Vision Research 165 (2019) 152–161

Contents lists available at ScienceDirect

Vision Research
journal homepage: www.elsevier.com/locate/visres

Visual attention modulates reading acquisition T

a,⁎ b a
Sylviane Valdois , Jean-Luc Roulin , Marie Line Bosse
a
Univ. Grenoble Alpes, CNRS UMR5105, LPNC, 38000 Grenoble, France
b
Univ. Savoie Mont Blanc, CNRS, LPNC, 73000 Chambéry, France

A R T I C LE I N FO A B S T R A C T

Keywords: The processing of letters within strings is challenging for beginning readers. Letter identification is affected by
Reading acquisition visual similarity, loss of information with eccentricity and interference from nearby letters. In contrast, visual
Visual attention attention enhances letter identification. We here explored whether visual attention resources for multi-element
Longitudinal study processing, as measured through tasks of visual attention span prior to literacy instruction, predicted reading
Multi-element parallel processing
fluency performance one year later. One hundred and twenty-four mainstream children were assessed in kin-
Prereaders
Models of reading
dergarten on their visual attention span abilities, phonological awareness, letter-name knowledge, early literacy
knowledge, verbal short-term memory and non-verbal IQ. The participants’ reading performance was measured
at the end of grade 1 using tasks of irregular word, pseudo-word and text reading. Results from regression
analyses showed that kindergarteners’ VA span predicted reading fluency for text, irregular words and pseudo-
words one year later, after controlling for age, non-verbal IQ, phonological skills, letter name knowledge and
early literacy skills. Path analyses carried out to estimate the differential contribution of VA span to the different
reading skills revealed a stronger contribution for pseudo-word reading than irregular word or text reading at the
end of Grade 1. These results suggest that pre-reading visual attention resources contribute to later reading
fluency, whatever the reading subskills and whatever the reading context (words in isolation or in sentences),
with higher involvement to pseudo-word reading. We propose a new conceptual model of the role of visual
attention in reading acquisition and argue that many aspects of the models are already supported by available
findings.

1. Introduction position of its constituent letters. Letters are the building blocks of
words, so that single letter identification is a first step towards word
Reading is both a visual and a linguistic task. However, most studies processing. Some letters share many of their visual features (R and P for
on reading acquisition have focused on the linguistic dimensions of example), thus increasing the probability of misidentification and letter
reading, emphasizing the importance of letter-sound mapping and prior confusion (Pelli, Burns, Farell, & Moore-Page, 2006); some differ by
phonological processing skills (Castles & Coltheart, 2004; Melby- visual properties that are not relevant for object identification. The
Lervåg, Lyster & Hulme, 2012). Some studies have even suggested that visual system of pre-readers and illiterates is tuned to identify visual
learning to read is not a visual skill (Goswami, 2015) and that visual objects regardless of their orientation (Kolinsky & Fernandes, 2014).
processing skills only have moderate impact on reading acquisition, if The identity of an animal, a person or an object is not affected by a
any (Share, 1999; Ziegler, Perry, & Zorzi, 2014). In the current paper, left–right or up-down reversal. But this does not hold true for letters.
we will focus on the visual mechanisms that are involved in reading and Differences in left–right or up-down orientation define letters that differ
how they can modulate reading acquisition. Our main hypothesis is that in identity (e.g., b and d or u and n). Reading acquisition thus requires
visual attention is critical for learning to read. We will argue that learning to discriminate letters of similar shapes but different orienta-
learning to read relies on the capacity to accurately identify the whole tions, which implies “unlearning” the general principle of mirror-in-
letters that form relevant orthographic units in the language under variance (Dehaene, Cohen, Morais & Kolinsky, 2015; Pegado et al.,
concern and that simultaneous letter identification within these units 2014). This is not straightforward. Some children show mirror-letter
depends on the amount of visual attention resources available for confusions at the beginning of literacy instruction (Dehaene et al.,
processing. 2010) and dyslexic children tend to process symmetrical letters as
In most languages, a written word is defined by the identity and identical (Lachman & Van Leeuwen, 2007). Reading acquisition thus

⁎
Corresponding author at: Université Grenoble-Alpes, LPNC, CNRS 5105, BSHM – 1251 Avenue Centrale, CS 40700, 38058, Grenoble Cedex 9, France.
E-mail address: sylviane.valdois@univ-grenoble-alpes.fr (S. Valdois).

https://doi.org/10.1016/j.visres.2019.10.011
Received 15 February 2019; Received in revised form 16 July 2019; Accepted 30 October 2019
0042-6989/ © 2019 Published by Elsevier Ltd.
S. Valdois, et al. Vision Research 165 (2019) 152–161

relies on functional specialization of the visual system through implicit defined as an attentional span of effective vision (Pollatsek, Bolozky,
visual perceptual learning (Gilbert, Sigman, & Crist, 2001). Well, & Rayner, 1981; Rayner, 2009), the two spans differ in their es-
Reading involves visual specialization for accurate letter processing, timated size, symmetry around fixation and foveal/parafoveal limits
but normal processing of isolated letters is not enough to guarantee (for a review, Frey & Bosse, 2018) but, more importantly, the VA span is
efficient reading. Additional visual mechanisms are at play for the assumed to be a purer measure of the visual attention mechanisms in-
processing of letters within strings. The visibility of letters within words volved in reading. Indeed, the VA span –which is measured through
is modulated by both the acuity gradient and crowding (Bernard & tasks that require the processing of meaningless material, as consonant
Castet, 2019; Grainger, Dufau, & Ziegler, 2016). Acuity is maximal for strings, digit strings or strings of unknown shapes (Lobier, Zoubrinetzky
the letter under fixation but the drop-off in visual acuity as a function of & Valdois, 2012; Valdois et al., 2003, Valdois, Lassus-Sangosse &
distance from fixation makes letter identification worse with eccen- Lobier, 2012)– does not reflect any lexical processing or linguistic skills.
tricity. Letter identification within strings is further limited by In contrast, the perceptual span that is measured in conditions of
crowding, i.e., lateral interference due to nearby letters (Bouma, 1970; meaningful text reading is affected by linguistic variables as word fre-
Whitney & Levi, 2011). Maximal interference is observed for letters quency or context effects (Rayner, 1998). The VA span mainly reflects
surrounded by a letter on each side, so that the inner letters of a word the quantity of visual attention available for processing and how at-
are more difficult to identify than the outer letters (i.e., the first and last tention distributes over the letter-string to modulate letter identity
letters; Scaltritti & Balota, 2013; Tydgat & Grainger, 2009). Crowding processing (see Ginestet, Phénix, Diard, & Valdois, 2019, for an im-
affects visual feature integration. The features of both the target letter plementation of visual attention in a model of word recognition). This
and nearby letters are combined, which may result in letter identity specificity of the VA span is supported by neuroimaging data. The su-
confusions and/or location errors (Pelli, Palomares, & Majaj, 2004; perior parietal lobules have been identified as the neural underpinnings
Whitney & Levi, 2011). Variations in crowding relates to variations in of the VA span, showing that VA span specifically relates to the dorsal
reading speed. Reading speed is limited by crowding in typical readers attentional network (Lobier, Peyrin, LeBas & Valdois, 2012b; Peyrin,
(Pelli, Tillman, Freeman, Su, Berger, & Majaj, 2007) and some dyslexic Démonet, Baciu, LeBas & Valdois, 2011; Peyrin et al., 2012; Reilhac,
children –but not all (Ziegler, Pech-Georgel, Dufau, & Grainger, 2010)– Peyrin, Démonet, & Valdois, 2013; Valdois, Lassus-Sangosse, Lallier,
show greater crowding than their peers, which affects their ability to Moreaud, & Pisella, 2019).
recognize words faster (Bouma & Legein, 1977; Atkinson, 1993; Different tasks of whole and partial report (Dubois et al., 2010;
Spinelli, De Luca, Judica, & Zoccolotti, 2002; Martelli, Di Filippo, Valdois et al., 2003; Zoubrinetzky, Bielle, & Valdois, 2014,
Spinelli, & Zoccolotti, 2009; Callens, Whitney, Tops, & Brysbaert, Zoubrinetzky, Collet, Serniclaes, Nguyen-Morel & Valdois, 2016), ca-
2013). Fortunately, the deleterious effects of crowding on letter pro- tegorization (Lobier, Zoubrinetzky et al., 2012) or visual 1-back
cessing within strings are attenuated when inter-letter spacing is in- (Lallier, Acha, & Carreiras, 2016) are relevant to measure VA span.
creased, which results in improved reading speed (Spinelli et al., 2002; Appropriate VA span tasks must meet certain standards. To ensure
Martelli et al., 2009; Zorzi et al., 2012). parallel processing, the multielement array has to be displayed for a
Thus, three factors – visual similarity between letters, the acuity short-enough time (≤200 ms) that allows deployment of attention
gradient and crowding –contribute to degrade letter identification in while avoiding useful eye movements. Inter-element spacing is in-
typically formatted words. How can we accurately identify the letters creased to avoid potential crowding effects. Even in partial report
that make words based on so degraded information on the input string? conditions, the task must force parallel processing of all the elements of
Eye movement studies provide important insights on this issue. Eye the array, so that no cue indicating the position or identity of the ele-
movements in reading are characterized by a succession of fixations and ment to be processed is provided prior to the array offset. Because we
saccades. Information about letter identity is acquired during fixations are interested in the amount of attention deployed in parallel for
and the primary function of a saccade is to move the eyes towards identity processing, identification tasks are used while avoiding loca-
another portion of the word or text for detailed letter processing (for a tion encoding tasks that more likely rely on serial processing. We must
review, see Rayner, 2009). There is now a consensus that visual at- further ensure that poor performance in multi-element processing is not
tention shifting precedes eye movement and that visual attention is just the consequence of a single element identification problem, so that
required for word identification (Besner et al., 2016; Lachter, Forster & a control task of single element processing efficiency is further pro-
Ruthruff, 2004; Rayner & Reichle, 2010; Risko, Stolz, & Besner, 2010). posed. Furthermore and because parallel processing of unknown items
Attention is allocated to one word (or a few short words, or a portion of is challenging for the visual system (Shovman & Ahissar, 2006; Ziegler
a word) at a time; it acts as a filter that enhances letter identification et al., 2010; Collis, Kohnen, & Kinoshita, 2013), familiar visual ele-
under the attentional focus, thus limiting the detrimental effects of ments are more likely to be used in identification tasks while categor-
acuity and crowding (Carrasco, 2011; Strasburger, 2005; Yeshurun & ization tasks that do not require precise target identification are more
Rashal, 2010). In compensating for acuity limitations and adjacent appropriate when using unfamiliar visual items.
letter interference, visual attention allows accurate parallel processing Strong relationships have been reported between VA span and
of letter identities under the attentional focus (Bundesen, 1990; reading in typical and dyslexic readers. Typical children with higher VA
Carrasco, 2011). As a result, the number of letters that can be accu- span make lesser fixations in text reading, thus processing more letters
rately processed in parallel within a letter string (i.e., a sublexical or- per fixation (Prado, Dubois, & Valdois, 2007). They are less prone to
thographic unit or a word) should vary depending on the amount of show length effects in word reading (van den Boer, de Jong, &
attention resources available for processing. Assuming that reading Haentjens-van Meeteren, 2013), show higher performance in irregular
speed is higher when processing longer orthographic units – thus, more word reading accuracy (Bosse & Valdois, 2009) and read words,
letters in parallel–, the amount of attention available for parallel pro- pseudo-words and texts more fluently (Lallier, Valdois, Lassus-
cessing should modulate how fast words (or more generally letter Sangosse, Prado, & Kandel, 2014; Lobier et al., 2013). Higher VA span
strings) can be processed. capacities associated to faster pseudo-word reading have also been re-
The amount of visual attention available for the parallel processing ported in young adults (Antzaka et al., 2017). Evidence from develop-
of letters within strings can be estimated through the measure of visual mental dyslexia further supports the VA span-reading relationship. A
attention (VA) span (Bogon, Finke & Stenneken, 2014; Lobier, Dubois, VA span deficit characterizes a subset of the dyslexic population (Bosse
& Valdois, 2013). More generally, this span corresponds to the number et al., 2007; Germano, Reilhac, Capellini, & Valdois, 2014;
of distinct visual elements that can be processed simultaneously in a Zoubrinetzky et al., 2014, 2016). The VA span deficit contributes to
multi-element configuration (Bosse, Tainturier, & Valdois, 2007). The poor reading accuracy and slow reading speed in this population,
VA span differs from the perceptual span. Although the latter was also whatever the type of items to be read (words, irregular words and

153
S. Valdois, et al. Vision Research 165 (2019) 152–161

pseudo-words) and regardless of the child’s phonological skills. Greater number: 0820589).
length effects in word reading are also reported in dyslexic children Note that in France, formal literacy instruction only starts from
with a VA span deficit (Valdois et al., 2003, 2011; van den Boer et al., Grade 1. In kindergarten, teachers mainly focus on oral language skills
2013). The overall findings offer a quite coherent pattern. Visual at- to develop oral vocabulary and communication abilities. Letter-name is
tention is deployed over the letter string to enhance letter identifica- systematically taught including letter recognition and discriminability.
tion. Higher visual attention capacity allows allocating enough atten- Letter-sound knowledge is often taught in Grade 1 only. With respect to
tion to a higher number of letters, thus increasing probability of phonological skills, teachers try to improve both the ability of pupils to
accurate parallel processing of the whole letters that make longer or- manipulate syllables and rhymes and their sensitivity to alliterations
thographic units. Parallel processing of longer orthographic units im- but they do not systematically train phoneme awareness, which is the
proves reading fluency and reduces length effects with potential addi- focus of Grade 1 (Ziegler & Goswami, 2005).
tional effects on reading accuracy, at least for the languages with long
graphemes and irregular words. Accurate reading of irregular words 2.2. Material and procedure
involves processing the word letter-string as a whole, which requires
allocating enough attention to all of the word letters simultaneously. 2.2.1. Assessment in kindergarten
Thus, irregular word reading should be more specifically sensitive to Visual Attention Span. Stimuli: Random 4-digit strings (e.g., 4 2 6 1)
the amount of visual attention available for processing; this is sup- were built up from 8 digits (1, 2, 3, 4, 5, 6, 7 and 8). Digits were pre-
ported by consistent reports of poor irregular word reading accuracy in sented in black on a white background (Arial, 7 mm high). The strings
individuals with poor VA span (Dubois et al., 2010; Valdois et al., 2003, contained no repeated digits. To avoid crowding effects, the inter-digit
2011). In a similar way, more attention is recruited for accurate iden- distance was increased (1 cm center-to-center). The array subtended an
tification of all the letters that make longer graphemic units, yielding angle of approximately 4.2° at a viewing distance of 50 cms.
grapheme parsing errors in dyslexic children with poor VA span skills Sixteen digit-strings were displayed in the whole report condition.
(Zoubrinetzky et al., 2014). Each digit appeared twice in each position. Thirty-two random 4-digit
Although a relationship between VA span and reading performance strings were presented in the partial report condition. Each digit oc-
has been reported from the end of first grade (Bosse & Valdois, 2009), curred four times in each position.
whether VA span develops with printed word exposure as a con- Procedure: At the start of each trial, a central fixation point was
sequence of reading acquisition (Goswami, 2015; see however Lobier & presented for 1000 ms followed by a blank screen for 50 ms. The digit
Valdois, 2015) or reflects visual attention capacity prior to formal lit- string was then presented centred on fixation for 200 ms. In the whole
eracy instruction remains an open issue. We here address this issue report condition, a white screen was displayed at the offset of the digit
through the longitudinal follow-up of a cohort of typical children from string, and participants were asked to report verbally all the digits they
kindergarten to first grade. Four digit-strings were used to assess pre- had identified (max = 64). The number of accurately identified digits
readers’ VA span through tasks of global and partial report. A variety of was scored independently of their position. For example, the target
cognitive skills known to relate with reading acquisition were further string 4 9 2 1 scored four points either reported as 4 9 2 1 or 9 1 4 2,
assessed in kindergarten, namely phonological skills (Bradley & Bryant, while reporting 4 9 1 only scored 3 points. In partial report, a vertical
1983), verbal short-term memory (de Jong and van der Leij, 2002), bar indicating the digit to be reported was presented for 50 ms, 1.1°
letter-name (Foulin, 2005; Georgiou, Torppa, Manolitsis, Lyytinen, & below the target digit, at the offset of the digit string. Each digit was
Parilla, 2012) and letter-sound knowledge (Caravolas et al., 2012; used as target once in each position. Participants had to report the cued
Hulme, Bowyer-Crane, Carroll, Duff, & Snowling, 2012). Following digit only (1 point per target digit, max = 32). The experimental trials
Castles and Coltheart (2004)’s recommendation, we further controlled were preceded by five training trials for which participants received
for early literacy knowledge to exclude any potential influence of early feedback. No feedback was given during the test trials. The participant’s
reading skills on VA span performance in kindergarten. Their reading responses were recorded by the experimenter on the numerical key-
skills were assessed one year later at the end of first grade through tasks board.
of irregular word, pseudo-word and text reading fluency. Regression It is worth noting that the partial report condition as the global
and path analyses were used to explore whether VA span in kinder- report condition requires parallel processing of the entire digit string
garten contributed significantly and independently to reading speed since the position of the target to be reported is unknown and cannot be
one year later. guessed at the time the string is displayed. Furthermore, as in global
report, the partial report condition mainly taps identity processing
2. Material and methods skills. Indeed, due to the appearance of the vertical bar immediately at
the offset of the digit string, the digit to be reported is perceived as
2.1. Participants being physically present above the vertical bar. Partial report perfor-
mance thus reflects how well the target is identified without requiring
One hundred and twenty-six children (56 males) were recruited any encoding of its relative position within the string.
from eight local Grenoble public schools. All children had normal Phonological awareness. Three tasks of rhyme judgement, syllable
hearing and normal or corrected-to-normal vision. First assessment took deletion and syllable reversal were administered to pre-readers. In the
place prior to formal literacy instruction (at T1) when the participants rhyme judgement task, 10 pairs of spoken words were presented and the
were 5 years 10 months (SD = 3.3 months) old. The second assessment child had to decide whether the words in each pair rhymed or not. The
(at T2) was carried out one year later at the end of Grade 1 (mean syllable deletion task was taken from the ODEDYS battery (Jacquier-
CA = 6:10 years, SD = 3.4 months). All participants were French Roux, Valdois, & Zorman, 2002). Children had to delete the first,
native speakers who attended school regularly. Their mean reading age median or final syllable of bi- and tri-syllabic spoken words (N = 12)
at T2 was 7:01 years (SD = 7.9). Their non-verbal IQ corresponded to a and pronounce the resulting pseudo-word. The syllable reversal task of
mean percentile of 56 (SD = 27). We ensured that all the participants the BELEC battery (Mousty, Leybaert, Alegria, Content, & Morais, 1994)
were able to name all the digits in kindergarten, which was required to required switching the syllables of 10 spoken bi-syllabic words or
assess their visual attention span abilities. The parents/legal tutors of pseudo-words (e.g., /doka/ →/kado/).
the pupils gave informed written consent for participation of their child Letter-name and letter-sound knowledge. Sixteen upper-case letters (A,
to the study. Ethics approval for the study was granted by the local F, R, S, M, D, L, B, J, I, N, V, O, P, T, E) were presented in a 4X4 table
Ethic Committee of the Grenoble-Alpes University and by the (Arial, 48). In two separate sessions, the experimenter asked the child to
“Commission Nationale de l’Informatique et des Libertés” (CNIL report the name or the sound of each letter in turn. Accurate responses

154
S. Valdois, et al. Vision Research 165 (2019) 152–161

Table 1
Mean, standard deviation (SD) and range (min–max) of the children’s skills and chronological age in kindergarten (T1) and at the end
of Grade 1 (T2).
Assessment in Kindergarten (T1; N = 126) Mean (SD) Range (min–max)

Chronological age (in months) 70.65 (3.34) 65–76

Visual Attention Span
Whole report performance (raw score, max = 64) 38.17 (8.11) 19–58
Partial report performance (raw score, max = 32) 18.17 (6.90) 2–30
Composite Visual attention span (%) 58.20 (15.09) 25–89
Phonological skills
Rhyme judgement (raw score, max = 10) 8.21 (1.76) 2–10
Syllable deletion (raw score, max = 12) 7.40 (3.35) 0–12
Syllable reversal (raw score, max = 10) 5.97 (3.47) 0–10
Composite phonological skills (%) 67.82 (22.36) 20–100
Letter name knowledge (raw score, max = 16) 12.78 (4.15) 1–16
Letter sound knowledge (raw score, max = 16) 7.33 (5.65) 0–16
Early word identification (raw score, max = 12) 3.87 (3.09) 0–12
Verbal short-term memory (raw score, max = 10) 6.95 (1.24) 4–10
Assessment in Grade 1 (T2; N = 126)
Chronological age (in months) 82.58 (3.40) 77–88
Raven Matrices (raw score, max = 36) 26.36 (5.49) 13–36
Text reading fluency (wpm) 40.79 (25.57) 2.5–113
Pseudo-word reading fluency (wpm) 14.97 (12.63) 0–63.5
Irregular word reading fluency (wpm) 12.11 (14.15) 0–76

were summed to derive scores of letter-name and letter-sound knowl- computed after control for age and IQ, with the main purpose to
edge. Letter name was proposed first to half participants. identify the T1 cognitive skills that related with reading knowledge at
Early word identification. Children were asked to read aloud 12 T2.
French words: le, la, une, ou, ri, lu, son, ami, école, papa, maman, noël Regressions and path analyses were then performed, focussing on
(respectively: the (male), the (female), a (female), or, laughed, read, the direct links between early skills at T1 (predictive factors) and
his, friend, school, daddy, mummy, Christmas). All were short frequent reading performance at T2, considering text reading, pseudo-word
words that usually occurred on posters in classrooms or in children’s reading and irregular word reading fluency as the outcome variables.
picture books (mean frequency in first grade books = 8006 per one All of the relevant variables were residualized for Age and IQ to ex-
million, from MANULEX: Lété, Sprenger-Charolles, & Colé, 2004). amine Age- and IQ-independent relationships between the variables.
Words were written (black upper case, Arial 28) in two columns on a The residual values were then used in all subsequent analyses (i.e.,
white sheet. Performance on the task was the number of words accu- partial correlations, regressions and path analyses). All the models were
rately named. tested using the R package Lavaan (Rosseel, 2012) and MPLUS (Version
Verbal short-term memory. The participants were administered the 7.2). Mahalanobis distance (D2) was used to detect multivariate out-
digit span task (forward and backward recall) of the WISC IV. A com- liers. Two participants with probability lower than 0.001 for D2 were
posite short-term memory score was calculated as the sum of the for- not included in the path analyses. Maximum likelihood parameter es-
ward and backward spans. timates with robust standard errors and a mean-adjusted chi-square test
statistic were used to account for the fact that some measures were not
normally distributed.
2.2.2. Grade 1 assessment
Fluid Intelligence. We administered the Raven’s colour matrices,
PM47 (Raven, Raven, & Court, 1998), to assess fluid intelligence. The 3. Results
raw score (max = 36) was used as a control variable in the analyses.
Text reading fluency. Children were asked to read a text aloud for 3.1. Descriptive statistics
2 min. The text from a child book for beginning readers consisted of 278
words (18 lines) printed (upper and lower case, times 14) in black on a Performance of the 126 participants in kindergarten (T1) and at
white sheet. An illustration of the story appeared below the text. Text Grade 1 (T2) is reported in Table 1. Theoretically-driven composite
reading performance was the number of words accurately read per scores of VA span and phonological skills were computed as the mean
minute. percentage of performance on the two tasks of global and partial report
Pseudo-word and irregular word reading. Two lists of 20 pseudo-words for the former and from the three tasks of rhyme judgement, syllable
(PWs) and 20 high frequency irregular words (mean frequency = 126 deletion and syllable reversal for the latter. Table 1 shows the absence
per million) from the ODEDYS battery (Jacquier-Roux et al., 2002) of either ceiling or floor effects on any of the measures. However as
were administered. The PWs were legal and pronounceable 1-to-3 syl- expected, the two tasks of letter-sound knowledge (median = 7) and
lable PWs. Words and PWs were matched in length (mean letter length: early word identification (median = 3) were rather difficult for kin-
5.75). They were presented by blocks printed (lower case, times 14) in dergarteners. Twenty percent children were unable to sound out any
column in black on a white sheet. Children were asked to read aloud letter, 13% did not read any word.
each item as quickly and accurately as possible. Reading fluency was
measured as the number of words or PWs accurately read per minute.
3.2. Correlation analyses

2.3. Design and analysis Table 2 provides correlations between performance on the tasks
measured in kindergarten (VA span, phonological awareness, letter-
We first provide descriptive information for each of the adminis- name and letter-sound knowledge, early word identification and verbal
tered tasks as well as correlations between all the variables in kinder- short-term memory) and correlations between the tasks administered in
garten and Grade 1 and across years. Partial correlations were then kindergarten and at Grade 1. Age correlated with performance in

155
S. Valdois, et al. Vision Research 165 (2019) 152–161

Table 2
Correlations (above the diagonal) and partial correlations (after control of age and IQ effects; below the diagonal) between predictive variables in kindergarten (T1)
and, text, pseudo-word (PW) and irregular word (IW) reading performance at Grade 1 (T2).
2 3 4 5 6 7 8 9 10 11

1. Age 0.14 0.35 0.40 0.09 0.25 0.19 0.20 0.19 0.14 0.11
2. Raven Matrices __ 0.37 0.52 0.14 0.29 0.30 0.37 0.25 0.29 0.23
3. T1 VA span __ __ 0.57 0.13 0.28 0.37 0.29 0.38 0.41 0.39
4. T1 Phonological awareness __ 0.41 __ 0.21 0.43 0.45 0.51 0.32 0.33 0.30
5. T1 Letter name knowledge __ 0.07 0.15 __ 0.56 0.50 0.10 0.34 0.28 0.30
6. T1 Letter sound knowledge __ 0.13 0.29 0.55 __ 0.70 0.27 0.29 0.33 0.35
7. T1 Early word identification __ 0.25 0.33 0.48 0.66 __ 0.25 0.50 0.53 58
8. T1 Verbal STM __ 0.13 0.37 0.04 0.15 0.14 __ 0.26 0.29 0.28
9. T2 Text reading fluency __ 0.29 0.18 0.31 0.21 0.44 0.17 __ 0.89 0.84
10. T2 PW reading fluency __ 0.33 0.20 0.25 0.25 0.48 0.19 0.88 __ 0.87
11. T2 IW reading fluency __ 0.32 0.20 0.27 0.29 0.55 0.21 0.83 0.86 __

In bold, significant correlations after a bonferroni correction (p < .0009)

phonological awareness and visual attention span in kindergarten. Non- variance respectively. The contribution of early literacy knowledge was
verbal IQ significantly related to the same two cognitive subskills but near significance for text reading fluency (1.7% of explained variance,
further to early word identification and verbal short-term memory. p = .09). Letter name knowledge was a significant predictor of text
Partial correlations controlling for age and IQ were thus computed. reading fluency only, with a unique contribution of 2.8% of explained
Results are presented on Table 2, below the diagonal. variance. Phonological awareness accounted for no additional amount
Phonological awareness, letter-name and letter-sound knowledge of unique variance whatever the reading subskills. On the contrary, VA
correlated with early word identification in kindergarten. The correla- span was a significant predictor of all the measures of reading fluency at
tion between letter-sound knowledge and early word identification was the end of Grade 1, with a unique contribution of 4.7, 6.3 and 4.6
very high (close to 0.70) suggesting that letter-sound knowledge, as a percent of explained variance, respectively for text, pseudo-word and
particular case of grapheme-phoneme mapping, mainly reflected early irregular word reading.
literacy skills (Hulme et al., 2012). A composite variable was thus A path diagram was created in which all three reading fluency
computed as mean performance of the two tasks, which resulted in a measures at grade 1 were included as separate endogenous variables
normally distributed variable. This score was used as an indicator of the with separate paths from all kindergarten predictors (see supplemen-
pre-readers’ literacy knowledge (“Early literacy knowledge” variable tary materials S1 and S2 for models including global report and partial
hereafter) in subsequent analyses. As expected, early verbal short-term report performance as separate predictors). After first fitting the satu-
memory correlated with phonological awareness in kindergarten but rated model, non-significant paths were removed step-by-step, until a
with none of the reading variables. More unexpectedly, phonological simplified model was obtained in which all remaining paths and cov-
awareness was found to correlate with VA span in kindergarten. ariances were statistically significant. The resulting simplified model,
At Grade 1, all the reading variables were highly correlated. illustrated in Fig. 1, fits the data very well [χ2(5, N = 124) = 1.013,
Further, significant correlations were seen between performance in p = .961, comparative fit index (CFI) = 1, root mean square error of
Kindergarten and Grade 1. Early word identification significantly cor- approximation (RMSEA) = 0 (90% confidence interval CI = 0–0,
related with all the reading measures at T2 but none of the correlations probability RMSEA < 0.05 = 0.98), Standardized Root Mean Square
between phonological awareness or letter knowledge in kindergarten Residual (SRMR) = 0.015].
and reading performance at Grade 1 were significant, except for letter Next, we examined whether the predictive power of VA span dif-
name knowledge that significantly correlated with text reading fluency. fered for the different reading subskills. For this purpose, we computed
Critically, VA span abilities in kindergarten correlated with irregular a new model in which the three path weights originating from VA span
word and pseudo-word reading fluency at the end of Grade 1. The re- were constrained to be the same. Then, the two models were compared
lationship was close to significance (r = 0.29, p = .0011; threshold of with a χ2 difference test adjusted with a Satorra-Bentler scaling cor-
significance at 0.0009 after Bonferroni correction) for text reading. rection. The Satorra-Bentler scaled χ2 difference test was significant (Δ
χ2 [6] = 32.12p < .001), suggesting that the VA span predictive
3.3. Regression and path analyses power differed for the three reading subskills. Results showed that VA
span was a stronger predictor of pseudo-word reading fluency than of
We used regression analyses to explore the contribution of basic text or irregular word reading fluency.
cognitive skills in kindergarten to reading performance one year later.
Verbal short-term memory was not included in the models as it corre- 4. Discussion
lated with neither early literacy knowledge nor with any of the reading
subskills at the end of Grade 1. Three regression analyses were com- The follow-up of a cohort of pupils from kindergarten to first grade
puted, one for each of the reading subskills, namely text reading, irre- shows that children who identified more digits in parallel in kinder-
gular word reading and pseudo-word reading fluency. The measures of garten showed higher reading fluency one year later, regardless of their
letter name knowledge, phonological awareness, VA span and the au- phonological skills, early literacy knowledge or letter-name knowledge
toregressive effect of early literacy skill were implemented as potential and after control for potential effects of age or IQ. As previously re-
predictors of reading performance at T2. Results are presented in ported (Caravolas et al., 2012; Castles & Coltheart, 2004; Shapiro,
Table 3 for the three reading measures (regression analyses computed Carroll, & Solity, 2013), phonological awareness and letter name
on raw scores while further including Age and IQ as regressors, are knowledge appear as early predictors of later reading skills but their
presented in Table S1, supplementary material). relationship to reading performance at the end of grade 1 was found
The total model R2 values ranged from 0.182 for text reading flu- indirect, via shared variance with other predictors. Several methodo-
ency to 0.216 for irregular word reading fluency (all ps < 0.001). logical differences with previous studies may explain this finding.
Early literacy skill was a significant predictor of pseudo-word and ir- Phonological skills in kindergarten were assessed through tasks re-
regular word reading fluency, accounting for 5.4% and 6.9% of unique quiring the manipulation of larger phonological units than phonemes.

156
S. Valdois, et al. Vision Research 165 (2019) 152–161

Table 3
Hierarchical regression results (standardized coefficients and z values, after control of age and IQ) of early literacy knowledge, letter-name knowledge, phonological
awareness and VA span in kindergarten as predictors of text reading fluency, pseudo-word and irregular word fluency at the end of Grade 1; the ΔR2 columns present
the incremental R2 for each significant factor when entered after all the others in the analysis, representing the unique contribution of that factor to the dependent
variable.
Dependent variables Text reading fluency Pseudo-word reading fluency Irregular word reading fluency
Equation results: R = 0.39; Adj.R2 = 0.182; R = 0.46; Adj.R2 = 0.214; R = 0.46; Adj.R2 = 0.216;
F(4,119) = 6.606 *** F(4,119) = 8.083 *** F(4,119) = 8. 201***

Factors β z ΔR2 β z ΔR2 β z ΔR2

Early literacy knowledge 0.169 1.66~ 0.017~ 0.297 2.78** 0.054** 0.337 2.81** 0.069**
Letter name 0.207 2.45* 0.028* 0.071 0.91 0.072 0.99
Phonological awareness -0.009 -0.09 -0.021 -0.26 -0.040 -0.41
Visual attention span 0.236 2.72** 0.047** 0.273 2.95** 0.063** 0.235 2.67** 0.046**

The use of more challenging phoneme awareness tasks in pre-readers Are unfamiliar visual characters more relevant than familiar char-
might reveal stronger and more direct relationships with later reading acters to estimate visual attention capacity? Previous studies have
skills (Melby-Lervåg et al., 2012). However, for the first time in the shown that single unfamiliar visual characters can be quite efficiently
current study, the contribution of phonological skills to later reading identified after only a few thousand exposures (Pelli et al., 2006).
performance was explored while controlling for both early literacy However and despite efficient single character identification, the par-
knowledge and VA span. Future studies are required to clarify whether ticipants were unable to process more than one or two of these un-
performance on more challenging phoneme awareness tasks in kin- familiar characters at a time (Pelli et al., 2006). Similar findings have
dergarten would significantly and more directly contribute to later been reported by Shovman and Ahissar (2006). They observed that
reading skills, beyond the influence of early literacy knowledge and expert readers of Hebrew who could successfully identify the central
visual-attention simultaneous processing skills. target of a Georgian (unknown alphabet) three-letter string were unable
The main contribution of the current study is to provide first evi- to accurately identify all three letters simultaneously. These results are
dence that VA span prior to formal literacy instruction predicts later quite in line with reports of poor symbol string identity processing in
reading speed performance, which would support a causal relationship. expert readers, showing close-to-chance-level identification of the
Our results indeed suggest that higher visual attention capacity in symbols outside fixation (Ziegler et al., 2010), poor signal to noise ratio
kindergarten allows children to identify more letters simultaneously, (Yeari et al., 2017) and high error rate (Collis et al., 2013). The overall
thus processing longer orthographic units and reading words and findings are consistent with a balance between object-based and space-
pseudo-words more fluently. based attention, so that increased difficulty for feature integration of
However, the relevance of VA span tasks to measure visual attention unfamiliar and complex visual shapes would proportionally reduce the
capacity when using alphanumeric material and/or verbal paradigms deployment of visual attention capacity over multiple shapes
has raised doubts (Banfi et al., 2017; Collis et al., 2013; Yeari, Isser, & (Elahipanah, Christensen, & Reingold, 2011; Khan et al., 2016).
Schiff, 2017; Ziegler et al., 2010). Performance on these tasks was in- The use of multiple familiar visual shapes (like letters or digits) in
terpreted as potentially reflecting verbal/phonological encoding skills VA span tasks was privileged to reduce the amount of visual attention to
instead of visual attention capacity, so that some researchers have be allocated to the processing of each single character within the array
privileged the use of unfamiliar visual characters (like symbols, pseudo- and rather focus on the amount of attention distributed over multiple
letters or letters of unknown alphabets) in the hope of more specifically elements for parallel processing. It is further noteworthy that the use of
tap visual attention resources. letters and digits is quite common in research on visual processing and

Fig. 1. Path diagram showing the longitudinal predictors of text reading, pseudo-word reading and irregular word reading fluency. Solid lines represent the sta-
tistically significant predictive relationships (standardized results). * p < .05; ** p < .01; ***p < .001.

157
S. Valdois, et al. Vision Research 165 (2019) 152–161

visual attention skills. Letters were used to measure the very early steps of individual letters is not fully automatized. Letters are mainly ana-
of visual processing (Sperling, 1960, 1967) and develop theories of lysed as combinations of visual features and attention is required to
visual attention (Bundesen, 1990; Laberge & Brown, 1989). Further- bind these features together into a single letter percept (Treisman &
more, tasks of whole and partial report are quite classical paradigms to Gelade, 1980; Wolfe & Cave, 1999). The identification of letters within
explore simultaneous perception of multiple visual elements (Allport, strings is further degraded depending on distance to fixation (the acuity
1968; Averbach & Sperling, 1968; Duncan et al., 1999; Habekost, 2015; gradient) and due to interference from adjacent letters (crowding ef-
Kyllingsbæk & Bundesen, 2006). Research has clearly established that fect). Additional visual attention capacity is thus required to compen-
letters within words and even within consonant strings are identified in sate for these acuity and crowding effects in order to enhance accurate
parallel (Adelman, Marquis, & Sabatos de Vito, 2010; Marzouki & letter identification (Carrasco, 2011; Risko et al., 2010). Overall, be-
Grainger, 2014). Parallel information on letter identity is available after ginning readers have to allocate a large amount of their visual attention
presentation durations of only a few tens of milliseconds, before any resources to each individual letter processing within strings, allowing to
verbal recoding can take place. For all positions within string, letter solve the binding problem in integrating relevant visual features to
identification improves when presentation duration increases but re- form the letter percept while excluding features from adjacent letters to
sponse accuracy curve shapes remain quite similar across presentation avoid misidentifications. Visual attention thus concentrates on each
durations, suggesting that visual encoding was not abandoned for letter in turn which improves discriminability and letter identification
verbal encoding at longer (yet short enough to force parallel proces- but results in letter-by-letter reading. With time and exposure to printed
sing) presentation durations (Marzouki & Grainger, 2014). With respect letters and words, visual expertise improves. Children more efficiently
to VA span tasks, Lobier, Zoubrinetzky et al. (2012) explicitly addressed discriminate the letters of the alphabet that are perceived as wholes,
the verbal encoding issue, exploring whether VA span estimation would thus requiring minimal attention (Pelli et al., 2006; Wiley, Wilson, &
differ depending on using letters, digits or unfamiliar visual characters. Rapp, 2016). More automatized individual letter identification makes
A categorization task procedure was used to avoid floor effects in the an amount of visual attention resources available for further processing.
condition requiring parallel processing of unfamiliar characters. Results Visual attention then deploys over adjacent letters, which allows
showed that children with higher performance in the conventional VA learning statistical letter cooccurrences (Egly, Driver, & Rafal, 1994;
span whole letter report paradigm also showed higher performance in Zhao, Cosman, Vatterott, Gupta, & Vecera, 2014). Attention acts as a
the categorization tasks regardless of character familiarity. In the same glue to encode multi-letter units, like bigrams and trigrams (Dehaene,
way, dyslexic children with a VA span deficit showed poor performance Cohen, Sigman, & Vinckier, 2005), or multi-letter graphemes. Attention
in the non-verbal categorization task and the deficit was of similar thus contributes to the precise and timely encoding of longer and longer
amplitude for the verbal (familiar) and non-verbal (unfamiliar) cate- orthographic units from bigrams and graphemes to syllables, mor-
gories. The available behavioural findings thus suggest that VA span phemes and words. The multi-letter units successfully encoded as per-
performance is visually-driven whether using familiar letters or digits ceptual units are stored in long-term memory (Lachter, Forster, &
or unfamiliar character strings. Ruthruff, 2004). The model thus predicts that children would initially
To definitely establish that VA span tasks tap visual attention pro- show an exaggerated length effect due to letter-by-letter reading but
cessing skills and discard any involvement of phonological/language that this effect would gradually decrease due to the processing of longer
skills, neuroimaging studies were carried out. The superior parietal and longer units as wholes. More specifically, variations in the length
lobules bilaterally were consistently identified as the neural under- effect are expected to depend on the child visual attention capacity.
pinnings of VA span, thus brain regions that are neither linguistic nor Although further research is needed to address this issue more in depth,
phonological but belong to the dorsal attention network (Peyrin et al., some available findings already support this prediction. van den Boer
2011, 2012; Reilhac et al., 2013; Lobier, Peyrin et al. (2012); Lobier, et al. (2013) showed that VA span is a unique predictor of individual
Peyrin, Pichat, Le Bas, & Valdois, 2014; Valdois et al., 2014; Valdois, differences in length effect in second grade children. Dyslexic children
Lassus-Sangosse, Lallier, Moreaud, & Pisella, 2019). Typical readers and young adults with a VA span deficit show an exaggerated length
who show higher performance on the whole report task, thus showing effect in reading and/or lexical decision (Juphard, Carbonnel, &
higher VA span, further show stronger activation of the superior par- Valdois, 2004; Valdois et al., 2003, 2011). Evidence from acquired
ietal lobules (SPLs) bilaterally. Dyslexic children with reduced VA span dyslexia also supports a relationship between visual attention capacity
behaviourally – as measured through tasks of whole and partial letter and the length effect in reading (Habekost, 2015). Another source of
report – show reduced activation of the SPLs (Peyrin, Démonet, Baciu, evidence comes from computational modelling showing how variations
Le Bas, & Valdois, 2011; Reilhac et al., 2013) while these same brain in the distribution of visual attention affect word length effect in lexical
regions are normally activated during multi-element parallel processing decision (Ginestet et al., 2019). As reading is speeded-up when pro-
in dyslexic individuals with a selective phonological deficit (Peyrin cessing longer units as wholes, a relationship between VA span and
et al., 2012). Moreover, as a mirror of the experimental findings reading speed is further expected. Such a relationship is supported by
showing similar VA span estimation regardless of character type, neu- data from typical readers (Bosse & Valdois, 2009; Lobier et al., 2013)
roimaging findings have shown similar involvement of the SPLs when and dyslexic children (Bosse et al., 2007; Zoubrinetzky et al., 2014).
using either familiar or unfamiliar character strings in categorization Another main result of the current study is that VA span measured
tasks (Lobier, Peyrin et al., 2012, 2014). Converging evidence from in kindergarten predicts future performance in different reading sub-
behavioural and neuroimaging research thus supports a visual attention skills (irregular word and pseudo-word reading) and in text reading.
interpretation of the results on VA span tasks, which leads to conclude This result is quite compatible with the conceptual model proposed
from the present results that children with higher visual attention ca- above, since the model postulates involvement of visual attention re-
pacity in kindergarten read more fluently one year later. sources whatever the letter-string to be processed. If we assume that a
How does visual attention capacity affect reading acquisition? We larger amount of visual attention resources allows processing longer
here propose a sketch of the role of visual attention in reading acqui- units as wholes earlier, then the processing of longer multi-letter gra-
sition that is largely inspired by the model initially proposed by Laberge phemes and longer syllables would improve pseudo-word reading speed
and Samuels (1974). For beginning readers, all printed words are un- and more generally sublexical processing. If we assume that words are
familiar letter strings. Attempting to deploy visual attention over the the largest units that can be processed as wholes then the same logic
entire word letter-string would result in diffuse attention deployment, applies and greater VA span should allow to process words by sight
thus allocating not enough attention to each letter for their accurate earlier, with a positive effect on irregular word reading speed. Our
identification. At the beginning of literacy instruction, children have results further suggest a relationship between VA span and text reading,
been familiarized with letter shapes and letter names but the processing which is quite compatible with the findings discussed above, assuming

158
S. Valdois, et al. Vision Research 165 (2019) 152–161

that more speeded reading of individual words would result in more 12, 231–232.
speeded text reading. However, text reading differs from individual Antzaka, A., Lallier, M., Meyer, S., Diard, J., Carreiras, M., & Valdois, S. (2017).
Enhancing reading performance through action video games: The role of visual at-
word reading in requiring serial shifting of attention and eye move- tention span. Scientific Reports, 7, 14563. https://doi.org/10.1038/s41598-017-
ments, and in relying on parafoveal preview (Franceschini, Gori, 15119-9.
Ruffino, Pedrolli, & Facoetti, 2012; Ghahghaei, Linnell, Fischer, Dubey Atkinson, J. (1993). Vision in dyslexic: Letter recognition acuity, visual crowding, con-
trast sensitivity, accommodation, convergence and sightreading music. In S. F.
& Davis, 2013; Ghahghaei & Linnell, 2018; Schotter, Angele, & Rayner, Wright, & R. Groner (Eds.). Facets of dyslexia and its remediation. Studies in visual
2012). The present findings suggest that VA resources that improve the information processing (pp. 125–138). Amsterdam, The Netherlands: Elsevier Science.
ability to process several adjacent letters as wholes remain as a sig- Averbach, E., & Sperling, G. (1968). Short term storage of information in vision. In R. N.
Haber (Ed.). Contemporary theory and research in visual perception (pp. 196–211). New
nificant contributor of reading speed independently of the reading York: Holt, Rinehart & Winston.
context (words in isolation or in sentences). Banfi, C., Kemény, F., Gangl, M., Schulte-Körne, G., Moll, K., & Landerl, K. (2017). Visuo-
Although contributing to reading fluency in general, the current spatial cueing in children with differential reading and spelling profiles. PLoS ONE,
12(7), e0180358. https://doi.org/10.1371/journal.pone.0180358.
findings suggest that VA span when measured in kindergarten more
Bernard, J. B., & Castet, E. (2019). The optimal use of non-optimal letter information in
strongly contributes to pseudo-word than to irregular word reading foveal and parafoveal word recognition. Vision Research, 155, 44–61. https://doi.org/
fluency. Interpretation of this finding is not straightforward. Previous 10.1016/j.visres.2018.12.006.
studies carried out on older participants rather suggested a privileged Besner, D., Risko, E. F., Stolz, J. A., White, D., Reynolds, M., O’Malley, S., & Robidoux, S.
(2016). Varieties of Attention: Their roles in visual word identification. Current di-
relationship between VA span and irregular word reading (Bosse & rections in Psychological Science, 25(3), 162–168. https://doi.org/10.1177/
Valdois, 2009). Available findings might suggest a stronger influence of 0963721416639351.
VA span to pseudo-word reading at the beginning of reading acquisition Bogon, J., Finke, K., & Stenneken, P. (2014). TVA-based assessment of visual attentional
functions in developmental dyslexia. Frontiers in Psychology, 5, 1172. https://doi.org/
but higher contribution to irregular word reading later on. If we assume 10.3389/fpsyg.2014.01172.
that the size of VA span, that is the amount of visual attention re- Bosse, M. L., Tainturier, M. J., & Valdois, S. (2007). Developmental dyslexia: The visual
sources, increases during primary school (Bosse & Valdois, 2009) and attention span deficit hypothesis. Cognition, 104, 198–230. https://doi.org/10.1016/
j.cognition.2006.05.009.
that only a few letters can be processed simultaneously at the end of Bosse, M. L., & Valdois, S. (2009). Influence of the visual attention span on child reading
kindergarten (2.33 letters on average in the current study), it follows performance: A cross-sectional study. Journal of Research in Reading, 32, 230–253.
that beginning readers would mainly process relatively short ortho- https://doi.org/10.1111/j.1467-9817.2008.01387.x.
Bouma, H. (1970). Interaction effects in parafoveal letter recognition. Nature, 226,
graphic units simultaneously. This VA span limitation would favour 177–178.
simultaneous processing of sublexical units as compared to whole Bouma, H., & Legein, C. P. (1977). Foveal and parafoveal recognition of letters and words
words. Later on, improvement of VA span abilities would allow the by dyslexics and by average readers. Neuropsychologia, 15(1), 69–80. https://doi.org/
10.1016/0028-3932(77)90116-6.
processing of larger and larger units which will enhance simultaneous
Bradley, L., & Bryant, P. (1983). Categorising sounds and learning to read – a causal
processing of larger sublexical units and whole words, with potential connection. Nature, 301, 419–421.
more sustained effect on irregular words. Bundesen, C. (1990). A theory of visual attention. Psychoogical Review, 97, 523–547.
https://doi.org/10.1037/0033-295X.97.4.523.
Callens, M., Whitney, C., Tops, W., & Brysbaert, M. (2013). No deficiency in left-to-right
5. Conclusion processing of words in dyslexia but evidence for enhanced visual crowding. The
Quarterly Journal of Experimental Psychology, 66(9), 1803–1817. https://doi.org/10.
Previous research has widely documented the existence of a re- 1080/17470218.2013.766898.
Caravolas, M., Lervåg, A., Mousikou, P., Efrim, C., Litavský, M., Onochie-Quintanilla, E.,
lationship between VA span and reading performance in typical readers ... Hulme, C. (2012). Common patterns of prediction of literacy development in
and dyslexic individuals. The current study for the first time establishes different alphabetic orthographies. Psychological Science, 23, 678–686. https://doi.
through a longitudinal study that VA span prior to literacy instruction org/10.1177/0956797611434536.
Carrasco, M. (2011). Visual attention: The past 25 years. Vision Research, 51, 1484–1525.
predicts reading fluency one year later, at the end of Grade 1. These https://doi.org/10.1016/j.visres.2011.04.012.
findings bring support for a theory that reading acquisition depends Castles, A., & Coltheart, M. (2004). Is there a causal link from phonological awareness to
critically on the amount of visual attention capacity available for multi- success in learning to read? Cognition, 91(1), 77–111.
Collis, N. L., Kohnen, S., & Kinoshita, S. (2013). The role of visual spatial attention in
element simultaneous processing. We propose a new conceptual model
adult developmental dyslexia. The Quarterly Journal of Experimental Psychology, 66(2),
of the role of visual attention span in learning to read. The model opens 245–260. https://doi.org/10.1080/17470218.2012.705305.
new perspectives on reading acquisition and word-specific orthographic de Jong, P. F., & van der Leij, A. (2002). Effects of phonological abilities and linguistic
comprehension on the development of reading. Scientific Studies of Reading, 6, 51–77.
learning that would trigger a number of new studies to validate or in-
Dehaene, S., Cohen, L., Morais, J., & Kolinsky, R. (2015). Illiterate to literate: Behavioural
validate all or part of its predictions. and cerebral changes induced by reading acquisition. Nature Reviews Neuroscience.
https://doi.org/10.1038/nrn3924.
Acknowledgments and funding sources Dehaene, S., Cohen, L., Sigman, M., & Vinckier, F. (2005). The neural code for written
words: A proposal. Trends in Cognitive Science, 9, 335–341. https://doi.org/10.1016/j.
tics.2005.05.004.
This study was carried out with the support of the French govern- Dehaene, S., Nakamura, K., Jobert, A., Kuroki, C., Ogawa, S., & Cohen, L. (2010). Why do
ment within the framework of the National Research Agency (Grant children make mirror errors in reading? Neural correlates of mirror invariance in the
visual word form area. NeuroImage, 49, 1837–1848.
ANR-12-BSH2-0013 to S.V.) and the e-FRAN PIA program operated by Dubois, M., Kyllingsbaek, S., Prado, C., Musca, S. C., Peiffer, E., Lassus-Sangosse, D., &
the “Caisse des Dépôts et Consignation”, grant attributed to Sylviane Valdois, S. (2010). Fractionating the multi-character processing deficit in develop-
Valdois (FLUENCE project). We thank the teachers and pupils of the mental dyslexia: Evidence from two case studies. Cortex, 46, 717–738. https://doi.
org/10.1016/j.cortex.2009.11.002.
schools of the Académie de Grenoble for their participation. Duncan, J., Bundesen, C., Olson, A., Humphreys, G., Chavda, S., & Shibuya, H. (1999).
Systematic analysis of deficits in visual attention. Journal of Experimental Psychology:
Appendix A. Supplementary data General, 128, 450–478.
Egly, R., Driver, J., & Rafal, R. D. (1994). Shifting visual attention between objects and
locations: Evidence from normal and parietal lesion subjects. Journal of Experimental
Supplementary data to this article can be found online at https:// Psychology: General, 123, 161–177.
doi.org/10.1016/j.visres.2019.10.011. Elahipanah, A., Christensen, B. K., & Reingold, E. M. (2011). Controlling the spotlight of
attention: Visual span size and flexibility in schizophrenia. Neuropsychologia, 49(12),
3370–3376.
References Foulin, J. N. (2005). Why is letter-name knowledge such a good predictor of learning to
read? Reading & Writing, 18, 129–155. https://doi.org/10.1007/s11145-004-5892-2.
Adelman, J. S., Marquis, S. J., & Sabatos de Vito, S. G. (2010). Letters in words are read Franceschini, S., Gori, S., Ruffino, M., Pedrolli, K., & Facoetti, A. (2012). A causal link
simultaneously, not left-to-right. Psychological Science, 21(12), 1799–1801. https:// between visual spatial attention and reading acquisition. Current Biology, 22,
doi.org/10.1177/0956797610387442. 814–819.
Allport, D. A. (1968). The rate of assimilation of visual information. Psychonomic Science, Frey, A., & Bosse, M. L. (2018). Perceptual span, visual span, and visual attention span:
Three potential ways to quantify limits on visual processing during reading. Visual

159
S. Valdois, et al. Vision Research 165 (2019) 152–161

Cognition. https://doi.org/10.1080/13506285.2018.1472163. https://doi.org/10.1037/a0026744.

Georgiou, G. K., Torppa, M., Manolitsis, G., Lyytinen, H., & Parilla, R. (2012). Mousty, P., Leybaert, J., Alegria, J., Content, A., & Morais, J. (1994). BELEC. Batterie
Longitudinal predictors of reading and spelling across languages varying in ortho- d’évaluation du langage écrit et de ces troubles. In Grégoire J., Piérart B., editors.
graphic transparency. Reading & Writing, 25, 321–346. https://doi.org/10.1007/ Evaluer les troubles de la lecture. Les nouveaux modèles théoriques et leurs im-
s11145-010-9271-x. plications diagnostiques. Bruxelles: De Boeck, 127–145.
Germano, G. D., Reilhac, C., Capellini, S. A., & Valdois, S. (2014). The phonological and Pegado, F., Nakamura, K., Braga, L., Ventura, P., Nunes, G., Jobert, A., ... Dehaene, S.
visual basis of developmental dyslexia in Brazilian Portuguese reading children. (2014). Literacy breaks mirror invariance for visual stimuli: A behavioral study with
Frontiers in Psychology, 5, 1169. https://doi.org/10.3389/fpsyg.2014.01169. adult illiterates. Journal of Experimental Psychology: General, 143, 887–894. https://
Ghahghaei, S., & Linnell, K. J. (2018). The effect of load on spatial attention depends on doi.org/10.1037/a0033198.
preview: Evidence from a reading study. Vision Research, 149, 115–123. https://doi. Pelli, D. G., Burns, C. W., Farell, B., & Moore-Page, D. C. (2006). Feature detection and
org/10.1016/j.visres.2018.05.003. letter identification. Vision Research, 46, 4646–4674. https://doi.org/10.1016/j.
Ghahghaei, S., Linnell, K. J., Fischer, M. H., Dubey, A., & Davis, R. (2013). Effects of load visres.2006.04.023.
on the time course of attentional engagement, disengagement, and orienting in Pelli, D. G., Palomares, M., & Majaj, N. J. (2004). Crowding is unlike ordinary masking:
reading. The Quarterly Journal of Experimental Psychology, 66(3), 453–470. Distinguishing feature integration from detection. Journal of Vision, 4(12),
Gilbert, C. D., Sigman, M., & Crist, R. E. (2001). The neural basis of perceptual learning. 1136–1169. https://doi.org/10.1167/4.12.12 http://journalofvision.org/4/12/12/.
Neuron, 31, 681–697. Pelli, D. G., Tillman, K. A., Freeman, J., Su, M., Berger, T. D., & Majaj, N. J. (2007).
Ginestet, E., Phénix, T., Diard, J., & Valdois, S. (2019). Modelling the length effect for Crowding and eccentricity determine reading rate. Journal of Vision, 7(2), 1–36.
words in lexical decision: The role of visual attention. Vision Research, 159, 10–20. https://doi.org/10.1167/0.0.1 20.1-36.
Goswami, U. (2015). Sensory theories of developmental dyslexia: Three challenges for Peyrin, C., Démonet, J. F., Baciu, M., Le Bas, J. F., & Valdois, S. (2011). Superior parietal
research. Nature Review Neuroscience, 16, 43–54. lobe dysfunction in a homogeneous group of dyslexic children with a single visual
Grainger, J., Dufau, S., & Ziegler, J. C. (2016). A vision of reading. Trends in Cognitive attention span disorder. Brain and Language, 118, 128–138. https://doi.org/10.1016/
Neuroscience, 20(3), 171–179. https://doi.org/10.1016/j.tics.2015.12.008. j.bandl.2010.06.005.
Habekost, T. (2015). Clinical TVA-based studies: A general review. Frontiers in Psychology, Peyrin, C., Lallier, M., Démonet, J. F., Baciu, M., Le Bas, J. F., & Valdois, S. (2012). Neural
6, 290. https://doi.org/10.3389/fpsyg.2015.00290. dissociation of phonological and visual attentional span disorders in developmental
Hulme, C., Bowyer-Crane, C., Carroll, J. M., Duff, F. J., & Snowling, M. J. (2012). The dyslexia: fMRI evidence from two case reports. Brain and Language, 120(3), 381–394.
causal role of phoneme awareness and letter-sound knowledge in learning to read: https://doi.org/10.1016/j.bandl.2011.12.015.
Combining intervention studies with mediation analyses. Psychological Science, 23, Pollatsek, A., Bolozky, S., Well, A. D., & Rayner, K. (1981). Asymmetries in the perceptual
572–577. https://doi.org/10.1177/0956797611435921. span for Israeli readers. Brain and Language, 14(1), 174–180.
Jacquier-Roux, M., Valdois, S., & Zorman, M. (2002). ODEDYS, un outil de dépistage des Prado, C., Dubois, M., & Valdois, S. (2007). The eye movements of dyslexic children
dyslexies. Grenoble: Laboratoire cogni-sciences, Université Grenoble-Alpes. during reading and visual search: Impact of the visual attention span. Vision Research,
Juphard, A., Carbonnel, S., & Valdois, S. (2004). Length effect in reading and lexical 47, 2521–2530. https://doi.org/10.1016/j.visres.2007.06.001.
decision: Evidence from skilled readers and a developmental dyslexic participant. Raven, J., Raven, J. C., & Court, J. H. (1998). Progressive Matrices Couleur (SPM).
Brain and Cognition, 55(2), 332–340. https://doi.org/10.1016/j.bandc.2004.02.035. Montreuil, France: Editions ECPA.
Khan, A. Z., Prost-Lefebvre, M., Salemme, R., Blohm, G., Rossetti, Y., Tilikete, C., & Rayner, K. (1998). Eye movements in reading and information processing: 20 years of
Pisella, L. (2016). The attentional fields of visual search in simultanagnosia and research. Psychological Bulletin, 124(3), 372–422.
healthy in- dividuals: How object and space attention interact. Cerebral Cortex, 26(3), Rayner, K. (2009). Eye movements and attention in reading, scene perception and visual
1242–1254. https://doi.org/10.1093/cercor/bhv059. search. The Quarterly Journal of Experimental Psychology, 62(8), 1457–1506. https://
Kolinsky, R., & Fernandes, T. (2014). A cultural side effect: learning to read interferes doi.org/10.1080/17470210902816461.
with identity processing of familiar objects. Frontiers in Psychology, 5, 1224. https:// Rayner, K., & Reichle, E. D. (2010). Models of the reading process. Cognitive Science, 1,
doi.org/10.3389/fpsyg.2014.01224. 787–799. https://doi.org/10.1002/wcs.68.
Kyllingsbæk, S., & Bundesen, C. M. (2007). Parallel processing in a multifeature whole- Reilhac, C., Peyrin, C., Démonet, J. F., & Valdois, S. (2013). Role of the superior parietal
report paradigm. Journal of Experimental Psychology: Human Perception and lobules in letter-identity processing within strings: fMRI evidence from skilled and
Performance, 33(1), 64–82. dyslexic readers. Neuropsychologia, 51(4), 601–612. https://doi.org/10.1016/j.
Laberge, D., & Brown, V. (1989). Theory of attentional operations in shape identification. neuropsychologia.2012.12.010.
Psychological Review, 96, 101–124. Risko, E. F., Stolz, J. A., & Besner, D. (2010). Spatial attention modulates fea- ture
Laberge, D., & Samuels, S. J. (1974). Toward a theory of automatic information proces- crosstalk in visual word processing. Attention, Perception and Psychophysics, 72(4),
sing in reading. Cognitive Psychology, 3, 293–323. 989–998. https://doi.org/10.3758/APP.72.4.989.
Lachman, T., & Van Leeuwen, C. (2007). Paradoxical enhancement of letter recognition in Rosseel, Y. (2012). Lavaan: An R Package for Structural Equation Modeling. J Statistical
developmental dyslexia. Developmental Neuropsychology, 31(1), 61–77. Software. 48, 1–36. http://www.jstatsoft.org/v48/i02/.
Lachter, J., Forster, K. I., & Ruthruff, E. (2004). Forty-five years after Broadbent (1958): Scaltritti, M., & Balota, D. A. (2013). Are all letters really processed equally and in par-
Still no identification without attention. Psychological Review, 111, 880–913. https:// allel? Further evidence of a robust first letter advantage. Acta Psychologica, 144,
doi.org/10.1037/0033 -295X.111.4.880. 397–410. https://doi.org/10.1016/j.actpsy.2013.07.018.
Lallier, M., Acha, J., & Carreiras, M. (2016). Cross-linguistic interactions influence Schotter, E. R., Angele, B., & Rayner, K. (2012). Parafoveal processing in reading.
reading development in bilinguals: A comparison between early balanced French- Attention, Perception, and Psychophysics, 74(1), 5–35. https://doi.org/10.3758/
Basque and Spanish-Basque bilingual children. Developmental Science, 19(1), 76–89. s13414-011-0219-2.
https://doi.org/10.1111/desc.12290. Shapiro, L. R., Carroll, J. M., & Solity, J. E. (2013). Separating the influences of pre-
Lallier, M., Valdois, S., Lassus-Sangosse, D., Prado, C., & Kandel, S. (2014). Impact of reading skills on word and nonword reading. Journal of Experimental Child Psychology,
orthographic transparency on typical and atypical reading development: Evidence in 116, 278–295.
French-Spanish bilingual children. Research in Developmental Disabilities, 35, Share, D. L. (1999). Phonological recoding and orthographic learning: A direct test of the
1177–1190. https://doi.org/10.1016/j.ridd.2014.01.021. self-teaching hypothesis. Journal of Experimental Child Psychology, 72, 95–129.
Lété, B., Sprenger-Charolles, L., & Colé, P. (2004). MANULEX: A lexical database from Shovman, M., & Ahissar, M. (2006). Isolating the impact of visual perception on dyslexics’
French readers. Behavioral Research Methods, Instruments and Computers, 36, 156–166. reading ability. Vision Research, 46, 3514–13425. https://doi.org/10.1016/j.visres.
Lobier, M., Dubois, M., & Valdois, S. (2013). The role of visual processing speed in reading 2006.05.011.
speed development. Plos One, 8. https://doi.org/10.1371/journal.pone.0058097. Sperling, G. (1960). The information available in brief visual presentations. Psychological
Lobier, M., Peyrin, C., Le Bas, J. F., & Valdois, S. (2012). Pre-orthographic character string Monographs74 (11, Whole No. 498).
processing and parietal cortex: A role for visual attention in reading? Sperling, G. (1967). Successive approximations to a model for short-term memory. Acta
Neuropsychologia, 50, 195–204. https://doi.org/10.1016/j.neuropsychologia.2012. Psychologica, 27, 285–292.
05.023. Spinelli, D., De Luca, M., Judica, A., & Zoccolotti, P. (2002). Crowding Effects on Word
Lobier, M., Peyrin, C., Pichat, C., Le Bas, J. F., & Valdois, S. (2014). Visual processing of Identification in Developmental Dyslexia. Cortex, 38(2), 179–200.
multiple elements in the dyslexic brain: Evidence for a superior parietal dysfunction. Strasburger, H. (2005). Unfocused spatial attention under- lies the crowding effect in
Frontiers in Human Neuroscience, 8, 479. https://doi.org/10.3389/fnhum.2014. indirect form vision. Journal of Vision, 5(11), 1024–1037. https://doi.org/10.1167/5.
00479. 11.8.
Lobier, M., & Valdois, S. (2015). Visual attention deficits in developmental dyslexia Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive
cannot be ascribed solely to poor reading experience. Nature Reviews Neuroscience, Psychology, 12, 97–136.
16(4), 225. https://doi.org/10.1038/nrn3836-c1. Tydgat, I., & Grainger, J. (2009). Serial Position Effects in the Identification of Letters,
Lobier, M., Zoubrinetzky, R., & Valdois, S. (2012). The visual attention span deficit in Digits, and Symbols. Journal of Experimental Psychology: Human Perception and
developmental dyslexia is visual not verbal. Cortex, 48(6), 768–773. https://doi.org/ Performance, 35(2), 480–498. https://doi.org/10.1037/a0013027.
10.1016/j.cortex.2011.09.003. Valdois, S., Bidet-Ildei, C., Lassus-Sangosse, D., Reilhac, C., N'Guyen, M. A., Guinet, E., &
Martelli, M., Di Filippo, G., Spinelli, D., & Zoccolotti, P. (2009). Crowding, reading, and Orliaguet, J. P. (2011). A visual processing but no phonological disorder in a child
developmental dyslexia. Journal of Vision, 9(4), 14, 1–18. https://doi.org/10.1167/9. with mixed dyslexia. Cortex, 47, 1197–1218. https://doi.org/10.1016/j.cortex.2011.
4.14. 05.011.
Marzouki, Y., & Grainger, J. (2014). Effects of stimulus duration and inter-letter spacing Valdois, S., Bosse, M. L., Ans, B., Carbonnel, S., Zorman, M., David, D., & Pellat, J. (2003).
on letter-in-string identification. Acta Psychologica, 148, 49–55. https://doi.org/10. Phonological and visual processing deficits can dissociate in developmental dyslexia:
1016/j.actpsy.2013.12.011. Evidence from two case studies. Reading and Writing, 16, 541–572.
Melby-Lervåg, M., Lyster, S. A. H., & Hulme, C. (2012). Phonological skills and their role Valdois, S., Lassus-Sangosse, D., Lallier, M., Moreaud, O., & Pisella, L. (2019). What does
in learning to read: A meta-analytic review. Psychonomic Bulletin, 138, 322–352. bilateral damage of the superior parietal lobes tell us about visual attention disorders

160
S. Valdois, et al. Vision Research 165 (2019) 152–161

in developmental dyslexia? Neuropsychologia, 130, 78–91. https://doi.org/10.1016/j. crowding and reduces the critical distance. Journal of Vision, 10(10), 16.
neuropsychologia.2018.08.001. Zhao, L., Cosman, J. D., Vatterott, D. B., Gupta, P., & Vecera, S. P. (2014). Visual sta-
Valdois, S., Lassus-Sangosse, D., & Lobier, M. (2012). Poor letter string processing in tistical learning can drive object-based attentional selection. Attention, Perception and
developmental dyslexia: What visual-to-phonological code mapping disorder. Psychophysics, 76, 2240–2248. https://doi.org/10.3758/s13414-014-0708-1.
Dyslexia, 18, 77–93. https://doi.org/10.1002/dys.1437. Ziegler, J. C., & Goswami, U. (2005). Reading acquisition, developmental dyslexia, and
Valdois, S., Peyrin, C., Lassus-Sangosse, D., Lallier, M., Démonet, J. F., & Kandel, S. skilled reading across languages: A psycholinguistic grain size theory. Psychology
(2014). Dyslexia in a French-Spanish bilingual girl: Behavioral and neural modula- Bulletin, 131(1), 3–29. https://doi.org/10.1037/0033-2909.131.1.3.
tions following a specific visual-attention span intervention program. Cortex, 53, Ziegler, J. C., Pech-Georgel, C., Dufau, S., & Grainger, J. (2010). Rapid processing of
120–145. https://doi.org/10.1016/j.cortex.2013.11.006. letters, digits, and symbols: What purely visual-attentional deficit in develop- mental
van den Boer, M., de Jong, P., & Haentjens-van Meeteren, M. (2013). Modeling the Length dyslexia? Developmental Science, 13, F8–F14. https://doi.org/10.1111/j.1467-7687.
Effect: Specifying the Relation With Visual and Phonological Correlates of Reading. 2010.00983.x.
Scientific Study of Reading, 17, 243–256. https://doi.org/10.1080/10888438.2012. Ziegler, J. C., Perry, C., & Zorzi, M. (2014). Modelling reading development through
683222. phonological decoding and self-teaching: Implications for dyslexia. Philosophical
Whitney, D., & Levi, D. M. (2011). Visual crowding: A fundamental limit on conscious Transactions of the Royal Society B, 369, 20120397. https://doi.org/10.1098/rstb.
perception and object recognition. Trends in Cognitive Science, 15(4), 160–168. 2012.0397.
https://doi.org/10.1016/j.tics.2011.02.005. Zorzi, M., Barbiero, C., Facoetti, A., Lonciari, I., Carrozzi, M., Montico, M., ... Ziegler, J. C.
Wiley, R. W., Wilson, C., & Rapp, B. (2016). The effect of alphabet and expertise on letter (2012). Extra- large letter spacing improves reading in dyslexia. Proceedings of the
perception. Journal of Experimental Psychology: Human Perception and Performance, National Academy of Sciences of the United States of America, 109(28), 11455–11459.
42(8), 1186–1203. https://doi.org/10.1037/xhp0000213. https://doi.org/10.1073/pnas.1205566109.
Wolfe, J. M., & Cave, K. R. (1999). The psychophysical evidence for a binding problem in Zoubrinetzky, R., Bielle, F., & Valdois, S. (2014). New insights on developmental dyslexia
Human vision. Neuron, 24, 11–17. subtypes: Heterogeneity of mixed reading profiles. Plos One, 9(6), e99337. https://
Yeari, M., Isser, M., & Schiff, R. (2017). Do dyslexic individuals present a reduced visual doi.org/10.1371/journal.pone.0099337.
attention span? Evidence from visual recognition tasks of non-verbal multi-character Zoubrinetzky, R., Collet, G., Serniclaes, W., Nguyen-Morel, M. A., & Valdois, S. (2016).
arrays. Annals of Dyslexia, 67(2), 128–146. https://doi.org/10.1007/s11881-016- Relationships between categorical perception of phonemes, phoneme awareness, and
0132-4. visual attention span in developmental dyslexia. PLoS One, 11(3), e0151015. https://
Yeshurun, Y., & Rashal, E. (2010). Precueing attention to the target location diminishes doi.org/10.1371/journal.pone.0151015.

161