Brain, Vol. 126, No. 9, E3,
September 2003
© 2003 Guarantors of Brain
doi: 10.1093/brain/awg218
Letters to the Editor |
Reply to: Visual magnocellular deficits in dyslexia
1 Interdisciplinary Center for Neural Computation and Departments of 2 Neurobiology and 3 Psychology, Hebrew University of Jerusalem, Israel, and MRC Institute of Hearing Research, Nottingham, UK
Correspondence to: Sygal Amitay. E-mail:sygal{at}ihr.mrc.ac.uk
The hypothesis of a mild magnocellular deficit in a large number of dyslexics has been challenged many times in the past. In fact, there have been more studies that failed to verify its predictions than those that successfully did. For example, a recent review by Skottun (2000
) describes many studies that failed to verify the theory’s main prediction that dyslexics will have poor contrast sensitivity for sinusoidal gratings of high temporal frequency and low spatial frequency. Still, we decided to perform a systematic evaluation of this magnocellular hypothesis. We chose to do so because we found the magnocellular hypothesis very appealing. Its main merit was in relating a plethora of complex and seemingly unrelated phenomena to a single deficient mechanism, and, within the visual system, to a concrete anatomical pathway. As such, this hypothesis yielded specific predictions that could be assessed with many different methodologies. To our disappointment, when we assessed its predictions with a relatively large sample population (n = 60; 30 dyslexics and 30 controls) and with a broad range of psychophysical tasks, these predictions were not confirmed.
Most of the dyslexics in our sample had no difficulties on magnocellular tasks compared to our control population. A small subgroup (6/30) did have difficulties. However, the psychophysical performance of this small subgroup was poorer than that of the controls on all the tasks we assessed. Naturally, we did not claim that these few individuals do not have a magnocellular deficit, but rather that their deficit is not specific to magnocellular stimuli. A ‘magnocellular deficit’ is not a proper description of their perceptual deficits since their impairments were much broader. Note that if we were to characterize the performance of this small dyslexic subgroup on the basis of magnocellular tasks alone, we might have concluded that their deficit is magnocellular. It is therefore essential to include tasks that are not magnocellular when examining the predictions of the magnocellular hypothesis.
However, an even more important aspect of our study was that the majority of dyslexic participants, who had no deficits in magnocellular tasks, showed greatly impaired performance in other perceptual tasks. In the visual domain, the majority of dyslexics had difficulties in a task that required accurate comparisons between the spatial densities of two sequentially presented gratings. The contrast of these gratings was well above all of our subjects’ contrast detection threshold and their presentation was not very brief. Based on these findings our conclusion was, as specified in the title of our paper, that ‘disabled readers suffer from visual and auditory impairments but not from a specific magnocellular deficit’. Given that the major focus of recent studies of dyslexics’ visual abilities is on assessing magnocellular functions, this is an important conclusion. The finding that most dyslexics perform magnocellular tasks well should therefore not be confused with an absence of visual deficits among the majority of dyslexics. A magnocellular deficit is not simply a useful description to capture these deficits among any of our dyslexic participants.
Several specific criticisms were raised, relating to three issues. (i) What is a reliable behavioural characterisation of magnocellular function? (ii) To what extent are the participants of our study representative of the general dyslexic population? (iii) What is the appropriate statistical approach when analysing data from a heterogeneous sample? We shall address these three issues in a slightly more general way since we believe that they relate to fundamental questions, which require careful evaluation.
(i) What would be a good behavioural assessment of magnocellular function? This turns out to be quite a difficult question, since the range of overlap between stimuli that activate magnocellular and parvocellular neurons is rather large. In principle, the most direct means for evaluating this question is to lesion the magnocellular layers of monkeys’ LGN and then assess its effect on psychophysical performance. Being remarkably difficult technically, very few such studies were actually conducted (Merigan and Maunsell, 1990; Merigan et al., 1991; Schiller et al., 1990). Taken together, these studies suggest that only contrast detection of very transient stimuli containing low spatial frequencies is hampered. A particularly clear deficit was found when stimuli did not contain any spatial information within the area of the visual field whose magnocellular representation had been lesioned, i.e. blobs were harder to detect than gratings (Merigan and Maunsell, 1990). The deficit in contrast detection probably underlies the somewhat degraded motion detection found when low contrasts are used (Merigan et al., 1991). To the best of our knowledge, there is no direct evidence of impaired coherent motion detection associated with magnocellular lesions. Given the monkey data, we used detection of whole screen flicker at temporal frequencies of 5–25 Hz as our most sensitive magnocellular task. In addition, we assessed detection of drifting gratings, speed discrimination of low contrast gratings and coherent motion with high luminance dots, mainly since these tasks were used in previous studies assessing dyslexics’ ‘magnocellular’ function (e.g. Borsting et al., 1996; Demb et al., 1998; Cornelissen, 1993).
Given the actual psychophysical monkey data, the basis for Chase and Stein’s statement, ‘we do not conclude that only the perception of brief stimuli would be affected’, is puzzling. Even more surprisingly, they continue this line of argument with the statement, ‘Actually the first evidence for an magnocellular impairment in dyslexics came from studies of contrast sensitivity using static displays (Lovegrove et al., 1980)’. Lovegrove et al. (1980) were the first to assess contrast sensitivity among dyslexics and report deficits, but the deficits they reported were not magnocellular, and they did not relate their findings to a transient system deficit. They measured sensitivity to static grating stimuli of 2–16 c/deg, presented for 40–1000 ms, in a group of 10 dyslexic children. No systematic data were then available to relate this range of stimuli to parvo- rather than magnocellular sensitivity ranges. Twenty-three years later we now know that the stimuli used in their study (with long presentation durations and intermediate or high spatial frequencies), are just the type of stimuli for which sensitivity is not critically dependent on magnocellular pathways. Interestingly, these are the conditions (500 ms presentations, 4 c/deg) where the most dramatic deficits were found in their sample. The conclusion Lovegrove et al. (1980) drew from their initial findings was that dyslexics’ visual deficits were not specific to written material. We do not disagree with this conclusion. However, Chase and Stein’s interpretation that these findings constitute ‘evidence’ for a magnocellular deficit is disturbing. It suggests that their operative definition of magnocellular functions is, at best, blurred and even somewhat circular. In other words, they claim that (a) dyslexics have a magnocellular deficit and (b) magnocellular tasks include all perceptual tasks that dyslexics perform poorly. At this level of blurring, the hypothesis fails to be useful, since it loses all its predictive power.
(ii) Is there a sampling bias in our study? Chase and Stein’s criticism is that the reading level of our participants is unusually severe while ‘at the same time achieving above average performance on Similarities and Block Design subtests’. The suggestion that our dyslexic participants did not show a magnocellular deficit because they are too dyslexic is in itself odd. In any event, we shall first show that our dyslexic participants had a ‘typical’ reading deficit, but Chase and Stein got the measures confused. We shall then show that similar findings were obtained in subsequent studies with different population samples; in adults with mild dyslexia (Ramus et al., 2003) and in a large-scale study of school children (with no prior selection; Talcott et al., 2002).
Regarding our dyslexic participants, Chase and Stein state, ‘We have never seen a sample of 30 dyslexic young adults ... who were on average six standard deviations below the mean of controls on reading measures’. Indeed, the average reading score of our dyslexic participants was six control standard deviations below our controls’ average. Chase and Stein must have confused the standard deviation of our control sample, which is small because the controls we selected are individuals with no reading difficulties, with that of the general population, which is much larger since it is based on a much broader range of reading abilities. We shall illustrate this point with an example taken from a recent paper by Stein and colleagues (France et al., 2002). Their sample consisted of 15 dyslexic young adults and 15 controls. The average reading scores were 13.6 (with a standard deviation of 0.7) for controls and 7.6 for dyslexics. Thus, their dyslexics were nearly 9 [(13.6–7.6)/0.7 = 8.6) control standard deviations below their controls’ average. We assume that Chase and Stein’s confusion stems from the standard deviation measure of the general population. In their notation, the average of the general population was 10 and the general standard deviation was 3. Thus, their dyslexics were 2 general population standard deviations below that of controls. To summarize, the reading scores of our dyslexic participants were typical of such studies (control versus dyslexic populations). The point concerning the orthographic test is interesting. In our sample, the relative difference (in measures of standard deviations) in orthographic abilities was larger than that for reading. This aspect probably characterises Hebrew script. Since spelling in Hebrew is relatively easy, good adult Hebrew readers hardly make any spelling errors, while dyslexics do. Thus, in relative terms, Hebrew spelling is a particularly sensitive measure for diagnosis of Israeli dyslexics.
Other recent studies also report negative findings regarding a specific magnocellular deficit. For example, Ramus et al. (2003) assessed students in University College London. They systematically evaluated the performance of 16 dyslexics and 17 controls on reading, auditory, cerebellar and magnocellular tasks. They found that, first, the majority of their dyslexic participants showed no deficit in magnocellular tasks. Second, two dyslexics (and two controls) showed deficits in magnocellular tasks. Third, for the dyslexics who did show magnocellular deficits, the deficits were not specific. These individuals also had difficulties in auditory tasks using both brief and longer stimuli.
Another study with similar findings was a large scale study (n = 350) among the general population of British children (Talcott et al., 2002). Here, the task used for assessment of magnocellular function was coherent motion discrimination. Thresholds of children with non-verbal intelligence higher than the 50th percentile and reading ability lower than the 25th percentile did not differ from that of an intelligence-matched group whose reading ability was between the 75th and 100th percentile. Thus, among children who would be roughly defined as dyslexic, no deficit was found in coherent motion discrimination. A small reading-related difference in motion discrimination thresholds was found among children with non-verbal intelligence lower than the 50th percentile. This is exactly what we would expect based on our findings. As stated in our paper, the small subgroup with non-specific deficits in psychophysical performance (including magnocellular tasks) had lower non-verbal intelligence scores (i.e. Block Design) compared with the other reading-disabled participants. We thus predict that had Talcott et al. (2002) assessed a broader range of perceptual tasks, they would have found that the small perceptual deficits of these children are not specific to magnocellular tasks.
Taken together, studies using different sampling strategies find similar non-specific magnocellular deficits among individuals with a specific reading disability.
(iii) What is the appropriate statistical analysis? Chase and Stein claim that the statistical methods we used were too conservative. They suggest that we failed to detect (i.e. we dismissed as noise) real, though exceedingly mild, magnocellular deficits among the majority of our dyslexic participants because (a) we used too low-power criteria (in other words, we preferred the risk of missing small, though potentially real, deficits to the risk of falsely accepting noisy results as real deficits) and (b) we calculated the average for magnocellular task performance with respect to all our participants rather than just the controls.
Our findings, as reported, are robust and not sensitive to the exact analysis chosen. In our paper, we performed two levels of analysis. First, at the whole group level [table 2 (Amitay et al. (2002)]. Here, as Chase and Stein note, we found that dyslexics as a group show significant deficits on some magnocellular tasks (e.g. flicker detection) and marginal deficits on others (e.g. coherent motion). These findings are consistent with many previous reports (e.g. Ridder et al., 1997). We then asked two more questions. First, we asked whether these deficits are specific to magnocellular tasks at the group level. Second, we asked whether these deficits are an outcome of small effects among the majority of dyslexics, or large effects among the minority (or a combination of both).
First, at the group level, using exactly the same statistical analyses and criteria, we find deficits on non-magnocellular tasks. Second, as shown in figure 6 and explained in the Discussion (Amitay et al., 2002), when outliers are removed the inter-group effect disappears. Even very liberal statistical tests fail to reach marginal significance levels. For example, figure 2B shows the contrast sensitivity function for flicker detection (i.e. the most sensitive magnocellular task) for the three subgroups: (1) controls, (2) the small subgroup of dyslexics with clearly poor magnocellular and generally poor perceptual performance, and (3) all other dyslexic participants (the majority). The performance of controls (1) and the majority of dyslexics (3) are almost overlapping. Could it be that the minute difference between groups 1 and 3, very far from significance, is real? Perhaps, though it seems unlikely given that using a large population and careful assessment procedures we could not reject the null hypothesis that there is no real difference. Nor are we aware of any other study that rejected the null hypothesis upon finding such a small difference.
Another statistical point raised in Chase and Stein’s critique is that our method of normalising the magnocellular Z-scores was erroneous, in that we calculated these scores in relation to both groups’ average rather than the control group’s alone. In their opinion, this is why we only found six dyslexics with substantial magnocellular deficits, and had we used a ‘control-normalized scale’, we would have found that ‘the vast majority [of the dyslexics] turn out to have impairments’. Using a control-normalized scale, seven rather than six dyslexics were identified as having a substantial impairment—hardly the vast majority. A t-test between the controls (group 1) and the dyslexics with magnocellular Z-scores in the control range (group 3) yielded no significant difference between means (P = 0.29; excluding one control participant with very poor magnocellular Z-score). Thus, even using a control-normalized scale, we cannot conclude that the vast majority of dyslexics have a magnocellular impairment.
In summary, while we do not presume to understand the fundamental perceptual deficit of dyslexics, we believe that revising the initial magnocellular hypothesis to an updated working hypothesis is warranted. A conservative revision would be to a ‘parietal-deficit’ hypothesis (e.g. Hari and Renvall, 2001). It follows the magnocellular hypothesis in terms of looking for specifically impaired anatomical structures. Yet it relates the deficit to high- rather than low-level mechanisms. A parietal deficit is consistent with some other characteristics of dyslexics’ cognitive profile [e.g. impaired performance in the Digit–Symbol (hand–eye coordination) subtest of WAIS-III, (Wechsler, 1997)] and can perhaps account for previous reports of letter localisation confusion (Enns et al., 1995; Cornelissen et al., 1998) and poor spatial resolution around fixation (Geiger and Lettvin, 1997). Our finding that dyslexics have particularly impaired perceptual memory for visual spatial evaluations does not refute this working hypothesis (see Greenlee et al., 2000).
Taken together, a ‘parietal-deficit hypothesis’ could be a useful tool in directing subsequent research. Note, however, that its predictions are quite different from those of the magnocellular hypothesis. A ‘parietal deficit’ hypothesis implies that rather than spending more time and effort on administering low-level tasks such as contrast detection and motion discrimination, we should examine higher-level mechanisms of perceptual attention and memory. Our accumulated recent data suggest that these are the domains we should study now.
References
Amitay S, Ben-Yehudah G, Banai K, Ahissar M. Disabled readers suffer from visual and auditory impairments but not from a specific magnocellular deficit. Brain 2002; 125: 2272–85.
Borsting E, Ridder WH 3rd, Dudeck K, Kelley C, Matsui L, Motoyama J. The presence of a magnocellular defect depends on the type of dyslexia. Vision Res 1996; 36: 1047–53.[CrossRef][Web of Science][Medline]
Cornelissen P. Fixation, contrast sensitivity and children’s reading. In: Wright SF, Groner R, editors. Facets of dyslexia and its remediation. Amsterdam: Elsevier; 1993. p. 139–62.
Cornelissen PL, Hansen PC, Gilchrist I, Cormack F, Essex J, Frankish C. Coherent motion detection and letter position encoding. Vision Res 1998; 38: 2181–91.[CrossRef][Web of Science][Medline]
Demb JB, Boynton GM, Best M, Heeger DJ. Psychophysical evidence for a magnocellular pathway deficit in dyslexia. Vision Res 1998; 38: 1555–9.[CrossRef][Web of Science][Medline]
Enns JT, Bryson SE, Roes C. Search for letter identity and location by disabled readers. Can J Exp Psychol 1995; 49: 357–67.[Web of Science][Medline]
France SJ, Rosner BS, Hansen PC, Calvin C, Talcott JB, Richardson AJ, et al. Auditory frequency discrimination in adult developmental dyslexics. Percept Psychophys 2002;64: 169–79.
Geiger G, Lettvin JY. Peripheral vision in persons with dyslexia. New Engl J Med 1987; 316: 1238–43.[Abstract]
Greenlee MW, Magnussen S, Reinvang I. Brain regions involved in spatial frequency discriminaiton: evidence from fMRI. Exp Brain Res 2000; 132: 399–403.[CrossRef][Web of Science][Medline]
Hari R, Renvall H. Impaired processing of rapid stimulus sequences in dyslexia. Trends Cogn Sci 2001; 5: 525–32.[CrossRef][Web of Science][Medline]
Lovegrove WJ, Bowling A, Badcock D, Blackwood M. Specific reading disability: differences in contrast sensitivity as a function of spatial frequency. Science 1980; 210: 439–40.
Merigan WH, Maunsell JHR. Macaque vision after magnocellular lateral geniculate lesions. Vis Neurosci 1990; 5: 347–52.[Web of Science][Medline]
Merigan WH, Byrne CE, Maunsell JHR. Does primate motion perception depend on the magnocellular pathway? J Neurosci 1991; 11: 3422–9.[Abstract]
Ramus F, Rosen S, Dakin SC, Day BL, Castellote JM, White S, et al. Theories of developmental dyslexia: insights from a multiple case study of dyslexic adults. Brain 2003;126: 841–65.
Ridder WH 3rd, Borsting E, Cooper M, McNeel B, Huang E. Not all dyslexics are created equal. Optom Vis Sci 1997; 74: 99–104. [CrossRef][Web of Science][Medline]
Schiller PH, Logothetis NK, Charles ER. Functions of the colour-opponent and broad-band channels of the visual system. Nature 1990; 343: 68–70.[CrossRef][Medline]
Skottun BC. The magnocellular deficit theory of dyslexia: the evidence from contrast sensitivity. Vision Res 2000; 40: 111–27.[CrossRef][Web of Science][Medline]
Talcott JB, Witton C, Hebb GS, Stoodley CJ, Westwood EA, France SJ, et al. On the relationship between dynamic visual and auditory processing and literacy skills; results from a large primary-school study. Dyslexia 2002; 8: 204–25.[CrossRef][Medline]
Wechsler D. WAIS-III – Administration and Scoring Manual. San Antonio: Psychological Corporation; 1997.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||