00?8~393290s300+000 c 1990 Pergamon Prcs plc

,‘,‘rurop\~choluq,a. Vol 28. No. 2. pp 1X1-198. 1990 Pnnkd I” Great Brnam

DIRECTED ATTENTION DICHOTIC LISTENING IN READING DISABLED CHILDREN: A TEST OF FOUR MODELS OF MALADAPTIVE LATERALIZATION* JOHN tontario

Institute

R. KERSHNERPQ:

for Studies m Education, Windsor, (Receired

and L. L. MORTON::

University of Toronto, Ontario, Windsor, Ontario. Canada

10 February

Canada;

and ZUniversity

of

1989: accepted 31 August 1989)

Abstract-In two experiments, critical u priori tests compared four competing LD theories of maladaptive language lateralization (Poor Structural Lateralization; Excessive RH Activation: Excessive LH Activation; Excessive LH or RH Activation). In Experiment 1, 32 auditory-linguistic LD children from public schools were compared to 32 high-achieving. age-matched controls. Experiment 2 replicated the main findings of Experiment 1 by comparing 14 LD children to 14 controls attending a private school. The effect of priming or car order was analyzed in a directed attention. dichotic paradigm with strings of quadruple digit names as stimuli. The order in which the ears were monitored (LE first or RE first) determined (1) in both experiments whether the LD children compared to controls were more weakly or strongly lateralized and (2) in Experiment 2 whether the REA correlated negatively of positively with reading and spelling achievement. The results indicate that the hemispheric processes involved in disengaging and shifting response strategies between dichotic channels may share common neuronal mechanisms with cognitive processes that are important to school achievement. Rather than LD children suffering from a fixed laterality deficit, the results are consistent with the view that a task-dependent attentional dysfunction may interfere with left hemisphere language processes by overengaging either hemisphere. However, pronounced asymmetry in the relative magnitude of the effects at each channel also suggests greater interference in LD children between the systems controlling auditory-linguistic attention and lateral orienting.

INTRODUCTION ACCORDING to BRYDEN [4] directed attention, dichotic listening is the preferred method for measuring receptive speech lateralization in learning disabled children. 11In contrast to freerecall, dichotic listening which imposes no constraints on attending and reporting bias, the directed attention, dichntk Wening procedure (DAD) requires exclusive surveillance and report from each ear in a predetermined sequence. Typically, ear order becomes a counterbalanced grouping factor as one-half of the subjects attend first to the right ear and

*This research was supported by a grant to the tirst author from the Social Sciences and Humanities Research Council of Canada. ficorrespondence to be addressed to: Dr John Kershner. The Ontario Institute for Studies in Education. 252 Bloor Street West, Toronto. Ontario. M5S IV6. Canada. :lA right ear advantage (REA) in dtchottc performance is Interpreted. tempered by certain methodological and theoretical qualifications, as an indication of lmguislic specialization of the left cerebral hemisphere. This inference is based on studies indicating that (al the left hemisphere is preprogrammed when activated differentially for attending selectively to the right side ofspace (i.e. right ear).(b) the contralateral pathways are prepotent over ipsilateral (same ear-hemisphere) auditory pathways. and (c) in most right-handed people the left hemisphere is specialized for linguistic processing. Thus. auditory-verbal Information perceived dichotically at the right ear is easier to code and process than the same information arrtving at the left ear.

JOHN R. KEKSHNEK and L. L. MORTON

1x7

one-half attend first to the left ear. Thus, the directed attention modification may produce a more valid estimate of lateralization by controiling for ambient and strategic fluctuations in attention. A well-founded conclusion that can be inferred from a careful review of studies that have used DAD with learning disabled children is that these children are impaired or at least different when compared to age-matched, normal controls in their strength of processing auditory receptive language unilaterally in the left hemisphere [32]. However, aside from a surface level of agreement among researchers, four very different, more specific learning disability (LD) models of maladaptive lateralization have beeh suggested [21]. Poor structurul

lateralization

model (PSL)

By far the oldest idea. dating back to the seminal work of ORToN [38], is that LD children are poorly lateralized. This seasoned model, which we refer to as the PSL model, implies: (a) that the dichotic test measures a hard-wired neural substrate for linguistic processing [23]: and (b) that language is represented more bilaterally at the level of cerebral structure in LD children. The PSL model is supported by one DAD study which has shown a reduced right ear advantage (REA) in LD children. The diminished REA was apparent in both left and right attended conditions. OBRZUT et al. [37] compared a group of auditory-linguistic disabled readers to a control group of above-average readers who were matched on chronological age (CA), gender and handedness and who were selected on the basis of teacher recommendation. The LD children were poorer in reporting CV syllables when precued to attend to the right dichotic channel and they were better in attending selectively to the left channel. As a result, both attended conditions contributed overall to their reduced REA. This specific combination of left ear and right ear effects has yet to be replicated; but in the absence of more critical experiments. the PSL model cannot be discounted [4]. Ricgclht/wmi.splwrc cwessire

activation

model (RHA )

A second viable model. related closely to the PSL model. has its basis in the hypothesis that LD children may fail to suppress the right hemisphere during verbal tasks [33]. This model. which ue refer to as the RHA model implies: (a) that the dichotic test measures hemispheric activation coexisting with the underlying structural system [27]; and (b) that. unlike the PSL model. the cortical structure for language in LD children is normal in its geographical distribution but weaker in its ability to inhibit attentional bias and subsequent right hemisphere enagagement in verbal processing. Like the PSL model the RHA model assumes that right hemisphere activation interferes with or undermines the eff%iency of language processing in the left hemisphere. This model forces a distinction between p~~~~r Itrrrwlixtior~ (PSL model) which entails a structurally fixed reduction in the degree of left hemisphere spcciaiiLatlon and NYLI~Itrtcwrlizutiort (RHA model) which means a dynamic difference in the balance of activation between hemispheres during linguistic processing. Although both the PSL and KHA models predict a smaller REA in LDchildren. the more lluid RHA framework predicts oniy that these children will demonstrate an increase in correct recall from the attended left car coupled with fewer intrusions from the right car. without necessarily interfering with selective report from the right channel. The RHA model finds support in two DAD studies. In both experiments [2. 341. LD children dcmonstratcd a lower REA than above-average rcadcrs. selected. again. on the basis 01‘teacher rccclnlmendation. However. in these studies there were no group differences in

POOR ATTENTlONAL

CONTROL

Ih‘ LD

IX?

recall from the right channel. The smaller REA was produced solely by the superior ability of the LD children to recall CV and CVC stimuli when directed left. Even though much of its justification is post hoc, after a thorough review of the extensive dichotic research carried out by him and his colleagues, OBRZUT [32] cautiously supported the RHA model as a leading contender among the field of promising possibilities. Left hemisphere

excessive

activation model (LHA)

An incompatible, alternative candidate is that the left hemisphere in LD children ma! become engaged excessively during linguistic proceesing [ 13,201. This model, which we refer to as the LHA model implies: (a) that the dichotic test measures hemispheric activation coexisting with an underlying structural system [27]; and (b) that the cortical structure for language in LD children is normal in its geographical distribution but an unusually high level of left hemisphere activation is required to perform language tasks. Supporters of the model have suggested that LD children may have to try harder at verbal tasks because of limitations in left hemisphere capacity [20] or because they suffer from left hemisphere pathology [ 131. In either case, in comparison with normal controls LD children are expected to generate heightened levels of left hemisphere engagement with a concomitant suppression of processing in the right hemisphere. In effect, the LD child may be over-lateralized. Thus, the LHA model is in marked opposition to the PSL and RHA models, advancing exactly opposite predictions for differences between LD children and high achieving children in dichotic listening. The LHA model finds some support in DAD studies that employed CV syllables as stimuli [35, 361. The LD children and controls were selected using similar criteria to the other directed attention studies. The LDs were poor in reading and arithmetic. The normals were high achievers who were matched with the LDs on CA, gender and handedness and recommended by their teachers. In both studies LD children demonstrated a stronger REA than controls. Unfortunately, these results are difficult to interpret because the LD children in both studies actually outperformed the normals in recalling CV stimuli from the left and right channels. Clearer evidence endorsing the LHA model was reported by AYLWARD [l] who found a larger REA in dysphonetic and dyseidetic, dyslexics that was produced primarily by their poorer ability to recall post-cued, digit names from the left channel. The LHA model has additional. suggestive support from an electrophysiological experiment [42] and from one regional blood flow study with adult alexic patients [16]. The LHA model was preferred as a working hypothesis by H~scoc~ and KINSBOURNE [ I33 in their recent critical review. Bidirectional

excessive

hemisphere

activation model (BHA )

A fourth contingency is that an instability in left hemisphere language processing in LD children may make them extra-vulnerable to bidirectional attentional demands which, in turn, can promote excessive activation in either the left or right hemisphere [ 18, 20, 2 1. 22, 361. We use the term “excessive activation” to include the possibility of a greater spread of activation which, together with-an increase in arousal intensity. would produce less efficient processing and make it more difficult to reallocate the focus of processing between separate anatomical areas [21]. This model, referred to as the BHA model, implies: (a) that the dichotic test measures hemispheric activation coexisting with an underlying structural system [27]; and (b) that the cortical structure for language in LD is normal in its

184

JOHV R. KEKSHSEK and L. L. Mmrm

geographical distribution but that hemispatial attentional and response requirements of verbal tasks can induce both weaker or stronger than normal lateralized processing. We assume that laterally displaced attention challenges activation mechanisms in the contralateral hemisphere [2,3.24,29,30] and that such attentional biases. to the extent that they may remain unchecked by a recurrent inhibitory mechanism, are accomplished by a reciprocal suppression of homologous neuronal units in the ipsilateral hemisphere [25], Moreover, we assume that the left hemisphere does key the linguistic processing in dichotic listening but reporting from the left channel requires some simultaneous. or at least synchronous, right hemisphere engagement [S]. Therefore, the BHA model. which we prefer (see Kershner [20,21] for a more detailed rationale) makes two related claims. LD children are expected to have difficulties shifting attention between channels (resulting from hemispatial response bias) and they are expected to be poor in left ear recall whenever attentional activation is incompatible with linguistic processing activation. The BHA model combines the right hemisphere and left hemisphere dysfunctions in attentional control of the RHA and LHA models into a more inclusive, dynamic framework. According to the BHA model, whether LD children will be less or more lateralized than normal will depend entirely on the asymmetrical attentional, processing and response demands ofthe task or. in this case, the dichotic procedure. Regrettably, the widespread use ofCV stimuli in previous studies precludes even a posr hoc consideration of the BHA model’s claims because CV stimuli (a) do not promote a strong hemispatial response bias [15] and (b) may involve nonlinguistic processes in the right hemisphere for their discrimination [28]. Thus, left channel recall may not be incompatible attentionally with processing CV stimuli. This may explain why LD children have been found to be superior to normals in recall when CVs are used as stimuli [2. 34, 35, 36. 371. Moreover. it is apparent that the directed attention research viewed collectively has little or no power of discrimination among the other competing theoretical models. Two design modifications, however, will permit a critical, a priori test among the PSL. RHA. LHA and BHA models. Research design and predictions

A well-known phenomenon in DAD research is priming bias [ 12. 141. Recalling stimuli from the ear attended second can be more difficult than the ear attended first. This difficulty in shifting attention from the first attended channel to the second attended channel is produced largely by attention acting on a late response organizing process [ 153. Thus, the dificulty of making a leftward shift in directed attention should be proportional to the relatively greater engagement of the left hemisphere in excitatory processing when attending right. Strongly lateralized individuals will have relatively more difficulty shifting attention to the left channel. Somewhat similarly, the difficulty in shifting rightward should be proportional to the relative level of engagement of the right hemisphere when attending left. Poorly and weakly lateralized individuals will have greater difficulty shifting to the right channel. Because the LD latcrality models make fundamentally different assumptions about hemispheric imbalances in responding to verbal stimuli, examination of the priming bias can be useful in testing the validity of the models. In directed attention studies, the effect of priming can be analyzed in Group by Order interactions. But. as we have already indicated, research with normal adults suggests that dichotic CV stimuli. which have been used almost exclusively in LD research, do not produce a priming effect [IS]. This may explain why priming bias has not been analyzed in most

POOR ATTENTIONAL

CONTROL

IN LD

185

studies. Only one LD study [7] reported a Group by Order analysis and no priming interactions were found. On the other hand, dichotic digit names do produce a priming bias. Priming appears to require encoding multiple stimuli that are localized easily from the ear of presentation [14]. Hence, we used digit strings, which are known to produce a priming effect. rather than CV stimuli in the present experiments. An REA for dichotic digit names is associated with left hemisphere specialization for speech whereas an LEA is associated with right hemisphere speech [23]. Finally, we use the term REA to refer only to between-ear differences or laterality coefficients that have been calculated by estimates of the likelihood of reporting from the right ear compared to reporting from the left ear. We do not consider ratios of correct stimuli to intrusions from a single channel as measures of laterality. Each model makes different predictions: 1. PSL model. LD children will show a reduced REA resulting from increased difficulty switching rightward after attending left and less dificulty switching leftward after attending right. LD children will be poorer than controls in reporting from the right ear-first and they may be better than controls at the left ear-first. Academic achievement will correlate positively with the REA. 2. RHA mode/. This model has the same predictions as the PSL model except that right ear recall when attending right-first need not be a contributing factor to the LD’s reduced REA. 3. LHA model. LD children will show a greater REA resulting from increasing djficult~ switching leftward after attending right and less dijjiculty switching rightward after attending left. LD children are not expected to be better than controls at the right ear-first because the model hypothesizes that LDs require higher left hemisphere arousal compared to normals to attain the same level of performance: they may be poorer than controls at the left ear-first. Academic achievement will correlate negatively with the REA. 4. BHA model. Group differences in laterality will interact with the order in which the ears are monitored. LD children will show a reduced REA when attending left-first but an exaggerated REA when attending right-first. This will result from the LD’s greater difficulty altering attentional and response focus between precued channels, regardless of direction. In addition, because the BHA model proposes both a response perseveration or priming bias and a leftward attentional deficit, the LD children should be poorer in recall from the left ear regardless of order. Finally. the BHA model predicts that academic achievement-laterality correlations in the normal and LD children, combined. will interact with ear order. In the left-first group, the REA should correlate positively with achievement; while in the right-first group the REA should correlate negatively with achievement. In the first experiment these predictions were tested with a group of LD children attending public schools, using a group testing, DAD procedure [22]. In the second experiment, the results of the first experiment were replicated with a second cohort of LD children from a private school. In the second study. reading, spelling, and arithmetic achievement scores were correlated with dichotic performance which was obtained in more traditional, individual testing.

EXPERIMENT

1

Method Suhject.s. Thirty-two LD children (27 males and five females: mean CA = 12.3 years. SD = I .4) were selected from special education classes for LD children. An age. and gender-matched group of 32 high-achieving children were selected by teacher recommendation from regular classes at the same schools (mean CA= 12.0, SD= 1.O). The LD children had been identified by a Special Education Identification, Placement and Review Committee (SEIPRC).

JOHU R. KEKSHUER and L. L. MORTON

1X6

accordmg to Ontario Ministry of Education Guidelmes. which are consistent with the Federal Register defimtlon.* They demonstrated low-average to average intellectual potential (WI%-R performance. mean =92.3. SD= 12.6) and a minimum deficit of2 years m school achievement in at least reading and arthmetic. Because of the desirabilit) for a high achieving normal control group. it is not unusual for LD children m dichotic studies to be somewhat lower on measures of IQ. However. studies wtth CV sttmuli [17] and with directed attention. diglts [31] indicate that IQ does not correlate with dichotic ear advantage. We recognize that this issue requires more detailed study; but m the present experiment we do not have controls for the effects of IQ. Because they already had been placed in special educatton classes. we can assume that this sample represented a group of severeI! disabled children with prlmariij auditory-lingurstic deficits. Left-handedness was found in five LD children and m none of the controls. Entermg handedness as a grouping variable in the design for the LDs revealed that there were no slgnilicant effects due to handedness. Fs < 1. Routine audiometric screening carried out by the schools had estabhshed that all children demonstrated normal pure-tone thresholds bilaterally at 20 dB (X0--X.ooO HZI. ,Marerru/.~. The tape of dichotic digits was recorded at the Phonetics Laboratory of the Umvcralt!, of Califorma. San Diego. Stimuli of 307 msec duration were produced from a computer tape of digitized natural speech (adult female voice) after amplitude and duration normalization. A total of 64 dlchotic pairs of monosyllable digits was arranged into 16 trials consisting of two blocks (8 trials each) of quadruple pairs. They were presented at the rate of 2 pairs per set within trials, with 14 set intervals between trials. The dichotlc tape was played on a TEAC 160 Stereo Cassette Deck-C47, through Realistic NOVA 40 headphones connnected in series to permit group texting. A Hewlett&Packard 427A voltmeter was used to measure output at each headphone. The ambient noise level was 40 dB. The average signal amplitude for each channel at each outlet was 65 dB. Proc,rdure. Dichotic testmg was done in a quiet room with group:, of four children by an E who was bhnd to the experimental hypotheses. One-half ofeach reading-level group were asslgned randomly to either a Left Ear-First or a Right Ear-First condition. Testing was preceded by 8 warmup trials ofdigit pairs presented in a free-recall format. The children wrote their test responses on a sheet of paper with 16 lines. each line containing a strmg of 4 blank spaces corresponding to the number ofdigits on each trial. We are aware that writing in dichotic free-recall can have a differential performance effect in LD children [?I]: but the directed attention modification eliminates this group bias 161, Students were instructed to record the numbers heard only from the designated ear. encouraged to report 4 dIgIts per trial and the target ear was touched by the E and by the students. Following 8 trials from one ear they were instructed to switch their attention to the other ear for 8 trials. Again. the target ear was touched b: the E and by the students Headsets were reversed between directed conditions to offset any channel differences In signal-to-noise ratio Achievement (LD. Normal) and Order (Left Ear-First, Right Ear-First)groups,wslthin a split plot design. recalled all the digIts that they could remember from the precued ear when Directed Left and when Directed Right. Performance was scored for the number of correct digits from the attended channel and the number of intrusions from the nonattended channel.

Results Analyses based on inverse sine (arc sine) transformations [44] did not differ from the resu!ts when raw scores were entered; so the statistical analyses are based on the raw scores. G‘t~t~rrc~lrrsulrs. Table 1 presents the mean percent reported by group, order, directed condition and ear. The nearly identical results from Experiments I and 2 are displayed together for easy comparison. A fixed effect, 2 groups (LD, Normal) x 2 ear orders (LF, RF) x 2 directed attention strategies (DL, DR) x 2 ears of recall (L. R) ANOVA was computed with repeated measures on the last two factors. Significant main effects for Directed Attention and Ear were precluded by a Group x Directed Attention x Ear interaction. F (I,60)= 11.41, P