veldre2013

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