Acta Psychologica 160 (2015) 23–34

Contents lists available at ScienceDirect

Acta Psychologica

j ourna l homepage: www.e lsev ie r .com/ locate /actpsy

The effects of crowding on eye movement patterns in reading

Emanuela Bricolo a,b,⁎, Carola Salvi a, Marialuisa Martelli c,d, Lisa S. Arduino e,f, Roberta Daini a,b

a Psychology Department, University of Milano-Bicocca, Milano, Italyb Milan Center for Neuroscience, Milan, Italyc Psychology Department, University of Rome “La Sapienza”, Rome, Italyd IRCCS Fondazione Santa Lucia, Rome, Italye Department of Human Sciences, University LUMSA, Rome, Italyf Institute of Cognitive Sciences and Technologies, ISTC-CNR, Rome, Italy

⁎ Corresponding author at: Dipartimento di PsicologiaEdificio U6, Piazza dell'Ateneo Nuovo 1, 20126 Milano, Ita

E-mail address: emanuela.bricolo@unimib.it (E. Bricol

http://dx.doi.org/10.1016/j.actpsy.2015.06.0030001-6918/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 17 April 2014Received in revised form 7 June 2015Accepted 8 June 2015Available online xxxx

Keywords:CrowdingReadingEye movementsSpace

Crowding is a phenomenon that characterizes normal periphery limiting letter identification when other letterssurround the signal. We investigated the nature of the reading limitation of crowding by analyzing eye-movement patterns. The stimuli consisted of two items varying across trials for letter spacing (spaced, unspacedand increased size), lexicality (words or pseudowords), number of letters (4, 6, 8), and reading modality (oraland silent). In Experiments 1 and 2 (oral and silent reading, respectively) the results show that an increase inletter spacing induced an increase in the number of fixations and in gaze duration, but a reduction in the firstfixation duration. More importantly, increasing letter size (Experiment 3) produced the same first fixationduration advantage as empty spacing, indicating that, as predicted by crowding, only center-to-center letterdistance, and not spacing per se, matters. Moreover, when the letter size was enlarged the number of fixationsdid not increase as much as in the previous experiments, suggesting that this measure depends on visual acuityrather than on crowding. Finally, gaze duration, ameasure ofword recognition, did not changewith the letter sizeenlargement. No qualitative differences were found between oral and silent reading experiments (1 and 2),indicating that the articulatory process did not influence the outcome. Finally, a facilitatory effect of lexicalitywas found in all conditions, indicating an interaction between perceptual and lexical processing. Overall, ourresults indicate that crowding influences normal word reading bymeans of an increase in first fixation duration,a measure of word encoding, which we interpret as a modulatory effect of attention on critical spacing.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

The relationship between eye-movements and reading has beenstudied for a long time. At the beginning of the 20th Century, Hueycalculated that,while reading a text, the eyesmove across thepage (sac-cadic eye movements) at a nearly constant rate and that fluent adultreaders make about four fixations per second (Huey, 1908). As a conse-quence, the reading rate was thought to be the product of the numberof fixations and the number of letters that could be acquired in eachfixation (Woodworth, 1938). Subsequently, O'Regan (1980) suggestedthat the amplitude of saccades in reading should be expressed as a num-ber of characters rather than as degrees of visual angle, and Morrisonand Rayner (1981) showed that the average saccade amplitude remainsconstant at 5–6 characters with increasing character size.

Recently, it has been shown that crowding, a decoding impairmentlimiting the number of letters that can be processed in parallel in aglimpse, predicts reading rate (Pelli, Tillman, Su, Berger, & Majaj, 2007).

, Università di Milano-Bicocca,ly.o).

Crowding is a well-studied operationally defined psychophysical phe-nomenon, whereby a letter is hardly identified when surrounded bynearby letters. The aimof this study is to show the eye-movementmarkerof crowding in functional reading.

1.1. Crowding

Beyond acuity, letter recognition is impaired by crowding (for areview see Pelli, Palomares, & Majaj, 2004; Levi, 2008; Whitney & Levi,2011). This phenomenon, first named by Stuart and Burian (1962),has been explained in terms of the failure of the feature integrationprocess within a spatial window (e.g. Parkes, Lund, Angelucci,Solomon, & Morgan, 2001; Pelli et al., 2004). This window has beenvariously termed recognition span, perceptual span, visual span oruncrowded window (Legge, Mansfield & Chung, 2001; O'Regan, 1990;Pelli et al., 2007; Rayner, 1986).

Pelli et al. (2007) showed that the visual span (i.e., the number ofletters that can be processed in a glimpse) corresponds to the size ofthe uncrowded window, namely, the letters that escape crowdingat a given retinal eccentricity. The crowding effect is in fact relatedto the critical spacing between letters that is needed to restore

24 E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

recognition. This spacing is roughly equal to half of the target viewingeccentricity (Bouma, 1970). Bouma's proportionality of critical spacingwith eccentricity means that feature integration failure is presentalmost always in the periphery. In fact, for the identification of a fovealletter, the integration field extends only through a few minutes of arc,which is close to the acuity threshold (Latham & Whitaker, 1996),while the amplitude of the integration field increases together with ec-centricity but independently from visual acuity.

Critical spacing is not linked to letter size per se nor to empty spacingper se, but it is center-to-center letter distancewhich limits letter recog-nition in crowding (Arditi, Knoblauch, & Grunwald, 1990; Pelli et al.,2004; Strasburger, Harvey, & Rentschler, 1991). With this in mind, weexamined whether the effect of interletter spacing on eye movementsduring reading could be attributed to crowding.

Indeed, when reading a text, some letters fall in the fovea, but mostletters are located in the periphery. Since critical spacing scales with ec-centricity, there will be a point beyond which it will not be possible toidentify the letters. The size of the uncrowdedwindow for reading shrinksas it moves away from the foveal region (Chung, Mansfield, & Legge,1998; Legge, Ahn, Klitz, & Luebker, 1997; Legge, Mansfield, & Chung,2001; Legge et al., 2007; Pelli et al., 2007). In a fixed gaze condition, a pro-portional increase in spacing starting from fixation allows crowding to beavoided because the letters pushed further into the periphery have pro-portionally increasing spacing needs. On the other hand, this proportionalincrease in spacing starting from fixation is not feasible in an ecologicalreading context in which the eyes move continuously. Because of this,up to now, crowding has been studied almost exclusivelywith fixed gaze.

We aimed to study the direct effect of crowding on the efficiency ofreading by measuring eye movements in conditions of free viewing. Inthis condition, one possibility for partially reducing crowding is constantspacing. Indeed,while reading, someof thewordswill be seenparafoveallyand increasing spacing at a constant ratewould slightlymove the crowdingimpairment towards the letters more in the periphery. Accordingly, itcould be predicted that, in functional reading, when the eyes are free tomove, an increase in letter spacing or letter size may similarly improveeye movement guidance by reducing the number of fixations or/and thefixation duration. Two studies suggested the involvement of crowding inthe effect of spacing on eye movements measures. McDonald (2006)found that a reduction of letter spacing, keeping constant the spatialwidth of word stimuli, increased fixation duration. Hautala, Hyona, andAro (2011) compared two different spacings given by proportional fontand monospaced font. They found that the former, where an increase inthe number of letters did not widen the word's spatial extent, induced anincrease in fixation duration and gaze duration with respect to the latter.Although Hautala et al. (2011) attributed this effect to the number of let-ters, both studies suggested a role of crowding in fixation duration.

1.2. Visual span, perceptual span, and the lexicality status of the stimuli

The visuo-spatial distribution of characters is relevant for thecalculation and the programming of sacades, and the manipulation ofboth interletter and interword spacing greatly influences reading andsaccadic eye movements (e.g., Paterson and Jordan, 2010; Pollatsek &Rayner, 1982). McConkie and Rayner (1975) elegantly demonstratedthat the amount of information that is used by the observer to guidesaccades while reading extends for up to 10 characters to the right offixation. However, when random letters are used, the span size isconsiderably lower than McConkie and Rayner's (1975) estimate.O'Regan (1990) proposed a distinction between the perceptual spanthat is obtained with words and that might be influenced by the lexicalknowledge of the stimuli, and the visual span that is obtained withrandom letters (see Rayner, 1986 but also Legge et al., 1997, 2001;Chung et al., 1998; Legge et al., 2007; Legge & Bigelow, 2011).

This suggests an interaction between perceptual and lexical com-ponents, during eye-movement guidance in reading. The first stepin reading aloud consists of the mapping of visual features onto

representations through the computation of a set of letters that aredisplayed in a horizontal spatial orientation (McClelland &Rumelhart, 1981, Rumelhart & McClelland, 1982). This computationis probably achieved in parallel and represents a major challenge forword recognition models that need to incorporate visual limitations,such as crowding (e.g., Coltheart, Rastle, Perry, Langdon, & Ziegler,2001; Plaut, McClelland, Seidenberg, & Patterson, 1996). In accor-dance to the dual route model of word recognition (DRC) proposedby Coltheart et al. (2001), while pseudowords are read via a slowgrapheme-to-phoneme conversion strategy (GPC route), words canbe read with both the GPC route and a less slow direct lexicalmatching (lexical route). According to the DRC model, readingaloud would be achieved in parallel using the lexical route and seri-ally using the grapheme-to-phoneme conversion rule (but see Zorzi,Houghton, & Butterworth, 1998). The lexicality advantage may thussuggest that during reading, acquisition letter processing is opti-mized through a reduction in the size of the integration fields witha consequent increase in the uncrowded window size expressed bya reduction in number of fixation. However, if the perceptual limita-tion set by crowding constitutes a rigid bottleneck one might expectthe same number of letters to be uncrowded when words andpseudowords are presented (Levi, 2008; Pelli & Tillman, 2008).

In this case subjects may use a guessing strategy for words(e.g., Paap, Newsome, McDonald, & Schvaneveldt, 1982), producingdifferent decoding times. In this vein, differences may be found in thefixation duration for these types of stimuli when crowding is relievedby increasing the spacing or size of letters.

In the present study, we conducted three experiments in order toanalyze the effects of interletter spacing, lexicality and number ofcharacters on eye movements during reading. We developed a newtwo items reading task that allowed the testing of the effects ofcenter-to-center letter distance (either bymanipulating the letter spac-ing within a word or the font size). As in functional reading, in this taskthe reading pattern of the second item (the only one analyzed) is influ-enced by a previous similar item andnot by afixed startingpoint (as in asingle item reading task).

In the first experiment, we recorded eye movements in normalreaders by manipulating spacing and stimulus length while observersread words and pseudowords aloud. Although investigation of thecomplexities of oral compared to silent reading is out of the scope ofthe present paper, in the second experiment, in order to exclude theinterference of time consuming articulatory processes, which couldhave slowed visual scanning, we asked new participants to performthe same task reading silently. The third experiment used the samestimuli and procedures as Experiment 2, but manipulated charactersize rather than spacing. We hypothesized that if the observed changesare due to crowding andnot to the insertion of empty interletter spacingper se, then manipulating size or spacing should lead to similar results.

In particular, it has been shown that increasing spacing inducesmore fixations, reduces fixation duration and does not influence gazeduration (e.g., Slattery & Rayner, 2013). We hypothesized that thenumber of fixations depends mostly on the string spatial extension.Thus, we predicted obtaining similar results on this parameter byincreasing the number of letters or the spaces between them. In con-trast, we conjectured that the decrease in fixation duration may reflectencoding and may be due to a release from crowding. In this vein, wepredicted the same reduction in fixation duration when increasingletter size or letter spacing. On the other hand, if the increase in thenumber of fixations and the decrease in fixation duration are due tospacing per se, the manipulation of size should not induce the sameeffects as the introduction of empty spacing.

2. Experiment 1: oral reading

The first experiment was designed to study the effect of interletterspacing on eye movements. We required participants to read aloud.

25E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

This allowed us to control for accuracy. We measured the effect oflexicality by comparing words and pseudowords. To maintain theexploratory behavior as close to functional reading as possible, wepresented two-word stimuli in a row and analyzed only the secondstimulus (target word). Finally, to allow correct identification of thelast fixation on the target word, we added a secondary task consistingin the identification of a letter presented at the right edge of the screen.Observers were instructed to perform the letter identification task asfast as possible after reading the stimuli.

2.1. Methods

2.1.1. ParticipantsSixteen students at the University of Milano-Bicocca participated in

the experiment (6 males and 10 females, mean age 22.4 ± 2.9 years).All participants had normal or corrected to normal vision (with contactlenses only), were native speakers of Italian and were skilled readers.Visual acuity was evaluated using the Lea SYMBOLS® charts(Hyvärinen, Näsänen, & Laurinen, 1980).

2.1.2. ApparatusParticipants' eye movements were recorded using a monocular

video-based eye tracking system (ASL MODEL 5000, Applied ScienceLaboratories Inc.). Horizontal and vertical coordinates of the eye lineof gazewere recorded at a 60Hz sampling rate and stored on a separatePC for offline analysis. Eye position was measured with a spatial resolu-tion of about 0.5 deg. Saccades were defined as movements of the eyesbetween fixations. Fixations were defined as periods when the line ofgaze remained within a 0.5 deg circle for at least 48 ms. Eye blinkswere detected as any abrupt “loss” of the eye position signal.

The stimuli were presented on a 19-in. Samsung SyncMaster 1200nfmonitor with a 1024 × 768 pixel resolution. Participants sat with theirhead supported by a chin rest and a forehead rest that was 67 cmfrom the screen. Stimulus presentations and response recordings werecontrolled using a PC running E-Prime (Psychology Software Tools,Inc., version 1.2).

2.1.3. Materials and designStimuli were drawn in white on a black background (see Fig. 1). In

each trial the stimuli consisted of two items (both words or bothpseudowords) rendered in Courier font with each character having afixed size of 21 × 21 pixels (0.70 × 0.70 deg). The two items wereseparated by a 1.71 deg empty space and presented horizontally inthe center of the screen so that one word was displayed on the rightside and the other word was displayed on the left side of the screen

Fig. 1. Example of the stimuli used in Experiments 1, 2 and 3: Small unspaced stimuliabove, small spaced stimuli in the middle and larger stimuli and pseudowords below.

(see Fig. 1). The 432 words were selected from the LEXVAR database(Barca, Burani, & Arduino, 2002). Pseudowords were constructed fromwords and at least two letters were changed in random positions topreserve pronunciation and minimize word similarity. Pseudowordswere matched with words in character spacing and number of letters.Words and pseudowords were composed of 4, 6 and 8 letters with144 items for each length. The stimuli were presented in two condi-tions: unspaced (standard) spacing (0 pixel between characters;21 pixels i.e. 0.70 deg center-to-center distance) and spaced (21 pixelsbetween characters; 42 pixels, i.e. 1.40 deg center-to-center distance).Stimuli with the same lexicality, length and spacing were paired in thesame trial. The word and pseudoword stimuli were presented in differ-ent blocks and the order of the two blocks was counterbalanced acrosssubjects. Spacing and length conditions were randomized in the twolists. Forty practice trials (20 for each lexicality condition)were includedat the beginning of each list and these trials were not considered in theanalyses.

2.1.4. ProcedureParticipants were informed that they would have to read words

or pseudoword stimuli aloud and perform an identification task (sec-ondary task) of a letter that appeared at the right border of the screenand was vertically aligned with the stimuli (see Fig. 1). Prior to thebeginning of the experiment, a 9-point calibration procedure wasperformed. Calibration was checked before each trial and was repeatedif necessary to reduce possible eye position measurement errors due tosubject repositioning movements.

After a successful calibration, the trials were presented. Subjectswere instructed to fixate on the middle-left point of the calibrationscreen that was colored in red. As soon as their gaze was on this point,the trial began with the appearance of the stimuli. When participantscompleted reading, they were required to identify a letter, either an Mor an N that was positioned at the extreme right side of the screen, bypressing one of two buttons. No response time limit was given to theparticipants. The experiment took approximately 25 min to perform.

2.2. Results

Our focus was on the saccades falling on the second word of eachstimulus pair. First fixation durations and landing coordinates for allfixations on the second word were extracted using an ad-hoc programin MATLAB. All trials with a loss of signal or a blink (12.6%; mean 27.3;SD 14), with m–n identification errors (1.5%; mean 3.2; SD 2.4) orwith refixations on the first word (2.8%; mean 6.1; SD 3.4) werediscarded. The distribution of such missing data was comparable acrossexperimental conditions. Reading errors were very few (3%; mean 6.6;SD 3.8). Overall, 93.3% (SD 2.97) of the errors were made while partici-pants were reading non-words. In this condition no significant differ-ences were found for unspaced (45.7%; SD 1.85) and spaced (47.6%;SD 1.84) non-words. Likewise no significant difference was found forspaced (3.8%; SD .59) and unspaced (2.9%; SD .41)words. Number of er-rors increased proportionally to the number of letters only in the non-word condition (respectively for the unspaced condition: 3.8%; SD .58for 4 letter words, 13.3%; SD .81 for 6 letter words and 28.6%; SD 1.41for 8 letter words; for the spaced condition: 4.8%; SD .6 for 4 letterwords, 12.4%; SD 1.28 for 6 letter words and 30.5%; SD 1.63 for 8 letterwords). None of the comparisons across spaced and unspaced numberof errors turn out significant.

The number of fixations, thefirstfixation duration and gaze durationfor the second word were used as dependent variables. We chose, as ameasure of word encoding, the first fixation duration because it iscomparable across different word lengths.

2.2.1. Number of fixationsFor each subject, the number of fixationswas computed and entered

into a repeated measures ANOVA with spacing (unspaced and spaced),

Fig. 2. Experiment 1: The number of fixations (±1 SE) as a function of item length for thereading aloud of words (black squares) and pseudowords (red circles). Results for thestandard spacing condition are shown as open symbols; results for the spaced conditionare shown as filled symbols. (For interpretation of the references to color in this figurelegend, the reader is referred to the web version of this article.)

26 E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

lexicality (words and pseudowords) and item length (4, 6 and 8 letters)as the main factors.

The number of fixations is reported in Table 1 and plotted in Fig. 2.All the main effects were significant. The main effect of spacing

[F(1,15) = 113.27, p b 0.001] showed a significantly larger number offixations for spaced (mean 2.61) compared to unspaced stimuli (mean1.90). Themain effect of lexicality [F(1,15)= 174.34, p b 0.001] showedmore fixations for pseudowords compared to words (mean 2.68 and1.83, respectively). The main effect of item length [F(2,30) = 120.65,p b 0.001] showed an increase in the number of fixations as the numberof letters increased (4 letters:mean 1.54; 6 letters:mean 2.30; 8 letters:mean 2.93). All two-way interactions were also significant. The interac-tion between spacing and lexicality [F(1,15)=40.78, p b 0.01] emergedbecause spacing caused a larger increase in fixations for pseudowords(unspaced: mean 2.19; spaced: mean 3.17) compared to words (1.61compared to 2.06, unspaced and spaced, respectively). The interactionbetween spacing and item length [F(2,30) = 22.05, p b 0.001], accord-ing to Scheffé post-hoc comparisons, emerged because spacing causeda larger increase in fixations for 6 (p b 0.001) and 8 (p b 0.001) letterwords compared to 4 (p b 0.001) letter words (spaced 4, 6, 8 items:1.74, 2.75, 3.35, respectively; unspaced 4, 6, 8 items: 1.35, 1.85, 2.50,respectively). The interaction between lexicality and item length[F(2,30)=65.90, p b 0.001]was also significant and reflected the highernumber of fixations as item length increased for pseudowordscompared to words. Scheffé post-hoc comparisons indicated that thenumber of fixations was higher for pseudowords compared to wordsfor each length (p b 0.005). No other interactions were significant.

2.2.2. First fixation durationsFor each subject, the average first fixation duration for each stimulus

was entered into a repeated measures 2 × 2 × 3 ANOVA with spacing(unspaced and spaced), lexicality (words and pseudowords) and itemlength (4, 6 and 8 letters) as the main factors.

The first fixation duration is reported in Table 2 and plotted in Fig. 3.A significant main effect of spacing [F(1,15) = 31.64, p b 0.001] was

found reflecting the shorter fixations for the spaced stimuli (188 ms)compared to the unspaced stimuli (220 ms). A significant main effectof lexicality [F(1,15)= 50.53, p b 0.001] was also found, which showedlongerfixations for pseudowords (217ms) than forwords (192ms). Noother factors or interactions reached significance.

2.2.3. Gaze durationFor each subject, gaze duration (the sum of the duration of all

fixations on the target word before the eyes leave that word) wascomputed and entered into a repeated measures 2 × 2 × 3 ANOVAwith spacing (unspaced and spaced), lexicality (words andpseudowords) and item length (4, 6 and 8 letters) as the main factors.

The gaze duration is reported in Table 3 and plotted in Fig. 4.All the main effects were significant. The main effect of spacing

[F(1,15)=29.47, p b 0.001] showed a significantly longer gaze duration

Table 1The table reports means ± SE for the number of fixations for all experimental conditions for a

Pseudowords

4 6 8

Experiment 1: Spacing aloudCrowded 225.3 ± 8.9 229.6 ± 10.0 247.7 ± 1Uncrowded 194.4 ± 9.2 202.6 ± 9.0 202.4 ± 1

Experiment 2: Spacing silentCrowded 222.5 ± 7.9 231.3 ± 8.7 212.1 ± 9Uncrowded 181.1 ± 7.6 169.2 ± 5.9 167.2 ± 5

Experiment 3: Size silentCrowded 220.7 ± 13.2 238.5 ± 13.8 225.3 ± 1Uncrowded 197.4 ± 11.1 180.2 ± 9.1 171.7 ± 8

for spaced (mean 470.92 ms) compared to unspaced stimuli (mean411.1 ms). The main effect of lexicality [F(1,15) = 167.38, p b 0.001]showed longer gaze duration for pseudowords compared to words(mean 543 ms and 339.01 ms, respectively). The main effect of itemlength [F(2,30) = 85.06, p b 0.001] showed an increase in the gaze du-ration as the number of letters increased (4 letters: mean 305.34 ms;6 letters: mean 440.98 ms; 8 letters: mean 576.71 ms). All two-wayinteractions were also significant. The interaction between spacingand lexicality [F(1,15) = 31.48, p b 0.001], according to Scheffé post-hoc comparisons, emerged because spacing induced longer gazeduration for pseudowords (unspaced: mean 493.47 ms; spaced: mean592.53 ms; p b 0.001), but not for words (328.72 ms compared to349.31 ms, unspaced and spaced, respectively; p = 0.27). The interac-tion between spacing and item length [F(2,30) = 5.31, p b 0.005],according to Scheffé post-hoc comparisons, emerged because spacingcaused a longer gaze duration for 6 (p b 0.001) and 8 (p b 0.005) butnot 4 (p = 0.23) letter stimuli compared to unspaced stimuli (spaced4, 6, 8 items: 321.61 ms, 484.94 ms, 606.21 ms, respectively; unspaced4, 6, 8 items: 289.01 ms, 397.02ms, 547.21ms, respectively). The inter-action between lexicality and item length [F(2,30) = 52.29, p b 0.001]was also significant and reflected, according to Scheffé post-hoc

ll experiments.

Words

4 6 8

2.4 196. 6 ± 9.0 208.6 ± 10.6 213.7 ± 10.00.5 180.1 ± 8.9 175.6 ± 7.6 174.9 ± 6. 7

.4 193.9 ± 7.2 193.0 ± 6.8 197.2 ± 7.4

.5 176.9 ± 6.3 170.7 ± 4.7 165.5 ± 6.0

1.3 179.3 ± 7.0 188.89 ± 7.0 195.0 ± 9.5.9 173.4 ± 8.3 169.1 ± 6.6 157.2 ± 7.4

Table 2The table reports mean ± SE first fixation duration for all conditions for all experiments.

Pseudowords Words

4 6 8 4 6 8

Experiment 1: Spacing aloudCrowded 225.3 ± 8.8 229.6 ± 10.0 247.6 ± 12.4 196.6 ± 9.0 208.6 ± 10.7 213.7 ± 10.0Uncrowded 194.4 ± 9.2 202.6 ± 9.0 202.4 ± 10.5 180.1 ± 8.9 175.5 ± 7.6 174.9 ± 6.7

Experiment 2: Spacing silentCrowded 222.5 ± 7.9 231.2 ± 8.7 212.1 ± 9.4 193.9 ± 7.2 193.0 ± 6.7 197.2 ± 7.4Uncrowded 181.1 ± 7.6 169.1 ± 5.9 167.2 ± 5.5 176.9 ± 6.3 170.7 ± 4.7 165.5 ± 6.0

Experiment 3: Size silentCrowded 220.7 ± 13.2 238.5 ± 13.8 225.3 ± 11.3 179.3 ± 7.0 188.9 ± 7.0 195.0 ± 9.5Uncrowded 197.4 ± 11.1 180.2 ± 9.0 171.7 ± 8.9 173.4 ± 8.3 169.1 ± 6.6 157.2 ± 7.4

27E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

comparisons, that the increase in gaze duration as item length increasedwas higher for pseudowords compared to words for each length(p b 0.001). No other interactions were significant.

3. Discussion

The results show that increasing spacing affected each one of thevariables considered in the study: the number of fixations and thegaze duration increased, while the first fixation duration decreased.

The number of fixations increased as a function of stimulus lengthboth in terms of the number of letters and spacing. The gaze durationfollowed the number of fixations. The first fixation duration decreasedwith the increase in spacing, to a value of around 20–25% less than theusual fixation duration obtained in most eye movement studies. Thislatter finding, other than being consistent with recent findings on theeffect of spacing on reading (Bai, Yan, Liversedge, Zang, & Rayner,2008, Kohsom & Gobet, 1997; Paterson & Jordan, 2010; Sainio, Hyönä,Bingushi, & Bertram, 2007), could be explained by the fact that spacinginduces a release in crowding,which is responsible for a reduction in thedecoding time.

An alternative explanation for the decrease in first fixation durationin the spaced condition could be that this variable is not independentfrom the number of saccades. If this was the case, we would have

Fig. 3. Experiment 1: Themean first fixation duration (±1 SE) as a function of item lengthfor the reading aloud of words (black squares) and pseudowords (red circles). Results forthe standard spacing condition are shown as open symbols; results for the spaced condi-tion are shown asfilled symbols. (For interpretation of the references to color in this figurelegend, the reader is referred to the web version of this article.)

observed complementary effects on the two measures in all conditionsand no gaze duration increase. However, the two measures behaveddifferently from each other in different conditions: the number offixations was sensitive to the item length while first fixation durationwas not. Both measures were sensitive to lexicality in the same direc-tion (both increasing for pseudowords), while they are affected byspacing in opposite directions.

Our results are in agreement with the idea that an increase in fixa-tion time may, at least in part, be due to greater task difficulty, whichis reflected in the oculomotor programming of saccades (Kowler &Anton, 1987). The longer first fixation duration for unspaced stimuli isconsistent with the idea that in this case encoding is limited bycrowding.

Finally, the increase in the number of fixations, as in gaze duration,with the increase in item length was higher for pseudowords than forwords.

The present results may represent the outcome of two differenttypes of processing, similar to the dual route model proposed byColtheart et al. (2001). In this case, the engagement of the nonlexicalroute, which is considered to be serial, would activate a different oculo-motor behavior to the parallel analysis that is involved in the use of thelexical route (see also Paap et al., 1982).

The absence of an interaction between lexicality and spacing in firstfixation duration suggests that the mechanisms underlying the twoeffects are independent. Words may be encoded faster because fewerfeatures are needed to identify the letters in words; first fixation dura-tion may be shorter for spaced compared to unspaced letters becausethe features are relieved from crowding. The lack of interaction mayimply that crowding imposes a hard limit on the visual content that isavailable for higher level processing, irrespective of stimulus type.

4. Experiment 2: silent reading

In Experiment 1, subjects read the stimuli aloud and the resultsshow that they were quite accurate. Nevertheless, it is possible thatthis modality lowered reading speed and altered the eye movementpattern. To exclude the articulatory component and make the taskmore similar to everyday reading, we replicated the conditions butasked the subjects to read silently. Previous research has found thatreading silently and reading aloud differ quantitatively but not qualita-tively in terms of parafoveal information facilitatory effects (Ashby,Yang, Evans, & Rayner, 2012). Accordingly, we expected quantitative,but no qualitative effects, on relevant variables such as lexicality andspacing.

4.1. Methods

4.1.1. ParticipantsSixteen students at the University of Milano-Bicocca participated

in the present experiment (2 males and 14 females, mean age

Table 3The table reports mean ± SE gaze duration for all conditions for all experiments.

Pseudowords Words

4 6 8 4 6 8

Experiment 1: Spacing aloudCrowded 326.3 ± 30.5 469.4 ± 32.0 684.7 ± 36.5 251.8 ± 20.4 324.7 ± 26.6 409.7 ± 27.6Uncrowded 378.6 ± 29.3 596.0 ± 24.6 803.0 ± 35.9 264.6 ± 19.8 373.9 ± 22.6 409.4 ± 19.7

Experiment 2: Spacing silentCrowded 263.8 ± 16.3 372.6 ± 25.4 483.0 ± 38.2 215.2 ± 10.3 257.4 ± 16.7 318.1 ± 25.5Uncrowded 284.6 ± 18.9 443.0 ± 34.6 553.4 ± 41.2 241.5 ± 13.8 321.7 ± 19.8 343.9 ± 14.1

Experiment 3: Size silentCrowded 270.1 ± 19.0 429.7 ± 41.6 483.4 ± 48.6 190.4 ± 8.0 236.4 ± 14.2 294.2 ± 27.7Uncrowded 281.6 ± 25.1 381.3 ± 33.5 482.7 ± 48.6 200.9 ± 12.6 242.4 ± 18.9 303.4 ± 24.9

28 E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

22.6 ± 4.5 years). All participants had normal or corrected to normalvision (with contact lenses only), were native speakers of Italian andwere skilled readers. Visual acuity was evaluated using the LeaSYMBOLS® charts (Hyvärinen, Näsänen & Laurinen, 1980).

4.1.2. ApparatusThe same apparatus as in Experiment 1 was used.

4.1.3. Materials and designThe stimuli were identical to those used in Experiment 1.

4.1.4. ProcedureThe only procedural difference from Experiment 1 was that partici-

pants were required to perform the reading task silently.

4.2. Results

All trials with a loss of signal or a blink (10.4%; mean 22.4; SD 13.2),in which subjects made m–n identification errors (1.1%, mean 2.4; SD2.9), or with refixations on the first word (2.4%; mean 5.1; SD 6), werediscarded. The distribution of such errors was comparable across exper-imental conditions. All remaining trials were considered valid and wereused for all subsequent analyses.

The number of fixations and the firstfixation duration for the secondword were used as dependent variables.

Fig. 4. Experiment 1: The mean gaze duration (±1 SE) as a function of item length for thereading aloud of words (black squares) and pseudowords (red circles). Results for thestandard spacing condition are shown as open symbols; results for the spaced conditionare shown as filled symbols. (For interpretation of the references to color in this figurelegend, the reader is referred to the web version of this article.)

4.2.1. Number of fixationsFor each subject, the number of fixations on the second item of each

stimuluswas computed and entered into a repeatedmeasures 2 × 2 × 3ANOVA with spacing (unspaced and spaced), lexicality (words andpseudowords) and item length (4, 6 and 8 letters) as the main factors.

The number of fixations is plotted in Fig. 5.All the main effects were significant. The main effect of spacing

[F(1,15) = 115.11, p b 0.001] showed a significantly larger number offixations for spaced (2.10) compared to unspaced stimuli (1.55). Themain effect of lexicality [F(1,15) = 24.10, p b 0.001] showed morefixations for pseudowords than for words (2.06 compared to 1.59,respectively). The main effect of item length [F(2,30) = 104.83,p b 0.001] showed an increase in the number of fixations as the numberof letters increased (1.33, 1.84, 2.30 for 4, 6 and 8 letter items, respec-tively). All two-way interactions were significant. The interactionbetween spacing and lexicality [F(1,15) = 12.10, p b 0.005] emergedbecause spacing caused a larger increase in fixations for pseudowords(unspaced: 1.71 spaced: 2.41) compared to words (unspaced: 1.38;spaced: 1.80). The interaction between spacing and item length[F(2,30)=13.82, p b 0.001], according to Scheffé post-hoc comparisons,

Fig. 5. Experiment 2: The number of fixations (±1 SE) as a function of item length for thesilent reading of words (black squares) and pseudowords (red circles). Results for thestandard spacing condition are shown as open symbols; results for the spaced conditionare shown as filled symbols. (For interpretation of the references to color in this figurelegend, the reader is referred to the web version of this article.)

Fig. 7. Experiment 2: The mean gaze duration (±1 SE) as a function of item length for thereading aloud of words (black squares) and pseudowords (red circles). Results for thestandard spacing condition are shown as open symbols; results for the spaced conditionare shown as filled symbols. (For interpretation of the references to color in this figurelegend, the reader is referred to the web version of this article.)

29E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

emerged because spacing caused an increase in fixations as item lengthincreased (unspaced: 1.16, 1.50, 1.98, for 4, 6 and 8 letter items, respec-tively; spaced: 1.51, 2.18, 2.63 for 4, 6 and 8 letter items, respectively).All Scheffé post-hoc comparisons were significant at p b 0.001. Theinteraction between lexicality and item length [F(2,30) = 21.70,p b 0.001] was also significant and reflected a larger number of fixationson pseudowords compared to words at item lengths 6 (pseudowords:2.06; words: 1.62; Scheffé p b 0.001) and 8 (pseudowords: 2.70;words: 1.90; Scheffé p b 0.001), but not at an item length of 4(pseudowords: 1.41; words: 1.25; Scheffé post-hoc p = 0.32). Thethree-way interaction was not significant.

4.2.2. First fixation durationsFor each subject, the average fixation duration on the second item of

each stimulus was computed and entered into a repeated measures2 × 2 × 3 ANOVA with spacing (unspaced and spaced), lexicality(words and pseudowords) and item length (4, 6 and 8 letters) as themain factors. The mean first fixation duration is plotted in Fig. 6.

As expected, a significant main effect of lexicality [F(1,15) = 10.82,p b 0.005] was found and showed longer fixations for pseudowords(197 ms) than for words (183 ms). More interestingly, a significantmain effect of spacing [F(1,15) = 47.67, p b 0.001] was also found,which reflected longer fixations on unspaced text (208 ms) than onspaced stimuli (172ms). The interaction between spacing and lexicalityalso reached significance [F(1,15) = 41.27, p b 0.001] and emergedbecause the spacing of letters had a bigger effect on the 1st fixationduration of pseudowords (unspaced: 222 ms; spaced: 172 ms) than ofwords (unspaced: 197 ms; spaced: 171 ms), so that pseudowords, butnot words, differed from each other. Neither the main factor of itemlength nor its interaction with the other factors reached statisticalsignificance.

4.2.3. Gaze durationFor each subject, gaze duration was computed as in Experiment 1

and entered into a repeated measures 2 × 2 × 3 ANOVA with spacing(unspaced and spaced), lexicality (words and pseudowords) and itemlength (4, 6 and 8 letters) as the main factors.

The gaze duration is reported in Table 3 and plotted in Fig. 7.

Fig. 6. Experiment 2: Themean first fixation duration (±1 SE) as a function of item lengthfor the silent reading of words (black squares) and pseudowords (red circles). Resultsfor the standard spacing condition are shown as open symbols; results for the spacedcondition are shown as filled symbols. (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of this article.)

All the main effects were significant. The main effect of spacing[F(1,15)=29.67, p b 0.001] showed a significantly longer gaze durationfor spaced (mean 364.68 ms) compared to unspaced stimuli (mean318.35 ms). The main effect of lexicality [F(1,15) = 26.73, p b 0.001]showed longer gaze duration for pseudowords compared to words(mean 400.05 ms and 282.98 ms, respectively). The main effect ofitem length [F(2,30) = 83.17, p b 0.001] showed an increase in thegaze duration as the number of letters increased (4 letters: mean251.27 ms; 6 letters: mean 348.67 ms; 8 letters: mean 424.56 ms).The two-way interaction between spacing and item length tended tosignificance [F(2,30) = 3.14, p = 0.057] and this, according to Scheffépost-hoc comparisons, was because spacing caused a longer gazeduration for 6 (p b 0.001) and 8 (p b 0.005) but not 4 (p = 0.61) letterstimuli compared to unspaced stimuli (spaced 4, 6, 8 items: 263.3 ms,382.37 ms, 448.65 ms, respectively; unspaced 4, 6, 8 items: 239.5 ms,314.98 ms, 400.55 ms, respectively). The interaction between lexicalityand item length [F(2,30) = 20.09, p b 0.001] was significant andreflected, according to Scheffé post-hoc comparisons, that the increasein gaze duration as item length increased was higher for pseudowordscompared to words for each length (p b 0.001). No other interactionswere significant.

4.3. Discussion

Overall, the use of silent reading in this experiment reduced thenumber of fixations (mean: 1.82) relative to reading aloud (mean:2.26) of Experiment 1 (t(30) = 3.37; p b 0.005), the gaze duration(mean: 341.5 compared to 441.0; t(30) = −5.034; p b 0.001) andthe first fixation duration (mean: 190.05 compared to 204.28;t(30) = −3.639; p b 0.005). A direct comparison between oral andsilent reading is out of the scope of thepresent paper, and several factorsmay potentially influence the difference across these two conditionswith respect to the processing needed for speech production or thecomplex mechanisms involved in the eye-voice span (e.g. Morton,1964; for a review see Carver, 1990). Nonetheless, with respect to theeffects of the experimental manipulations, Experiment 2 replicatedthe results of Experiment 1, which suggests that qualitatively thearticulatory component had no effect on eyemovements and that silentreading was as reliable as reading aloud both in terms of accuracy andoculomotor behavior. For this reason, in the following experiment weasked participants to read silently.

Fig. 8. Experiment 3: The number of fixations (±1 SE) as a function of item length for thesilent reading of words (black squares) and pseudowords (red circles). Results for thesmall letter size condition are shown as open symbols; results for the large letter sizeare shown asfilled symbols. (For interpretation of the references to color in this figure leg-end, the reader is referred to the web version of this article.)

30 E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

5. Experiment 3: silent reading and size manipulation

The previous two experiments showed that an increase in spacingreduces fixation time, while the number of saccades and their ampli-tude grow. In this third experiment, we manipulated the center-to-center letter spacing by increasing the text size up to a spacing thatwas equal to the spacing used in Experiments 1 and 2 (Fig. 1).

Only crowding matters in the comparison between the small andlarge size conditions. It is well known that increasing font size increasesenergy (i.e. squared contrast times the letter area, that depends on thesize of the tip marker, see Pelli & Farell, 1999). Nevertheless, efficiencyfor single letter identification is largely independent of size in the sizerange used here, thus we can rule out letter visibility (Pelli, Burns,Farell, & Moore-Page, 2006). Additionally, our text is high contrast,while reading time, fixation time, and number of saccades increaseonly when contrast is reduced below 10% (Legge et al., 1997). Finally,we can rule out masking since it extends beyond the target a distanceof only 1.4 times acuity, and in our case even the small size conditionaround fixation is about 6 times above acuity (Song, Levi, & Pelli,2014). Thus, crowding and not visibility nor masking should affect thedifference across the two size conditions.

On the basis of the crowding phenomenon we would predict sizeand spacing to produce similar effects on eye movement guidance.

5.1. Methods

5.1.1. ParticipantsSixteen students at the University of Milano-Bicocca participated in

the experiment (5 males and 11 females, mean age 22.6 ± 2.7 years).All participants had normal or corrected to normal vision, were nativespeakers of Italian andwere skilled readers. Visual acuity was evaluatedusing the Lea SYMBOLS® charts (Hyvärinen et al., 1980).

5.1.2. ApparatusThe apparatus was identical to the apparatus that was used in

Experiments 1 and 2.

5.1.3. Materials and designThe samematerials and design as in Experiments 1 and 2were used.

The only differencewas that Experiment 3 used sizemanipulation rath-er than spacing manipulation.

Stimuli were presented in two conditions: small font size (the samestandard font size and spacing of Experiments 1 and 2) and large size(35 × 35 pixels, i.e. 1.16 × 1.16 deg; the space between characters was7 pixels, i.e. 0.23 deg; the center-to-center distance was 42 pixels, i.e.1.40 deg). The size used was within the plateau range for maximumreading speed, but the reading rate decreases for larger sizes (seeIntroduction). The center-to-center distance between the characters inthe spaced conditions of Experiments 1 and 2 was the same as thelarge-sized stimuli in this experiment (see Fig. 1).

5.1.4. ProcedureThis experiment used the same procedure as Experiment 2.

5.2. Results

All trials with a loss of signal, a blink (19.8%; mean 42.7; SD 29.2), inwhich subjects made m–n identification errors (1.4%; mean 2.9, SD 3),or with refixations on the first word (1.5%; mean 3.3; SD 3.3) werediscarded. The distribution of suchmissing data was comparable acrossexperimental conditions. All remaining trials were considered valid andwere used for all subsequent analyses.

The number of fixations, the firstfixation duration and gaze durationfor the second word were used as dependent variables.

5.2.1. Number of fixationsFor each subject, the number of fixations on the second item of each

stimuluswas computed and entered into a repeatedmeasures 2 × 2 × 3ANOVA with font size (small or large), lexicality (words andpseudowords) and item length (4, 6 and 8 letters) as the main factors.

The number of fixations are reported in Table 1 and plotted inFig. 8.

Themain effect of size [F(1,15)=102.18, p b 0.001] showed a signif-icantly larger number of fixations for large-sized compared to small-sized stimuli (1.80 compared to 1.56, respectively). The main effect oflexicality [F(1,15) = 50.82, p b 0.001] showed more fixations forpseudowords than for words (1.95 compared to 1.40, respectively)and the main effect of item length [F(2,30) = 76.79, p b 0.001] showedthat the number of fixations increased with the number of letters (1.23,1.71, 2.11, for 4, 6 and 8 letter items, respectively). We found a signifi-cant interaction between size and lexicality [F(1,15) = 9.27, p b 0.01]because a large size caused a larger increase in fixations forpseudowords (small: 1.81; large: 2.10) than for words (small: 1.32;large: 1.50). The interaction between size and item length [F(2,30) =4.85, p b 0.05] indicated that an increase in the number of lettersincreased the effect of size (small-sized: 1.56, 1.61, 1.93, for 4, 6 and 8letter items, respectively; large-sized: 1.30, 1.82, 2.28 for 4, 6 and 8letter items, respectively). Scheffé post-hoc comparisons showed asignificant difference between large and small stimuli at item lengths6 and 8 (p b 0.001), but not at an item length of 4 (p = .14). The inter-action between lexicality and item length [F(2,30) = 29.27, p b 0.001]was also significant and reflected the higher number of fixations asitem length increased in pseudowords (1.34, 2.04, 2.49 for 4, 6 and 8 let-ter items, respectively) compared to words (1.11, 1.38, 1.72 for 4, 6 and8 letter items, respectively). All Scheffé post-hoc comparisons weresignificant [p b 0.001]. The three-way interaction was not significant.

5.2.2. First fixation durationsFor each subject, the duration of each fixation on the second item of

each stimulus was computed and the average duration of each initialfixation per word was calculated. This data was entered into a repeatedmeasures ANOVA with the same factors used above: size (small or

Fig. 10. Experiment 3: Themean gaze duration (±1 SE) as a function of item length for thereading aloud of words (black squares) and pseudowords (red circles). Results for thestandard spacing condition are shown as open symbols; results for the spaced conditionare shown as filled symbols. (For interpretation of the references to color in this figure

31E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

large), lexicality (words and pseudowords) and item length (4, 6 and 8letters). The mean first fixation duration is plotted in Fig. 9.

A significant main effect of size [F(1,15) = 103.58, p b 0.001] wasfound and reflected longer fixations for small stimuli (208 ms) com-pared to enlarged stimuli (175ms). A significantmain effect of lexicality[F(1,15) = 35.25, p b 0.001] showed longer fixations for pseudowords(206 ms) than for words (177 ms). The interaction of lexicality by size[F(1,15)=18.14, p b 0.001]was significant and reflected a bigger differ-ence in 1st fixation duration for pseudowords (large-sized: 228; small-sized: 183 ms) than for words (large-sized: 188; small-sized: 166 ms)as character size increased. The interaction between size and itemlength [F(2,30) = 8.94, p b 0.001] was also significant and Scheffépost-hoc comparisons indicated that the two sizes were different at 6and 8 letter items (both p b 0.001), but not at 4 letter items. However,comparisons showed a significant difference only in the large print be-tween 4 and 8 letter items (185 ms at 4 letter items compared to164 at 8 letter items; p b 0.05). None of the other comparisons weresignificant.

Neither the main factor of item length nor the triple interactionreached statistical significance.

legend, the reader is referred to the web version of this article.)

5.2.3. Gaze durationFor each subject, gaze duration was computed as in previous exper-

iments and entered into a 2× 2×3 repeatedmeasuresANOVAwith fontsize (small and large), lexicality (words and pseudowords) and itemlength (4, 6 and 8 letters) as the main factors.

The gaze duration is reported in Table 3 and plotted in Fig. 10.Themain effect of lexicality [F(1,15)=46.60, p b 0.001] showed lon-

ger gaze duration for pseudowords compared to words (mean388.14 ms and 244.6 ms, respectively). The main effect of item length[F(2,30) = 43.03, p b 0.001] showed an increase in the gaze durationas the number of letters increased (4 letters: mean 235.76ms; 6 letters:mean 322.44 ms; 8 letters: mean 390.92 ms).

No effect of font size emerged, either as main effect or in the two-way interactions. The interaction between lexicality and item length[F(2,30) = 19.28, p b 0.001] was significant and reflected, according toScheffé post-hoc comparisons, the fact that the increase in gaze durationas item length increased was higher for pseudowords compared to

Fig. 9. Experiment 3: Themean first fixation duration (±1 SE) as a function of item lengthfor the silent reading ofwords (black squares) and pseudowords (red circles) for the smalland large letter size (open and filled symbols, respectively). (For interpretation of the ref-erences to color in this figure legend, the reader is referred to the web version of thisarticle.)

words for each length (p b 0.001), No other interactions weresignificant.

5.3. Discussion

Experiment 3 replicated the results obtained in Experiment 2 (andExperiment 1), in terms of the number of fixations and the first fixationduration, but not in terms of gaze duration. Our observation that theincrease in letter distance (by the introduction of spaces or an enhance-ment of text size) reduced the first fixation duration and increased thenumber of saccades confirms that the two manipulations are equiva-lent. The direction of the first effect is coherentwith the idea that closelyplaced letters are jumbled andhard to identify. An increase in center-to-center letter spacing reduced fixation duration that we ascribe to areduction in crowding. However, our manipulation also caused anincrease in stimulus spatial extension, which may be responsible forthe increase in the number of fixations.

Finally, the font size enlargement did not change gaze duration andthis result suggests either that such a measure is not completely drivenby the number of fixations or that it is due to the global processing ofword recognition (see General discussion).

5.4. A comparison between letter spacing and letter size

The figures from Experiments 2 and 3 reveal a quantitative differ-ence between letter spacing and character size manipulations in termsof the number of fixations (Figs. 5 and 8, respectively), but not for thefirst fixation duration (Figs. 6 and 9, respectively). A possible explana-tion of this difference could relate to visual acuity. In particular, in thecase of letter spacing, small eccentric characters more easily fall belowthe acuity threshold, whereas in the case of large letters, size compen-sates for poor visual acuity.

It follows that different predictions can be made, based on crowdingand visual acuity: the former implying the same result for the twomanipulations, the latter not.

We added two analyses in order to show that the first fixationduration result follows the prediction of the crowding hypothesis,while the number of fixations follows the acuity hypothesis.

In order to test our hypothesis we confined our analysis to the 8-letter condition. From an acuity stand point, 8-letter strings fall furtherin the periphery, more so in the spaced condition, and the last lettersmay be close to acuity when size is small. As for crowding, 8-letterstrings, relative to 4-letter strings, more likely exceed the size of the

32 E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

uncrowded window around fixation. However, since changing size orspacing is equivalent for crowding (i.e. the only relevant variable iscenter-to-center letter spacing) the two manipulations should affectcrowding and, according to our results, first fixation durations similarly.

We performed two ANOVAs, one on each variable, directly compar-ing the performance of the two different groups in the two experiments(2 and 3) at the greater length (8 letters), which is the length at whichthe effects might be maximized (see Fig. 11).

The number of fixations was entered into a mixed ANOVA with tworepeatedmeasures: center-to-center distance (crowded or uncrowded)and lexicality (words and pseudowords) and a between factor: manip-ulation (space or size).

Significant main effects for center-to-center distance [F(1,30) =105.88, p b 0.001] and lexicality [F(1,30) = 66.12, p b 0.001] and forthe interaction between the two repeated measures [F(1,30) = 14.38,p b 0.05]were found, as in the analysis of fixation duration. Additionally,the interaction between manipulation and center-to-center distancewas significant [F(1,30) = 9.22, p b 0.005], suggesting that size andspacing are not comparable for this dependent measure. In particular,in the standard (crowded) condition (which was exactly the same inExperiments 2 and 3), the parameter did not differ between the twogroups of subjects (1.672 compared to 1.575 for words and 2.285compared to 2.288 for pseudowords), while the number of fixationsin Experiment 2 (space manipulation) was higher than that inExperiment 3 (size manipulation), both for words (2.135 compared to1.865), and pseudowords (3.115 compared to 2.701). This result iscoherent with the acuity hypothesis which predicts such a differenceand its direction.

No other differences reached significance.The first fixation duration was also entered into a mixed ANOVA

with two repeated measures: center-to-center distance (crowded oruncrowded) and lexicality (words and pseudowords) and a betweenfactor: manipulation (space or size).

We found a significant main effect for center-to-center distance[F(1,30) = 125.68, p b 0.001] and lexicality [F(1,30) = 43.63,p b 0.001], and a significant interaction between the two repeatedmeasures [F(1,30) = 51.86, p b 0.001], as in the two separate experi-ments; A significant interaction between lexicality and manipulationwas also found [F(1,30)= 4.75, p b 0.05]. Scheffé post-hoc comparisonsindicated that the difference between pseudowords and words waspresent in both manipulations, but the one of size was bigger thanthat of space manipulation (206 ms 177 ms, p b 0.001; and 197 ms vs183 ms, p b 0.05, respectively). No significant interaction betweencenter-to-center distance and manipulations emerged, suggesting thatwhat matters for fixation duration is center-to-center letter spacing,not size per se or spacing per se.

0

0,5

1

1,5

2

2,5

3

3,5

4

num

ber

of f

ixat

ions

space

crowded pseudowuncrowded pseudocrowded wordsuncrowded words

Mansize

wordsowords

nipulation

Fig. 11.Mean number of fixations (±1 SE) and fixations duration (±1 SE) for silent reading ofwords (black circles), in Experiment 2 (spacemanipulation) and 3 (sizemanipulation). (For intversion of this article.)

6. General discussion

We performed three experiments to contrast the effects of letterspacing, letter size, lexicality and stimulus length in reading silentlyand reading aloud. Our experiments, all together, showed that eitherthe introduction of spaces or the enlargement in the text determineda higher number of fixations and a reduction in the first fixationduration, whereas gaze duration was affected by our manipulationsonly in Experiments 1 and 2. The main effect of spacing was in linewith the literature about text reading, confirming that our paradigm isreliable. Indeed,when readers' eyemovementswere recorded, previousinvestigators (Bai et al., 2008; Paterson & Jordan, 2010) have found adifference between normal unspaced text and textwith spaces betweencharacters. In particular, spaces between characters induced morefixations of shorter duration in both Chinese (Bai et al., 2008) andEnglish (Paterson & Jordan, 2010). We found the same results in theItalian language and we replicated these results by increasing fontsize, which affected the center-to-center letter distance withoutdisrupting the Gestalt of theword by abnormally segregating the lettersin the word. Overall, the novelty of this study is linking the pattern ofeye movements during reading to predictions based on crowding, inparticular, contrasting the effect of stimulus spatial extension (numberof letters) and its size.

The center-to-center distance has been shown to be the onlyvariable that is important for crowding (Pelli et al., 2007). We showedthat fixation duration was not dependent on the introduction ofempty space between letters per se since an increase in letter sizeproduced the same decrease in fixation duration. On the other hand,the number of fixations depended on the portion of the visual fieldthat was covered by the stimulus, and was affected similarly whetherthis area was generated by an increase in interletter spacing, text sizeor item length.

Can our pattern of results be explained by a trade-off between thenumber of fixations and their duration? It is unlikely, since our experi-mental manipulations affected the number and the duration differentlyacross the different conditions. The number of fixations and the firstfixation duration were both sensitive to the lexicality of the stimulus,increasing for pseudowords relative to words. Nevertheless, whileincreasing the spatial extension by adding letters affected the numberbut not the duration of fixations, adding spaces or increasing size gener-ated opposite effects on the two variables. Moreover, direct comparisonbetween Experiments 2 and 3 showed that while the first fixationdurationwas exactly the same, independent of the type ofmanipulation(spacing or increasing size), the number of fixationswas affected by thevisual differences of the two manipulations, probably due to changes invisual acuity.

crowded (open symbol) and uncrowded (filled symbols), pseudowords (red square) anderpretation of the references to color in this figure legend, the reader is referred to theweb

33E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

This evidence is contrary to the hypothesis that the shorter firstfixation duration observed with uncrowded items could be a mereconsequence of more frequent fixations.

We propose that thefirstfixation duration reduction observed in thespaced/large-sized condition is due to relief from crowding. Conversely,regarding the increase in the number of fixations for the spaced/largewords we believe that it is likely to be due to the larger spatial extentof the target words and is not a consequence of crowding (i.e., theeffects on number and duration are independent). Thus, releasingfrom crowding by increasing interletter spacing or enlarging the stimu-lus reduces the first fixation duration, while increasing the spatialextent by adding letters increases the number of fixations with no effecton duration.

Fig. 12 shows that the number of fixations increased as a function ofthe region of the visual field that was covered by the words andpseudowords by considering both the number of letters and theinterletter distance (Experiments 2 and 3). The difference betweenExperiments 2 (circles) and 3 (squares), particularly for bigger exten-sions, could be due to visual acuity.

An increase in the extent of the stimulus by the addition of spaces orletters increased the required number of fixations.

Gaze duration was affected by spacing as much as the number offixations, but it was not affected by font size enlargement, suggestingthat it is not due to extension per se. Spacing, aside from extendingthe stimulus in the periphery, breaks up the perceptual grouping ofletters, whereas increasing font size does not. Vinckier, Qiao, Pallier,Dehaene, and Cohen (2011) found that spacing letters, but not increas-ing font size, slows down reading and, according to the local combina-tion detectors model (Dehaene, Cohen, Sigman, & Vinckier, 2005),they suggested that spacing disrupt the parallel processing of lettersand letter grouping. Our eye movement results are coherent with thishypothesis and with the interpretation of gaze duration in terms of amarker for word recognition (e.g. Reichle, Pollatsek, & Rayner, 2006).

The first fixation duration on a word, the number of fixationsand gaze duration depended on lexicality and word predictability,which are both central factors in lexical processing (Kliegl, Grabner,

Fig. 12. Results from Experiments 2 (circles) and 3 (squares). The figure shows thenumber of fixations as a function of the extent of the visual field in degrees (deg) thatwas covered by the stimulus (words: black, open; pseudowords: red,filled). (For interpre-tation of the references to color in this figure legend, the reader is referred to the webversion of this article.)

Rolfs, & Engbert, 2004). Rayner, Fischer, and Pollatsek (1998) showedthe same differences in the oculomotor behavior by comparing low-frequency and high-frequency words. The lexicality effects on the firstfixation duration may imply that words are decoded faster thanpseudowords because fewer features are needed to identify letters.The higher rate of increase in the number of fixations with stimulusextension suggests that eye guidance is influenced by lexicality, whichinteracts with the size of the visual span.

7. Conclusion

We conclude that the number of fixations on a word in readingdepends on the spatial extension of the stimulus, which is due in partto its physical distance (e.g., a spaced 4-letter stimulus covers a largerpart of the visual field than an unspaced 4-letter stimulus) and in partto the number of letters in the string (e.g., 8-letter stimuli cover morespace than 4-letter stimuli). Moreover, this measure is sensitive tovisual acuity (i.e. spaced 8-letter pseudowords requires more fixationsthan a large 8-letter pseudowords).

The number of fixations also varies according to the type of process-ing triggered by the reading task (serially read pseudowords are parsedin smaller chunks relative to words).

Gaze duration is not affected by crowding or stimulus extension butis sensitive to interletter spatial distance, corroborating the hypothesisthat gaze duration reflects more global processing underlying wordrecognition.

The encoding time, as expressed by the first fixation duration, isinfluenced by integration field limitations. These limitations correspondto the letter visibility impaired by crowding. Overall our study suggeststhat eye movement guidance reflects crowding limitations through thefirst fixation duration. McDonald (2006) suggested the same relation-ship between crowding and the effect of space reduction on fixationduration, although he had not disentangled the effect of spacing per sefrom crowding.

A speculative explanation of how longer fixation duration can helpreadingwhen letters are crowded can beproposed in terms of attention.Some authors, in fact, have suggested that critical spacing can beinfluenced and modulated by top-down mechanisms, such as spatialattention (e.g., Intriligator & Cavanagh, 2001; Yeshurun & Rashal,2010). Specifically, while focusing on a smaller area of visual spacewould require more time, it could increase the spatial resolution ofthe feature integration process, at least up to a structural limit. On thebasis of this, we might predict that an attentional deficit could affectreading by improving crowding. The same prediction might be appliedto conditions other than neglect, which is what studies on patientswith posterior cortical atrophy (PCA) seem to suggest. PCA patientsshow reading deficits associated with the alteration of both perceptualand attentional mechanisms (e.g., Mendez, Shapira, & Clark, 2007;Saffran & Coslett, 1996). Indeed, most PCA patients develop both visualagnosia and simultagnosia (Mendez & Perryman, 2002). Recently,crowding has been taken into account in explaining the nature of thereading deficits in PCA patients. Crutch and Warrington (2007, 2009)state that the reading deficit in their PCA patients is caused by an abnor-mal and pathological crowding, which induces an altered integration ofthe perceptual characteristics of verbal stimuli (letters) presentedsimultaneously. “Confusability” of characteristics would make it impos-sible to correctly recognize stimuli, not only in the periphery but also inthe fovea. Although Crutch andWarrington's interpretation of crowdingand of the reading disorder in PCA patients had been given as purelybottom-up, an attentional interpretation is quite likely. Furtherevidence should be collected in order to support this interpretation.

In conclusion we found an effect in first fixation duration duringnormal reading that we interpreted as due to crowding, being indepen-dent from stimulus extension at least in term of number of letters andequally affected by interletter distance and letter size. We suggestedthat the increase in first fixation duration is due to the role of attention

34 E. Bricolo et al. / Acta Psychologica 160 (2015) 23–34

and we speculate that this accounts for substitution errors in unilateralspatial neglect as well as in posterior cortical atrophy.

Finally, word recognition, as expressed by gaze duration, appearsindependent from the effect of crowding. These results confirm thosemodels that postulate different levels and different mechanisms invisual processing of orthographic stimuli (e.g. Engbert, Nuthmann,Richter, & Kliegl, 2005; Reichle et al., 2006), and specifically a level ofword encoding of individual letter detection, sensitive to crowding,and one of word recognition, characterized by the global processing ofthe stimulus. Each of these levels appears to be influenced by lexicality.

Acknowledgments

This research was supported by a PRIN (Research Programmes ofNational Interest) (FMRPEN_002) 2007 Grant to the authors. All of theexperiments were conducted in accordance with the ethical standardsof the 1964 Declaration of Helsinki. The authors declare that they haveno conflict of interest.

References

Arditi, A., Knoblauch, K., & Grunwald, I. (1990). Reading with fixed and variable characterpitch. Journal of the Optical Society of America. A, Optics and Image Science, 7(10),2011–2015.

Ashby, J., Yang, J., Evans, K.H.C., & Rayner, K. (2012). Eye movements and the perceptualspan in silent and oral reading. Attention, Perception, & Psychophysics. http://dx.doi.org/10.3758/s13414-012-0277-0.

Bai, X., Yan, G., Liversedge, S.P., Zang, C., & Rayner, K. (2008). Reading spaced andunspaced Chinese text: Evidence from eye movements. Journal of ExperimentalPsychology. Human Perception and Performance, 34(5), 1277–1287.

Barca, L., Burani, C., & Arduino, L.S. (2002). Word naming times and psycholinguisticnorms for Italian nouns. Behavior Research Methods, Instruments, & Computers, 34,424–434.

Bouma, H. (1970). Interaction effects in parafoveal letter recognition. Nature, 226,177–178.

Carver, R.P. (1990). Reading rate: A review of research and theory. San Diego, CA, US:Academic Press.

Chung, S.T., Mansfield, J.S., & Legge, G.E. (1998). Psychophysics of reading: XVIII. The effect ofprint size on reading speed in normal peripheral vision. Vision Research, 38, 2949–2962.

Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: A dual routecascaded model of visual word recognition and reading aloud. Psychological Review,108(1), 204–256.

Crutch, S.J., & Warrington, E.K. (2007). Foveal crowding in posterior cortical atrophy: A spe-cific early-visual-processing deficit affecting word reading. Cognitive Neuropsychology,24(8), 843–866.

Crutch, S.J., & Warrington, E.K. (2009). The relationship between visual crowding and let-ter confusability: Towards an understanding of dyslexia in posterior cortical atrophy.Journal of Cognitive Neuropsychology, 26(5), 471–498.

Dehaene, S., Cohen, L., Sigman, M., & Vinckier, F. (2005). The neural code for writtenwords: A proposal. Trends in Cognitive Science, 9, 335–341.

Engbert, R., Nuthmann, A., Richter, E., & Kliegl, R. (2005). SWIFT: A dynamical model ofsaccade generation during reading. Psychological Review, 112, 777–813.

Hautala, J., Hyona, J., & Aro, M. (2011). Dissociating spatial and letter-based word lengtheffects observed in readers' eye movement patterns. Vision Research, 51, 1719–1727.

Huey, E.B. (1908). The psychology and pedagogy of reading. New York: Macmillan.Hyvärinen, L., Näsänen, R., & Laurinen, P. (1980). New visual acuity test for pre-school

children. Acta Ophthalmologica, 58(4), 507–511.Intriligator, J., & Cavanagh, P. (2001). The spatial resolution of visual attention. Cognitive

Psychology, 43, 171–216.Kliegl, R., Grabner, E., Rolfs, M., & Engbert, R. (2004). Length, frequency, and predictability

effects of words on eye movements in reading. European Journal of Cognitive Psychol-ogy, 16, 262–284.

Kohsom, C., & Gobet, F. (1997). Adding spaces to Thai and English: Effects on reading.Proceedings of the Cognitive Science Society, 19, 388–393.

Kowler, E., & Anton, S. (1987). Reading twisted text: Implications for the role of saccades.Vision Research, 27, 45–60.

Latham, K., & Whitaker, D. (1996). A comparison of word recognition and reading perfor-mance in foveal and peripheral vision. Vision Research, 36(17), 2665–2674.

Legge, G.E., Ahn, S.J., Klitz, T.S., & Luebker, A. (1997). Psychophysics of reading— XVI. Thevisual span in normal and low vision. Vision Research, 37(14), 1999–2010.

Legge, G.E., & Bigelow, C.A. (2011). Does print size matter for reading? A review offindings from vision science and typography. Journal of Vision, 11(5), 8 (1–22).

Legge, G.E., Cheung, S. -H., Yu, D., Chung, S.T.L., Lee, H. -W., & Owens, D. (2007). The casefor the visual span as a sensory bottleneck in reading. Journal of Vision(2), 9 (1–15).

Legge, G.E., Mansfield, J.S., & Chung, S.T.L. (2001). Psychophysics of reading: XX. Linking let-ter recognition to reading speed in central and peripheral vision. Vision Research, 41,725–743.

Levi, D.M. (2008). Crowding — An essential bottleneck for object recognition: A mini-review. Vision Research, 48(5), 635–654.

McClelland, J.L., & Rumelhart, D.E. (1981). An interactive activation model of contexteffects in letter perception: Part 1. An account of basic findings. PsychologicalReview, 88, 375–407.

McConkie, G.W., & Rayner, K. (1975). The span of the effective stimulus during a fixationin reading. Perception & Psychophysics, 17, 578–586.

McDonald, S.A. (2006). Effects of number-of-letters on eye movements during readingare independent from effects of spatial word length. Visual Cognition, 13(1), 89–98.

Mendez, M.F., & Perryman, K.M. (2002). Neuropsychiatric features of frontotemporal de-mentia: Evaluation of consensus criteria and review. The Journal of Neuropsychiatryand Clinical Neurosciences, 14(4), 424–429.

Mendez, M.F., Shapira, J.S., & Clark, D.G. (2007). “Apperceptive” alexia in posterior corticalatrophy. Cortex, 43, 264–270.

Morrison, R.E., & Rayner, K. (1981). Saccade size in reading depends upon characterspaces and not visual angle. Perception & Psychophysics, 30, 395–396.

Morton, J. (1964). The effects of context upon speed of reading, eye movements and eye-voice span. Quarterly Journal of Experimental Psychology, 16(4), 340–354.

O'Regan, J.K. (1990). Eye movements and reading. In E. Kowler (Ed.), Eye movements andtheir role in visual and cognitive processes, vol 4. (pp. 395–453). Amsterdam: Elsevier,North-Holland.

O'Regan, J.K. (1980). The control of saccade size and fixation duration in reading: Thelimits of linguistic control. Perception & Psychophysics, 28(2), 112–117.

Paap, K.R., Newsome, S.L., McDonald, J.E., & Schvaneveldt, R.W. (1982). An activation–ver-ification model for letter and word recognition: The word-superiority effect.Psychological Review, 89(5), 573–594.

Parkes, L., Lund, J., Angelucci, A., Solomon, J.A., &Morgan,M. (2001). Compulsory averaging ofcrowded orientation signals in human vision. Nature Neuroscience, 4(7), 739–744.

Paterson, K.B., & Jordan, T.R. (2010). Effects of increased letter spacing on word identifica-tion and eye guidance during reading. Memory & Cognition, 38, 502–512.

Pelli, D.G., Burns, C.W., Farell, B., & Moore-Page, D.C. (2006). Feature detection and letteridentification. Vision Research, 46(28), 4646–4674.

Pelli, D.G., & Farell, B. (1999). Why use noise? JOSA A, 16(3), 647–653.Pelli, D.G., Palomares, M., & Majaj, N.J. (2004). Crowding is unlike ordinary masking:

Distinguishing feature integration from detection. Journal of Vision, 4(12),1136–1169.

Pelli, D.G., & Tillman, K.A. (2008). The uncrowded window of object recognition. NatureNeuroscience, 11(10), 1129–1135.

Pelli, D.G., Tillman, K.A., Su, M., Berger, T.D., & Majaj, N.J. (2007). Crowding and eccentric-ity determine reading rate. Journal of Vision, 7, 1–36.

Plaut, D.C., McClelland, J.L., Seidenberg, M.S., & Patterson, K. (1996). Understandingnormal and impaired word reading: Computational principles in quasi-regulardomains. Psychological Review, 103(1), 56–115.

Pollatsek, A., & Rayner, K. (1982). Eye movement control in reading: The role of wordboundaries. Journal of Experimental Psychology. Human Perception and Performance,8(6), 817–833.

Rayner, K. (1986). Eye movements and the perceptual span in beginning and skilledreaders. Journal of Experimental Child Psychology, 41(2), 211–236.

Rayner, K., Fischer, M.H., & Pollatsek, A. (1998). Unspaced text interferes with both wordidentification and eye movement control. Vision Research, 38(8), 1129–1144.

Reichle, E.D., Pollatsek, A., & Rayner, K. (2006). E-Z Reader: A cognitive-control, serial-attention model of eye-movement behavior during reading. Cognitive SystemsResearch, 7, 4–22.

Rumelhart, D.E., & McClelland, J. -L. (1982). An interactive activation model of context ef-fects in letter perception: Part 2. The contextual enhancement effect and some testsand extensions of the model. Psychological Review, 89, 60–94.

Saffran, E.M., & Coslett, H.B. (1996). “Attentional dyslexia” in Alzheimer's disease: A casestudy. Cognitive Neuropsychology, 13, 205–228.

Sainio, M., Hyönä, J., Bingushi, K., & Bertram, R. (2007). The role of interword spacing inreading Japanese: An eye movement study. Vision Research, 47(20), 2575–2584.

Slattery, T.J., & Rayner, K. (2013). Effects of intraword and interword spacing on eyemovements during reading: Exploring the optimal use of space in a line of text.Attention, Percpetion & Psychophysics, 75, 1275–1292.

Song, S., Levi, D.M., & Pelli, D.G. (2014). A double dissociation of the acuity and crowdinglimits to letter identification, and the promise of improved visual screening. Journal ofVision, 14(5), 3.

Strasburger, H., Harvey, L.O., & Rentschler, I. (1991). Contrast thresholds for identificationof numeric characters in direct and eccentric view. Perception & Psychophysics, 49(6),495–508.

Stuart, J.A., & Burian, H.M. (1962). A study of separation difficulty: Its relationship tovisual acuity in normal and amblyopic eyes. American Journal of Ophthalmology, 53,471–477.

Vinckier, F., Qiao, E., Pallier, C., Dehaene, S., & Cohen, L. (2011). The impact of letterspacing on reading: A test of the bigram coding hypothesis. Journal of Vision, 11, 1–21.

Whitney, D., & Levi, D.M. (2011). Visual crowding: A fundamental limit on consciousperception and object recognition. Trends in Cognitive Sciences, 15(4), 160–168.

Woodworth, R.S. (1938). Experimental psychology. New York: Holt.Yeshurun, Y., & Rashal, E. (2010). Precueing attention to the target location diminishes

crowding and reduces the critical distance. Journal of Vision, 10(10), 16.Zorzi, M., Houghton, G., & Butterworth, B. (1998). Two routes or one in reading aloud? A

connectionist “dual-process” model. Journal of Experimental Psychology. HumanPerception Performance, 24(4), 1131–1161.