non-verbal tones presented at various rates. The present paper describes two experiments which investigated the effects of using a verbal labelling strategy on the ART, in both normal and dyslexic readers. Explicit encouragement to assign verbal labels to the stimuli facilitated the performance of normal readers. In contrast, the ART scores of dyslexic children did not improve with a verbal strategy. The results suggest that children's propensity to label stimuli in a 'non-verbal' task influences performance. The extent to which performance is facilitated by such a verbal strategy may be determined by underlying language ability.

Low-level auditory processing in compensated adult dyslexics investigated with whole-head magnetoencephalography (MEG): Preliminary findings
Alison Fisher, Gareth Barnes, Arjan Hillebrand, Ian Holliday and Ian Richards

Neurosciences Research Institute, Aston University, Birmingham, UK.

A tone-pair frequency discrimination task was used to examine auditory processing in compensated adult dyslexics. The MEG evoked responses and psychophysical performance of dyslexics and matched controls to the tone-pair stimuli were measured. We set out to examine the effect of varying the frequency separation and timing of the stimuli. Preliminary analysis (Fisher et al. BDA. 2001) suggests that low-level auditory processing differences exist between groups. However we find no evidence that the rate at which stimuli are presented has a detrimental effect on the performance of dyslexics relative to controls. The nature of a possible low-level auditory processing deficit will be discussed.

Correlations between FM detection and pseudo-word reading: Do they depend on statistical outliers?
Caroline Witton and Joel Talcott
University Laboratory of Physiology, University of Oxford, Oxford, UK.

Several studies of sensory processing in dyslexia have used correlational analyses to examine relationships between auditory sensitivity and phonological decoding skills. However it is important to know if significant correlations rely upon the poor performance of a small number of statistical outliers. In this paper, we systematically examine our published data relating detection thresholds for 2 Hz FM to pseudo-word reading ability, and account for the effects of statistical outliers. We show that correlations between 2 Hz FM sensitivity and phonological decoding skills are robust to the removal of statistical outliers.

Auditory event related potentials associated with phonological processing in dyslexic children
Putter-Katz H., Kishon-Rabin L., Sachartov E., Weiz R., Gadoth N., Pratt H.
Department of Communication Disorders, Tel Aviv University Speech and Hearing Center, Sheba Medical Center, Israel.

Phonological awareness, i.e. the knowledge that spoken words can be decomposed into phonological segments, is impaired in some dyslexic readers. This impairment is presumed to underlie their difficulty to map alphabetic characters onto the spoken word. “Oddball” Auditory Event Related Potentials (AERPs) were compared between dyslexic children and a control group of skilled readers in tasks involving auditory discrimination between vowels (/i/-/u/) or syllables (/da/-/ga/). Syllable discrimination is considered a more difficult task since it involves processing of rapidly changing auditory cues. Brain processing patterns differed significantly between groups, with dyslexic readers showing slower brain processing with increased phonological demand. Latencies of AERPs endogenous components, particularly P3, were significantly longer with syllable discrimination. The same trend was observed for reaction time differences between groups. Moreover, P3 amplitudes over the anterior scalp were significantly higher in dyslexics. These results are compatible with phonological impairment in dyslexia associated with a more frontal auditory processing in this impairment.

Recent MEG studies of early (< 200 ms) letter-string processing in dyslexic and non-dyslexic adults
Piers Cornelissen¹, Riitta Salmelin², A. Tarkiainen² and P. Helenius^2
1Department of Psychology, University of Newcastle, Newcastle-upon-Tyne, UK
²Brain Research Unit, Low Temperature Laboratory, Helsinki University of Technology, Finland.

In the last 5 years, MEG (magnetoencephalography) studies have revealed much about the nature, location and timing of letter-string processing in the brain. In non-dyslexic people, the following distinct patterns of cortical activity are observed during single word reading: 1) Earliest responses occur in the occipital midline (90-130 ms latency) and are related to non-specific processing of all visual stimuli. 2) At ~150 ms latency, non-dyslexic subjects show bilateral inferior occipito-temporal activity. The left hemisphere (LH) component responds preferentially to letters, compared to equivalent length symbol strings, for example. The right hemisphere component (RH) responds equivalently to letters and symbol strings. However, unlike the LH component, the RH response does appear to be sensitive to the position of items in a string. Of particular interest is the finding that, in samples of Finnish dyslexics at least, the LH occipito-temporal response to letter-strings is either absent or diminished. 3) Lexicality effects (i.e., significant differences between random consonant strings and words) are absent before ~200 ms, but are found in the sustained response of predominantly left perisylvian cortex 300 ms after stimulus onset.

Behavioural and neuroimaging data suggest a dissociation between left and right hemispheric processing of the global and local attributes of complex visual stimuli. As a working hypothesis, we suggest that early processing of letter strings in the brain may involve a combined identity space code that is computed in parallel. Accordingly, the left hemisphere may handle the local attribute of letter identity while the right hemisphere may deal with the global attribute of spatial-position.