Elucidating the component processes involved in dyslexic and non-dyslexic reading fluency: An eye-tracking study

Abstract

The relationship between rapid automatized naming (RAN) and reading fluency is well documented (see Wolf, M. & Bowers, P.G. (1999). The double-deficit hypothesis for the developmental dyslexias. Journal of Educational Psychology, 91(3), 415–438, for a review), but little is known about which component processes are important in RAN, and why developmental dyslexics show longer latencies on these tasks. Researchers disagree as to whether these delays are caused by impaired phonological processing or whether extra-phonological processes also play a role (e.g., Clarke, P., Hulme, C., & Snowling, M. (2005). Individual differences in RAN and reading: A response timing analysis. Journal of Research in Reading, 28(2), 73–86; Wolf, M., Bowers, P.G., & Biddle, K. (2000). Naming-speed processes, timing, and reading: A conceptual review. Journal of learning disabilities, 33(4), 387–407). We conducted an eye-tracking study that manipulated phonological and visual information (as representative of extra-phonological processes) in RAN. Results from linear mixed (LME) effects analyses showed that both phonological and visual processes influence naming-speed for both dyslexic and non-dyslexic groups, but the influence on dyslexic readers is greater. Moreover, dyslexic readers’ difficulties in these domains primarily emerge in a measure that explicitly includes the production phase of naming. This study elucidates processes underpinning RAN performance in non-dyslexic readers and pinpoints areas of difficulty for dyslexic readers. We discuss these findings with reference to phonological and extra-phonological hypotheses of naming-speed deficits.

Introduction

Although reading represents one of the more complex acquired skills of the human brain, when mastered it is normally executed with considerable ease and fluency. For some individuals, however, effortless, fluent reading is never fully realized, despite normal IQ and no obvious socioeconomic factors that would impede reading progress. This cognitive profile is known as developmental dyslexia. In this paper we use eye-tracking methodology to investigate the role of extra-phonological factors in reading fluency, and whether dyslexic readers display extra-phonological deficits, which are proposed to underpin naming-speed deficits in dyslexia (Wolf & Bowers, 1999). In the following section we outline competing causal accounts of dyslexia.

There is broad agreement concerning the language-related behavioral characteristics of dyslexia. Dyslexic readers are generally agreed to have reading and spelling problems, difficulty learning rhymes, and often omit and reverse the orientation of letters when writing. They are also generally slower to read and write (see Snowling, 2000). A considerable controversy surrounds the neurological and cognitive cause of dyslexia, however. One area of disagreement concerns whether dyslexia should be classified as a purely phonological deficit, or whether attentional and perceptual processes (visual and auditory) are also causally implicated.

Under one influential account, the phonological deficit hypothesis, dyslexia is caused by a difficulty in consolidating and accessing phonological representations (Elbro, 1996, Hulme and Snowling, 1992, Snowling, 2000). This account therefore identifies a causal link between the cognitive impairment (phonological processing) and the behavioral outcome (difficulty learning to read) (Ramus, 2003): Dyslexia is caused by a language-specific deficit within the phonological system that arises from difficulty in processing the speech stream (e.g., Bradley and Bryant, 1978, Frith, 1985, Stanovich, 1988). Consistent with this account, extensive research has demonstrated that dyslexic children are impaired compared to normally developing readers on tasks involving non-word repetition (e.g., Elbro et al., 1998, Snowling et al., 1986), phonological learning (Aguiar and Brady, 1991, Wimmer et al., 1998) phonemic awareness (e.g. Bradley and Bryant, 1978, Griffiths and Snowling, 2002, Morris et al., 1998), picture naming (Snowling, van Wagtendonk, & Stafford, 1988) and verbal short term memory (e.g., Griffiths and Snowling, 2002, Nelson and Warrington, 1980).

This approach can be contrasted with theories of dyslexia that view phonological deficits as one proximal cause or symptom of an independent, primary causal impairment. Proponents of such approaches argue that existing research has not demonstrated unequivocally that difficulties in reading arise from impaired phonological processes per se (e.g., Castles & Coltheart, 2004). Although these approaches are united in questioning the centrality of phonological deficits, they differ in their characterization of the primary causal impairment. Some approaches emphasize potential cognitive impairments in dyslexic readers, such as reduced working memory capacity (see Ramus & Szenkovits, 2008). Other theories identify a low-level, neurological cause to account for the broad and multi-modal deficits characteristic of dyslexia. The cerebellar deficit hypothesis (see Nicolson, Fawcett, & Dean, 2001), for example, suggests that a range of deficits such as impaired information processing speed, memory, motor skill and balance in addition to phonology can be explained in terms of impaired cerebellar activity that causes multi-modal attentional deficits at the cognitive level. Although it is contentious whether the cerebellum is the precise neurological locus of dyslexic readers’ difficulty (e.g., Ivry & Justus, 2001), there is extensive evidence that dyslexic readers experience difficulties in processing multiple stimuli across auditory, tactile and visual modalities (e.g., Hawelka and Wimmer, 2005, Tallal, 1980; see Wolf & Bowers, 1999 for review).

An alternative sensorimotor theory, the magnocellular deficit hypothesis, typically ascribes reading difficulties to irregularities in magnocellular neurons projecting visual information to the primary visual cortex (Livingstone, Rosen, Drislane, & Galaburda, 1991). It is thought that some dyslexic readers have impaired functioning of magno cells in the lateral geniculate nucleus that are responsible for detecting transient movement information in the visual field (Galaburda & Livingstone, 1993). Impaired magnocell function may cause unstable fixations and concomitant difficulty in discriminating and processing orthographic information (e.g., Stein and Talcott, 1999, Stein and Walsh, 1997). It could also adversely affect parafoveal processing of upcoming stimuli (Dacey, 1994) and processing of multiply presented visual items (Omtzigt, Hendriks, & Kolk, 2002).

Despite numerous demonstrations of impaired visual attention in dyslexia, magnocellular impairment is not always concurrent (e.g., Roach & Hogben, 2004; see Ramus (2003) and Skottun (2005) for reviews), and the precise locus of a visual attentional impairment is therefore open to debate (see e.g., Pammer & Vidyasagar, 2005). However, recent studies have identified specific conditions under which dyslexic readers experience visual processing difficulties. Hawelka and Wimmer (2005) found higher recognition thresholds for dyslexic readers compared with non-dyslexic readers on 4- and 6- digit arrays, but not on 2-digit arrays. This finding suggests that dyslexic readers do not have a significant difficulty recognizing a letter in isolation, but that the presence of other stimuli in the array does significantly impair their ability to identify a target item.

Similar findings have also been found in crowding: a ubiquitous visual phenomenon, defined as the “deleterious influence of nearby contours on visual discrimination” (Levi, 2008). Crowding is reliably correlated with reading speed (Levi et al., 2007, Pelli et al., 2007), and consistently discriminates non-dyslexic from dyslexic reading groups: Both dyslexic and non-dyslexic readers are impaired in identifying a target item when it is flanked by other items, compared to when it is presented in isolation, but dyslexic readers show particularly poor discrimination levels (Bouma and Leigen, 1977, Pernet et al., 2006). Pernet et al. argued that dyslexic readers’ increased difficulty in identifying a target in these contexts may reflect a parafoveal processing difficulty: Information from the parafoveally presented target item impairs identification of the foveally presented target. This conclusion converges with findings by Geiger and colleagues, who found that (child and adult) dyslexic readers are more accurate at identifying letters in the periphery than non-dyslexic readers (e.g., Geiger and Lettvin, 1987, Lorusso et al., 2004). Geiger, Lettvin, and Zegarra-Moran (1992) suggested that dyslexic readers do not effectively suppress information from the parafovea and periphery, leading to confusion and unclear perception of the target item. Dyslexic readers might therefore find it particularly difficult to identify target items when they are flanked by other items in the visual field. Further, results from crowding studies suggest that interference could occur from information presented either to the left or right of the target (e.g., Pernet et al., 2006), and is more likely to occur when flanking items are similar in shape and size to the target (Kooi et al., 1994, Nazir, 1992).

Visual processing difficulties may explain some of the difficulties that dyslexic readers experience with multiple-item presentations, but related effects in tasks requiring explicit naming of a target that is flanked by other items may also reflect dyslexic readers’ difficulty in binding the visual stimulus to a phonological representation for production. For example, recent evidence suggests that dyslexic readers are impaired at identifying a target from a number of visually presented distractors only when the task requires an articulatory response (e.g., Hawelka and Wimmer, 2005, Hawelka and Wimmer, 2008) suggesting that retrieval of the appropriate label from a selection of competitors is a crucial component of the difficulty elicited by multiply presented items. If this is correct, we would predict that dyslexic readers would be particularly impaired at identifying a target from among other items in the array when the task involves articulation of the target’s name. Whilst dyslexic readers may experience difficulty on perceptual aspects of text processing (e.g., recognition of letters and words), therefore, the evidence suggests that retrieving orthographic and phonological codes for production is also impaired.

In sum, researchers continue to debate whether dyslexia is caused by a purely phonological deficit or whether alternative theories are better able to explain the range of difficulties characteristic of the reading disability. A significant number of studies suggest that attentional, perceptual and retrieval processes are impaired in dyslexia. An interesting and consistent feature of these findings is that presentation of multiple items often leads to dyslexic readers’ significantly decreased performance levels. It is not currently clear, however, how attentional and perceptual processes relate to reading ability, and how impairment in these extra-phonological processes may translate to a reading impairment in dyslexia. In this paper we demonstrate how this issue can be examined using detailed measurement of online events in the rapid automatized naming task.

Wolf and Bowers (1999) argued that RAN is a cognitive task that provides a window on the low-level processes involved in reading. RAN consists of a visually presented array of high frequency items (for example, letters, digits, colours or common objects) repeated multiple times in a randomized order across several rows (in standard RAN studies, 5 items are repeated 10 times across 5 rows). The participant’s task is to name each of these target stimuli from left to right across and down the page as quickly as possible.

There is extensive evidence that children and adults with dyslexia are significantly slower than their peers when performing RAN (e.g., Denckla and Rudel, 1976a, Denckla and Rudel, 1976b; see Wolf & Bowers, 1999, for a review), but there is considerable controversy concerning which processing requirements engender this group difference and which components of RAN are critical in determining performance levels. RAN is traditionally considered an index of phonological retrieval, or “retrieval of codes from a long-term store” (Wagner, Torgesen, Laughon, Simmons, & Rashotte, 1993). By the same argument, dyslexia is considered to be a difficulty in retrieving phonological codes. Moreover, impaired phonological retrieval is proposed to limit the resources available to effectively inhibit stimuli that have already been named so that subsequent stimuli can be processed (Clarke, Hulme, & Snowling, 2005).

However, there is substantial evidence that RAN predicts reading ability independently of phonological skill (Bowers, 1993, Powell et al., 2007, Savage et al., 2007). Specifically, RAN uniquely predicts expressiveness, reading efficiency and reading speed of text (Young & Bowers, 1995) in addition to predicting later reading fluency (Manis et al., 2000, Wolf and Obregón, 1992). It also remains an independent predictor of reading, even when prior reading ability has been partialled out (Parilla, Kirby, & McQuarrie, 2004). Bowers and Swanson (1991) further demonstrated that continuous versions of RAN, in which items are presented simultaneously in a grid format, are better predictors of reading fluency than discrete formats, in which items are presented and named individually. This finding suggests that processes other than phonological retrieval of each item name influences naming-speed performance.

Wolf and Bowers (1999) argued that RAN is a reliable predictor of reading fluency, which is largely independent of phonological ability. Under this account, phonological retrieval is a crucial component of the naming process but performance on RAN is also influenced by a large number of extra-phonological processes, including attention to the letter stimulus; bi-hemispheric visual processes responsible for feature detection; matching of feature and pattern encoding to stored orthographic representations; integration of visual information with phonological information; and motor activation leading to articulation. Because RAN is typically measured by the total time it takes to name the items in a whole trial, these processes must also be activated and synchronized quickly and accurately.

RAN is therefore a deceptively complex task, requiring rapid visual recognition and matching to orthographic and phonological representations. Wolf and Bowers (1999) argued that precise timing across operations is the critical factor underlying successful performance on RAN. In normal reading, for example, repeated execution of these processes in response to print leads to highly practised (automatized) identification and naming of the letter. They therefore proposed that slower naming-speeds characteristic of dyslexia reflects a speed of processing deficit in one or more of these processes, leading to impaired automaticity in naming. Impaired speed of processing might be restricted to one domain, such as vision. This would be consistent with the evidence reviewed above for a visual processing impairment in dyslexia (e.g., Geiger et al., 1992). A visual deficit would impair discrimination of visually presented stimuli with concomitant effects on identifying patterns with clusters of frequently co-occurring letters. It would also lead to difficulty in binding the visual stimulus to information from other domains, such as phonology (Stein & Walsh, 1997). Alternatively, perceptual processing per se might be intact, but timing mechanisms responsible for synchronizing information from different domains, such as visual and phonological processes, might be impaired (Breznitz, 2003, Wolf and Bowers, 1999).

Surprisingly little is currently known about the nature of the processes that underpin task performance on the RAN (Georgiou, Parrila, & Kirby, 2006). This is largely due to the methodology in which RAN is used: Response times in a given trial are typically summed, and averages across trials are then correlated with reading measures. Although correlations can associate components of naming with naming-speed performance, such studies cannot in isolation elucidate any direct influence of these components on RAN naming times. Indeed, Wolf, Bowers, and Biddle (2000) argued that “The more precise specification of what constitutes the underlying components of naming-speed and how disruption of each component in naming might affect specific reading skills is one of the most important future directions in this line of research” (p. 400).

In a bid to understand the processes underlying RAN, some researchers have sought to distinguish articulation times of the stimulus name from pause times between articulation of names. Obregón (1994) and Neuhaus, Foorman, Francis, and Carlson (2001) found that whilst pause times were longer for dyslexic readers than for non-dyslexic readers, articulation times were not. Obregón and Wolf and Bowers (1999) suggested that longer inter-stimulus pauses for dyslexic readers may reflect impaired ability to disengage with an already-named stimulus in order to focus on processing the current stimulus. Similarly, Hari and Renvall (2001) proposed that dyslexic readers exhibit ‘Sluggish Attentional Shifting’ (SAS) that impairs their ability to disengage from processing one item in order to engage with another item. Such explanations therefore associate dyslexic readers’ longer pause times in RAN with their independently observed difficulties in processing multiply presented items. This conclusion is currently speculative, however: Segmentation of the speech stream alone does not directly pinpoint the source of difficulty to items surrounding a target item. Neither can we tell at which point the pause times reflect disengagement from a previous stimulus versus processing of an upcoming stimulus in the RAN array.

Other research has focused on separating processes associated with phonological retrieval of the letter name from other processes required for the task. Manipulating the visual versus phonological demands of the task, for example, provides a means of examining whether processes other than phonology (e.g., visual processing) influence the naming process. Thus far, a relatively small number of studies have attempted to experimentally manipulate visual and phonological information in the context of RAN. Compton (2003) compared dyslexic and non-dyslexic children on RAN letter sets in which the target letters were visually or phonologically similar, or visually and phonologically similar, and hence potentially confusable. Note that confusability with respect to a particular processing domain (e.g., phonology) is informative about which domains influence naming-speed. He found that although phonologically confusable sets were better predictors of later reading speed for text, only letter sets containing visually confusable items impaired dyslexic readers compared with non-dyslexic readers. Such findings suggest a role for visual processing in the naming-speed difficulty characteristic of dyslexia, but the absence of a baseline condition renders these findings somewhat less reliable. It is possible, for example, that the findings were an artefact of the letters used in the study, rather than any specifically visual or phonological effects. Wile and Borowsky (2004) also showed that whole word reading or phonological awareness skills predict RAN performance depending on whether the task emphasizes retrieval of letter names versus retrieval of the letter’s phonological features.

The studies outlined above are helpful for beginning to unpick the processes underpinning RAN, but it is difficult to make further progress in this area using offline methodology. Reaction times that are restricted to summed, whole-trial times or even pause times in the speech stream are arguably too coarse a measure for parsing the intricately linked visual and phonological processes involved in RAN. Recent work has shown that more fine-grained online measures can yield new insights into these processes. Breznitz and colleagues provided EEG (electroencephalography) evidence that dyslexic readers have a speed of processing deficit that causes asynchronous processing of visual and auditory information, resulting in impaired temporal binding of information across modalities (Breznitz and Misra, 2003, Meyler and Breznitz, 2005; see also Breznitz, 2006): Even highly compensated (university educated) dyslexic adults, for example, showed processing delays between orthographic and phonological processes. Such an impaired temporal synchrony of visual and phonological information is a potential cause of slowed naming times in the RAN-letters task.

RAN measures are therefore informative about the processes that contribute to reading fluency, and they consistently discriminate dyslexic from non-dyslexic readers. However, very little work has successfully elucidated the nature of the processes that underlie rapid naming. Clearly, it is important to understand the components of naming-speed; but perhaps more importantly, addressing this question also casts light on the more fundamental question of whether dyslexic readers’ impairment is restricted to phonological processing, or whether they also have impaired extra-phonological (attentional and/or perceptual) processes.

We now report a study that investigated which component processes of RAN influence task performance (as an index of reading fluency), and which of these processes, when impaired, contribute(s) to slower naming-speeds characteristic of dyslexia. Specifically, we compared the influence of phonological and visual processing domains on both non-dyslexic and dyslexic readers’ RAN performance; we chose visual processing as representative of an extra-phonological processing requirement in RAN. We compared groups of age-matched, high-functioning adult dyslexic and non-dyslexic readers on versions of the RAN-letters task in which we manipulated characteristics of the array. Our groups were selected in order to provide a conservative indication of the impaired processes underlying reading dysfluency. Further, using a high-functioning (university undergraduate) adult dyslexic group minimized the risk of including participants with co-morbid reading and other difficulties. We chose the RAN-letters task because of its enduring relationship with reading ability into adulthood (e.g., Bowers, 1993, Wolf et al., 1994).

Our study is novel in using eye-tracking methodology to investigate the processes underpinning naming-speed. We recorded an online measure of production times for individual items within the RAN array. The advantage of measuring eye-movements in RAN in addition to the speech stream is that the eye movement data indicates at which point the eye lands on a specific item in the array. From this point, it is possible to measure the delay until the participant articulates the item’s name. We can therefore obtain an online indication of the participant’s production time for a particular item (see Griffin and Bock, 2000, Morgan and Meyer, 2005, for a similar approach to naming objects from a language production perspective). Our study also contrasts with most previous RAN studies because it directly compared dyslexic and non-dyslexic readers’ performance on experimentally manipulated versions of the RAN rather than correlating performance on RAN with other cognitive or component reading measures.

Our experiment manipulated whether the RAN array was either phonologically or visually difficult to process. Our phonological manipulation was to present ten pairs of letters in RAN that were either ‘confusable’ (similar) or ‘non-confusable’ (dissimilar), such as b and v (identical rimes) or k and q (identical onsets); note that both rime and onset information have been hypothesized to be important in reading development (e.g., Goswami and Bryant, 1990, Hulme et al., 2002). Our visual manipulation was to present ten pairs of letters in RAN that were either visually ‘confusable’ (similar) or visually ‘non-confusable’ (dissimilar). Confusable letters in the visual set were mirror images of one another on the vertical axis: (e.g., p and q). These visually similar letters are often confused by dyslexic readers (e.g., Terepocki, Kruk, & Willows, 2002). Non-confusable equivalents for phonological letter sets comprised identical letters, but in a different order (in which ‘confusable’ items were not adjacent). Non-confusable equivalents for the visual letter set comprised letters in an identical sequence (to control phonological output), but the letters were in upper case (e.g., P and Q) to avoid visual similarity. In each case, given a target letter in position n, we were concerned with the effects of presenting a confusable letter in the position that preceded the target letter (an already-named item in position n − 1) versus the position that succeeded the target letter (an upcoming item in position n + 1).

In RAN, already-named items need to be rapidly suppressed following processing in order to engage with the next item (Wolf & Bowers, 1999). This requires (a) disengagement from previous stimuli, (b) moving attention to the next stimulus, and (c) spotlighting the new area of text (Wolf, 2007). This orienting network of attention implicates the parietal lobes, superior colliculi, and thalamic regions, respectively (e.g., Fisher, 2006, Pasik et al., 1966, Vivas et al., 2006), and dyslexic readers’ are proposed to execute these processes within prolonged time windows (Hari & Renvall, 2001). Visual and phonological processing of the text is then supervised by the executive attentional system (Wolf, 2007), which is also proposed to be impaired in dyslexia (e.g., Smith-Spark & Fisk, 2007). Thus, a confusable item in the previous position may delay dyslexic readers’ orienting attentional network or impede the executive system’s ability to process the target. Whilst previous items involve suppression of information, processing of upcoming items involves partial activation of information, even in the parafovea, which may be beneficial to reading (or naming) times (e.g., Sereno & Rayner, 2000). However, as we have discussed, extra-foveal information can also impair target detection for dyslexic readers (e.g. Geiger et al., 1992).

Our experiment therefore investigates the role of multi-item processing in RAN, and by extension reading fluency; in particular, it examines whether the necessity of rapidly processing multiple items in RAN contributes to slower naming-speeds characteristic of dyslexia. The phonological and visual manipulations discriminate which processing demands imposed by items that either precede or succeed the target item influence RAN performance for fluent (non-dyslexic) readers, and which are crucially involved in dyslexic readers’ naming-speed difficulty.

We use several dependent measures in this experiment. A whole-trial naming time measure includes the time taken to name items in a whole trial and is therefore comparable to previous RAN studies (Denckla and Rudel, 1976a, Denckla and Rudel, 1976b). Of greater interest in our study are two measures that distinguish (at the individual letter level) how visual and phonological information might affect the initial uptake of information from the way in which they might affect overall response time (i.e., uptake of information and time taken to initiate the articulatory response). Hence we first measure processing time, or the time during which the participant’s eyes fixate a target letter. Previous research suggests that we can assume that overt attentional resources are allocated to the stimulus during this fixation (Just & Carpenter, 1980); moreover, longer fixations are associated with cognitive difficulty (Inhoff & Rayner, 1986). We further distinguish between the current processing time, which measures how long the participant’s eyes fixate the target item before saccading to the next item, and total processing time, which measures how long in total the participant’s eyes fixate the target item (including any regressions). This allows us to examine whether uptake of visual and/or phonological information influences RAN speed parafoveally (current processing time) or whether effects only emerge in a measure including parafoveal and foveal time (total processing time). We also measure the time from when the participant’s eyes first fixate on the target item to the time when they initiate their articulatory response; we term this the “eye-voice span”. This measure therefore explicitly includes production time, and so allows us to examine whether effects of confusability are associated in some way with the requirement to produce an articulatory response.

Previous research has shown that reaction times on RAN underpin its relationship to reading ability (e.g., Denckla and Rudel, 1976a, Denckla and Rudel, 1976b). The current study therefore investigates the effect of increased processing load in phonological and visual domains on RAN speed. If confusable items that tap a particular domain (e.g., vision) lead to longer processing times than non-confusable items, this implies a link between the processing domain in which the effect was observed, and RAN processing times. Further, it is well established that dyslexic readers yield generally longer RAN times compared with non-dyslexic readers (Denckla and Rudel, 1976a, Denckla and Rudel, 1976b). If a processing domain is involved in the RAN difficulty characteristic of dyslexia, then confusability in that domain should affect (slow down) dyslexic readers’ processing/naming-speed to a greater extent compared with the non-dyslexic group. Thus, with respect to the dyslexic group, we measure whether the confusable letter pairs in visual and/or phonological processing domains extend dyslexic readers’ latencies to a greater extent compared to their own performance on non-confusable conditions, and to a greater extent than confusable letter pairs in visual and/or phonological processing domains extend non-dyslexic readers’ latencies. For the predictions, we will summarize this effect as impairment.

We therefore expect that if phonological and/or visual information influences non-dyslexic readers’ whole-trial naming times, for example, they will show longer latencies when trials are phonologically and/or visually confusable than when they are non-confusable. If phonological and/or visual information influences dyslexic readers’ whole-trial naming times to a greater extent than it influences the non-dyslexic group, we expect impaired performance in the dyslexic group on this measure.

We now outline specific predictions for current and total processing times and the eye-voice span measure, which are of primary interest in this experiment. We begin with the non-dyslexic group. We predicted that in both the processing time and eye-voice span measures, the non-dyslexic group would be able to effectively suppress visual and phonological information from items immediately preceding the target item. Wolf and Bowers (1999) suggested that rapid disengagement of attention from an already-named item in order to allocate resources to the next (target) item is an important component of RAN; there is no benefit to continued activation of information associated with an already-named item.

For upcoming items (immediately succeeding the target), however, we expected that in the processing time measure, non-dyslexic readers might be sensitive to visual and phonological information, and that this might well occur in the parafovea (current processing time measure). Unlike activation of an already-named item, partial activation of an upcoming item may be beneficial to RAN performance. When we read text, visual and phonological properties of an upcoming word are processed in the parafovea (e.g., Sereno & Rayner, 2000). Picture-naming studies also show that participants retrieve visual and phonological information from the parafovea (Morgan & Meyer, 2005). Hence processing of a target and an upcoming item overlap to some extent. We therefore expected that confusable (versus non-confusable) information in the item succeeding the target would lead to longer processing times on the target for the non-dyslexic group. Morgan and Meyer also found that information from an upcoming item modulated naming times. Hence, we expect a similar pattern of results in the eye-voice span measure.

We now turn to predictions for the dyslexic group. For this group, we make explicit predictions for visual uptake of the target (processing times) versus target production times (eye-voice span time) in visual and phonological domains. We expected that if dyslexic readers experience difficulty in visual uptake of the letters, effects would emerge in the processing time measures. Specifically, if dyslexic readers have a deficit in processing visual aspects of letters (e.g., Stein & Walsh, 1997), they should show impaired (significantly longer) total processing times in response to visually confusable stimuli. Moreover, this perceptual interference would occur from items preceding and succeeding the target (Kooi et al., 1994, Pernet et al., 2006). If this effect also occurs in extra-foveal regions (here, the parafovea) (Geiger et al., 1992), it should also occur in the current processing time measure.

If dyslexic readers are impaired by uptake of phonological information, we would expect a similar pattern of results on phonological (rime and/or onset) letter sets (e.g., Clarke et al., 2005) as outlined above for visual letter sets: The time taken to recognize a phonological item should be impaired in the presence of an item with similar phonology in the previous position. By the same argument, activation of an upcoming item with similar phonology might interfere with visual uptake of the target. Again, if this effect occurs in the current processing time measure, we have evidence to suggest that this difficulty occurs in parafoveal processing.

Recent evidence also suggests that dyslexic readers are only impaired in multi-item identification when the task involves a naming component (Hawelka and Wimmer, 2005, Hawelka and Wimmer, 2008). If it is the case that retrieval of item names is problematic, we would expect the dyslexic group to be significantly impaired on confusable conditions in the measure that includes articulation, the eye-voice span measure. Specifically, if failure to suppress visual information or a specific orthographic code from an already-named item leads to delayed retrieval of appropriate (orthographic or phonological) codes, we would expect delayed (impaired) articulation. If competition from an upcoming visually confusing letter leads to confusion in retrieving the appropriate orthographic code for the target, we would also expect impaired (delayed) articulation. Similarly, if longer naming speeds arise from impaired retrieval of phonological codes (e.g., Wagner et al., 1993), we would expect delayed articulation in these letter sets. Failure to suppress phonological information from a previously named item or competition from a similar upcoming phoneme might significantly delay articulation time in the eye-voice span measure.