Change detection is impaired in children with dyslexia
Jacqueline S. Rutkowski 1,
David P. Crewther 2 and
Sheila G. Crewther 3
The severe deficits in rapid automatized naming demonstrated by children with developmental dyslexia has usually been interpreted in terms of a deficit in speed of access to the lexicon rather than as a possible deficit in speed of visual object recognition.  Yet fluent reading requires rapid visual recognition and semantic interpretation of new letters and words appearing in successive fixations of the eyes.
Thus we wondered whether change detection performance was related to reading ability. We investigated whether children with developmental dyslexia (DD) were less able to detect change in a simple display-gap-display paradigm than normal reading (NR) children of the same age and children with impaired reading and mentation (LD). In a first experimental phase, the DDs required a longer initial exposure of four letter items in order to detect change of a single letter at a level of 71% correct, compared with NRs performing at the same level. Thus the deficit in reading in DD is associated with a deficit in early processes associated with visual recognition. In a second experimental phase (using the individual target display exposures measured in the first phase), cues appeared during the 250 ms gap for a period of either 0 (no cue), 50 or 200 ms immediately prior to the presentation of the second (comparison) display. Children of all groups showed dependence on the presence of the cue to help make a judgement of change (versus no change), with the NRs least affected. When change was detected in the presence of a cue, the NRs were better able to identify the new letter than either of the other groups. However, only about 50% of the correct detections were accompanied by a correct identification. Despite published reports of a mini-neglect for left visual field in dyslexic adults, none of our groups showed such an effect. However, a significant upper visual field (UpVF) advantage in change detection performance was found across groups, which we interpret in terms of the interactions of the ventral and dorsal streams.
dyslexia magnocellular change detection posterior parietal cortex upper visual field advantage
Developmental dyslexia (DD) is an impairment in the acquisition of literacy skills despite normal intelligence, an absence of physical or psychological problems, and adequate formal education (DSM-IV, 1994), which is estimated to affect approximately 5–10% of school-aged children (Habib, 2000). Such reading and spelling problems limit career choices and professional opportunities (Snowling, 2000; Winner, von Karolyi, Malinsky, French, Seliger, Ross, et al., 2001). The neural basis or bases of developmental dyslexia is currently unknown. While the causes may be diverse, most dyslexic children demonstrate difficulty in phonological processing (Tallal, 1980) and rapid automatized naming (Denkla & Rudel, 1976).
The Neural Basis of Dyslexia: Competing Hypotheses
Competing hypotheses for the neural basis of developmental dyslexia include: deficits in the rapid temporal processing of both auditory and visual stimuli, dysfunction in the magnocellular visual pathway, cerebellar dysfunction, and abnormalities in transient attention.
The temporal processing hypothesis derives from evidence indicating that dyslexics have difficulty in rapidly processing sequential information resulting in problems with phonological decoding, and hence reading (Tallal, 1980). Rapid temporal dot counting is more difficult for children with dyslexia than is spatial dot counting (Eden, Stein, Wood, & Wood, 1995). Longer interstimulus intervals (ISIs) are also needed to make temporal order judgements vs. spatial location judgements in poor readers (May, Williams, & Dunlap, 1988). In rapid serial visual presentation (RSVP) protocols the cognitive recovery time after target recognition is some 30% longer for dyslexic versus normal adults, when stimuli are presented in quick succession, indicating that processing speed and time to disengage attention seem compromised (Hari, Valta, & Uutela, 1999). More recently, both auditory gap detection and visual double flash detection performance has been shown to be inferior in dyslexic compared with normal reading children of the same age, indicative of a general, cross-modality temporal processing deficit in dyslexia (Van Ingelghem, van Wieringen, Wouters, Vandenbussche, Onghena, & Ghesquiere, 2001).
The magnocellular hypothesis proposes an anatomical and functional abnormality in the magnocellular (M) visual pathway from retina to brain as a cause of dyslexia. During the early 1980’s Lovegrove and colleagues proposed that individuals with dyslexia have visual impairments affecting the transient visual system (Lovegrove, Bowling, Badcock, & Blackwood, 1980). The impairment was identified on the basis of deficits in the contrast thresholds for low spatial frequency achromatic stimuli. These observations, coupled with the lowered motion and motion coherence sensitivity (Cornelissen, Richardson, Mason, Fowler, & Stein, 1995; Talcott, Hansen, Assoku, & Stein, 2000) as well as reduced brain activation in V5/MT+ to moving stimuli (Demb, Boynton, & Heeger, 1998; Eden, VanMeter, Rumsey, Maisog, Woods, & Zeffiro, 1996), led to the emergence of the magnocellular deficit theory (reviewed in Habib, 2000; Stein & Walsh, 1997). Recently this interpretation has been criticized by (Skottun, 2000), who notes that little has been made of the fact that the M stream projects to both the dorsal and ventral cortical streams.
Visual evoked potential (VEP) studies have not supported pre-cortical impairment of the M-pathway in dyslexics (Johannes, Kussmaul, Munte, & Mangun, 1996; Victor, Conte, Burton, & Nass, 1993) (but see Lehmkuhle, Garzia, Turner, Hash, & Baro, 1993); nor has direct measurement of the M-pathway contribution to the multi-focal VEP (Crewther, Crewther, Klistorner, & Kiely, 1999).
The cerebellar hypothesis proposes that the failure to learn to read fluently is representative of a generalized failure of automatization and is parsimoniously explained by cerebellar dysfunction. Children with dyslexia automatize temporal skills more slowly (Nicolson & Fawcett, 1993) and show neurological signs indicative of vestibulo-cerebellar dysfunction (Fawcett & Nicolson, 1999). Neuroimaging tests also indicate that dyslexia is associated with cerebellar impairment (reviewed in Nicolson, Fawcett, & Dean, 2001).
The parietal attention hypothesis links dyslexia with a deficit in transient and spatial attention. In performing visual search tasks, dyslexics tend to show longer response times (Eskenazi & Diamond, 1983), impaired accuracy (Casco & Prunetti, 1996) and a tendency not to focus visual attention as much as normal readers (Facoetti, Paganoni, & Lorusso, 2000a). Serial search strongly activates posterior parietal cortex (PPC) (Corbetta, Shulman, Miezin, & Petersen, 1995) and search speed is slowed by transcranial magnetic stimulation to this region (Ashbridge, Walsh, & Cowey, 1997). Search performance in dyslexics correlates with motion coherence thresholds (Iles, Walsh, & Richardson, 2000), suggesting a connection between lowered search capability and magnocellular dysfunction.
There is a strong overlap between the attentional hypothesis and the magnocellular hypothesis (at least in terms of visual attention), due to the fact that the magnocellular pathway is the major visual input to the dorsal cortical stream, including parietal cortex, which is one of the major sites of activation in attention-related tasks (Corbetta, Akbudak, Conturo, Snyder, Ollinger, Drury, et al., 1998), and that magnocellular neurons are characterized by transient response characteristics.
Change Detection and Change Blindness
As reading is a spatio-temporal process, involving the sequential decoding of spatially arranged visual symbols, ability on spatio-temporal tasks such as change detection may have important implications for reading, but have yet to be examined in children with dyslexia. Tasks assessing change detection have recently emerged in the search literature in an attempt to systematically uncover the mechanisms underlying ‘change blindness’ (Rensink, O'Regan, & Clark, 1997). The nature of stimuli used in change detection experiments is wide-ranging, from simple geometrical figures to realistic dynamic scenes. However, even for simple shapes, a considerable degree of change blindness can be induced whenever there are more than a few items in the display (Rensink, 2002). Inserting a transient such as a flicker or a blank as the change is taking place, removes the salience of this target change, inducing ‘change blindness’ (O’Regan, Rensink, & Clark, 1999). Change detection rates are greatly improved when the target to be changed is cued during the blank ISI between the two pictures to be compared for change. Also, detecting the presence or absence of a change alone is less effortful than identifying the exact nature of the change (Becker, Pashler, & Anstis, 2000).
The neural correlates of change detection and change blindness have been recently identified with functional magnetic resonance imaging (fMRI) (Beck, Rees, Frith, & Lavie, 2001). Change detection activated parietal and right dorso-lateral prefrontal cortex as well as category-selective extrastriate cortex. Change detection is best distinguished from change blindness by enhanced activity bilaterally in parietal lobe and right dorsal-lateral pre-frontal cortex. The level of activation was highest in the right intraparietal sulcus (IPS) when change was consciously detected as opposed to when change was not detected (Beck et al., 2001).
Change and Memory in Dyslexia
While there have been no published reports of change detection in DD children, several studies involving dyslexic individuals have used comparison for difference between two displays, but mainly for the purpose of estimating memory performance. Thus Koenig et al (Koenig, Kosslyn, & Wolff, 1991) used visualization of remembered patterns in order to estimate spatial overlap. Dyslexic participants showed difficulty with letter forms, but as the authors point out the subjects are “integrating visual information stored in long-term memory”. Similarly Nelson and Warrington showed an impairment in dyslexia cf normal readers for verbal long-term memory functions (Nelson & Warrington, 1980). Allegretti and Puglisi used both immediate and remembered comparisons in a letter-search task, probing whether a letter in the first presentation matched any in a second presentation (Allegretti & Puglisi, 1986), again not a classical change detection task, requiring identification.
Visual Field Biases and Neglect
There is a continuing debate as to whether dyslexics show visual field (left/right) asymmetries on tasks of a visual spatial nature (Geiger & Lettvin, 1987; Klein, Berry, Briand, D'Entremont, & Farmer, 1990; Stein & Walsh, 1997). Recent evidence for a possible deficiency in right PPC functioning in dyslexia comes from findings of left inattention and right over-distractibility in recent visual flanker and reaction time tasks (Facoetti & Molteni, 2001; Facoetti & Turatto, 2000). Also, for line-motion and temporal two-dot judgements across the midline, dyslexics show a statistically significant right-sided bias (Hari, Renvall, & Tanskanen, 2001) leading to the terminology “mini-neglect” of the left visual field in poor readers.
Lower visual field (LVF) biases in normal human for reaching (Danckert & Goodale, 2001) and attentional resolution (He, Cavanagh, & Intriligator, 1996) have been related to the dorsal cortical stream and magnocellular dominance of peripersonal (LVF) space ((Previc, 1990; Previc, 1998) — reviewed in Danckert & Goodale, in press). Indeed, in the primate visual system some dorsal areas (e.g. V6A) are strongly devoted to LVF (Galletti, Fattori, Kuntz, & Gamberi, 1999). However, both visual search in normal adults (Christman & Niebauer, 1997) and change detection in normal-reading children (Rutkowski, Crewther, & Crewther, 2002) show an upper visual field (UpVF) advantage, presumably indicative of ventral pathway requirements for these tasks.
We aimed to investigate change detection performance in developmental dyslexic, learning disabled, and normal reading children, and to ascertain whether the provision of cues as an indicator of the position of likely change would be utilized to the same extent in the three groups. It was hypothesized that if there is a magnocellular-pathway/attentional dysfunction associated with dyslexia, then dyslexics would show impaired performance on a change detection task compared with normal readers. It was also suggested that the children with dyslexia would have greater difficulty utilizing brief cues, and that even when dyslexics detect change, they would identify a lower percentage of the changed items than do normal readers. In addition, dorsal pathway dysfunction should be accompanied by alterations in visual field detection biases.
86 children aged 7–16 drawn from three regions — city, suburban and rural — voluntarily participated in the current study (mean age ± standard error = 11.8 ± 0.1yr). The children were recruited from a wider subject pool involved in ongoing research into visual and attentional processes in reading and reading disorders. The Institutional Ethics Committee approved the study and informed consent was obtained from parents before testing commenced with any of the children. Children were screened for visual abnormalities and were excluded if any uncorrected binocular or refractive errors were present.
Chronological and Reading Ages for the Experimental Groups.
The reading age of some of the normal reading children had already been assessed by the Reading Progress Test (Vincent, Sadowsky, Saunders, & Reeves, 1977). All other children were administered the Neale Test of Reading Analysis (Neale, 1988). The two tests showed a high degree of overlap when correlated with a computerized measure of reading speed (“FastaReada” — coded in Authorware Professional, Macromedia) and hence reading ages were taken from either instrument without adjustment. Reading accuracy and the number of errors (mispronunciations, substitutions, refusals, additions, omissions and reversals) were quantified and compared to Australian age norms to estimate reading age. Children with a 2-year or more lag in reading for their chronological age were termed “Poor Readers”. Of these, two groups were formed: children with a mentation score within one standard deviation of the mean were defined as the Developmental Dyslexic (DD) group, while children with a mentation score below one standard deviation of the mean, formed the Learning Disability (LD) group. All children easily recognized the letters of the alphabet. Mentation scores were determined by performance on The Raven’s Coloured Progressive Matrix Test (Raven, Court, & Raven, 1990), a widely used measure of non-verbal intelligence, and which comprises 3 sets of 12 matrix puzzles of increasing difficulty. In each matrix, a segment is missing and the children are required to choose from 6 possibilities which segment best completes the pattern. Functional brain imaging demonstrates that the Raven’s Test activates many areas of the brain comprising a network of working memory areas (Prabhakaran, Smith, Desmond, Glover, & Gabrieli, 1997), with either left or right hemisphere activations dominating depending on whether the actual tasks involved analytic or figural reasoning or simple pattern matching. A group of children with normal reading skills for age (NR) was used, chronologically age-matched to the (DD) group. Preliminary data on the visual field preferences for change detection of 61 children with reading skill commensurate with age (across the range 7 – 13 years) have previously been published (Rutkowski et al., 2002)
The Change Detection task was custom programmed using Authorware 2.2 (Macromedia, Redwood City, USA), and was presented via an Apple iMac Computer with a 15 inch display monitor, running at 95 Hz screen refresh rate. The stimuli were placed at an eccentricity of 3.5° from the fixation cross, when viewing distance was 57 cm. Michelson contrast of the letters was 94%.
The task was based on that of Becker et al. (2000) which used 6 letter elements in a study of adults. We chose to use four elements in a square array (Figure 1), as our pilot studies indicated that children found the assimilation of information from 6 potential targets too difficult. The letter stimuli were sequentially drawn from the first 20 letters of the alphabet and could appear with equal probability at any of the four locations.
Experimental stimuli. In the first phase of the experiment the cue was not present. Participants viewed the first image of four letters in circular placeholders for a duration P1-Time. The stimulus was removed and replaced with a fixation cross for a period of 250 msec. Then the second image of four letters in circular place holders P2 was displayed until the participant clicked on a button to indicate Change or No Change (same/different). The P1-Time was adjusted so that participants attained 71% correct performance. In the second experimental phase, the same stimuli were used and P1-Time was adjusted to the value established for each participant in the first phase. In the second phase, a cue consisting of a short line element appeared in some trials, either 200msec or 50 msec prior to the appearance of P2.
To enable a direct comparison of change detection performance across diagnostic and age groups a staircase parameter estimation by sequential testing (PEST) procedure was used to determine the threshold duration for the first display (P1-Time) which would allow each child to detect change at approximately 71% correct. The threshold P1-Time was taken as that after 6 reversals and was used within the second phase of the experimental paradigm, that examined cueing effects and possible asymmetries in change detection performance. Change detection response was recorded by a 2-alternate-forced-choice-box (Same/Different) that appeared on the right-hand side of the screen simultaneously with P2. The second stimulus remained on the screen until a response was made. If ‘Same’ was selected, the next trial began; if ‘Different’ was selected, the stimulus array disappeared and subsidiary questions were posed. Change identification response was recorded by a 4-alternate-forced-choice-response box (“What was the new letter?”) followed by another 4-alternate-forced-choice-response box (“What was the letter before it changed?”). These data were not used in the later analysis of change identification because they were gathered using variable P1-Times.
To examine the effects of cueing of position of change on change detection performance, 48 trials, 16 of each of 3 intermixed cue conditions (Cue 200, Cue 50, No Cue), were presented in a completely randomized order which included 50% null trials (no change). Exposure time for the first stimulus (P1-Time) was fixed for each individual to the value found in Phase 1. In Change/Cue trials, Cues were presented at 200 ms or 50 ms preceding the second presentation (P2), and always pointed to the location of the item to be changed (Figure 1). In NoChange/Cue trials, location of the cue was random. Change detection and identification performance for P1 and P2 were recorded as noted for Phase 1.
Children were seated at the computer prior to the experimenter giving instructions and demonstrating the task. Emphasis was placed on the importance of accuracy of detecting change, not reaction time after the appearance of the second stimulus, and children were informed that as performance improved, the task would get faster, making it harder to see whether any changes were being made to the letters. Children were closely supervised throughout the PEST component to ensure understanding of the task and compliance with instructions. If children were responding at random, reinforcement was given and the experiment was restarted. After finishing Phase 1, children were allowed a few minutes rest, and the instructions for the second phase of the experiment were given. Emphasis again was placed on the accuracy of detecting change rather than reaction time and children were clearly informed that if a cue appeared they needed to attend only to the cued location.
Data were screened prior to the commencement of analysis for outliers and errors in data entry. Normality and homogeneity of variance tests were conducted to ensure the assumptions underlying the use of analysis of variance were met. There were no violations, so data analysis proceeded without transformations.
The duration of exposure (P1-Time) of the first display necessary for threshold detection of change between displays is shown in Figure 2 for the three groups. Analysis of variance (ANOVA) for P1-Time indicates a significant main effect for group (F (2, 83)= 8.25, P = .0005). Comparisons between groups revealed that the NRs required significantly shorter presentation times to detect change when compared with DDs (Fisher’s PSLD, P = .008) and LDs (Fisher’s PLSD, P = .0003), all groups performing at the same level of accuracy (71 % correct).
Presentation time of the first stimulus (P1-time) for which change detection performance yielded about 71% for each participant. Data is presented as means (with error bars indicating 1 SEM) for the four experimental groups. Normal readers required less initial presentation time to incorporate the identities of the letters to a level required for change detection than did the other groups.
Effect of Cue on Change Detection
For the second phase of the experiment, with P1-Time for each individual set to the value found in Phase 1, the effect of cue on change detection was investigated. Overall, despite the expectation of at least 71% correct detection, the presence of a cue did not manifestly increase the overall detection performance (Figure 3A). More strikingly, for the No Cue condition mean correct detection was 44 ± 3 %, against an expectation of 71% (a significant difference for each of the experimental groups, single sample t-test, P < .001 in each case). The provision of a cue 200 ms before the presentation of the second stimulus gave no advantage for change detection over a cue appearing only 50 ms before, for any of the groups, and thus these two cue conditions were combined for the purpose of analysis. Repeated measures ANOVA between Cue and No Cue conditions showed significantly reduced change detection performance (in the no cue condition) as illustrated in Figure 3A (F(1, 83) = 46.5, P < .0001), with a significant interaction between Group and Cue conditions (F(2, 83) = 4.7, P =.01). The DDs performed worse overall in both cued and uncued conditions, and post-hoc comparisons revealed a significant difference in the performance of the DDs and NRs (Fisher’s PLSD, P = .0057).
The effect of cue on change detection. A. Trials in which there was a letter change. Under three interleaved conditions (Cue 200, Cue 50 and NoCue), there was an overall reduction of change detection performance compared with expectation (71%). In addition, there appeared to be a strong reliance on cue trials especially for the LD and DD groups, with performance on NoCue trials close to chance for these groups. B. Trials with no change. Detection of no change (Correct Rejection) was uniformly high, around 93% for the experimental groups, whether or not there was a cue.
Performance for all groups was highly reliable under conditions when there was no change (Figure 3B) as illustrated by a mean overall correct rejection rate of 0.93 with no differences between groups.
For trials in which a change was correctly detected, children were asked to indicate the identity of the new letter (P2-ID) and that of the letter that had changed (P1-ID). Repeated measures (Cue/No Cue) ANOVA on the first identification (P1-ID) demonstrated no significant effects for Cue or for Group, but showed a significant interaction (F(2, 72) = 3.24, P = .04). A similar analysis for the second identification (P2-ID) showed significant main effects for Cue (F(2, 72) = 3.69, P = .03) and experimental group (F(1, 72) = 4.29, P = .04). As Figure 4 illustrates, post-hoc testing showed that the NRs identified letters more accurately than either of the other groups and this was significant for the cued conditions (P1-ID, Fisher’s PLSD, NR vs DD, P =.04; P2-ID, Fisher’s PLSD, NR vs DD, P < 0.005, NR vs LD, P <.02).
View larger version:
In this page In a new window
Download as PowerPoint Slide
Mean performance for experimental groups for the correct identity of the letter that changed (P1-ID) and for the letter that it changed to (P2-ID), with trials filtered for correct detection. Overall, identification was better for the second letter (most recently seen) with normal readers performing at a level above other groups.
Effect of Visual Field
Change detection performance was investigated across the four stimulus locations for the three experimental groups. Because of low trial numbers and randomized location presentation, two variables were created for each participant — an UpVF Bias (mean UpVF detection — mean LoVF detection) and RVF Bias (mean RVF detection — mean LVF detection) for both Cue and NoCue conditions, prior to analysis. A failure to correctly detect change at any one location excluded the data for that individual from further analysis, because visual field effects were calculated according to a difference equation which could not be calculated in the presence of an empty cell. Thus, only the data of 53 children were utilized in the analysis of visual field effects. A repeated measures ANOVA revealed that there was a significant main effect for visual field (F(3, 50) = 4.69, P = .004), however there was no significant main effect of experimental group.
We addressed the question of a possible left ‘minineglect’ in dyslexia (Hari et al., 2001) by testing whether the mean RVF Bias was different from zero (single value t-Test, Cued RVF Bias: Mean = −.001, t(63) = −0.03, p = .98; NoCue RVF Bias: Mean = −.05, t(63) = −0.98, p = .33). Non-significant figures indicated that there was no change detection biases to either the left or right hemifield for the cued or un-cued trials.
Similarly, on the basis of our findings of an upper visual field advantage in a group of normal reading children across a wider age range (Rutkowski et al., 2002), we tested the UpVF Bias variable and found a significant upper visual field bias for all groups in both cued and un-cued conditions (Cued condition: single value t-Test, mean = .119, t(63) = 3.36, p = 0.001; NoCue, mean = .124, t(56) = 2.63, p =.01).
This is the first time that change detection has been directly assessed in children with developmental dyslexia. In terms of procedure, the paper of Allegretti and Puglisi (1986) is perhaps closest, particularly in the immediate presentation condition. However, at no stage were their subjects asked whether a change occurred — it always did, with three letters being replaced by one, or vice versa. They were instead asked whether a letter in the first presentation matched any in a second presentation — an identity matching task, also having an element in common with visual search. Similarly, while Stanley and Hall’s early paper (Stanley & Hall, 1973) was indicative of early visual processing differences, the nature of the study was more of integration or impletion of letters than the detection of change, perhaps relating to the literature on visible persistence (Di Lollo, Hanson, & McIntyre, 1983; Lovegrove, Billing, & Slaghuis, 1978; Stanley, 1975).
Summary of Results
Our experimental results indicate that developmental dyslexia is characterized by poor change detection. Children with dyslexia require substantially longer to detect change than chronological age-matched normally reading controls.
Change identity performance was considerably worse than change detection performance for all groups of children, especially in correct identification of the letter that had changed (P1-ID), giving a clear indication that detection of change rather than identification substantially determined the threshold for P1-Time. Performance for P2-ID was probably inflated because the second display remained on screen awaiting subject response. Thus correct P2-ID only required correct spatial localization of the changed item in order to determine the identity of the new item.
Finally, all groups demonstrated an upper visual field advantage for change detection.
Poor Change Detection in DD Is Not Due to an Inability to Decode Letters
One might query the choice of letter targets for a comparison between groups, one of which exhibits reading disability. It is clear from our population data, however, with mean reading age of the DDs being 7.4 years, and from direct observation of each individual, that recognition of single letters, per se, was not a problem with the DDs. Also, the idea that problems of dyslexic children are specific to words or even letters is not supported in the literature on the rapid automatized naming test (RAN) (Denkla & Rudel, 1976). Anderson et al (1984) showed both vocalization time and pause time means were significantly longer for the dyslexics on each of the four RAN subtests. Similarly, Fawcett & Nicolson (1994) showed lower naming speed for dyslexic children compared with age and IQ matched normal readers for all stimulus categories tested (colours, digits, letters, pictures), whether or not they required grapheme-to-phoneme conversion.
Children With Dyslexia Are Less Sensitive to Change
The discovery that dyslexics are less sensitive than the normal readers to change was hypothesized on the basis of a magnocellular/parietal dysfunction. In order to perform change detection at the same level as NRs, DDs required a longer time to process the first image of the four letter targets sufficiently to detect change. This raises the question of whether time to recognition is affected in DDs or whether the deficit is in the pathways sub-serving the alerting function. fMRI evidence points to dorsal pathway (as well as dorso-lateral pre-frontal cortex) activation in change detection (Beck et al., 2001). The magnocellular input via the dorsal pathway accounts for the great majority of the visual information projecting to the PPC which appears to be necessary for alerting of visual attention prior to the conscious detection of change (Beck et al., 2001). The notion that change detection is controlled through parietal cortex receives support from the finding of longer conjunction search times under conditions of trans cranial magnetic stimulation of right parietal cortex (Ashbridge et al., 1997). Rensink has clearly drawn parallels between change detection and visual search through experiments investigating whether the mechanisms for change detection is related to the attentional processes used in search for complex static patterns (Rensink, 2000).
Developmental Dyslexics Don’t Use Cues Effectively
The data presented in Figure 3 indicate that when subjects did not know whether or not to expect a cue, their change detection performance was strongly dependent on the appearance of a cue. Of the three groups, NRs were least cue-dependent, as performance was relatively stable across cued and un-cued trials. The change detection rate was expected to be approximately 71% correct for un-cued trials on the basis that the presentation time for the first stimulus in Phase 2 was identical to the P1-Time at change detection threshold found for each subject in Phase 1. Thus, higher levels of change detection performance were expected when cues were provided. This expectation was not borne out by the data. Cued performance around 70% correct for NRs and LDs (60% for the DDs) was observed. Presuming the cue direction was accurately perceived, chance performance would be 50% correct detection for a forced choice decision of change or no change. Considered in this way, the DDs were performing close to chance while both LDs and NRs benefited significantly. Overall, NoCue performance was worse than cued (ranging from around 60% correct for NRs to a little more than 30% for DDs and LDs). We suggest that the more complex experimental structure of the Phase 2 accounts for this lower than expected performance. Multiple strategies (Cue versus NoCue) are required in the second phase experiment compared with the first . If there was a cue, a rapid shift of attention in the direction of the cue would increase the chance of successful change detection, while if there was not a cue, then attention has to be distributed over the four letters to maximize the chance of success. The presence of a cue is likely to improve performance relative to Phase 1, while the higher cognitive load due to the dual strategy is likely to lower performance overall. The especially poor performance of the DD and LD groups for NoCue trials thus suggests a strategic reliance on the likelihood of a cue appearing, and an inability to rapidly switch attention or strategy. The situation is probably exacerbated by the fact that all of the subjects would be described as “novice” in the terminology of Braun (1998), who showed that novices perform poorly compared with expert or trained observers under conditions of increased cognitive load or dual task.
The cue created a transient disturbance, capturing attention in a way that the children may have had difficulty suppressing. It is possible that the observed effect was not a problem with cue utilization per se, but an inability to adequately monitor the required location for a change . This is consistent with Hari et. al’s finding that dyslexics take significantly longer to release attention after the recognition of a target in an attentional blink paradigm (Hari et al., 1999). In addition, dyslexics can attend to and perform recognition tasks as well as normal readers if given cues of longer durations, presumably because there is time enough to disengage attention (Facoetti, Paganoni, Turatto, Marzola, & Mascetti, 2000b). Encoding stimuli draws attention, and while the letters were not necessarily encoded to the point of conscious identification (on average, only 35% of trials where change was correctly detected were both of the letters correctly identified), it was proposed that under the conditions of this experiment, letter change cannot be captured globally, but rather requires a degree of local attention. If the dyslexic children were not able to shift attention from the cue to the intended location rapidly enough they would have been able to detect whether the cued letter changed.
No Evidence of Left Mini-Neglect for Change Detection in Dyslexia
Contrary to expectation, hemifield analysis on the change detection task failed to reveal any evidence consistent with the ‘mini-neglect’ finding in adult dyslexics (Hari et al., 2001). Change detection performance was not reduced in the left hemifield relative to the right for the dyslexic children (nor for the other groups). The resolution of this apparent conflict may lie in a fundamental difference between the mechanisms underlying temporal order judgment and change detection. In the temporal order judgment task of Hari, both of the elements are detected but the one lying in left hemifield suffers a 15 ms lag compared with right hemifield in adult dyslexics. In change detection the problem is one of detection, wherein a possible timing lag may not affect detection performance.
An Upper Visual Field Detection Advantage for All But the Dyslexics
The discovery of an UpVF advantage for change detection in all groups of children is consistent with our previous findings of an UpVF bias in children reading at normal levels (Rutkowski et al., 2002). This conforms with a similar UpVF bias for complex visual search (Christman & Niebauer, 1997; Previc & Naegele, 2001). Previc originally proposed that the lower hemifield was concerned with near vision (peripersonal space) and that in a complimentary fashion the upper hemifield showed more ventral characteristics and was concerned with far vision (extrapersonal space) (Previc, 1990). We suggested (Rutkowski et al., 2002) that as the dorsal cortical stream receives greater input from LoVF and has greater attentional resolution there (Danckert & Goodale, 2001; He et al., 1996), visual masking by the simultaneous reappearance of the four letters and their place-holders may also be greater in LoVF, allowing ventral mechanisms associated with letter recognition to perform better for upper visual field presentation. Unfortunately, the fMRI study of Beck et al, while demonstrating the requirement of parietal activation for change detection, sheds no light on the question of relative activation to targets in UpVF compared with LoVF.
Change Detection and Reading
We proposed that change detection required an element of pre-conscious attention related to the coding of some attribute such as shape, but generally prior to letter identification in the reading process. To adequately detect change, one may have to, from a global perspective, quickly adapt to a local processing mode, to process the nature of the change. The dyslexics were unable to do this rapidly in the change detection task. When children learn to read, they have to decode a series of new letter images, without the benefit of much context. Following each saccadic eye movement, these images falling on the fovea, appear suddenly and with unknown identity. The more rapidly a child can process these changes, the more rapidly they are likely to both read and to perform change detection. Thus there emerges a possible rationale for the relationships found between reading ability and change detection.
 Poorer change detection performance was demonstrated by the developmental dyslexic and learning disabled populations compared with the normal reading population, with age controlled between the groups. This gives an indication that there may be a closer relationship between fluent reading and rapid visual processing, as exhibited in change detection performance, than between reading and mentation.
Do children with developmental dyslexia have an implicit learning deficit?
S Vicari1, A Finzi1, D Menghini1, L Marotta1, S Baldi1, L Petrosini2
Objective: The purpose of this study was to investigate the effects of specific types of tasks on the efficiency of implicit procedural learning in the presence of developmental dyslexia (DD).
Methods: Sixteen children with DD (mean (SD) age 11.6 (1.4) years) and 16 matched normal reader controls (mean age 11.4 (1.9) years) were administered two tests (the Serial Reaction Time test and the Mirror Drawing test) in which implicit knowledge was gradually acquired across multiple trials. Although both tests analyse implicit learning abilities, they tap different competencies. The Serial Reaction Time test requires the development of sequential learning and little (if any) procedural learning, whereas the Mirror Drawing test involves fast and repetitive processing of visuospatial stimuli but no acquisition of sequences.
Results:  The children with DD were impaired on both implicit learning tasks, suggesting that the learning deficit observed in dyslexia does not depend on the material to be learned (with or without motor sequence of response action) but on the implicit nature of the learning that characterises the tasks.
Conclusion: Individuals with DD have impaired implicit procedural learning.
A varying percentage of children have difficulty learning to read. In many of these children, the reading impairment is secondary to a global cognitive deficit, as in the case of mentally retarded individuals, such as children with Down’s1,2 or Williams’3 syndrome.  However, in most instances reading disorders are observed in children with normal intelligence and no learning difficulties linked to factors such as sensory acuity deficits, neurological impairment, or socioeconomic problems. This disorder is called developmental dyslexia (DD). Its prevalence in the school population varies across countries and languages. It is higher (4–12%) in languages characterised by non-transparent orthography, such as English, and lower (3–8%) in those characterised by strict grapheme–phoneme correspondence, such as Italian.
In spite of these epidemiological data, many researchers believe that DD is a linguistic disorder and, more specifically, the consequence of a phonological disorder. In fact, clinical evidence strongly supports this hypothesis (for review see Demonet et al5 and Vellutino et al6). Children with DD usually have great difficulty analysing and processing phonological characteristics of spoken words.7,8 Thus, for example, dyslexic children may have problems generating rhymes9 or subdividing a word into its single phonemes.10,11 The results of recent neuroimaging studies provide further support for these findings. Indeed, adults with DD show an atypical pattern of activation in the brain regions usually involved in phonological processing.12–15
Although it is generally believed that DD is based on a phonological disorder, other hypotheses have also been advanced. In particular, several researchers consider DD to be the consequence of a disorder in visual processing. Stein et al16 reported visual search difficulty caused by reduced ability to correctly control ocular movements. Furthermore, individuals with dyslexia are less sensitive than normal readers to some variables, such as contrast sensitivity and visual stimulus persistence.17 Consistent with these findings, functional neuroimaging studies in individuals with DD confirm the impairment in visual processing linked with the transient or magnocellular visual subsystem.18
Another hypothesis is that DD may be caused by deficits in visual attention.19,20 Individuals with DD have been reported to have reduced ability to find a target on a confusing background21 and to have visuospatial disorders in orienting and maintaining attention on a visual stimulus.22 Impaired information processing speed has also been reported in dyslexic individuals.23–27 According to Hari and Renvall,28 this disorder is because of difficulty in quickly shifting attention from one stimulus to the next, regardless of the sensorial modality in which the stimulus is presented. In this hypothesis, the attentional slow down is also responsible for impaired phonological representation and inaccurate visual searching.
Others suggest that DD deficits might be linked to the impaired ability to acquire and automatise new cognitive procedures. Thus, the acquisition and automatisation of competencies, such as elementary articulatory and auditory skills, eye movement processing, and letter recognition may be severely compromised during development.29,30 As a result, individuals with DD not only have difficulty reading, but also have difficulty in other functions such as gross motor coordination, balance, and speed of information processing.31
In a recent study, Vicari et al32 explored implicit learning abilities in dyslexic individuals and similar age controls by means of a modified version of the Serial Reaction Time (SRT) task, originally developed by Nissen and Bullemer,33 and demonstrated the presence of a specific implicit learning deficit in individuals with DD. However, Kelly et al34 drew contrasting conclusions when investigating implicit sequence learning in dyslexic individuals by means of a modified version of the SRT task.35 Furthermore, Waber et al36 found no evidence of an association between poor reading abilities and deficits in sequential learning in a study on a large sample of children with “heterogeneous learning problems”. These conflicting results may be due to differences in methodology. Indeed, the various studies used different SRT tasks and the ages of the individuals included were also different, as were the criteria adopted to define reading disorders. Furthermore, at least in some cases, the suspicion that explicit awareness may be involved cannot be definitively excluded.
Given the complexity of this issue, we again approached it, looking at performance of individuals with DD on different implicit learning tests. Implicit memory functions include skill learning (acquisition of general task procedure with practice), learning repeated sequences of events, habit learning (stimulus–response association), repetition priming (item-specific learning), and classical conditioning. We aimed to investigate the effects of specific types of tasks on the efficiency of implicit procedural learning in the presence of DD. For this purpose, we tested individuals with DD and matched controls using two tests (the SRT test33 and the Mirror Drawing (MD) test37–40) in which implicit knowledge was gradually acquired across multiple trials. Although both tests analyse implicit learning abilities, they tap different competencies. The SRT task requires the development of sequential learning and little (if any) procedural learning, whereas the MD involves the establishment of fast and repetitive processing of visuospatial stimuli but no acquisition of sequences. To avoid ambiguous “heterogeneous learning problems”, all individuals with DD included in the present study were clearly diagnosed as having dyslexia, based on the DSM-IV criteria.41
A total of 32 children and adolescents participated in the study. Of these, 16 children or adolescents (12 boys; mean (SD) chronological age 11.6 (1.4) years) were recruited at the Children’s Hospital Bambino Gesù in Santa Marinella, Italy, where they had been clinically diagnosed as having DD. The diagnosis was based on the standard exclusion criterion of children with normal or above normal intelligence (intelligence quotient (IQ) of 90 or more), without neurobehavioural, sensorial, or socioeconomic problems, whose reading abilities were at least two standard deviations below their chronological age. The mean (SD) IQ, measured by the Italian version of the Wechsler Intelligence Scale for Children—Revised,42 was 99.4 (5.4). Moreover to avoid any possible familiarity with implicit learning tasks by the participants, we did not include any children and adolescents in this study from among those already enrolled in our previous study.32
The control group consisted of 16 normal readers (11 boys) matched with the dyslexic group for chronological age (mean 11.4 (1.9) years) and socioeconomic level. None of the controls had difficulty reading, evidence of cognitive impairment, attention deficit, or a hyperactivity disorder.
We obtained informed consent from the children and their parents.
The Serial Reaction Time33
The SRT33 was administered on a portable computer (Compaq LTE 5280), which controlled stimulus presentation and reaction times (RTs), and stored data online. The children sat facing the screen on which a bar with four empty squares (length 3.3 cm) appeared. During the task, one of the four squares was coloured red. The children were instructed to put their left middle and index fingers on the C and V keys of the keyboard, respectively, and to put their right index and middle fingers on the B and N keys, respectively, and to press the key corresponding to the red square when it appeared on the screen. They were asked to respond as quickly and accurately as possible. When a child pressed a key the red square disappeared, and after an interval of 0.667 ms it appeared again in a new position. The position of red square changed according to a pseudorandom pattern or according to a pre-established sequence. Total randomness was limited to the extent that the red square could not appear in the same position twice in a row. Six blocks of 54 stimulus–response pairs were given. Although coloured (red) square presentation was random in blocks I and VI, in blocks II–V a nine item sequence (CVCNVBNCB) of stimuli was repeated six times in each block. The children were not told about the existence of the repeating pattern. To verify whether they had gained declarative knowledge of the sequence, at the end of the sixth block they were asked whether the red square presentation had a pattern or not. In addition, each child was invited to reproduce the sequence on the keyboard. The degree of declarative knowledge gained was evaluated by calculating the percentage of items in the ordered sequence correctly reproduced. There was no difference between groups in the percentage of items reproduced (p>0.10). We analysed the data by computing RT and response accuracy in each trial.
This was calculated as latency between stimulus appearance on the screen and key pressing, regardless of the correctness of the key pressed. Whether or not the children implicitly learned the order in which the items alternated on the screen, their RTs in the ordered sequences should have gradually decreased with respect to the first random sequence and, more importantly, should have greatly increased during the last random sequence.
This was evaluated as the percentage of correct key pressing during the single blocks of each trial.
Mirror Drawing test37
All the children who performed the SRT task subsequently performed the MD task by looking at the model, their hand, and the trace through a mirror that inverted the image. The apparatus consisted of a mirror (35×35 cm2) mounted on a 5° vertically orientated metallic support and a wooden support (40×20 cm2) that prevented the children from having a direct view of the model and of their drawing movements. The task required the children to trace a line between the double outline (0.4 cm) of a five pointed star while they looked at the star and their hand only in the mirror. The five pointed star consisted of ten segments (each 3.5 cm). A crossbar on the star perpendicular to the base of the mirror indicated the point where the drawing had to start and stop.
The children were instructed to draw the model as rapidly and accurately as possible. They had four 10 minute sessions in which they drew as many stars as they could. The first two sessions were one after the other and the third session was at least a half hour later. The final, fourth session took place on the following day to verify long term learning.
To evaluate performance, we calculated two indexes:
speed of tracing—defined as the number of star segments traced in 10 minutes (the duration of each session)
response inaccuracy—defined as the number of pathway line crossings. This index was computed for each star segment reproduced.
An increase in the number of segments reproduced throughout the four sessions and a decrease in the number of errors from session I to IV represented reliable measures of the occurrence of procedural learning.
We compared metric units of each group’s results with one or two way analysis of variance (ANOVA) with repeated measures. Where appropriate, post hoc comparisons were made with Tukey’s test.
Serial Reaction Time test
Averages of the median reaction times of the two groups in the six consecutive blocks of the SRT test are shown in fig 1. The performance of the controls was affected by the presence of an ordered item presentation sequence (one way ANOVA: F5,75 = 6.54, p<0.0001), but the response pattern of the dyslexic children was not modulated by item presentation (one way ANOVA: F5,75 = 1.67, p = 0.11). We analysed these data by means of a 2×6 (group × block). The main effect of group was not significant (F1,30 = 2.04, p = 0.16). The block effect (F5,150 = 6.4, p<0.0001) and the group × block interaction (F5,150 = 2.8, p = 0.02) were significant, demonstrating a different pattern of RT changes in the two groups across blocks.
View larger version: In a new window Download as PowerPoint Slide
Reaction times (RTs) of dyslexic (circles) and control (squares) children in a task of serial learning acquisition. *Random blocks.
Critical for the aims of the study, the two groups’ RTs differed significantly (Tukey’s test) passing from the fifth to the sixth block. This difference, usually considered the most reliable measure of visual-motor sequence learning, was in fact highly significant in the controls (p = 0.0002) but not in the dyslexic children (p = 1). It is worth noting that the group effect was not significant. In particular in the first experimental block, when the participants did not know anything about the task, the mean RTs of the two groups were quite similar (controls = 453.7 and dyslexic children = 477.4; no significant difference (p = 0.98)).
In conclusion, although the two groups did not show RT differences at the beginning of the task (block I), two different RT curves were drawn throughout the blocks. Whereas normal readers exhibited the “U shaped” learning curve usually observed in this type of task, the dyslexic children performed similarly in both randomised and ordered blocks, thus failing to exhibit a learning curve.
We also analysed response accuracy, calculated as the number of errors. A 2×2 (group × block) ANOVA showed a significant effect of the group variable (F1,30 = 5.5, p<0.03), due to the larger number of errors made by the dyslexic children (mean 7.7 (3.3)) than the controls (mean 5.2 (3.1)). The type of block (random or repeated) was not significant (F1,30 = 1.2, p = 0.4), given the similar number of errors made in ordered or random blocks. The interaction was also non-significant (F1,30 = 2.2, p = 0.2), indicating a similar distribution of errors in the two groups passing from repeated to random blocks (fig 2).
View larger version: In a new window Download as PowerPoint Slide
Errors of dyslexic (circles) and control (squares) children in a task of learning acquisition. *Random blocks.
Mirror Drawing test
The average number of elements reproduced in the four sessions by the two groups is shown in fig 3. One way ANOVAs demonstrated highly significant learning effects in both groups (controls: F3,45 = 16.2; p<0.00001; dyslexic children: F3,45 = 76.5; p<0.00001).
View larger version: In a new window Download as PowerPoint Slide
Number of elements reproduced by dyslexic children (circles) and controls (squares) in each session of the Mirror Drawing test.
We analysed these data by means of a 2×4 (group × session) ANOVA. The main effect of group was significant (F1,30 = 13.2, p = 0.001), with normal readers producing more elements than dyslexic children. The session effect was also significant (F3,90 = 37.1, p<0.00001), as was the interaction (F3,90 = 3.2, p = 0.027). Post hoc comparisons demonstrated that although dyslexic children and controls produced a similar number of elements in the first, second, and third sessions (always p>0.4), the former produced fewer elements than normal readers in the fourth session (mean 162.8 (51.5) and 301.1 (195.8), respectively, p = 0.0001).
We then analysed the number of errors. Since the participants reproduced different numbers of star elements, the children who drew more segments should have made more errors (quicker but more inaccurate children). To avoid this bias, the ratio between the number of errors and the number of segments was computed for each participant in every session. The index was analysed by means of a 2×4 (group × session) ANOVA (fig 4). The main effect of group was not significant (F1,30 = 1.6, p = 0.2), nor was the interaction (F3,90 = 0.3, p = 0.8). The session effect was highly significant (F3,90 = 79.6, p<0.00001).
View larger version: In a new window Download as PowerPoint Slide
Ratio between number of errors and number of segments in the Mirror Drawing test.
Procedural learning is a heterogeneous phenomenon including cognitive, perceptual, motor, and other skills. Previous studies have reported variable results in individuals with dyslexia on the basis of different implicitly learned tasks.32,34,36,43 One reason for this variability may be the multifaceted complexity of procedural learning, tapped differently by the demands of different tasks. Our results indicate that the children with DD were impaired on both implicit learning tasks—that is, SRT test results indicated a deficit in sequential learning in children with DD who displayed similar responses in randomised and ordered blocks. The deficit observed in the SRT task in the present study is consistent with the results obtained in a previous study, conducted in a different group of children with dyslexia, demonstrating an implicit sequential learning deficit in children with DD.32 In that study, the dyslexic children had impaired implicit learning of a visual stimuli sequence. Note that no motor sequence was demanded by the task. Conversely, the visuomotor task of the present study required implicit learning of sequential stimuli accompanied by a complex motor pattern. In this task, as well as in the previous one, only the dyslexic children showed evident deficits in the implicit knowledge of stimuli serial order. Thus, an indirect comparison between studies suggests that the learning deficit observed in dyslexic individuals does not depend on the material to be learned (with or without motor sequence of response action) but on the implicit nature of the learning characterising both tasks.
In contrast with the findings of these two studies, Kelly et al34 reported similar patterns of implicit learning in dyslexic and normally reading university students, although the dyslexic students had slower RTs than normal readers. However, the dyslexic individuals included in this study were university students and thus rather successful academically. It is reasonable to hypothesise some degree of recovery of the cognitive processes involved in reading in these young adults. In fact, the impact of dyslexia can be modified by the availability of resources such as semantic knowledge,44 use of context,45 visual memory,46 and verbal ability,47 which can compensate for phonological deficits. In fact, neuroimaging findings have demonstrated that a large number of ancillary systems representing the neural correlates of these compensatory processes are present in adulthood.48 Thus, it is likely that cognitive processing in children is different from that of dyslexic adults.
The SRT task adopted in a study of a large sample of children with “heterogeneous learning problems”36 showed no direct association between poor reading abilities and defective sequential learning. However, since the criteria for reading disorders were extremely loose, the study may have included not only individuals with dyslexia but also people affected by completely different learning disabilities.
In the MD test, although the dyslexic and normal readers made a similar number of errors (see fig 4), the children with DD were always slower than normal readers in terms of number of segments reproduced. Even if both the groups improved as the sessions went by, we observed significant differences between the groups in the fourth session (see fig 3), which took place 24 hours after the preceding one. Behavioural studies on monkeys and humans have shown that procedural learning consists of at least two stages: an early stage and a late stage.49 Changes in learned behaviour, depending on the learning stages, and differential contribution of multiple brain areas and specific brain states, such as sleep,36 to the early and late stages have been demonstrated.50,51 In fact, in the initial stage of learning, performance asymptotically improves, and when the training has ended, the initially formed memory trace continues to be processed. Consequently, when tested later, performance is markedly improved even without any intervening training session. This late component of learning seems to depend critically on sleep. The observation that in dyslexic individuals the influence of sleep on the late component of procedural learning is much less robust than in controls suggests different processes of consolidation of procedural abilities through off-line practice during the night. At the system level, during sleep there is an experience-dependent reactivation of cortical areas reflecting the reprocessing of elaborated information contained in the learned material.52,53 This cortical reactivation is proportional to the level of performance achieved at the end of the training session.54 By interpreting our data in line with these studies, it can be hypothesised that in the dyslexic children the experience based cortical reactivation during post-training sleep was in some way impaired. Furthermore, it cannot be excluded that the level of implicit learning achieved prior to sleep by the dyslexic children was insufficient to modulate the cortical activation during sleep.
The results of the SRT and MD tests provide evidence of reduced procedural abilities in the dyslexic children, suggesting a general deficiency of implicit learning in DD. In this regard, the reported difficulty in dyslexia to process literacy as fast as in normal readers might be considered one facet of the more general impairment in implicitly learned procedures. Actually, implicit learning abilities allow acquiring and executing new motor, perceptual, and cognitive skills, and presumably also influence reading processes, thus leading to automatisation of the mechanisms reading is based on. Further, in DD automatisation deficits may influence phonological processing and interfere with the ability to automatise elementary articulatory and auditory skills.30,31,43,55,56 In the early stages of development, implicit learning deficits may affect the maturation of successive abilities. This may explain the scattered difficulties exhibited by individuals with DD such as phonological failure, visual processing inadequacy, or attentional deficits.
We are aware that any attempt to identify the brain structures specifically involved in the implicit learning impairment displayed by the children with DD would be entirely speculative. However, it is worth noting that functional neuroimaging studies have demonstrated cortical57–59 as well as cerebellar60,61 and striatal62 activation during implicit sequential learning,62–65 suggesting that all these structures have a role in the implicit acquisition of sequential information. The involvement of neuronal loops comprising the basal ganglia and cerebellum,49 the former for reward based evaluation and the latter for timing processing, has been proposed. It has been reported that the anterior striatum and the putamen are related to the acquisition of sequential learning and memory storage, respectively.66 The restructuring of neural response patterns of striatal neurones occurs as a result of procedural learning, culminating in task related activity emphasising the beginning and the end of the automatised procedure.67 Experimental data support the involvement of cerebellar circuits in the acquisition of spatial procedural competencies.68–70
Interestingly, the changes in activity in both the striatum and the cerebellum have been observed at different stages of motor sequential learning. The cerebellum has been considered as the structure that detects and corrects the errors71 mostly made in the initial stages to adjust movement to incoming sensory input and to produce accurate motor output.72,73 Conversely, striatal activation increases with practice reaching its maximum once learning is achieved.74 In this framework, striatal regions are critical for the long term storage of well learned movements.74,75
According to the clinical, experimental, and behavioural findings reported above, the procedural learning difficulties observed in individuals with dyslexia suggest that cerebellar and striatal activities are impaired. Rae et al76 demonstrated biochemical and morphological abnormalities in the cerebellar areas of dyslexic adults. In a positron emission tomography study on dyslexic adults, Nicolson et al30 described abnormal cerebellar activation in response to both learned and novel motor sequential tasks. In a functional magnetic resonance imaging study, Georgiewa et al77 documented a lower level of putamen activation in dyslexic children than in controls, hypothesising greater familiarity of the latter with reading performances and reduced retrieval of implicit knowledge in the former. In the light of the present findings, the possible role of striatal and cerebellar areas in the process of reading acquisition is intriguing and worthy of further investigation.
Finally, the present data suggest that evaluation of procedural abilities is a useful clinical approach for studying this developmental disorder. In fact, since procedural learning is acquired at the early stages of development,78 a deficit in the acquisition of procedural competencies may be a sign of future reading difficulty in pre-school children.