This article was downloaded by: [University of Massachusetts, Amherst] On: 29 September 2014, At: 07:59 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK The Quarterly Journal of Experimental Psychology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/pqje20 Lexical quality and eye movements: Individual differences in the perceptual span of skilled adult readers a Aaron Veldre & Sally Andrews a a School of Psychology, University of Sydney, Sydney, NSW, Australia Published online: 25 Aug 2013. To cite this article: Aaron Veldre & Sally Andrews (2014) Lexical quality and eye movements: Individual differences in the perceptual span of skilled adult readers, The Quarterly Journal of Experimental Psychology, 67:4, 703-727, DOI: 10.1080/17470218.2013.826258 To link to this article: http://dx.doi.org/10.1080/17470218.2013.826258 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014 Vol. 67, No. 4, 703–727, http://dx.doi.org/10.1080/17470218.2013.826258 Lexical quality and eye movements: Individual differences in the perceptual span of skilled adult readers Aaron Veldre and Sally Andrews Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 School of Psychology, University of Sydney, Sydney, NSW, Australia Two experiments used the gaze-contingent moving-window paradigm to investigate whether reading comprehension and spelling ability modulate the perceptual span of skilled adult readers during sentence reading. Highly proficient reading and spelling were both associated with increased use information to the right of fixation, but did not systematically modulate the extraction of information to the left of fixation. Individuals who were high in both reading and spelling ability showed the greatest benefit from window sizes larger than 11 characters, primarily because of increases in forward saccade length. They were also significantly more disrupted by being denied close parafoveal information than those poor in reading and/ or spelling. These results suggest that, in addition to supporting rapid lexical retrieval of fixated words, the high quality lexical representations indexed by the combination of high reading and spelling ability support efficient processing of parafoveal information and effective saccadic targeting. Keywords: Reading; Eye movements; Individual differences; Lexical quality; Perceptual span. Skilled reading of text depends upon a dynamic interaction of low-level oculomotor and perceptual processes with higher level lexical and comprehension processing. Each of these sets of processes occurs, at least to some extent, in parallel for both foveal and parafoveal information. The question of precisely how foveal and parafoveal information contribute to skilled comprehension is central to current theories of eye movement control in reading (see Radach & Kennedy, 2013; Schotter, Angele, & Rayner, 2012, for reviews). To contribute to further refinement of these theories, the two experiments reported here investigated whether and how individual differences amongst skilled readers modulate the use of parafoveal information during sentence processing. The reading perceptual span Critical evidence about the use of parafoveal information during sentence reading is provided by investigations of the perceptual span: the area of text from which information is extracted during a single eye fixation. Evidence for the size of the perceptual span comes from the gaze-contingent moving-window paradigm (McConkie & Rayner, 1975) in which the reader is provided with only a fixed window of text around their point of fixation, which moves with their eyes; text outside the window is masked. The size of the perceptual span is inferred by determining the size of the window at which the rate of reading is equivalent to that achieved with a full line of text. Previous research has established that, for skilled readers of English, the perceptual span is approximately 3–4 character spaces to the left and up to 15 characters to the right of the point of fixation (see Rayner, 2009, for review). The asymmetry of the perceptual span is a function of the left-toright reading direction of English: In languages Correspondence should be addressed to Aaron Veldre, School of Psychology, University of Sydney, Sydney, NSW 2006, Australia. E-mail: [email protected] # 2013 The Experimental Psychology Society 703 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS such as Hebrew, where the normal direction of reading is right to left, the span is asymmetric to the left (Pollatsek, Bolozky, Well, & Rayner, 1981), and when English is read right to left, the span asymmetry reverses (Inhoff, Pollatsek, Posner, & Rayner, 1989). The perceptual span for a particular language is also modulated by concurrent processing load and goals. Difficult foveal processing is associated with a smaller perceptual span (Henderson & Ferreira, 1990), and the perceptual span extends further to the left before a regressive eye movement (Apel, Henderson, & Ferreira, 2012). Such evidence demonstrates that the size of the perceptual span is, at least in part, a function of cognitive/linguistic processes. However, perceptual limitations on visual acuity mean that the nature of the extracted information is not uniform across the perceptual span (Rayner, 2009). The high spatial frequency information required for letter identification is only available within the foveal region (∼2 degrees; 7–8 characters around fixation; Häikiö, Bertram, Hyönä, & Niemi, 2009). The reduced visual acuity of the parafoveal region (2–5 degrees around fixation) only allows extraction of low-level features, such as word length, word shape, spaces, and beginning letters (Rayner, 2009). These differences in the nature of the information extracted from foveal and parafoveal regions lead to differences in the role that each information source plays in governing decisions about when and where readers move their eyes. It is generally assumed that the linguistic information extracted from foveal vision determines when to move the eyes, while the low spatial frequency information about word spacing and letter shape extracted from the parafovea primarily determines where to move the eyes (Schotter et al., 2012). The distinction between when and where to move the eyes is clearly somewhat arbitrary, because a “where” decision to move the eyes to a new location can be difficult to distinguish from a “when” decision to terminate processing of a particular item. Nevertheless, consistent with neurophysiological evidence that separate neural pathways control the temporal and spatial programming of movement (e.g., Findlay & Walker, 704 1999), most models of eye movement control assume that “where and when decisions can be considered separate dimensions of the reading process with a small degree of overlap” (Schotter et al., 2012, p. 13). Individual differences in the eye movement patterns of skilled readers might arise from either of these processes, or from interactions between them. Individual differences amongst skilled adult readers Research on eye movements, like other experimental psycholinguistic literature, has been dominated by a tacit “uniformity assumption” (Andrews, 2012): that all skilled readers read in the same way. Although there are a number of detailed computationally implemented theories of reading that have been systematically validated against empirical data from eye-tracking paradigms (e.g., Engbert, Nuthmann, Richter, & Kliegl, 2005; Reichle, Rayner, & Pollatsek, 2003; Reichle, Warren, & McConnell, 2009), the benchmark phenomena used to assess the models consist primarily of average data obtained from relatively small (n = 20–25) samples of university students. As elaborated below, recent evidence from both single word and sentence comprehension tasks has challenged the validity of conclusions based on average data by demonstrating systematic differences in both word recognition and sentence processing within samples of skilled, university student readers (see Andrews, 2012, for review). The extent to which eye movements during reading are systematically mediated by individual differences amongst skilled adult readers is also an area of growing interest. Jared, Levy, and Rayner (1999) found that more skilled readers, as measured by the comprehension subsection of the Nelson–Denny Reading Test, made significantly shorter fixations than poorer readers. There was also some evidence of differences in word identification processes between the two groups, suggesting that poorer readers relied more on phonological activation than better readers. Further evidence of differences in reading strategy were provided by Ashby, Rayner, and THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS Clifton’s (2005) finding that better readers, again indexed by the Nelson–Denny Reading Test, showed smaller frequency effects than poorer readers in neutral sentence contexts and different patterns of interaction between frequency and context. While good readers showed additive effects of frequency and context, poor readers showed an interaction—stronger context effects for low-frequency words over both target and posttarget regions of the sentence—suggesting that they relied more heavily on context for word identification. More recent research, using a larger battery of measures of verbal and cognitive skills to predict skilled readers’ eye movements during sentence reading, found that word identification and tests of rapid letter and digit naming (Wolf & Bowers, 1999) were the most robust and pervasive predictors of both early and late eye movement measures (Kuperman & Van Dyke, 2011). These results converge with Ashby et al.’s (2005) findings in supporting an interactive model of reading skill that assumes that highly skilled reading relies on rapid, autonomous lexical retrieval processes that place minimal reliance on context for word identification and conserve attentional resources for comprehension (Perfetti, 1992; Stanovich, 2000). The role of lexical quality in skilled reading The view that optimally efficient reading relies on autonomous, context-free lexical retrieval is consistent with Perfetti’s (1992, 2007) lexical quality hypothesis of reading skill, which attributes highly skilled reading to the development of “high-quality lexical representations”, defined by their precision, redundancy, and coherence (Perfetti, 2007). Orthographic precision refers to the specificity and completeness of the information stored in lexical memory about both the identity and order of the letters defining a particular word, which ensures that a particular written word directly activates its lexical representation with little interference from other similar words. Highquality representations are also defined by strong connections between these precise orthographic forms and the word’s phonology and semantics, so that presentation of a word triggers consistent, redundant patterns of activity that yield coherent, synchronous activation across the orthographic, phonological, and semantic codes that define a word’s identity. According to this view, individual differences in reading skill amongst adult readers arise from differences in the average quality of readers’ lexical representations. Highly skilled readers have fully specified lexical representations of most words, enabling rapid, automatic word identification, which supports effective comprehension (Andrews, 2008, 2012). However, relying on passage reading comprehension tests alone to assess reading skill is not sufficient to identify lexical quality. Reading comprehension is a necessary but not sufficient predictor of lexical quality because readers can use contextually based strategies to compensate for their imprecise lexical knowledge (Andrews, 2012). Spelling ability has been suggested to provide the most reliable index of lexical precision because it requires specific word-form knowledge (Perfetti, 1992). Although reading and spelling ability are quite highly correlated in large samples of typical readers (e.g., Malmquist, 1958), many competent readers are poor spellers, even within university populations (Andrews, 2012). Frith (1980) attributed this profile to use of a contextually guided reading strategy that relies on “partial visual cues” rather than “bottom-up” analysis of the full orthographic form of words required to develop highquality lexical representations. Recent evidence from studies employing the masked priming paradigm has shown that spelling predicts unique variance in orthographic priming. Andrews and Hersch (2010) showed that the absence of masked priming from one-letter-different neighbours for target words from dense lexical neighbourhoods that is typically reported for skilled readers (e.g., Forster, Davis, Schoknecht, & Carter, 1987) is due to averaging the results of good and poor spellers: Good spellers showed inhibition from a higher frequency orthographic neighbour prime (e.g., note NODE), whilst poor spellers showed facilitation. These results suggest that the precise lexical representations indexed by measures of spelling ability support fast activation of the THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 705 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS representations of briefly presented primes, which rapidly inhibit orthographically similar representations, including that for the target word. Andrews and Lo (2012) extended these findings by showing that individuals with high spelling ability relative to their level of reading comprehension and vocabulary showed stronger inhibitory priming from transposed letter (TL) prime–target pairs, which differ only in letter order (e.g., colt CLOT), suggesting that they are particularly sensitive to competition between the representations of words that share all of their constituent letters. By contrast, individuals with poor spelling relative to their reading and vocabulary levels showed facilitatory priming from both neighbours and TL words, suggesting that they are relatively insensitive to letter order and benefit from sublexical overlap between primes and targets. However, they do not appear to activate the prime representations quickly enough to trigger lateral inhibition of similar words (Perry, Lupker, & Davis, 2008). This evidence that masked orthographic priming effects are modulated by spelling ability implies that the speed of lexical retrieval is not solely determined by the linguistic properties of words, but depends on the precision of the reader’s representations of these words. The broad goal of the present research is to investigate whether and how these differences in lexical quality influence the information that readers extract during sentence reading. As well as supporting rapid, autonomous retrieval of fixated words (Ashby et al., 2005), precise lexical representations might facilitate the extraction and use of parafoveal information to “guide eye movements … to the optimal viewing position for full-form word recognition” (Kuperman & Van Dyke, 2011, p. 56). To provide direct evidence about readers’ extraction and use of parafoveal information during reading, we used the moving-window paradigm (McConkie & Rayner, 1975), described earlier, to investigate individual differences in the perceptual span amongst skilled readers. Individual differences in perceptual span Developmental investigations of the perceptual span have found that young children have smaller 706 perceptual spans than older children (Häikiö et al., 2009) and adults (Rayner, 1986). Underwood and Zola (1986) found no difference in the size of the perceptual span between good and poor fifth-grade readers. However, rather than using the moving-window paradigm, they estimated span by replacing particular letters in parafoveal vision. There is a growing body of research investigating the size of the perceptual span amongst skilled adult readers. Kuperman and Van Dyke (2011) interpreted their finding of differential word length effects on gaze duration for poor relative to good readers as being consistent with skillbased differences in the size of perceptual span. More direct evidence using the moving-window paradigm was provided by Rayner, Slattery, and Bélanger (2010) who split their sample of skilled adult readers on reading speed, assessed when no moving window was present. They found that the reading rate of slow readers reached asymptote with a two-word-rightward window whereas the reading rate of fast readers did not reach asymptote with a three-word window. Similarly, Ashby, Yang, Evans, and Rayner (2012) found a larger disruption in silent reading rate with one-word than threeword windows for fast than for slow readers. However, whilst speed is certainly a component of reading ability, fast readers are not necessarily effective readers (Perfetti, 2007). Differences in reading speed may also arise because some fast readers adopt a lower comprehension threshold and read in a careless fashion. Consistent with this possibility, Hyönä, Lorch, and Kaakinen’s (2002) investigation of individual differences in reading strategy found that the fastest readers did not produce better summaries of a text than readers classified as “topic structure processors”, who read the text significantly more slowly and with more regressive saccades. The minimal comprehension requirements of the relatively simple comprehension questions used in most eye-movement experiments potentially encourage a fast, careless reading style in sentence-reading tasks (Wotschack & Kliegl, 2013). Two recent studies provide evidence that reading strategy may influence the perceptual THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS span. Rayner, Castelhano, and Yang (2009) compared the perceptual span of elderly readers to that of college-aged readers. Younger readers showed an increase in reading rate when provided with two rather than one word to the right of the currently fixated word, while older readers showed no such benefit. However, older readers benefited from having the word to the left of the currently fixated word available, while younger readers did not. The authors argued that older readers may compensate for their slower foveal processing and less efficient extraction of parafoveal information by adopting “riskier”, more contextually guided reading strategy. An interesting recent addition to the evidence about individual differences in perceptual span is provided by Bélanger, Slattery, Mayberry, and Rayner’s (2012) comparison of perceptual span between skilled and less-skilled deaf readers and skilled hearing readers. Skilled deaf readers had a larger perceptual span than skilled hearing readers due to both a greater benefit from large window sizes (18 characters) and greater disruption from a small window size (6 characters). These findings are consistent with evidence from a range of nonreading tasks showing that individuals who experience severe deafness from early in life are more efficient at processing parafoveal and peripheral information than hearing individuals (e.g., Dye & Bavelier, 2010). Contradicting Dye, Hauser, and Bavelier’s (2008) speculation that this might disrupt reading by interfering with the processing of foveal information, Belanger, Slattery, et al. (2012) found that early deaf readers’ “wider distribution of attention … [allowed them to] take in more visual information within a fixation than do hearing readers matched on reading level” (p. 823). To contribute to understanding the basis of these individual differences in perceptual span amongst skilled readers, the present experiments used the moving-window paradigm to evaluate whether and how individual differences in lexical quality amongst skilled readers modulate the extraction and use of parafoveal information during sentence reading. To capture the orthographic precision that defines high-quality lexical representations, participants were assessed on measures of both reading comprehension and spelling. If the larger perceptual span demonstrated by faster adult readers (Ashby et al., 2012; Rayner et al., 2010) reflects the efficient lexical processing afforded by precise lexical representations, individual differences in spelling should make an additional contribution to predicting perceptual span, over and above reading comprehension and speed. Consistent with most previous investigations of perceptual span, Experiment 1 focused on window size to the right of fixation, while Experiment 2 also manipulated the amount of leftward information. EXPERIMENT 1 In Experiment 1, participants read sentences for meaning, which were presented either in their entirety (“full-line” condition), or as a series of a gaze-contingent moving windows. In the moving-window conditions, participants always saw 4 characters to the left of the point of their fixation and 3, 7, 11, or 15 characters to the right. Characters outside of this window were replaced by upper-case Xs. In addition to manipulating window size, we compared conditions in which the spaces between the words outside the window were retained or filled (see Figure 1). Previous research has shown that word boundary information is critical to eye movement targeting (Pollatsek & Rayner, 1982; Rayner, Fischer, & Pollatsek, 1998). It was therefore important to determine whether sensitivity to word spacing was a source of individual differences. Sentence difficulty was also manipulated by embedding matched triplets of high- and low-frequency words (e.g., fresh afternoon wind vs. brisk twilight gust) into identical sentence frames. This type of sentence difficulty manipulation has previously been found to produce differences in reading rate and fixation durations that persist beyond the critical string (Slattery, Pollatsek, & Rayner, 2007). Interactions with sentence difficulty provide evidence about whether individual differences in perceptual span are affected by the THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 707 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS Figure 1. Examples of the Experiment 1 critical stimuli (a) in the full-line condition; (b) with a 7-character moving window and word spaces retained; (c) with a 7-character moving window and word spaces filled. The point of fixation is represented by the asterisk (*). difficulty of foveal processing (e.g., Henderson & Ferreira, 1990). As well as assessing whether the overall size of readers’ perceptual span, as indexed by reading rate, is modulated by reading and/or spelling ability, measures of fixation duration and forward saccade length were extracted to provide insight into the basis of any observed differences in perceptual span. The lexical quality hypothesis predicts that the precise lexical representations indexed by the combination of high levels of reading and spelling should support more efficient lexical processing of words in foveal vision. Increased perceptual span might be a direct consequence of more efficient foveal processing because reduced “foveal load” may increase the resources available for parafoveal processing. Individual differences that are due to these “secondary” effects of foveal processing efficiency should interact with sentence difficulty. Lexical quality may also directly influence parafoveal processing by increasing the efficiency of lexical processing of parafoveal information and/ or facilitating the efficiency of oculomotor planning processes (Kuperman & Van Dyke, 2011). The latter effects would be expected to manifest primarily in measures of saccade length. Method Participants Forty-five undergraduate students from the University of Sydney (33 female; mean age 20.1) participated in Experiment 1 in exchange for 708 course credit. One participant was excluded from the analysis due to excessive blinking, leaving a final sample of 44. All had normal or correctedto-normal vision and reported English as the first language they learned to read. Materials The critical stimuli were 64 pairs of sentences averaging 10 words in length. One sentence within each pair contained a string of three high-frequency words (e.g., A fresh afternoon wind blew across the choppy harbour). In the other sentence, this string was replaced with three low-frequency synonyms (e.g., A brisk twilight gust blew across the choppy harbour). The critical string always occurred early in the sentence, beginning between the second and sixth word. The mean length of the three words comprising the critical string was matched within each pair, and, as far as possible, the individual synonyms were matched on length in a pairwise fashion. The critical sentences were rotated across a 5 (window size) by 2 (spacing) design. Because word spacing outside the moving window cannot be manipulated in the full-line condition, the design was not fully factorial. Therefore, all sentences appeared in all conditions over nine counterbalanced lists. The sentences were validated on a separate sample of 29 undergraduate students who did not participate in either Experiment 1 or Experiment 2. Maintaining low contextual predictability was important to ensure that effects were due primarily to the manipulated factors. To assess this, THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS participants were given a list of sentence fragments, half finishing prior to the second critical word and the other half before the third critical word, and they were asked to generate the word they thought was most likely to come next in the sentence. Two counterbalanced lists, containing either the two- or the three-word fragment from each critical string, were each read by approximately half the sample. Three third-position critical words had cloze predictability of 30% or more. For these sentences, the predictable word was replaced with a less frequently generated synonym for the final list of experimental items. Participants then rated the overall plausibility of the complete sentences on 7-point Likert scale. Each participant rated one sentence from each pair, with all sentences appearing over two counterbalanced lists. The critical word string characteristics and validation data for the final list of sentences are presented in Table 1. Measures of reading and spelling ability All participants completed several measures of reading and spelling ability. Reading ability. The Nelson–Denny Reading Test (Brown, Fishco, & Hanna, 1993) was used to assess participants’ reading ability. The test contains a vocabulary section of 80 items and a separately timed reading comprehension section of 38 items relating to seven passages of text.1 The raw scores from the reading comprehension section were standardized to provide a measure of reading ability. Spelling ability. Two measures of spelling ability were used: spelling dictation and spelling recognition. The spelling dictation test consists of 20 low-frequency words selected from a list compiled by Burt and Tate (2002), which vary in difficulty. Each word Table 1. Stimulus characteristics for the critical word string across sentence difficulty Sentence difficulty Characteristic Mean frequency per million Mean word length (characters) Cloze prediction: Word 2 Cloze prediction: Word 3 Sentence rated plausibility Easy Hard 124.9 6.5 .01 .05 6.4 8.7 6.7 .00 .01 6.2 Note: Frequency and length data from the CELEX database (Baayen, Piepenbrock, & van Rijn, 1993) using Davis’s (2005) N-Watch software. was read aloud twice, once alone and once in a sentence, after which the participant was required to write down the word. The spelling recognition task comprises 88 items: 44 correctly spelt and 44 incorrectly spelt words. Participants were given unlimited time in which to circle all incorrectly spelt words. The highly correlated spelling dictation and recognition scores (Experiment 1: r = .78; Experiment 2: r = .80) were standardized and averaged to form a single measure of spelling ability. The standardized measures of reading and spelling were only moderately correlated (Experiment 1: r = .34; Experiment 2: r = .48). This may reflect the restricted range of written language proficiency amongst university students, given that the majority of participants (95%) had scores above the 50th percentile for U.S. firstyear college students.2 Apparatus An SR Research EyeLink 1000 eye tracker was used to record participants’ eye movements as they read the sentences. Viewing was binocular but fixation location was monitored from the right eye. Participants were seated 60 cm from the monitor with their head position stabilized by a chin and forehead rest. Stimuli 1 The Nelson–Denny Reading Test also includes a measure of reading speed. When this variable was included in the analysis, the pattern of significant results was virtually identical. However, it was decided not to include this measure due to concerns with the validity of self-reported reading speed measures (Carver, 1985). 2 The relatively low correlations between reading and spelling do not reflect ceiling effects in the assessment instruments. No participant in either experiment achieved a perfect score on the Nelson–Denny Reading Test. In Experiment 1, total scores ranged from the 23rd to the 99th percentile for U.S. first-year college students, with mean percentile ranks of 63 and 92 for the lower and upper half of the sample. In Experiment 2, participants’ scores ranged from the 46th to the 99th percentile, and the average percentile rank of the upper and lower halves of the sample were 69 and 92, respectively. THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 709 VELDRE AND ANDREWS Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 were presented on a 19′′ Dell CRT monitor with a refresh rate of 85 Hz, and, at this distance, 3 characters equalled 1 degree of visual angle. Procedure Participants were tested individually in a single session that lasted approximately 1.5 hours. The session began with written and verbal instructions to read the sentences for meaning. This was followed by a three-point calibration procedure. Participants were recalibrated after the practice trials and then after every 16 experimental trials. At the beginning of each trial participants were required to fixate on a circle in the location of the sentence initial character. Once the participant made a stable fixation on this point, the sentence was displayed. If necessary, a new calibration procedure was initiated. Sentences were presented on a single line in black, 12-point Courier New font against a light-grey background and remained on the screen until the participant pressed the space-bar to indicate they had finished reading. Participants read all 128 sentences once in two counterbalanced blocks, separated by the battery of reading and spelling ability measures. The easy and hard version of each sentence appeared in separate blocks but within each block, half the sentences were easy, and half were hard. The first three trials were practice sentences followed by the experimental sentences, which were presented in a randomized order. On all practice trials and approximately 30% of experimental trials, the sentence was followed by a three-option multiplechoice comprehension question that required a moderate level of understanding of the sentence in order to make a relatively simple inference. Questions were designed so that a participant could not simply guess based on visual recognition of the answer with a word from the sentence. For example, the sentence: She noticed the modern building from several blocks away was followed by the question: The building was probably very … a) large, b) small, c) old. Results and discussion Mean comprehension accuracy was very high (95%), indicating that participants read the sentences for meaning. Fixations below 80 ms were merged with fixations within one character space (0.5% of fixations), and remaining fixations below 80 ms or above 1000 ms were deleted (3.3% of fixations). Trials in which a participant made two or more blinks during sentence reading were eliminated (8.4% of trials). Of the remaining trials, 31.7% contained one blink. Consistent with previous moving-window experiments, the major dependent variable was reading rate, measured by words per minute (WPM). Reading rate is influenced by both the duration of fixations and the length of saccades, so analyses of average fixation duration (FD) and average forward saccade length (SL) are also reported to provide more refined insight into how readers respond to constraints on their perceptual span. The results were analysed by testing linear mixed effects (LME) models using the lme4 package (Bates, Maechler, & Bolker, 2012) in R (R Development Core Team, 2011), with subjects and items as crossed random effects. Statistical significance was assessed by Markov chain Monte Carlo simulation using the “pvals.fnc” function from the languageR package (Baayen, 2011). Graphics were produced with ggplot2 (Wickham, 2009) based on LME model-adjusted values generated by the “remef” function (Hohenstein, 2011). To examine the independent contribution of reading and spelling ability to the eye movement record, the measures were entered as separate predictors and were allowed to interact with the experimental effects in the models. Trial order and experimental list were also included in the models as fixed effects.3 The sentence difficulty and word spacing effects were coded as sum contrasts, and successive difference contrasts 3 In the analysis of both experiments, models that included random slopes for subjects and items corresponding to each of the fixed effects, interactions, and correlations between these effects (i.e., maximal random effect structures) failed to converge. Simpler models that removed random slopes for the interactions did converge but some of the correlations were 1, indicating that their inclusion would risk overfitting. A model specifying subject random slopes for the window factor converged but also yielded correlations very close to 1. The data reported are therefore for models that only include subject and item random intercepts. The pattern of significant effects did not differ between the final model and the converged models that specified random slopes. 710 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Note: 3R = 3-character-rightward window; 7R = 7-character-rightward window; 11R = 11-character-rightward window; 15R = 15-character-rightward window; FL = full-line condition; WPM = reading rate measured in words per minute; FD = average fixation duration measured in ms; SL = forward saccade length measured in character spaces; #FF = number of forward fixations; #RF = number of regressive fixations. 202 213 8.3 9.4 3.0 197 224 8.3 9.5 2.7 195 224 7.9 9.5 2.8 174 226 7.2 9.9 2.7 116 267 6.6 10.9 2.8 193 229 8.3 9.4 2.6 193 225 7.4 9.7 2.6 169 234 6.5 10.1 2.2 103 272 5.3 10.8 2.4 238 205 8.6 8.6 2.6 220 218 8.4 8.8 2.5 216 216 8.1 9.0 2.6 193 223 7.3 9.4 2.2 129 259 6.4 10.3 2.7 223 223 8.4 8.7 2.4 217 220 7.6 9.0 2.1 117 267 5.5 10.7 2.4 WPM FD SL #FF #RF 190 223 6.5 9.5 2.0 15R 11R 11R 7R 11R 11R 7R 3R Measure Spaces filled 15R 3R 7R Spaces retained 15R FL 3R Spaces filled 15R 3R 7R Spaces retained Hard sentences Easy sentences Average data The average data were reasonably consistent across the four measures: In general, an increase in reading rate (WPM) was associated with a decrease both in average fixation duration (FD) and in regression count (REG) and an increase in average forward saccade length (SL). Results of the four measures are therefore reported together, noting divergent findings where applicable. All four measures significantly improved from the 3-character window to the 7-character window (WPM: t = 30.7, p , .001; FD: t = –27.5, p , .001; SL: t = 12.4, p , .001; REG: t = –2.8, p = .004). Regression count was not affected by further increases in window size (REG: all ts , 2). The remaining three measures improved from 7 to 11 characters (WPM: t = 11.8, p , .001; FD: t = –4.2, p , .001; SL: t = 12.6, p , .001). There was no significant change in either reading rate or fixation duration between the 11-character and 15-character window (both ts , 2) but saccade length increased significantly between these window sizes (SL: t = 7.2, p , .001). Only reading rate and fixation duration showed significant improvement between the 15-character FL (Venables & Ripley, 2002) were tested across the levels of window size (i.e., 3 vs. 7; 7 vs. 11, 11 vs. 15; 15 vs. full line, FL) to assess the effect of each increment in window size. Likelihood tests revealed that including the two-way interactions between all manipulated and individual difference variables significantly improved model fit for all measures, all χ 2(36) . 53.39, p , .031. However, adding the three-way interactions yielded no further improvement in fit, for any of the dependent variables, all χ 2(16) , 15.27, p . .505, so they were not included in the final models reported below. Means for the dependent variables as well as mean number of forward and regressive fixations in each condition are presented in Table 2, and the LME analyses for each dependent measure are summarized in Appendix A. We first report the results based on average data to allow comparison with previous studies and then describe how they were modulated by the individual difference measures. Table 2. Eye movement measures for each moving-window condition in Experiment 1 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 711 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS window and the full line (WPM: t = 3.9, p , .001; FD: t = –9.5, p , .001; SL: t , 1). There were significant main effects of sentence difficulty on all measures because easy sentences were read faster, and were associated with longer saccades and fewer regressions, than hard sentences (WPM: t = 15.8, p , .001; FD: t = –8.0, p , .001; SL: t = 2.5, p = .014; REG: t = –3.8, p , .001). The lack of word spacing information outside the moving window was associated with a significant average reduction in reading rate, saccade length, and regressions (WPM: t = –2.1, p = .034; FD: t = 2.9, p = .003; SL: t = –8.2, p , .001; REG: t = –3.5, p , .001), but a significant interaction between spacing and the 3- versus 7-character window contrast revealed that this was almost entirely due to a reduction in reading rate and saccade length for the 3-character window condition when word spacing information was unavailable (WPM: t = 2.1, p = .037; SL: t = 2.0, p = .049). There were no significant effects of word spacing for window sizes of 7 or more (all ts , 2) except that saccade length was significantly more affected by a lack of word spacing at the 11-character than at the 15-character window (SL: t = 3.5, p , .001). Thus, consistent with previous research, the average data revealed a perceptual span of approximately 15 characters to the right of fixation. There was evidence of an unexpected additional benefit from presenting the full line of text, which contrasts with the much-replicated finding of no improvement in reading rate beyond 14–15-character windows (Rayner, 2009) but, as discussed below, this benefit was restricted to high-ability readers. The effects of eliminating spaces between words outside the window were generally confined to the 3-character window size, as would be expected if parafoveal word length information is used primarily to programme the next saccade. Individual differences There were significant main effects of both reading ability (WPM: t = 2.4, p = .002) and spelling ability (WPM: t = 2.0, p = .008) on reading rate, reflecting faster average reading by better readers and spellers. Neither variable reliably predicted average fixation duration or saccade length (all ts , 2). 712 Increased reading ability was associated with significantly fewer regressions (REG: t = –2.0, p = .019). To evaluate how individual differences modulate the “when” and “where” components of eye movements, we consider each of the measures separately. Reading rate. Reading rate yielded significant interactions between reading ability and the difference between the 3-character and 7-character window sizes (WPM: t = 4.2, p , .001) and between the 15-character window and full line (WPM: t = 4.0, p , .001). These effects are illustrated in Figure 2a, in which reading ability has been categorized into above- and below-average ability groups. Poorer readers were less sensitive to window size than better readers, showing significantly less reduction in reading rate than better readers for the smallest window size and less increase in reading rate than better readers beyond the 11-character window size. A similar pattern occurred for spelling ability, but it manifested as a significant interaction with the difference between the 7-character and 11character sizes (WPM: t = 2.7, p = .008). As shown in Figure 2b, this reflected a cross-over interaction whereby better spelling was associated with slower reading for windows sizes up to 7 characters, but a faster reading rate for windows of 11 characters or more. Spelling ability also significantly modulated the effect of sentence difficulty: The effect of difficulty on reading rate was reduced from approximately 30 WPM amongst poor spellers to 15 WPM amongst better spellers (WPM: t = –3.1, p = .002). Reading and spelling ability also jointly interacted with sentence difficulty and the 15- versus 11-character contrast on reading rate (WPM: t = 2.0, p = .043). To summarize this interaction, Figure 3 presents the data for median splits on both reading and spelling. These data show that the combination of above-average reading and spelling ability was associated with an increased benefit in reading rate beyond the 11-character window, which was restricted to easy sentences. Fixation duration.. The interactions of reading and spelling ability with window size revealed by the THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS Figure 2. (a) Reading rate (words per minute, WPM) over window sizes in Experiment 1 for low- and high-ability readers. (b) Reading rate over window sizes in Experiment 1 for low- and high-ability spellers. THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 713 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS Figure 3. Reading rate (words per minute, WPM) over window sizes in Experiment 1 for low- and high-ability readers in easy and hard sentences, split by spelling ability. reading rate measures were generally paralleled in the analyses of fixation duration. Better reading was associated with significantly longer fixations in the 3-character than in the 7-character window condition (FD: t = –3.4, p , .001), and better spellers showed significantly less reduction in average fixation durations when moving from a 3- to a 7character window than did poorer spellers (FD: t = 2.6, p = .014). The combination of high reading ability and high spelling ability was associated with the greatest inflation of fixation duration for the 3- than for the 7-character window (FD: t = –2.4, p = .017). Paralleling the reading rate measure, both higher reading ability (FD: t = –2.0, p = .047) and higher spelling ability (FD: t = –2.2, p = .025) were associated with a significantly greater decrease in fixation duration between the 15-character window and full-line condition. Saccade length. The only interactions involving individual differences on saccade length involved spelling ability, either alone or in combination with 714 reading ability. Spelling ability yielded a significant interaction with window size because better spellers showed a greater increase in saccade length beyond the 11-character window than poorer spellers (SL: t = 2.6, p = .009). Reading and spelling ability also interacted jointly with the difference between the 11- and the 15-character window (SL: t = 2.7, p = .010) because the combination of higher reading and higher spelling was associated with the largest increase in saccade length for windows larger than 11 characters (see Figure 4). Regression count. Spelling ability interacted significantly with sentence difficulty because the regression counts of better spellers were less affected by difficulty than those of poorer spellers (REG: t = 3.5, p , .001). Overall, the average reading data are consistent with the previous estimates of the perceptual span for skilled readers (e.g., McConkie & Rayner, 1975; Rayner, 2009). However, the individual difference measures show that these average values are modulated by both reading and spelling THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS Figure 4. Forward saccade length over window sizes in Experiment 1 for low- and high-ability readers, split by spelling ability. ability. Although there were no significant main effects of either measure on any variable, better readers and spellers showed more benefit from larger window sizes, reflected in faster reading rates and shorter fixation durations for the full line than a 15-character window. The combination of above-average reading comprehension and spelling was also associated with longer saccades with larger window sizes. The findings confirm Kuperman and Van Dyke’s (2011) speculation that more proficient lexical processers are more efficient at extracting the foveal information required for lexical retrieval and suggest that earlier evidence of increased perceptual spans amongst faster readers (e.g., Ashby et al., 2012; Rayner et al., 2010) probably reflects the efficiency of lexical processing rather than reading speed, per se.4 The more novel finding of Experiment 1 is that, as well as benefiting more from a larger window, more proficient reader/spellers also showed slower reading rates and longer fixations for the 3-character window than did lower proficiency individuals. The disruptive effects on fixation duration were strongest for those high in both reading comprehension and spelling, consistent with the view that they arise from lexical skills that are not effectively captured by reading comprehension alone. In general, this pattern indicates that less proficient readers adjust their fixation durations less according to the amount of perceptual information available to them—as well as failing to take advantage of the larger window sizes, they also show less sensitivity to being deprived of parafoveal information. However, the disruptive effect of small window sizes on higher proficiency individuals also implies that individuals who are more efficient at lexical processing make greater use of parafoveal information for saccadic planning. Although spelling ability was associated with a reduced overall effect of sentence difficulty, the disruptive effects of small window sizes shown by higher proficiency reader/spellers were unaffected by difficulty, suggesting that they reflect disruptions to oculomotor planning processes rather than being a byproduct of better readers’/spellers’ more efficient foveal processing. Similarly, although the spacing manipulation affected average reading rate for small window sizes, the effect of filling word spaces did not interact with reading or spelling ability. This implies that all readers use word 4 We also conducted an analysis in which average reading rate from the full-line condition was a continuous predictor in the models, mirroring Ashby et al. (2012) and Rayner et al. (2010). The results of this analysis were clear: Reading ability and spelling ability no longer predicted average reading speed but all of the other significant effects remained. That is, even when reading speed was controlled for, lexical processing ability significantly modulated the size and use of the perceptual span. THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 715 VELDRE AND ANDREWS boundary information for saccade targeting but that better readers and spellers make use of additional information to optimize saccadic targeting (Kuperman & Van Dyke, 2011). Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 EXPERIMENT 2 The aims of Experiment 2 were to replicate the critical findings of Experiment 1 and to investigate whether individual differences in reading and spelling ability also modulate use of information to the left of the point of fixation. Previous investigations of skilled readers’ perceptual span have shown that they only use 3–4 characters to the left of fixation (e.g., McConkie & Rayner, 1975). Consistent with this conclusion, Rayner et al. (2009) found that, even with a limited window of one word to the right, college-age participants showed little benefit of being presented with an additional word to the left. Such data suggest that variations in skilled readers’ perceptual span will be limited to the region to the right of fixation and that individual differences will have no impact on the use of information to the left of fixation. In contrast, Jordan, McGowan, and Paterson (2013) recently found that reading was disrupted when interword spaces were filled as far as three words to the left of the currently fixated word, suggesting that readers utilize as much information to the left as to the right. These contradictory findings may arise from individual differences in the use of leftward information during reading. There are two possible alternative hypotheses about how individual differences might modulate the use of information to the left of fixation. Firstly, reliance on a top-down reading strategy may be associated with a more symmetric perceptual span. As noted earlier, Rayner et al. (2009) attributed the smaller and more symmetric span shown by elderly than by college-aged readers to use of a more contextually guided reading strategy to compensate for their slower foveal processing and less efficient parafoveal processing. They argue that this is a “riskier” reading strategy characterized by longer saccades and a higher probability of both skipping and regressions that leads older readers to “need more information available to the left of fixation to offset the limitations in processing information to the right of fixation” (Rayner et al., 2009, p. 759). Ashby et al.’s (2005) data, reviewed earlier, showed that poorer college-aged readers were Figure 5. Examples of the Experiment 2 critical stimuli with (a) 3-character-leftward and 3-character-rightward windows; (b) 3 left and 9 right; (c) 3 left and 15 right; (d) 6 left and 3 right; (e) 9 left and 3 right. The point of fixation is represented by the asterisk (*). 716 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS more reliant on sentence context than were better readers. If this strategy is associated with a more symmetric perceptual span, poorer readers, and particularly poorer spellers who have less precise orthographic knowledge and therefore need to rely more on context (Andrews & Bond, 2009; Hersch & Andrews, 2012), may make more use of leftward information to offset their less efficient use of rightward information. By this view, less proficient readers’/spellers’ reduced sensitivity to the 3-character rightward window in Experiment 1 might reflect greater use of information to the left of fixation. Recent evidence suggests that the perceptual span extends further to the left before regressive saccades (Apel et al., 2012). The increased rate of regressions shown by poorer readers/spellers in Experiment 1 may, therefore, further contribute to greater reliance on leftward information. Alternatively, higher proficiency readers may benefit more from information to the left of fixation, reflecting a generally broader attentional spotlight (Schad & Engbert, 2012), like that shown by deaf individuals in nonreading tasks. In Experiment 1, the amount of information available to the left in the moving-window conditions was always 4 characters. If better readers have both a larger rightward span and a larger leftward span, the constraint on leftward information in Experiment 1 may have impacted their reading and contributed to the disruptive effects of the smallest rightward window. To distinguish these alternatives, Experiment 2 orthogonally manipulated the size of the gaze-contingent moving window exposed on both the right (3, 9, or 15 characters) and the left (3, 6, or 9 characters) of fixation (see Figure 5). Method Participants Fifty-two (30 female; mean age 19.8 years) undergraduate students from the University of Sydney participated in Experiment 2 in exchange for course credit. Four additional participants were excluded from analysis due to a software error causing the moving window to display incorrectly on a number of their trials. Materials The same sentences were used as those in Experiment 1. The critical sentences were rotated across a 3 (leftward window size) by 3 (rightward window size) by 2 (sentence difficulty) design. All sentences appeared in all conditions over nine counterbalanced lists. Apparatus The stimuli were presented on a 21′′ ViewSonic G225fb CRT monitor with a refresh rate of 150 Hz, and 2.7 characters equalled 1 degree of visual angle. The same eye tracker was used as that in Experiment 1. Procedure The procedure was identical to that in Experiment 1. Results and discussion Again, analyses of reading rate, fixation duration, forward saccade length, and regression count are reported. Fixations shorter than 80 ms were merged with nearby fixations (0.4% of fixations), and remaining fixations shorter than 80 ms and longer than 1000 ms were deleted (2.8% of fixations). Trials in which a participant made more two or more blinks during sentence reading were eliminated (3.4% of trials). Of the remaining trials, 24.4% contained one blink. Mean comprehension accuracy was again very high (95%), indicating that the sentences were read for meaning. The results were analysed in the same way as in Experiment 1. Again, successive difference contrasts were tested over the levels of rightward window size. Two forward difference contrasts on leftward window size tested the additional benefit from (a) having more than 3 characters to the left and (b) more than 6 characters to the left. As in Experiment 1, likelihood tests revealed that including the two-way interactions significantly improved model fit for all measures, all χ 2 (32) . 46.30, p , .049, but adding the three-way interactions yielded no further improvement in fit, for any of the dependent variables, all χ 2 (16) , 17.29, p . .367, so they were not included in the final models reported below. Means for the dependent THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 717 718 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Note: 3R = 3-character-rightward window; 9R = 9-character rightward window; 15R = 15-character rightward window; WPM = reading rate measured in words per minute; FD = average fixation duration measured in ms; SL = forward saccade length measured in character spaces; #FF = number of forward fixations; #RF = number of regressive fixations. 233 219 6.9 8.6 2.5 159 246 5.5 11.9 3.0 240 219 7.7 7.8 2.9 235 221 6.9 8.4 2.4 162 249 5.3 11.8 2.7 235 230 7.5 7.9 3.1 223 229 6.8 8.8 2.8 150 255 5.3 11.9 3.1 276 212 7.9 7.3 2.5 255 213 6.9 8.1 2.2 181 240 5.4 10.8 2.3 272 215 7.8 7.3 2.5 258 214 7.0 8.0 2.3 179 241 5.4 11.0 2.3 260 223 7.7 7.2 2.5 164 251 5.2 11.5 2.6 WPM FD SL #FF #RF 245 222 6.9 8.1 2.4 9R 15R 9R 3R Measure 9R 15R 3R 9R 15R 3R 9R 15R 3R 9R 15R 3R 6 Left 3 Left 9 Left 6 Left 3 Left Average data The average data with respect to rightward perceptual span were highly consistent with the results of Experiment 1. Hard sentences were read more slowly than easy sentences (WPM: t = 16.0, p , .001; FD: t = –7.8, p , .001; REG: t = –7.5, p , .001). The effect of sentence difficulty on saccade length was in the same direction as the significant effect observed in Experiment 1, but failed to reach significance in Experiment 2 (SL: t = 1.8, p = .074). Moving from a 3- to a 9-character rightward window improved reading rate and saccade length and reduced regressions significantly (WPM: t = 41.3, p , .001; FD: t = –25.3, p , .001; SL: t = 34.4, p , .001; REG: t = –3.8, p , .001), as did moving from a 9- to a 15-character window (WPM: t = 5.8, p , .001; FD: t , 1; SL: t = 18.0, p , .001). However, regression rates were significantly higher for 15- than for 9-character windows (REG: t = 4.2, p , .001). The additional reading rate benefit of a 15-character window was significantly greater for easy than for hard sentences (WPM: t = 2.0, p = .048). This finding is consistent with previous investigations on the effect of foveal load on perceptual span (e.g., Henderson & Ferreira, 1990). This interaction was not significant in the average data in Experiment 1, perhaps because the smaller incremental change in window size in that experiment (4 characters vs. 6 characters) made the effect more difficult to detect. The average data with respect to use of leftward information showed that having more than 3 characters to the left resulted in a significant improvement on all three measures (WPM: t = 6.7, p , .001; FD: t = –10.5, p , .001; SL: t = 3.3, p , .001; REG: t = –3.5). However, having more than 6 characters to the left resulted in no significant benefit to reading (all ts , 1). This is consistent with findings that the use of information to the left does not extend beyond the currently fixated Table 3. Eye movement measures for each moving-window condition in Experiment 2 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 3R 9 Left 15R variables, as well as number of forward and regressive fixations, in each condition are presented in Table 3, and the LME analyses for each of the dependent measures are summarized in Appendix B. 242 222 7.7 7.9 2.9 VELDRE AND ANDREWS Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS Figure 6. Reading rate (words per minute, WPM) over rightward window sizes in Experiment 2 for low- and high-ability readers, split by spelling ability. word (Rayner, Well, & Pollatsek, 1980). This would usually equate to fewer than 3 characters but may extend further to the left, particularly in long words and words that receive multiple fixations. Individual differences In Experiment 2, reading comprehension was associated with higher average reading rate (WPM: t = 2.5, p = .014) but, in contrast to Experiment 1, there were no significant main effects of spelling on any measure (all ts , 1). However, as summarized below, spelling ability did interact significantly with other variables, consistent with the findings of Experiment 1. Reading rate. Paralleling the finding in Experiment 1 of greater disruption at small window sizes amongst good readers, there was a significant interaction between reading ability and the 3- vs. 9character window conditions because better readers were significantly slower reading with a 3character window than were poorer readers (WPM: t = 3.8, p , .001). Higher spelling ability was again associated with a reduced effect of sentence difficulty (WPM: t = –2.2, p = .027). There was also a significant interaction between reading ability, spelling ability, and the difference between the 9- and 15-character windows (WPM: t = 2.3, p = .025; see Figure 6). The form of this interaction was that the combination of above-average reading and above-average spelling was associated with the greatest additional benefit from having a 15-character window. There were no significant interactions involving leftward window size and either reading or spelling ability (all ts , 2). Fixation duration. The greater disruption of highability readers at the smallest window was reflected in a larger inflation of fixation duration (FD: t = –3.3, p = .001). Good spellers showed less reduction in fixation duration than did poor spellers when moving from the 3- to 9-character window (FD: t = 3.3, p = .001). Good spellers’ reduced effect of sentence difficulty was also reflected in fixation duration (FD: t = 2.6, p = .008). Again, there were no significant interactions between leftward window size and either of the individual differences measures (all ts , 2). Saccade length. Good readers showed a smaller increase in saccade length when moving from a 3to a 9-character window than poor readers (SL: t = –2.7, p = .006). Conversely, higher spelling THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 719 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS ability was associated with greater disruption to saccade length at the 3-character window (SL: t = 6.7, p , .001). Mirroring Experiment 1, the combination of high reading ability and high spelling ability was associated with the greatest improvement to saccade length between the 9and 15-character windows (SL: t = 2.3, p = .017). Reading and spelling ability also jointly interacted with the difference between the smallest window sizes (SL: t = 3.3, p = .002). Amongst poor spellers, good readers showed less disruption to saccade length than poor readers by the presence of a 3-character window and less improvement than poor readers to saccade length at the 9-character window, while, amongst good spellers, good readers showed greater disruption than poor readers to saccade targeting at the smallest window size. There were no significant interactions involving leftward window size and reading or spelling ability (all ts , 2). Regression count. Reading ability interacted with the 3- versus 9-character contrast because good readers showed higher regressions at the smallest window than poor readers but the reverse was true at the 9-character window (REG: t = –2.4, p = .019). Spelling ability showed the opposite pattern of interaction with the 3- versus 9-character contrast (REG: t = 2.8, p = .005): Good spellers showed fewer regressions at the small window than poor spellers but more regressions than poor spellers at the 9-character window. There were no significant interactions involving leftward window size and reading or spelling ability (all ts , 2). In summary, the results of Experiment 2 provide further evidence that skilled adult readers’ perceptual span varies with individual differences in reading and spelling. Using a less fine-grained manipulation of window size, the results confirmed Experiment 1 by showing that the combination of high levels of reading and spelling ability that we identify with lexical quality was associated with the biggest improvement in saccade length and reading rate for the largest rightward window and with significantly greater disruptive effects of the smallest rightward window on saccadic planning. 720 Also consistent with Experiment 1, higher reading and spelling ability were also associated with parallel, but independent, disruptive effects of the smallest rightward window size on fixation duration. However, reading and spelling were associated with different patterns of regressions for the small and intermediate window size manipulation of Experiment 2, perhaps reflecting somewhat different reading strategies in response to reduced parafoveal information. Finally, there was no evidence of individual differences in use of information to the left of fixation beyond the fixated word. It therefore appears that lexical processing ability does not determine the leftward extent of the perceptual span in skilled adult readers. GENERAL DISCUSSION The experiments reported here used the movingwindow paradigm to investigate whether previous findings of differences in perceptual span on the basis of reader age (Rayner et al., 2009), grade level (Häikiö et al., 2009; Rayner, 1986), and reading speed (Ashby et al., 2012; Rayner et al., 2010) might arise from individual difference in lexical quality (Perfetti, 2007). To capture the orthographic precision that is central to the critical construct of lexical quality, both reading comprehension and spelling ability were assessed (Andrews, 2012). Consistent with previous estimates of the perceptual span of skilled readers, the average data showed that readers used approximately 15 characters to the right of the point of fixation, but no more than 6 characters to left of the fixated word. Hard sentences were read more slowly than easy sentences, and removing parafoveal word boundary information by filling word spaces reduced saccade length for smaller window sizes. However, these average data obscure systematic individual differences in perceptual span. Both experiments provided clear evidence that more proficient readers and spellers were more sensitive to the availability of parafoveal information than less skilled individuals. This was manifested in two THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS ways. First, good readers and spellers benefited from a wider rightward perceptual span. In Experiment 1, the less skilled half of our sample showed no improvement in reading rate beyond an 11-character window but the higher skilled half showed enhanced benefit from the full line of text relative to a 15-character window.5 Rayner et al. (2010) found a similar result amongst their fast reader group using word-based windows: The reading rate of fast readers did not asymptote between a three-word window and the full line. They attributed this to variability in word length because, although their three-word windows contained an average of 14.9 characters, they could be as small as 8 characters. The use of characterbased windows in the present experiments rules out this possibility and shows that, at least some of the time, highly skilled readers use more than 15 characters to the right of fixation. It may be argued that the benefit from the full line of text reflects the superior integration abilities of highly skilled readers rather than more efficient lexical processing. However, the enhanced perceptual span was primarily restricted to sentences in which the foveal processing load was low and most evident amongst lexical experts—that is, those high in both reading comprehension and spelling ability (Hersch & Andrews, 2012). The larger rightward perceptual span therefore appears to be a function of the precise lexical representations indexed by spelling. The benefit for lexical experts was principally due to increases in forward saccade length with increasing window size, which were most strongly predicted by spelling ability, indicating that more precisely specified lexical representations allow readers to extract more parafoveal information during a single fixation, supporting more effective saccadic planning. Interestingly, this finding parallels the results for Bélanger, Slattery, et al.’s (2012) skilled deaf readers who had larger perceptual spans than skilled hearing readers, principally due to differences in saccade length. The second, less predictable, manifestation of individual differences in skilled readers’ perceptual span demonstrated by the present data is that lexical experts were more disrupted by the restriction of parafoveal information at small windows than were low-proficiency individuals. The association of higher proficiency with greater sensitivity to upcoming information is consistent with Chace, Rayner, and Well’s (2005) findings of increased parafoveal preview benefit amongst higher ability readers, which is attributed to faster lexical retrieval of the fixated word, allowing attention to shift to the next word before it is fixated. However, in both of the present experiments, better readers and spellers actually showed less efficient eye movements— longer fixations and shorter saccades—than poor readers/spellers when they were denied access to parafoveal information from the upcoming word. Furthermore, this did not depend on foveal processing load or word spacing, suggesting that lexical experts use parafoveal orthographic information for efficient saccade targeting. These findings differ from those of Ashby et al. (2012) and Rayner et al. (2010), who found that, although faster readers showed relatively more disruption at small windows, their absolute performance was better than that of slower readers. Therefore, the detrimental disruptive effect of denying parafoveal orthographic information appears to be specifically associated with lexical processing ability rather than speed. The efficient pattern of eye movements supported by precise lexical representations appears to depend on having access to information beyond the currently fixated word, suggesting that lexical experts extract information from multiple words during a single fixation. Again, this finding was paralleled in Bélanger, Slattery, et al.’s (2012) skilled deaf readers who were more disrupted at small window sizes than skilled hearing readers. The parallels between the data patterns of the lexical experts in the present study and the skilled deaf readers in Bélanger, Slattery, et al. (2012) are 5 It is possible that the use of upper-case Xs outside the moving window in the two experiments may have contributed to this finding. This type of mask may have been more perceptually salient in the periphery than, for example, a random letter mask. Nevertheless, the use of upper-case X masks is consistent with the majority of the moving-window literature. THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 721 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS suggestive of similarities in reading strategy between these groups. Skilled deaf readers’ masked priming performance is sensitive to orthographic but not phonological similarity (Bélanger, Baum, & Mayberry, 2012), suggesting that skilled deaf readers may learn to map orthography directly to meaning and acquire stronger orthographic representations through print exposure (Harris & Moreno, 2004). Similarly, the masked priming data reviewed in the introduction suggest that better spellers’ precise orthographic representations rapidly activate matching words and inhibit the representations of orthographically similar words (Andrews & Hersch, 2010; Andrews & Lo, 2012). The present results demonstrate that the predictive power of spelling extends to sentence reading. The precise orthographic representations indexed by the combination of spelling and reading ability support rapid identification of fixated words, which allows effective extraction and use of parafoveal information to guide eye movements. This is consistent with Kuperman and Van Dyke’s (2011) proposal that high-quality lexical representations support “a reading strategy that targets the eyes … (to) the optimal viewing position for full form word recognition” (p. 56), rather than the “piecemeal”, sublexical processing strategy relied upon by poorer readers. Such a possibility is supported by recent evidence of individual differences in masked morphological priming, which suggest that better spelling may be associated with the development of whole-word representations for morphologically complex words (Andrews & Lo, 2013). Finally, we found no evidence that the modulation of the perceptual span by lexical processing ability extends to information to the left of fixation. Rayner et al. (2009) found decreased asymmetry in the perceptual spans of elderly readers relative to college-aged readers. It is possible that the less efficient parafoveal processing of elderly readers is due, at least in part, to visual acuity limitations in the periphery (e.g., Cerella, 1985), which may necessitate a reading strategy based on partial visual information. This type of compensatory top-down reading strategy may therefore be a specific response to the sensory declines associated with age. 722 CONCLUSION These experiments demonstrate that lexical quality, indexed by combining measures of spelling ability and reading comprehension, is associated with a larger perceptual span reflected in both greater benefits from the provision of more rightward information and greater cost of very limited parafoveal information. These results confirm that skilled readers vary in the extent to which they extract and use parafoveal information and show that individual differences in lexical expertise modulate both the when and where of eye movements during sentence reading. As well as enhancing how quickly readers access fixated words (e.g., Ashby et al., 2005), high-quality representations facilitate the extraction and use of parafoveal information to support effective saccadic planning. The results support Kuperman and Van Dyke (2011)’s claim that “the eye movement strategy that a reader adopts is … an affordance of the overall quality of his or her lexical representations” (p. 56) and generalize previous evidence (Andrews & Bond, 2009; Andrews & Hersch, 2010; Andrews & Lo, 2012, 2013; Hersch & Andrews, 2012) showing that spelling ability is a useful index of the orthographic precision that defines lexical quality to a more naturalistic sentence reading task. Original manuscript received 12 March 2013 Accepted revision received 9 June 2013 First published online 26 August 2013 REFERENCES Andrews, S. (2008). Lexical expertise and reading skill. Psychology of Learning and Motivation, 49, 249–281. Andrews, S. (2012). Individual differences in skilled visual word recognition and reading: The role of lexical quality. In J. Adelman (Ed.), Visual word recognition (pp. 151–172). Hove: Psychology Press. Andrews, S., & Bond, R. (2009). Lexical expertise and reading skill: Bottom-up and top-down processing of lexical ambiguity. Reading & Writing: An Interdisciplinary Journal, 22, 687–711. THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 LEXICAL QUALITY AND EYE MOVEMENTS Andrews, S., & Hersch, J. (2010). Lexical precision in skilled readers: Individual differences in masked neighbor priming. Journal of Experimental Psychology: General, 139, 299–318. Andrews, S., & Lo, S. (2012). Not all skilled readers have cracked the code: Individual differences in masked form priming. Journal of Experimental Psychology: Learning Memory and Cognition, 38, 152–163. Andrews, S., & Lo, S. (2013). Is morphological priming stronger for transparent than opaque words? It depends on individual differences in spelling and vocabulary. Journal of Memory and Language, 68, 279–296. Apel, J. K., Henderson, J. M., & Ferreira, F. (2012). Targeting regressions: Do readers pay attention to the left? Psychonomic Bulletin & Review, 19, 1108–1113. Ashby, J., Rayner, K., & Clifton, C. (2005). Eye movements of highly skilled and average readers: Differential effects of frequency and predictability. Quarterly Journal of Experimental Psychology, 58, 1065–1086. Ashby, J., Yang, J., Evans, K. H. C., & Rayner, K. (2012). Eye movements and the perceptual span in silent and oral reading. Attention, Perception and Psychophysics, 74, 634–640. Baayen, R. H. (2011). languageR: Data sets and functions with “Analyzing Linguistic Data: A practical introduction to statistics”, (R package version 1.4) [Computer software]. Retrieved from http://CRAN. R-project.org/package=languageR Baayen, R. H., Piepenbrock, R., & Van Rijn, H. (1993). The CELEX lexical database [CD-ROM]. Philadelphia, PA: University of Pennsylvania, Linguistic Data Consortium. Bates, D., Maechler, M., & Bolker, B. (2012). lme4: Linear mixed-effects models using S4 classes (R package version 0.999999-0) [Computer software]. Retrieved from http://CRAN.R-project.org/ packages=lme4 Bélanger, N. N., Baum, S. R., & Mayberry, R. I. (2012). Reading difficulties in adult deaf readers of French: Phonological codes, not guilty!. Scientific Studies of Reading, 16, 263–285. Bélanger, N. N., Slattery, T. J., Mayberry, R. I., & Rayner, K. (2012). Skilled deaf readers have an enhanced perceptual span in reading. Psychological Science, 23, 816–823. Brown, J. I., Fishco, V. V., & Hanna, G. (1993). Nelson– Denny Reading Test. Rolling Meadows, IL: Riverside Publishing Company. Burt, J. S., & Tate, H. (2002). Does a reading lexicon provide orthographic representations for spelling? Journal of Memory and Language, 46, 518–543. Carver, R. P. (1985). How good are some of the world’s best readers? Reading Research Quarterly, 20, 389–419. Cerella, J. (1985). Age-related decline in extrafoveal letter perception. Journal of Gerontology, 40, 727–736. Chace, K. H., Rayner, K., & Well, A. D. (2005). Eye movements and phonological parafoveal preview: Effects of reading skill. Canadian Journal of Experimental Psychology, 59, 209–217. Davis, C. J. (2005). N-Watch: A program for deriving neighborhood size and other psycholinguistic statistics. Behavior Research Methods, 37, 65–70. Dye, M. W. G., & Bavelier, D. (2010). Attentional enhancements and deficits in deaf populations: An integrative review. Restorative Neurology and Neuroscience, 28, 181–192. Dye, M. W. G., Hauser, P. C., & Bavelier, D. (2008). Visual skills and cross-modal plasticity in deaf readers: Possible implications for acquiring meaning from print. Annals of the New York Academy of Sciences, 1145, 71–82. Engbert, R., Nuthmann, A., Richter, E. M., & Kliegl, R. (2005). SWIFT: A dynamical model of saccade generation during reading. Psychological Review, 112, 777–813. Findlay, J. M., & Walker, R. (1999). How are saccades generated? Behavioral and Brain Sciences, 22, 706–713. Forster, K. I., Davis, C. J., Schoknecht, C., & Carter, R. (1987). Masked priming with graphemically related forms: Repetition or partial activation? The Quarterly Journal of Experimental Psychology, 39, 211–251. Frith, U. (1980). Unexpected spelling problems. In U. Frith (Ed.), Cognitive processes in spelling (pp. 495–515). London: Academic Press. Häikiö, T., Bertram, R., Hyönä, J., & Niemi, P. (2009). Development of the letter identity span in reading: Evidence from the eye movement moving window paradigm. Journal of Experimental Child Psychology, 102, 167–181. Harris, M., & Moreno, C. (2004). Deaf children’s use of phonological coding: Evidence from reading. Journal of Deaf Studies and Deaf Education, 9, 253–268. Henderson, J. M., & Ferreira, F. (1990). Effects of foveal processing difficulty on the perceptual span in reading: Implications for attention and eye movement THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 723 Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 VELDRE AND ANDREWS control. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 417–429. Hersch, J., & Andrews, S. (2012). Lexical quality and reading skill: Bottom-up and top-down contributions to sentence processing. Scientific Studies of Reading, 16, 240–262. Hohenstein, S. (2011). Function remef (REMove EFfects) (Version 0.6.6) [Computer software]. Retrieved from http://read.psych.uni-potsdam.de Hyönä, J., Lorch, R. F., & Kaakinen, J. K. (2002). Individual differences in reading to summarize expository text: Evidence from eye fixation patterns. Journal of Educational Psychology, 94, 44–55. Inhoff, A. W., Pollatsek, A., Posner, M. I., & Rayner, K. (1989). Covert attention and eye movements during reading. The Quarterly Journal of Experimental Psychology, 41A, 63–89. Jared, D., Levy, B. A., & Rayner, K. (1999). The role of phonology in the activation of word meanings during reading: Evidence from proofreading and eye movements. Journal of Experimental Psychology: General, 128, 219–264. Jordan, T., McGowan, V. A., & Paterson, K. B. (2013). What’s left? An eye movement study of the influence of interword spaces to the left of fixation during reading. Psychonomic Bulletin & Review, 20, 551–557. Kuperman, V., & Van Dyke, J. A. (2011). Effects of individual differences in verbal skills on eye-movement patterns during sentence reading. Journal of Memory and Language, 65, 42–73. Malmquist, E. (1958). Factors related to reading disabilities in the first grade of the elementary school. Stockholm: Malmquist & Wiskell. McConkie, G. W., & Rayner, K. (1975). The span of the effective stimulus during a fixation in reading. Perception & Psychophysics, 17, 578–586. Perfetti, C. A. (1992). The representation problem in reading acquisition. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 145–174). Hillsdale, NJ: Lawrence Erlbaum Associates. Perfetti, C. A. (2007). Reading ability: Lexical quality to comprehension. Scientific Studies of Reading, 11, 357–383. Perry, J. R., Lupker, S. J., & Davis, C. J. (2008). An evaluation of the interactive-activation model using masked partial-word priming. Language and Cognitive Processes, 23, 36–68. Pollatsek, A., Bolozky, S., Well, A. D., & Rayner, K. (1981). Asymmetries in the perceptual span for Israeli readers. Brain and Language, 14, 174–180. 724 Pollatsek, A., & Rayner, K. (1982). Eye movement control in reading: The role of word boundaries. Journal of Experimental Psychology: Human Perception and Performance, 8, 817–833. R Development Core Team (2011). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, Retrieved from http://www.R-project.org/ Radach, R., & Kennedy, A. (2013). Eye movements in reading: Some theoretical context. Quarterly Journal of Experimental Psychology, 66, 429–452. Rayner, K. (1986). Eye movements and the perceptual span in beginning and skilled readers. Journal of Experimental Child Psychology, 41, 211–236. Rayner, K. (2009). Eye movements and attention in reading, scene perception, and visual search. The Quarterly Journal of Experimental Psychology, 62, 1457–1506. Rayner, K., Castelhano, M. S., & Yang, J. (2009). Eye movements and the perceptual span in older and younger readers. Psychology and Aging, 24, 755–760. Rayner, K., Fischer, M. H., & Pollatsek, A. (1998). Unspaced text interferes with both word identification and eye movement control. Vision Research, 38, 1129–1144. Rayner, K., Slattery, T. J., & Bélanger, N. N. (2010). Eye movements, the perceptual span, and reading speed. Psychonomic Bulletin & Review, 17, 834–839. Rayner, K., Well, A. D., & Pollatsek, A. (1980). Asymmetry of the effective visual field in reading. Perception & Psychophysics, 27, 537–544. Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-Z Reader model of eye-movement control in reading: Comparisons to other models. Behavioral and Brain Sciences, 26, 445–526. Reichle, E. D., Warren, T., & McConnell, K. (2009). Using E-Z Reader to model the effects of higher level language processing on eye movements during reading. Psychonomic Bulletin and Review, 16, 1–21. Schad, D. J., & Engbert, R. (2012). The zoom lens of attention: Simulating shuffled versus normal text reading using the SWIFT model. Visual Cognition, 20(4–5), 391–421. Schotter, E. R., Angele, B., & Rayner, K. (2012). Parafoveal processing in reading. Attention, Perception and Psychophysics, 74, 5–35. Slattery, T. J., Pollatsek, A., & Rayner, K. (2007). The effect of the frequencies of three consecutive content THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) LEXICAL QUALITY AND EYE MOVEMENTS Wickham, H. (2009). ggplot2: Elegant graphics for data analysis. New York, NY: Springer. Wolf, M., & Bowers, P. G. (1999). The doubledeficit hypothesis for the developmental dyslexias. Journal of Educational Psychology, 91, 415–438. Wotschack, C., & Kliegl, R. (2013). Reading strategy modulates parafoveal-on-foveal effects in sentence reading. The Quarterly Journal of Experimental Psychology, 66, 548–562. Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 words on eye movements during reading. Memory and Cognition, 35, 1283–1292. Stanovich, K. E. (2000). Progress in understanding reading: Scientific foundations and new frontiers. New York, NY: Guilford Press. Underwood, N. R., & Zola, D. (1986). The span of letter recognition of good and poor readers. Reading Research Quarterly, 21, 6–19. Venables, W. N., & Ripley, B. D. (2002). Modern applied statistics with S (4th ed.). New York, NY: Springer. THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) 725 VELDRE AND ANDREWS THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 726 APPENDIX A Contrasts tested in linear mixed effects models in Experiment 1 Reading rate (Log) average fixation duration Forward saccade length b SE t b SE t b SE 182.8 7.6 24.1 5.4E+0 1.7E–2 321.2 7.5 0.2 Reading (R) Spelling (S) Reading × Spelling (X) 20.3 15.9 6.8 8.3 8.0 8.2 2.4 2.0 0.8 5.8E–5 –2.4E–3 –1.6E–2 2.0E–2 1.9E–2 1.9E–2 0.0 –0.1 –0.8 0.1 0.3 0.0 0.2 0.2 0.2 Difficulty Spacing Window(3 vs. 7) Window (7 vs. 11) Window (11 vs. 15) Window (15 vs. FL) 11.4 –1.6 65.0 24.2 2.6 9.9 0.7 0.7 2.1 2.1 2.1 2.5 15.8S –2.1 30.7R 11.8S 1.2 3.9R –1.5E–2 5.7E–3 –1.6E–1 –2.2E–2 9.9E–3 –6.3E–2 1.9E–3 1.9E–3 5.6E–3 5.4E–3 5.3E–2 6.7E–3 0.1 –0.2 0.9 0.9 0.5 0.1 0.0 0.0 0.1 0.1 0.1 0.1 Difficulty × Spacing Difficulty × Window (3 vs. 7) Difficulty × Window (7 vs. 11) Difficulty × Window (11 vs. 15) Difficulty × Window (15 vs. FL) Spacing × Window (3 vs. 7) Spacing × Window (7 vs. 11) Spacing × Window (11 vs. 15) 0.9 2.5 2.1 1.1 2.0 4.5 2.5 –0.3 0.7 2.1 2.0 2.0 2.5 2.1 2.1 2.1 1.3 1.2 1.0 0.5X 0.8 2.1 1.2 –0.2 –2.5E–4 –3.9E–3 –8.1E–4 1.7E–3 –2.5E–3 –4.6E–3 –4.9E–3 6.8E–3 1.8E–3 5.6E–3 5.3E–3 5.3E–3 6.7E–3 5.6E–3 5.3E–3 5.3E–3 0.0 0.0 0.0 0.0 0.1 0.1 0.1 0.2 0.0 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.4 0.3 0.3 0.2 0.9 2.0 1.9 3.5 440.6 1911.2 2228.0 21.0 43.7 47.2 NA NA NA 4.1E–4 1.0E–2 1.3E–2 2.0E–2 1.0E–1 1.1E–1 0.1 1.0 2.3 0.3 1.0 1.5 NA NA NA Model parameter (Intercept) Random effects Item Subject Residual t –8.0 2.9 –27.5R,S,X –4.2 1.9 –9.5R,S –0.1 –0.7 –0.2 0.3 –0.4 –0.8 –0.9 1.3 NA NA NA Regression count b SE t 44.0 2.4 0.2 12.9 0.6 1.8 0.3 –0.4 –0.2 0.1 0.2 0.2 0.2 –2.0 –0.9 0.5 –0.1 –0.1 –0.2 0.1 0.1 0.2 0.0 0.0 0.1 0.1 0.1 0.1 –3.8S –3.5 –2.8 1.7 1.1 1.9 0.0 –0.1 0.0 –0.0 0.0 0.0 –0.1 0.1 0.0 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.6 –1.3 0.6 –0.1 0.2 0.0 –0.7 1.0 0.1 1.3 3.1 0.3 1.1 1.8 NA NA NA 2.5 –8.2 12.4 12.6 7.2S,X 0.8 Note: Contrasts significant at p , .05 indicated in bold. Significant interactions (p , .05) with reading (R), spelling (S), and Reading × Spelling (X) indicated by superscript. Contrasts tested in linear mixed effects models in Experiment 2 (Log) average fixation duration Reading rate b SE t b SE 220.5 7.4 29.6 5.4E+0 1.7E–2 Reading (R) Spelling (S) Reading × Spelling (X) 18.7 –1.8 2.0 7.6 7.7 7.1 2.5 –0.2 0.3 –2.8E–3 –1.4E–2 –1.6E–4 1.9E–2 2.0E–2 1.8E–2 Difficulty Right (3 vs. 9) Right (9 vs. 15) Left . 3 Left . 6 12.0 76.5 10.7 10.8 –0.7 0.8 1.9 1.8 1.6 1.8 16.0S 41.3R 5.8X 6.7 –0.4 –1.4E–2 –1.1E–1 2.3E–3 –4.1E–2 –2.7E–3 1.8E–3 4.5E–3 4.5E–3 3.9E–3 4.5E–3 Difficulty × Right (3 vs. 9) Difficulty × Right (9 vs. 15) Difficulty × Left . 3 Difficulty × Left . 6 Left . 3 × Right (3 vs. 9) Left . 6 × Right (3 vs. 9) Left . 3 × Right (9 vs. 15) Left . 6 × Right (9 vs. 15) 3.0 3.6 2.8 1.2 –0.5 –2.4 –4.6 3.1 1.8 1.8 1.6 1.8 3.9 4.5 3.9 4.5 1.6 2.0 1.7 0.7 –0.1 –0.5 –1.2 0.7 –5.3E–3 3.9E–5 –1.0E–3 –2.5E–3 6.4E–4 –8.7E–3 –5.0E–3 1.1E–2 4.5E–3 4.5E–3 3.9E–3 4.5E–3 9.6E–3 1.1E–2 9.5E–3 1.1E–2 487.2 1902.0 2799.0 22.1 43.6 52.9 NA NA NA 3.6E–4 1.1E–2 1.6E–2 1.9E–2 1.1E–1 1.3E–1 Model parameter (Intercept) Random effects Item Subject Residual t Forward saccade length b SE t 311.1 6.6 0.2 –0.2 –0.7 –0.0 0.2 0.0 0.1 0.2 0.2 0.2 0.0 1.5 0.8 0.1 0.0 0.0 0.0 0.0 0.0 0.0 –1.2 0.0 –0.3 –0.6 0.1 –0.8 –0.5 1.0 0.1 0.0 0.0 –0.0 –0.1 –0.1 0.1 0.1 0.0 0.0 0.0 0.0 0.1 0.1 0.1 0.1 NA NA NA 0.1 1.3 1.4 0.2 1.1 1.2 –7.8S –25.3R,S 0.5 –10.5 –0.6 Regression count b SE t 35.8 2.5 0.2 15.7 0.8 0.1 0.3 –0.3 –0.0 0.0 0.2 0.2 0.2 –1.7 –0.0 0.2 1.8 34.4R,S,X 18.0X 3.3 0.2 –0.2 –0.3 0.3 –0.2 0.0 0.0 0.1 0.1 0.1 0.1 –7.5 –3.8R,S 4.2 –3.5 0.6 1.3 0.4 0.3 –0.3 –1.1 –0.9 1.0 1.4 0.1 –0.1 0.0 –0.1 0.1 –0.2 0.2 0.2 0.1 0.1 0.1 0.1 0.1 0.2 0.1 0.2 1.6 –1.3 0.2 –1.1 0.7 –1.0 1.1 1.0 0.1 0.9 2.9 0.3 1.0 1.7 NA NA NA NA NA NA Note: Contrasts significant at p , .05 indicated in bold. Significant interactions (p , .05) with reading (R), spelling (S), and Reading × Spelling (X) indicated by superscript. 727 LEXICAL QUALITY AND EYE MOVEMENTS THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014, 67 (4) Downloaded by [University of Massachusetts, Amherst] at 07:59 29 September 2014 APPENDIX B