Forced-Attention Dichotic Listening With University Students With Dyslexia

Participants

The sample was composed of 107 students, 50 males and 57 females, ranging in chronological age (CA) from 18 to 55. All participants except three with dyslexia wrote with their right hand and indicated a right-hand preference in manual activities. The dichotic performance of the left-handed participants was unexceptional, so they were included in the analysis. The participants were divided into three experimental groups. A total of 34 good readers, 6 males and 28 females (CA M = 33.5, SD = 9.3), volunteered from graduate courses at the Ontario Institute for Studies in Education (OISE), University of Toronto, Canada. The participants with dyslexia were 73 undergraduate and graduate students selected from a larger pool of consecutive admissions to the Adult Study Skills Clinic at OISE. The participants with disability had an absence of hearing impairment, no history of neurological or attention-deficit disorder, and no history of mental illness or affective disturbances. Audiometric testing was done to ensure hearing within normal limits for frequencies 500, 1,000, 2,000, and 3,000 Hz. and interaural acuity differences less than 15 dB. A two phase screening was used for the participants with dyslexia and assignment into the moderate and severe groups. Each participant selected indicated that inability to read was the primary reason for applying to the clinic for help. Second, all had to score below the mean on the Woodcock Word Attack Test–Revised, a measure of pseudoword decoding that is dependent on phonological processing. Research has shown this test to be a valid assessment tool in the identification of dyslexia and that it can replace more extensive testing without a loss of reliability (Ackerman, Paal, Holloway, & Dykman, 1992). Participants were then assigned to lower and higher functioning subgroups based on whether their phonological deficit was moderate (less than one but more than two SDs below the mean) or severe (more than 2 SDs below the mean). A childhood diagnosis of reading disability had been made in 34 participants, 14 in the moderate group and 20 in the severe group. The moderate group was composed of 42 participants, 21 males and 21 females (aged M = 26.0, SD = 6.5). The severe group was composed of 31 participants, 19 males and 12 females (aged M = 27.2, SD = 7.9). Because the nondisabled were older, F(2,104) = 9.42, p < .01, and composed of more females, χ²(1) = 14.91, p < .01, analyses were conducted to verify that these differences were not threats to validity. In a preliminary ANOVA on the dichotic scores, gender failed to interact with group or ear order. Also, previous research with close approximations to this testing paradigm has not reported gender by ear effects (Hiscock & Beckie, 1993; Hiscock & Stewart, 1984). Finally, Pearson product–moment correlations between CA and dichotic performance were nonsignificant, failing to meet a minimal regression requirement for CA to be considered as a covariate. Therefore, gender and CA can be discounted as influential factors in the results.

The Experimental Paradigm

The dichotic tape was prepared by Auditec of St. Louis, consisting of 30 pairs of consonant–vowel syllables (CVs) differing only in the initial consonant sound (ba, da, pa, ka, ta, ga) The CVs included all possible nonidentical pairings that were stratified in presentation so that each randomly selected pair occurred once at each ear in each set of trials. The CVs were presented with an interval of 6 s via Realistic Nova 40 headsets on a TEAC 160 Stereo Cassette Deck-C 47. A Hewlett-Packard 427A voltmeter was used to measure output at each headphone. The average signal amplitude for each channel was 70 dB, and the ambient noise level was 30 dB. Prior to testing the experimenter described the stimuli and task. Each participant was shown a written list of the stimuli and was required to pronounce each one. Participants with a disability were tested individually, and the nondisabled participants were tested in groups of five or six. Participants wrote their responses with paper and pencil and were encouraged to guess if unsure. Testing was done in one session in a quiet room. Each participant received 65 trials, consisting of 5 warm-up trials of free recall followed in counterbalanced order by 30 trials with attention forced to the LE and 30 trials of attention forced to the RE. Attention was forced to each ear prior to each block of trials by the experimenter touching the designated ear of each participant and instructing to “attend to this ear and write down only what you hear from this ear.” Headsets were reversed during a 2-min interval between ear orders to offset any channel differences in signal-to-noise ratio. Performance was scored for the number correct (C) from the attended ear and the number of intrusions (I) from the unattended ear. This resulted in two forced ear conditions (FL, FR) for C responses (FL-LE and FR-RE) and two forced ear conditions for I responses (FL-RE and FR-LE). This paradigm resulted in a Group (nondisabled, moderate disability, severe disability) by Order (LF, RF) by Ear (LE, RE) design with repeated measures on the last factor.

Preliminary Analysis

A preliminary analysis was performed on the raw score percentage correct and intrusions shown in Table 1. A three-way ANOVA (Group × Order × Ear) for correct responses resulted in a main effect for Group, F(2, 101) = 17.59, p < .01, produced by the good readers performing at a superior level compared to both groups of those with dyslexia, F(1, 101) = 35.16, p < .01, whereas those with moderate and severe dyslexia performed at the same level, F < 1. A main effect for Ear, F(1, 101) = 18.69, p < .01, was produced by an REA in all groups. A significant Order effect was precluded by a three-way interaction of Group, Order, Ear. Post hoc tests showed that in the LF Order the good readers produced a stronger REA in comparison to both groups with dyslexia, F(1, 101) = 8.60, p < .01. In the RF Order the good readers showed a lower REA, but only in comparison with those with severe dyslexia, F(1, 31) = 4.77, p < .05. More specific testing between groups by Ear and Order would be unwarranted in view of the good readers’ overall superior performance. Moreover, the theoretical import of the study has no grounding for predictions for group differences in performance at a specific ear. The focus of the study is to examine group differences in their relative recall between ears as indicated by the REA in each Order. A three-factor ANOVA was also computed on the intrusion responses. Intrusions made up 82% of the errors for the good readers and 70% of the errors for both groups with dyslexia. Unlike the correct analysis, no main effect occurred for Group and an Ear effect was precluded by a Group by Ear interaction, F(2, 101) = 6.10, p < .01. Post hoc tests showed that those with moderate dyslexia were the only group not showing an REA. There were no other significant effects. Thus the correct raw score results were consistent with predictions, but the intrusion analysis was theoretically uninformative. However, the raw score analysis was seriously confounded by the overall superior performance of the good readers, which can artifactually influence raw score results (Hiscock & Decter, 1988). Thus the raw scores were transformed into the Lambda laterality index recommended by Bryden and Sprott (1981).

OPEN IN VIEWER

Table 1.

Percentage Correct and Intrusions for Consonant–Vowel Syllables in the Left Ear First and Right Ear First Orders.

	Left ear first		Right ear first
Group and Attended Direction	LE	RE	LE	RE
Severe dyslexia, DL	C	I	C	I
M	.29	.40	.34	.37
SD	.09	.11	.14	.13
Severe dyslexia, DR	I	C	I	C
M	.35	.32	.25	.48
SD	.12	.17	.09	.14
Moderate dyslexia, DL	C	I	C	I
M	.30	.39	.38	.32
SD	.13	.17	.09	.08
Moderate dyslexia, DR	I	C	I	C
M	.41	.32	.29	.44
SD	.15	.13	.11	.12
Good readers, DL	C	I	C	I
M	.39	.41	.51	.34
SD	.10	.09	.20	.15
Good readers, DR	I	C	I	C
M	.26	.57	.27	.52
SD	.11	.18	.15	.20

Note. C = correct; DL = directed left; DR = directed right; I = intrusion; LE = left ear; RE = right ear.

The log odds ratio or Lambda = In (RC)/LC). This coefficient adjusts for group differences in performance and yields an overall laterality score including correct and intrusions for each ear order (LF, RF). This index, In [(RC)(LI)/(LC)(RI)], compares the likelihood of recalling from the RE when attending right with the likelihood of recalling from the LE when attending left. In this more precise analysis, the ability to shift channels (priming) can be estimated more directly by examining the REA in each order.

Dichotic Listening

Lambda means and standard error of the means are shown in Figure 1. To directly address the main hypotheses of the study, two planned comparisons were computed in each ear order (LF, RF) comparing the good readers with each group with dyslexia. In the LF order, the good readers produced a significantly greater REA compared to those with moderate dyslexia, F(1, 39) = 21.69, p < .05, and those with severe dyslexia, F(1, 30) = 5.71, p < .05. Cohen’s d was used to express the effect size (ES) for the significant pairwise comparisons. The ES in the good readers versus moderate dyslexia comparison was 1.7, which indicates that 95% of the good readers would be above the average REA of those with moderate dyslexia. In the good reader versus those with severe dyslexia comparison ES was 0.67, which indicates 75% of the good readers would be above the average REA of those with severe dyslexia. In the RF order, the good readers produced a marginally lower REA, but only in comparison to those with severe dyslexia, F(1, 31) = 2.87, p < .10. The planned comparison between the good readers and those with moderate dyslexia was not significant, F < 1. In the ES comparison of interest, those with severe dyslexia versus good readers, ES was 0.94, indicating that 83% of those with severe dyslexia would be above the average REA of the good readers.

OPEN IN VIEWER

Figure 1.

Lambda laterality index for the good readers and those with moderate or severe dyslexia in the left first and right first orders.

Categorical, Nonparametric Analysis

To further explore the data for corroboration of the parametric testing which treats the dichotic scores as continuously distributed, binomial tests compared the number of participants showing an LE advantage (LEA) or REA in both orders (LF, RF). These results displayed on Table 2 show that in the LF order, the good readers were the only group to demonstrate a significant number of participants with an REA. This quantifies and corroborates the dichotic index results. In the RF order, however, the effect of borderline significance in the index was clearly significant. Those with moderate dyslexia and the good readers were similar in not demonstrating a significant number of participants with an REA. Both group’s performance at the RE when attended first was low relative to their facility in shifting attention to the LE. Consequently, a greater proportion of participants in the good reader and those with moderate dyslexia groups demonstrated LEAs. In contrast, those with severe dyslexia maintained a significant number of participants with REAs, caused primarily by inability to switch attention to the LE. Greater REAs were demonstrated in the LF order by the good readers and in the RF order by those with severe dyslexia. In addition, one-tailed Mann–Whitney U tests comparing each group’s performance across the two orders indicated that the good readers (U = 93, p < .05) and those with severe dyslexia (U = 86, p < .05) produced significant alterations of the REA between orders. Finally, Hiscock and Chipuer (1993) suggested calculating an “unbiased” REA by comparing only the ears attended first. This measure is intended to be a more accurate estimate of the LH hardwired specialization for verbal processing and may even be a more reliable estimate of baseline lateralization than including a free-recall condition. In our lab at the University of Toronto, we found that dichotic free recall, that is, not controlling for attention, produced greater variability in responses without significantly influencing the REA. The unbiased first ear attended scores suggest that those with severe dyslexia, in the absence of attention switching effects, are highly lateralized, scoring +.999, compared to the good readers score of +.797.

OPEN IN VIEWER

Table 2.

Binomial Tests of Significance for the Number of Ss With an Left Ear Advantage (LEA) Versus Right Ear Advantage (REA) in Left Ear First and Right Ear First Orders.

	Left ear first			Right ear first
Group	LEA	REA	p	LEA	REA	p
Good readers	1	16	<.01	7	10	ns
Moderate dyslexia	11	13	ns	8	10	ns
Severe dyslexia	6	9	ns	3	13	<.01

Note. p values determine whether the incidence of REAs versus LEAs is significantly different from chance by binomial test. ns = nonsignificant difference.

Dichotic Listening

The sheer magnitude of multimodal environmental stimuli surrounding us in our daily activities requires the recruitment of neuronal screening devices (Broadbent, 1958) to block irrelevant stimuli while selecting goal-directed targets for sequential processing. This executive action, the cognitive control of attention, is essential for survival in the broadest sense. More specifically, such top-down recruitment of attentional resources is a primary ingredient for academic success, especially in processing speech and printed text (Stevens & Bavelier, 2012). Without such successful management, the scores of stimuli competing for attentional priority in selection and computational processing in school tasks or intellectual assignments would pose formidable restraints to successful learning. The FA dichotic paradigm provides a narrow window into an important aspect of the functioning of this neurocognitive control mechanism, providing insights into the ability to select among competing stimuli and the ability to shift the focus of attention, essential for processing multiple events over time.

Dichotic research with adults with dyslexia addresses what is estimated conservatively at approximately 3% to 5% of the general population (National Adult Literacy Survey, 1992). No comparable Canadian estimates are available. The prevalence among university students is unknown and difficult to establish. On one hand, the rigors of competing academically at upper education levels would discourage some with reading disabilities from attempting higher studies. But on the other hand, once accepted and attending university classes, such students are more likely to come forward to be identified. Of importance, the behavioral findings from dichotic studies can now be synthesized with neuroimaging (fMRI) research, which has vast potential to enhance the scope and power of interpretation (e.g., Kompus et al., 2012). Neuroimaging initiated a new era in cognitive science and neuroscience. Measuring the magnitude of differences between oxygenated and deoxygenated blood, blood oxygenated level-dependent (BOLD) contrast, noninvasively identifies local metabolic activity of the brain in response to specific tasks and goal-driven behavior.

Structural Lateralization

Analysis of the LE and RE scores when each ear was attended first suggested that these adults with dyslexia were not lateralized atypically. Indeed, as shown on Table 1, those with severe dyslexia were the only group unable to overcome the inherent REA when directed to switch attention to the nondominant LE after recall from the RE. We recognize that the unbiased REA is not entirely free of attentional bias. The RE stimulus is more salient due to structural and/or arousal advantage. Thus, forced LE recall promotes a conflict between the stronger RE stimulus and instructions to attend left. Furthermore, LE and RE recall have been found to engage different neurological and cognitive networks (Kompus et al., 2012). Nevertheless, with these caveats in mind, the present results show no evidence for less lateralized language systems in the adults with dyslexia. Also, the results showed that the REAs of both the good readers and those with severe dyslexia were notably modifiable across orders. This finding further supports the antistructural view that the measured REA in dichotic studies can be modulated in nondisabled and participants with dyslexia by attentional requirements. The net result, however, clearly revealed significant group differences. The greater REA in the RF order of those with severe dyslexia is a pattern that was found in previous research to correlate negatively with reading and spelling (higher REA: poorer reading/spelling). Conversely, both groups with dyslexia produced low REAs in the LF order, the order in which higher REAs correlate positively with reading and spelling (Kershner & Morton, 1990). In addition, the poorer ability to shift attention to the RE of those with dyslexia remains unaltered by a reading-level match (Kershner & Graham, 1995) and in the present study by a reading-level control. Therefore, attentional modulation of the REA in dyslexia appears to be dysfunctional and a liability to school achievement. In good readers, attentional modulation of the REA appears to reflect more efficient cognitive control.

Right-Ear First Order

The severely impaired adults with dyslexia were lateralized highly in comparison to the nondisabled students and the moderately impaired with dyslexia, produced mainly by their poorer ability to switch attention from the RE to the LE. That this effect was related to their level of reading suggests that their difficulty switching left is a secondary function of their reading level, a developmental delay, as opposed to qualifying as a root cause of dyslexia. An alternative possibility is that this component of their dichotic performance represents a time-locked position along a dimensional model of reading with their performance fixed at the lower end of a normal distribution (Giger, Boreecki, Smith, DeFries, & Pennington, 1996). In both cases, however, the incapacity of those with severe dyslexia to switch left does not imply a neurobiological deficit. The underlying neurobiology of this behavioral finding, nonetheless, is worth pursuing. Recently, Kompus et al. (2012) presented a seminal, theoretical model for elucidating the underlying neuronal events recruited during the FA dichotic procedure. In an fMRI experiment, closely mirroring the present dichotic procedure, hemodynamic BOLD neuronal activation patterns were examined while nondisabled adults recalled CV syllables during the FA dichotic test. Recall was forced to the LE or RE, with CVs blocked into 30 trials to each ear counterbalanced by order. Forced recall from the LE produced left hemisphere inferior frontal cortex (IFC) and caudate activations (Brodman’s areas 44/45 and extending posteriorly to the parietal-occipital sulcus). In contrast, when recall was forced to the LE or RE the right hemisphere IFC and caudate were activated bidirectionally. This instructional dichotic model shows that in nondisabled adults the left IFC is activated when recalling from the LE, but deactivated when recalling from the RE. Related fMRI research has shown that the left IFC is centrally involved in phonological processing (Bookheimer, 2002; Scott & Johnsrude, 2003) and in normally developing children between 5 and 11 years of age, gray matter thickening in the left IFC is associated, independently of general maturation, with improving phonological processing (Lu et al., 2007). Thus, considerable evidence indicates that the left IFC plays a critical role in both dichotic listening and phonological processing. Since disengaging attention right to attend left in the RF order requires initial engagement of the left IFC, we would expect the region of the left IFC (Broca’s territory) in students with dyslexia to show atypical activation for their CA, that is, failure to activate.

Neuroimaging research with children and adults with dyslexia using a variety of phonological and other language tasks has found reduced IFC activation (Georgiewa et al., 1999), but also enhanced activation (MacSweeny, Brammeo, Waters, & Goswmi, 2009). A study by Shaywitz et al. (1998) suggests that adults with dyslexia may fail to efficiently temper the engagement of the left IFC. They reported that adults with dyslexia compared to nondisabled adults failed to systematically modulate left IFC activation as phonological tasks became increasingly more difficult. Those with dyslexia demonstrated IFC hyperactivation to the smallest increases in difficulty. Therefore, the present results showing inferior ability only in those with severe dyslexia to shift attention left suggest lack of control in efficiently engaging the left IFC. However, this inference, which is based on only a few studies, has to be seen as preliminary speculation pending further research. Nonetheless, the difficulty shifting left of those with severe dyslexia does not appear to be exceptional for their reading level or a causal factor of dyslexia.

Left-Ear First Order

Altering recall strategy to the RE after serial trials from the LE requires (a) continued activation of the right IFC and caudate and (b) a new requirement for deactivation of the left IFC and caudate (Kompus et al., 2012). Cognitive control of attention appears to be mediated by a frontostriatal network. Those with dyslexia in comparison to the nondisabled students in the present study experienced poorer performance orchestrating the switch to the RE, again implicating functions of the left IFC. However, unlike its suggested role in modulating active left IFC engagement, to facilitate switching right the left IFC must be modulated downward to decrease the level of activation. Replication of this dichotic empirical finding in adults with dyslexia shows this to be a lifetime problem, suggesting it may reflect a neurobiological aspect of dyslexia with etiological importance.

Neuroimaging studies provide some support for the involvement of the IFC in this effect and, furthermore, suggest that left IFC overactivation may signal a disconnection to other LH language areas. In studies with adults with dyslexia that have examined hemodynamic activation in the left IFC, the most frequent finding is greater activation, not less activation (MacSweeny et al., 2009). In a directly relevant study, Shaywitz et al. (1998) reported that adults with dyslexia, while performing phonological tasks, showed overactivation in the left IFC region, while at the same time showing underactivation in interconnected LH posterior areas, viewed as reflecting a phonological impairment. Attenuated activity was found in the superior temporal gyrus (Wernicke’s territory), angular gyrus, and striate cortex. But in a study with adults with dyslexia, which attests to the complexity of interpreting fMRI activation patterns as they relate to academic performance, Hoeft et al. (2007) found that reading level accounted for significant variability in a complex of excessive activation patterns, including left IFC overactivation. Only reduced activation of left posterior language areas was unaffected by reading level. One interpretation of this finding is that the association of reading level to numerous left hyperactivations may be produced by those with dyslexia expending greater effort when performing language tasks. However this issue turns out, suggestive research indicates that the left IFC may be recruited in language processing to form articulatory programs for speech (phonemic segmentation), whereas the left posterior temporal and angular gyrus may map phonological representations to semantics (McCrory, 2004). On the basis of this line of reasoning, it has been proposed that dyslexia may result from a disconnection between these LH regions. In further support of this notion, Van der Mark et al. (2011), in a study with normal children and those with dyslexia, found a disruption in connectivity in those with dyslexia between left IFC and both the ventral occipitotemporal language system and inferior parietal language area. Connectivity in the nondisabled occurred with beginning reading instruction and showed an anterior to posterior gradient of activation. There was relatively less left IFC activation as reading skills improved. Adding to the evidence for such a disconnection underlying dyslexia, Lopez-Barroso, Cantani, Pipollies, Dell’Aequa, and Rodrequez-Funella (2013) combined diffusion tractography and fMRI scans with nondisabled adults. They found that individual variability in the anatomical integrity and functional strength of the direct connections between Broca’s and Wernicke’s territories were correlated with better ability to learn phoneme segmentation skills. At the neuroanatomical level, the pathway between left IFC and the left posterior superior temporal gyrus is the left arcuate fasciculus. This is the communication link that may be poorly interconnected in dyslexia. In another recent study with adults with dyslexia, Boets et al. (2013) found significantly reduced functional connectively and reduced white matter integrity of the left arcuate fasciculus. Furthermore, they found that the phonetic representations of those with dyslexia were intact and of high quality, thus suggesting a very different orientation to the thrust of current remedial programming. Therefore, there appears to be a persuasive body of research attesting to the importance of the left IFC and its communication pathways to and from posterior language areas as causal factors in dyslexia. The left IFC appears to be integral to a distributed neural network, with overactivation in dyslexia reflecting impoverished connectivity with the left posterior superior temporal region. Therefore, the incapacity of adults with dyslexia to switch attention form the LE to the RE in the current study may have resulted from a disconnection between left anterior and posterior language zones, compromising the cognitive control mechanism required to modulate decrements in left IFC activation.

Remedial Implications

There is a paucity of theoretically based remedial programs for adults with dyslexia, and this shortcoming is even more severe for university students (see Swanson, 2011, for a comprehensive review and meta-analysis of adult reading disability issues in assessment and practice). The present dichotic results would appear to be consistent with the notion that a primary cause of dyslexia may be a disconnection between LH phonological and semantic processing networks: difficulty accessing otherwise intact linguistic representations. This concept, pending further confirmation, will require a reappraisal of the orthodox, primary emphasis in remediation on word recognition and phonics. The results suggest that a key ingredient to such a reorientation should be activities to improve the efficiency of top-down attentional control. Other programming ingredients should be based on the recognition that the cognitive problems in dyslexia, especially university students with above average IQ, extend well beyond phonological-semantic processing to include memory, vocabulary, and naming speed (e.g., Swanson, 2011). Detailed suggestions, of course, are beyond this article. However, although future research studies are needed to more carefully examine and refine the present findings, the theoretical model of dyslexia that is presented here has the potential to inform new approaches to remediation. Recent encouraging research demonstrates that genetic and epigenetic mechanisms interacting with learning environments can alter brain architecture and interconnections (Zatorre, 2013).

In summary, William James (1886, p. 379) may have been one of the first psychologists to recognize the importance of stimulus priming effects on cognition when he wrote, “The phenomena of ‘summation of stimuli’ in the nervous system prove that each stimulus leaves some latent activity behind which only gradually passes away.” And, as demonstrated by Saetrevik and Hugdahl (2007), such negative priming can be mitigated by engaging efficient cognitive control mechanisms. This priming study with adults with dyslexia replicates previous dichotic listening studies with children with dyslexia (Kershner & Graham, 1995; Kershner & Morton, 1990) who were also found to be less successful in recall, showing a negative priming effect, when forced to switch from one ear to the other. But inferior performance switching to the LE was found in previous research and in the present study to be a function of poorly developed reading ability. Inferior switching to the RE, however, occurred relatively independently of their reading level. Functional asymmetry between ear orders suggests that the direction of the shift in attention may engage different cognitive and neurobiological control networks, providing behavioral support for the fMRI dichotic findings of Kompus et al. (2012). Therefore, the behavioral switching effect does not merely reflect less agility in altering attention per se. Indeed, replication with adults with dyslexia suggests that such findings may be important in the identification of the neuronal substrates that produce the disorder. There was no evidence to suggest less or weaker LH lateralization. Opposed to this traditional concept, the present study demonstrated negative priming effects in those with dyslexia in both ear orders, producing significant malleability in the REA and implicating a dysfunction of the frontostriatal attention network. The poorer facility to shift leftward of those with severe dyslexia appears to be a reading-level incapacity to efficiently engage processes in the left IFC. In contrast, inferior ability to shift rightward implicates failure to modulate excessive left IFC activations, which may reflect poor connectivity with LH posterior language areas.

Finally, although the underlying neurobiology of this effect presents a coherent theoretical account, due caution is warranted. The functioning of the brain’s sheer complexity of interconnections currently is understood in only the most rudimentary form. The brain works as a whole with coordinated networks of billions of neurons and neuronal complexes. The interaction of multiple attentional systems in modulating processing regions could not be more consequential to advances in cognitive science and education; but it is important to acknowledge that our knowledge base has a long way to go and invariably will become more complex (e.g., Petersen & Posner, 2012).