Into the Looking Glass: Literacy Acquisition and Mirror ... · Into the Looking Glass: Literacy...

Into the Looking Glass: Literacy Acquisition and Mirror Invariance inPreschool and First-Grade Children

Tania FernandesUniversidade de Lisboa

Isabel LeiteUniversidade de Evora

Regine KolinskyUniversite Libre de Bruxelles and Fonds de la Recherche Scientifique-FNRS

At what point in reading development does literacy impact object recognition and orientation processing? Is itspecific to mirror images? To answer these questions, forty-six 5- to 7-year-old preschoolers and first gradersperformed two samedifferent tasks differing in the matching criterion-orientation-based versus shape-based(orientation independent)-on geometric shapes and letters. On orientation-based judgments, first graders out-performed preschoolers who had the strongest difficulty with mirrored pairs. On shape-based judgments, firstgraders were slower for mirrored than identical pairs, and even slower than preschoolers. This mirror costemerged with letter knowledge. Only first graders presented worse shape-based judgments for mirrored androtated pairs of reversible (e.g., b-d; b-q) than nonreversible (e.g., e-) letters, indicating readers difficulty inignoring orientation contrasts relevant to letters.

Learning to read is a gateway to culture and educa-tion, and most impressively, it profoundly changesthe brain and mind. Literacy acquisition leads tothe emergence of a neural network in the ventraloccipitotemporal cortex tuned to the processingof written strings, reproducible across literate peo-ple, independently of script (e.g., in Japanese andFrench readers; Dehaene, Nakamura, et al., 2010;e.g., for kana and kanji in Japanese readers; Naka-mura, Dehaene, Jobert, Le Bihan, & Kouider, 2005)and age of acquisition (i.e., in childhood for early lit-erate adults or in adulthood for late literate adults;Dehaene, Pegado, et al., 2010). Notably, learning toread also impacts on evolutionarily older systems,including visual object recognition (e.g., Dehaene,Nakamura, et al., 2010; Dehaene, Pegado, et al.,

2010; Pegado, Nakamura, Cohen, & Dehaene, 2011;Pegado, Nakamura, et al., 2014). This agrees withthe neuronal recycling hypothesis (Dehaene, 2009),which holds that the ventral occipitotemporalregions, originally devoted to object recognition,were partially recycled to accommodate literacy,with spillover effects on the older function.

In fact, probably as a consequence of the inten-sive perceptual training that it requires, literacyacquisition alters early visual responses in theoccipital cortex, including in areas involved in veryearly processing (i.e., primary visual cortex, V1;e.g., Dehaene, Pegado, et al., 2010; Pegado, Comer-lato, et al., 2014). The impact of literacy can thus befound on several visual tasks outside the writtendomain. For instance, visual integration is enhancedin readers, as shown by early and late literateadults superior capacity (compared to illiterates) inconnecting local elements into an overall shape(Szwed, Ventura, Querido, Cohen, & Dehaene,2012). The visual properties of the script itself havealso a moderator role. For example, Chineseand Korean children learning to read visuallycomplex and demanding scripts outperform Israeliand Spanish children on visuospatial skills

Tania Fernandes and the work presented in this article aresupported by the IF 2013 Program of FCT, Portugal (ref IF/00886/2013/CP1194/CT0002) and by the Research Center forPsychological Science (CICPSI). Regine Kolinsky is ResearchDirector of the Fonds de la Recherche Scientifique-FNRS, Bel-gium, and her work is supported by the Fonds de la RechercheScientifique-FNRS under Grant FRFC 2.4515.12 and by anInteruniversity Attraction Poles Grant IAP 7/33, Belspo. Wethank Vanessa Neves for the assistance in experimental testing.We are also very grateful to Ana Raposo, one anonymousreviewer, and Jon A. Du~nabeitia for their comments on prior ver-sions of this manuscript.

Correspondence concerning this article should be addressed toTania Fernandes, Faculdade de Psicologia, Universidade de Lis-boa, Alameda da Universidade, 1649-013 Lisboa, Portugal. Elec-tronic mail may be sent to [email protected].

2016 The AuthorsChild Development 2016 Society for Research in Child Development, Inc.All rights reserved. 0009-3920/2016/8706-0026DOI: 10.1111/cdev.12550

Child Development, November/December 2016, Volume 87, Number 6, Pages 20082025

(McBride-Chang et al., 2011, Experiment 1), and theimportance of reading to spatial skills is strongerthan that of spatial skills to reading, as shown bymultiple regression analyses run at two testingmoments in a 1-year longitudinal study with Chi-nese children (McBride-Chang et al., 2011, Experi-ment 2).

Learning to read in scripts such as the Latinalphabet also requires quite specific adaptations,especially when considering original properties ofthe visual system that may oppose to literacy acqui-sition. One such property is mirror-image general-ization or mirror invariance: Lateral mirror images(180 flip outside the image plane, e.g., and ) areprocessed as equivalent percepts by humans andother animals (e.g., Dehaene, Nakamura, et al.,2010; Logothetis, Pauls, & Poggio, 1995; Pegadoet al., 2011; Tarr & Pinker, 1989). Yet, to learn ascript with mirrored symbols, one must discrimi-nate mirror images, which collides with mirrorinvariance.

Besides the Latin alphabet (which is used inmore than 400 languages; e.g., b and d), otherscripts such as the Japanese hiragana (e.g., and) and the Cyrillic alphabet (e.g., e and ) alsoinclude mirrored or quasi-mirrored symbols. In anyof these scripts, mirrored symbols are just a smallproportion, but this is sufficient to trigger the abil-ity to discriminate them, which transfers to nonlin-guistic categories (e.g., Dehaene, Nakamura, et al.,2010; Kolinsky et al., 2011; Pegado et al., 2011),either novel (i.e., blob-like and geometric shapes) orfamiliar (e.g., pictures of tools or cloths). Indeed, incontrast to readers of the Latin alphabet, illiterateadults present poor mirror discrimination (Fernan-des & Kolinsky, 2013; Kolinsky et al., 2011), andreaders of Tamil, a script with no mirrored sym-bols, have the same difficulties as illiterates (Danzi-ger & Pederson, 1998; Pederson, 2003). Therefore, itis not learning to read in general that causes peopleto become able to discriminate mirror images. Thetrigger is learning a script with mirrored symbols.

Importantly, Pegado, Nakamura, et al. (2014; seealso Kolinsky & Fernandes, 2014) recently showedthat mirror discrimination also extends to situationswhere orientation processing is irrelevant and evenharmful to the task at hand. In a samedifferent,identity-based, and orientation-independent task,requiring a same response to both exact matches(henceforth, identical pairs) and mirrored pairs ofthe same object, only illiterate adults had as goodperformance for mirrored as for identical pairs. Incontrast, adult readers of the Latin alphabet showeda mirror cost, that is, worse performance for

mirrored than identical pairs of linguistic (i.e., pseu-dowords) and nonlinguistic materials (i.e., false-fontstrings, composed of letter-like characters; picturesof objects and faces). Thus, with literacy acquisition,mirror discrimination seems to become part of visualobject recognition, as readers of the Latin alphabetare unable to ignore mirror-image differences evenwhen this hinders performance. This specific impactof literacy could also explain the weaker primingeffect found for targets preceded by mirrored ratherthan by identical primes in short-term priming stud-ies with adult readers (e.g., Dehaene, Nakamura,et al., 2010; Pegado et al., 2011).

Although these studies have shown that learninga script with mirrored symbols enhances sensitivityto mirror images, to the best of our knowledge nostudy has hitherto examined when, during literacyacquisition, mirror discrimination becomes part ofvisual object recognition. Examining this questionin children differing on reading skills (preschoolerswith no reading skills vs. beginning readers at theend of the first grade) was the main aim of the pre-sent study. Single letters and geometric shapes wereused to investigate the effects of literacy acquisitionon linguistic and nonlinguistic material.

Mirror discrimination probably develops earlierfor letters and for nonlinguistic stimuli visually sim-ilar to letters, such as geometric shapes, than forother nonlinguistic categories. Indeed, even beforechildren are able to read, their letter knowledgealready predicts attention to text (as measuredthrough eye movements; Evans, Saint-Aubin, &Landry, 2009) and stronger responsiveness of theleft occipitotemporal region to letters than to othervisual categories (Cantlon, Pinel, Dehaene, & Pel-phrey, 2011). Consistently, illiterate adults who areunable to decode but have high letter knowledgeprocess letters differently than nonletters (Fernan-des, Vale, Martins, Morais, & Kolinsky, 2014).Regarding mirror-image processing, Kolinsky andFernandes (2014) recently showed that, whereas forpictures of familiar objects illiterate adults did notpresent any mirror cost on identity-based judg-ments (as in Pegado, Nakamura, et al., 2014), forgeometric shapes they did present a mirror cost.This sensitivity to mirror-image differences in geo-metric shapes by illiterate adults is consistent withthe hypothesis that the highly reproducible locationof the brain regions devoted to letter and visualword recognition is due (at least partially) to thefact that these neurons are specifically tuned toshape features similar to those of letters (Hannagan,Amedi, Cohen, Dehaene-Lambertz, & Dehaene,2015). Under this view, the closer the features of

Literacy Acquisition and Mirror Invariance 2009

nonlinguistic stimuli to those of letters, the earlierthe consequences of literacy on visual (nonlinguis-tic) processing. Notably, rudimentary reading seemsenough, as late literate adults present the same mir-ror costs as early literates on both linguistic andnonlinguistic materials (Kolinsky & Fernandes,2014; Pegado, Nakamura, et al., 2014).

However, late literate adults are not comparableto young children. Note that even explicit mirrordiscrimination, that is, when orientation is criticalto the task, seems to develop slowly in childhood.Children often present mirror errors in reading andwriting during the first 2 years of literacy instruc-tion (Cornell, 1985). Compared with fluent adultreaders, first-grade children fixate more and forlonger time on distractors differing from the target-word on two mirrored letters (e.g., meter, lettersunderlined were mirrored in the distractor) than oncontrol distractors (e.g., matar; Du~nabeitia, Dim-itropoulou, Estevez, & Carreiras, 2013). For nonlin-guistic material, in contrast to literate adults, 7- to8-year-old children present similar short-term prim-ing effects when pictures of familiar objects (e.g.,animals) are preceded by mirrored and by identicalprimes (Wakui et al., 2013).

We thus decided to investigate the impact oflearning to read on mirror-image processing of let-ters and geometric shapes in two tasks where orien-tation was either critical or irrelevant for successfulperformance using a within-participants design.Furthermore, it is still unclear whether the impact ofliteracy on orientation processing is restricted to oris at least stronger for mirror images than for otherorientation contrasts. Thus, we contrasted the pro-cessing of mirror images with the processing ofrotations in the image plane (henceforth, plane rota-tions; e.g., 180 clockwise rotation: and ). Indeed,as highlighted by Gibson, Pick, Osser, and Gibson(1962), both mirror images and plane rotations dis-tinguish letters of the Latin alphabet, for example,d-b and d-p, respectively. Given that letters are acategory of expertise for readers (McCandliss,Cohen, & Dehaene, 2003), dimensions that maxi-mally distinguish letters, like orientation, wouldbecome enhanced through perceptual learning (e.g.,Folstein, Palmeri, & Gauthier, 2013). Literacy shouldthus impact both mirror-image and plane-rotationprocessing. Yet, the neuronal recycling hypothesispredicts that this impact should be stronger for theformer contrast (Dehaene, 2009) because the visualsystem is originally sensitive to plane rotations butnot to mirror images (e.g., Logothetis et al., 1995).Consistently, both 4- to 6-year-old children and illit-erate adults find it harder to explicitly discriminate

mirror images than plane rotations of nonlinguisticobjects (Fernandes & Kolinsky, 2013; Gregory, Lan-dau, & McCloskey, 2011). Whether a similar patternwould be found when orientation is irrelevant tothe task is not clear. In Pegado, Nakamura, et al.(2014), mirror images were the only orientation con-trast examined. In Kolinsky and Fernandes (2014),although for identity-based judgments of familiarobjects illiterate adults presented no orientationcosts, for geometric shapes both illiterate and liter-ate adults presented stronger interference forrotated than mirrored pairs.

Therefore, to examine orientation processingwhen critical versus irrelevant to the task, childrenperformed two samedifferent tasks, on which theydecided in each trial whether the second stimulus(S2) was the same or not as the first one (S1). Asillustrated in Figure 1, the two tasks were per-formed separately on geometric shapes and singleletters, and had the same four trial types: fullydifferent trials (on which S2 differed from S1 onshape and on orientation; e.g., b-u), identical trials(S2 had the same shape and same orientation as S1;e.g., b-b), mirrored trials (S2 was a mirror image ofS1; e.g., b-d), and rotated trials (S2 was a planerotation of S1; e.g., b-q). The mirror-image andplane-rotation contrasts differed from the standardstimulus (S1) by the same 180 difference and pre-served all object-based properties (global shape,parts, and relation between parts). Thus, any differ-ence in performance between mirrored and rotatedtrials would not be due to low-level factors.

Both tasks examined mirror-image and plane-rotation processing, and differed only on the match-ing criterion: Orientation was either irrelevant orcritical for successful performance. In the shape-basedtask, children were asked to classify a stimulus pairas same if S2 had the same shape as S1; orientationwas thus irrelevant to the task, and hence not onlyidentical but also mirrored and rotated pairs shouldbe classified as same. In this task, orientation pro-cessing would hinder performance, leading to anorientation cost on mirrored or rotated trials com-pared to identical trials, which were used as base-line. In contrast, in the orientation-based task,orientation was the critical dimension: Childrenwere asked to classify a stimulus pair as same onlyif S2 was identical to S1same shape and same ori-entationand to classify as different both the fullydifferent pairs and the mirrored and rotated pairs.Orientation processing was assessed by examiningthe performance drop on trials on which only orienta-tion varied (mirrored and rotated trials) relative tofully different trials.

2010 Fernandes, Leite, and Kolinsky

Given the original property of mirror invarianceof the ventral visual system (Dehaene, 2009; Logo-thetis et al., 1995; Tarr & Pinker, 1989), we expectedpreschoolers to be better able to tolerate, that is, toclassify as same, the mirrored pairs in the shape-based task than to discriminate them in the orienta-tion-based task, whereas they would be as able totolerate as to discriminate plane rotations. Indeed,in the shape-based task, preschoolers would exhibitno mirror cost at all. Conversely, in the orientation-based task, they would present the worst perfor-mance for mirrored pairs, even when comparedwith rotated pairs. This pattern of results wasexpected for both materials. If mirror discriminationtransferred to nonlinguistic categories early on inreading acquisition, first graders would presentgood mirror discrimination of both letters and geo-metric shapes in the orientation-based task. If itbecame part of visual object recognition, theywould also present a mirror cost for both materialsin shape-based judgments. Given the importance ofplane-rotation contrasts in letter discrimination, weexpected first graders to be less able to tolerateplane rotations in the shape-based task than to dis-criminate them in the orientation-based task. There-fore, the strongest difference between preschoolersand first graders was expected for the mirroredpairs. Note that by using the two normalizedindices (i.e., the orientation cost and the perfor-mance drop), we ensured that any difficulty to be

found on explicit mirror discrimination bypreschoolers could not be due to overall differencesbetween groups.

We also examined the mirror cost for reversibleletters (i.e., differing only by orientation; e.g., u-n)and nonreversible letters, for which orientation con-trasts do not map onto different representations(e.g., e-). As mirror discrimination is most rele-vant to reversible letters (Perea, Moret-Tatay, &Panadero, 2011), the orientation cost on shape-based judgments of first graders should be stron-ger for these letters, but no difference wasexpected for preschoolers due to their limited letterknowledge.

Finally, to assess whether literacy-related skills(i.e., letter knowledge in preschoolers, and readingskills and phonological awareness in first graders)were associated with mirror discrimination or ori-entation processing in general, we conducted corre-lation analyses in each group for each material.

Method

Participants

Twenty-eight preliterate preschoolers (17 males;Mage = 65.9 months, SD = 3.2) and 24 first graders(7 males; Mage = 82.7 months, SD = 3.6), all Por-tuguese native speakers, from schools in Lisbonand Evora, Portugal, with no known history of

Figure 1. Experimental material. (A) Sequence of events in each experimental trial and illustration of the four trial types. The presenta-tion of the first stimulus (S1) and the second stimulus (S2) was separated by a mask to ensure no involvement of iconic memory in per-formance. S2 was presented until response or for the maximum of 2.5 s if no response was given, after which another trial began. Thetwo geometric shapes presented as S2 in the fully different and identical trials are the new figures used in this study (see text). (B) Thetwo letter types (reversible and nonreversible) organized by the four trial types.


developmental and/or neurological disorders, par-ticipated voluntarily (the study followed the ethicalguidelines of the Declaration of Helsinki). Datawere collected between March and June 2011, andMarch and June 2013. Due to the end of the schoolyear, six preschoolers did not perform the orienta-tion-based task for geometric shapes and three didnot perform it for letters. These children wereexcluded, as well as those who performed at thechance level on the fully different and identical tri-als, which led to same responses in the two tasks(for geometric shapes: two preschoolers and onefirst grader; for letters: two other preschoolers andthe same first grader). The final sample thusincluded 20 preschoolers for geometric shapes and23 for letters, plus 23 first graders for both materi-als.

Table 1 presents childrens results in fivedomains: nonverbal IQ (Colored Progressive Matri-ces of Raven, Portuguese Version; Sim~oes, 2000),visuospatial working memory (Corsi block test,Wechsler Memory Scale, 3rd ed.; Wechsler, 1997),phonological awareness (i.e., samedifferent task onthe target unitthe phoneme, the rhyme, or thesyllableof two words; 6 practice trials plus 16 tri-als per unit), letter knowledge (naming and recog-nition of lower- and upper-case letters of thePortuguese alphabet), and reading skills for firstgraders only, that is, reading fluency of isolateditems (3DM Battery, Portuguese Version; Reis,Faisca, Castro, & Petersson, 2013) and reading com-prehension (Lobrot L3 test, Portuguese adaptation;Sucena & Castro, 2008).

The phonological awareness task was examinedusing signal detection theory (SDT) d scores(Macmillan & Creelman, 2005). The reading indexwas the summed result across the 3DM subtestsand the Lobrot L3 test, given their high correla-tions, r(21)s > .85, ps < .001.

In the Portuguese educational system, literacyinstruction starts only at Grade 1. There are no offi-cial directives concerning literacy-related activitiesin preschool years, and hence, usually no (or onlylimited) instruction on letter knowledge is given,explaining the low letter knowledge of thesepreschoolers (see Table 1), and the independencebetween their letter knowledge and phonologicalawareness, r(18) = .26, p = .13. In contrast, for firstgraders, phonological awareness was significantlyassociated with reading skills, r(21) = .59, p < .005.In both groups, visual working memory was signif-icantly associated with nonverbal IQ (visuospatialabilities): preschoolers, r(18) = .46, and first graders,r(21) = .40, both ps .03.

Material

Two types of asymmetrical black-line materialwere used: nine geometric shapes and eight letters(see Figure 1). The geometric shapes were thoseused by Fernandes and Kolinsky (2013), except fortwo stimuli that were replaced by those presentedin Figure 1A. As shown in Figure 1B, half of theletters were nonreversible and the others werereversible (for b and p both orientation contrastscorresponded to real letters, but not for m and u).

For each material, three versions were createdwith irfanview (www.irfanview.com): the standard,its mirror image (180 lateral reflection), and itsplane rotation (180 clockwise rotation). For eachstandard stimulus, four pairs were prepared to cre-ate the four trial types (S1 was the standard):

Table 1Age and Average Performance in the Ancillary Tests

Preschoolers(n = 23)

First graders(n = 23)

Age (in months) 66.04 (3.34)[64.60, 67.49]

82.65 (3.64)[81.08, 84.22]

Nonverbal IQ:Raven testa

17.70 (2.99)[16.40, 18.99]

26.74 (5.37)[24.41, 29.06]

Visuospatial workingmemory: Corsi blocksb

6.61 (2.71)[5.44, 7.78]

13.17 (2.39)[12.14, 14.20]

Phonologicalawareness: d score

1.51 (1.40)[0.91, 2.12]

5.19 (0.92)[4.79, 5.58]

Letter knowledgec 24.65 (13.67)[18.74, 30.56]

65.26 (7.45)[62.04, 68.48]

Reading performance3DMdhigh-frequencywords

17.13 (11.22)[12.28, 21.98]

3DMlow-frequencywords

10.87 (7.40)[7.67, 14.07]

3DMpseudowords 11.52 (6.29)[8.80, 14.24]

Lobrot L3e 9.39 (6.45)[6.60, 12.18]

Reading index(summed score)

48.91 (29.85)[36.00, 61.82]

Note. SD in parentheses; 95% CI in brackets. aTotal of correctresponses out of 36 in the Colored Progressive Matrices ofRaven. bNumber of trials correctly performed in forward and inbackward sequences. cTotal number of correct responses out of68 items, that is, 2 (naming and recognition tasks) 9 22 upper-case letters of the Portuguese alphabet (excluding letters H, K,W, Y), plus 2 (naming and recognition tasks) 9 12 lower-case let-ters (i.e., b, d, p, q, f, g, r, s, i, o, m, x). dNumber of items readcorrectly per list in 30 s. eSilent reading test with 5-min timelimit, on which participants select the word that correctly com-pletes each sentence (out of five possible words). Performancecomputed as number of items correctly completed (total of 36sentences).


http://www.irfanview.com

identical trials (S2 was the same as S1), mirroredtrials (S2 was the mirror image of S1), rotated trials(S2 was the plane rotation of S1), and fully differenttrials (S2 was a different standard stimulus).

Procedure

Children were tested in a quiet room of theirschool. They performed the two samedifferenttasks for the two materials in four sessions. Theshape-based task was performed first to ensure thatany orientation cost to be found would not be dueto prior performance of the orientation-based task.Sequence of events in experimental trials was thesame for each task and material (see Figure 1A),and was controlled by E-Prime 2.0 (www.pstnet.-com/eprime). Children sat at a distance of ~70 cmof the computer screen (resolution: 640 9 480 pix-els; refresh rate: 60 Hz) and were asked to performa samedifferent judgment on S2 in each trial (ineach task, half of the trials were expected to lead toa same response). Instructions were given orallywith six demo trials using animals as stimuli. Next,to ensure that children understood the task, theyperformed 12 practice trials (six with animals, sixwith the experimental material; half of the trialsleading to a same response), with feedback onresponse accuracy.

In the shape-based task, on each trial childrenwere asked to decide as accurately and quickly aspossible whether S2 had the same shape as S1,independently of orientation, by pressing one of thetwo keys of the response box (same response givenwith the right index finger). It was emphasized thatstimulis name was irrelevant to the task; S2 shouldbe classified based on shape and not on name. Notethat at least for letters, especially for reversibleones, an identity-based criterion (same identity,same name) would induce an incorrect response(e.g., d and b are different letters, with differentnames, but have the same shape). In the orienta-tion-based task, children were asked to decidewhether S2 was an exact match of S1. They shouldrespond different (using the left key, left index fin-ger) if S2 had a different orientation than S1 even ifthey had the same shape. Accuracy and reactiontimes (RTs; measured from S2 onset to responseonset) were collected in each trial.

Children performed 108 trials for geometricshapes in each task (i.e., shape-based task: 54 fullydifferent, 18 identical, 18 mirrored, and 18 rotatedtrials; orientation-based task: 54 identical, 18 fullydifferent, 18 mirrored, and 18 rotated trials). For let-ters, they performed 96 trials per task (for each

letter type: in the shape-based task, 24 fully differ-ent, 8 identical, 8 mirrored, and 8 rotated trials; inthe orientation-based task, 24 identical, 8 fully dif-ferent, 8 mirrored, and 8 rotated trials).

Results

The mean accuracy and correct RTs (after the trim-ming of outliers 2.5 SD above or below the grandmean RT for each participant by material and task;< 3% data excluded) were examined separately foreach material, with group (preschoolers, first gra-ders; between-participants), task (shape- vs. orienta-tion-based), and trial type (fully different, identical,mirrored, rotated) as factors, plus letter type (re-versible vs. nonreversible) for the analyses run onletters. We also checked that a similar pattern ofstatistical significance was found when analyseswere run on SDT d scores adapted for samediffer-ent designs (i.e., hits correspond to proportion ofcorrect responses on different-response trials, andfalse alarms correspond to the proportion of incor-rect responses on same-response trials; cf. Macmil-lan & Creelman, 2005), with group, task, andcondition (fully different; mirrored; rotated) as fac-tors, plus the letter-type factor in the analyses runon letters.

Geometric Shapes

The three-way interaction between all factors attest was significant on both accuracy, F(3, 123) = 3.72,p = .013, g2p :08 (d scores, F(2, 82) = 3.97, p = .022, g2p :09) and RTs, F(3, 123) = 2.94, p = .036,g2p :07. Consistent with our predictions, as shownin Figure 2 (see also Table 2), whereas preschoolerswere immune to mirror-image differences, first gra-ders were sensitive to them even if harmful for per-formance. Specifically, preschoolers were perfectlyable to tolerate (i.e., responding same to) the mirroredpairs in the shape-based task (Figure 2A) and had thestrongest difficulty in discriminating them in the ori-entation-based task (Figure 2B), whereas first graderspresented a mirror cost on shape-based judgmentsand were quite able to explicitly discriminate the mir-rored pairs.

Shape-Based Task

In the shape-based, orientation-independent task(Figure 2A), preschoolers had a similar overall per-formance level as first graders on both accuracy,F(1, 41) = 2.02, p = .16 (d scores, F < 1), and RTs,


http://www.pstnet.com/eprimehttp://www.pstnet.com/eprime

F = 1; we thus directly compared their perfor-mance. Notably, it was only for mirrored trials thatfirst graders were significantly slower thanpreschoolers by 118 ms on average, t(41) = 1.65,p = .05 (accuracy: t(41) = 1.40, p = .10; d scores,t < 1) other ts < 1. Furthermore, although firstgraders showed a significant mirror cost, withslower performance on mirrored than on identicaltrials, F(1, 22) = 10.05, p = .004 (accuracy, F < 1),preschoolers did not show any mirror cost (accu-racy and RTs: Fs < 1). For plane rotations, no

difference was found between groups, ts < 1, asboth presented a rotation cost, with worse andslower performance on rotated than on identical tri-als: preschoolers, F(1, 19) = 10.85 and 39.36, respec-tively, both ps < .005; first graders, F(1, 22) = 9.34and 19.84, respectively, both ps < .010. On d scores,whereas preschoolers were not affected by orienta-tion, with similar d scores for mirrored, rotated,and fully different conditions, Fs 1, first graderswere affected by orientation, F(2, 44) = 6.20,p = .004, with higher d scores in the fully different

Figure 2. Mean performanceaccuracy on the top, reaction times on the bottomof preschoolers and first graders for geometricshapes. (A) Performance in the shape-based task. (B) Performance in the orientation-based task. Error bars represent the SEM - standarderror of the mean in each condition.


(which did not differ from the mirrored condition,F < 1), than in the rotated condition, F(1, 22) = 9.09,p = .006 (see Table 2).

Orientation-Based Task

As illustrated in Figure 2B, preschoolers had aspecific difficulty in discriminating mirrored pairs,presenting the worst and slowest performance forthese trials compared with either fully different tri-als, F(1, 19) = 23.15 (d scores, F(1, 19) = 14.47) andF(1, 19) = 16.93, respectively, ps .001, or rotatedtrials, F(1, 19) = 4.73 (d scores, F(1, 19) = 16.11) andF(1, 19) = 4.55, respectively, ps < .05. Preschoolerswere also less accurate and slower on rotated thanon fully different trials, F(1, 19) = 16.91 (d scores,F(1, 19) = 4.40) and F(1, 19) = 7.05, respectively,ps .05. Furthermore, on mirrored trials, preschool-ers were also worse than first graders, t(41) = 5.97(d scores), t(41) = 10.44, ps < .001, but not slower,t = 1.20, p > .10.

Although first graders were still less accurateand slower in discriminating mirror images thanplane rotations, F(1, 22) = 14.18 (d scores, F(1,22) = 19.65) and 4.97, respectively, ps < .05, theywere quite able to discriminate any orientation con-trast, with average accuracy above 80% (see Fig-ure 2B). Additionally, they were as accurate onrotated as on fully different pairs, F < 1 (d scores,F(1, 22) = 2.42, p = .13; see Table 2), albeit slowerfor the former pairs, F(1, 22) = 17.84, p < .001.

In contrast to what happened in the shape-basedtask, in the orientation-based task first graders wereoverall more accurate than preschoolers, F(1,41) = 37.84 (d scores, F(1, 41) = 122.11), ps < .001,but not faster, F < 1. As shown in Figure 2B, firstgraders were especially better than preschoolers formirror images (Group 9 Trial Type: accuracy, F(3,123) = 7.94, p < .001; RTs, F < 1; d scores, F(2,82) = 3.00, p = .05).

To ensure that this result was not merely due tothe overall difference between groups, we next used

a normalized index to compare preschoolers withfirst graders on their performance drop for mir-rored and rotated trials relative to fully different tri-als: [(x y)/(x + y)] 9 100, where x is theproportion of correct responses on fully differenttrials, and y is the proportion of correct responseson either mirrored or rotated trials (cf. Fernandes &Kolinsky, 2013). The higher the performance drop,the stronger the relative difficulty to discriminatethe pair on the basis of only orientation (on mir-rored or rotated trials) rather than on the basis ofboth shape and orientation (on fully different trials).For mirror images, the performance drop was sig-nificantly stronger in preschoolers than in first gra-ders (M = 11.89%, [SEM] = 4.19 vs. M = 6.06%,SEM = 2.19, respectively), F(1, 41) = 6.74, p = .013;for plane rotations, the difference between groupsdid not reach the conventional level of significance,but preschoolers tended to present a larger perfor-mance drop (M = 4.31%, SEM = 2.76 vs.M = 1.37%, SEM = 1.46, respectively), F(1,41) = 3.57, p = .07.

Comparison Between the Two Tasks and the TwoGroups

The comparison between tasks revealed thatpreschoolers found it harder to explicitly discrimi-nate (in the orientation-based task) than to tolerate(in the shape-based task) the mirrored pairs: accu-racy, d scores, and RTs, F(1, 19) = 14.90, 14.11, and20.78, respectively, all ps .001. For the other trialtypes (including the rotated trials), they were as fastand as accurate (with similar d scores) on shape-based as on orientation-based judgments, all Fs < 1(see Figure 2). Yet, the association between perfor-mance in the two tasks was not significant, for mir-rored trials (accuracy: r(18) = .22, p = .18; RTs:r(18) = .35, p = .12), for rotated trials (accuracy andRTs: r(18) = .05 and .13, ps > .25), or across trials(accuracy: r(18) = .26, p = .13; RTs: r(18) = .30,p = .10; one-tailed t tests).

Table 2Mean d Scores of Preschoolers and First Graders for Geometric Shapes in the Three Conditions at Test (Mirrored, Rotated, and Fully Different) inthe Shape-Based and Orientation-Based Tasks

Preschoolers First graders

Mirrored Rotated Fully different Mirrored Rotated Fully different

Shape-based task 2.96 (.23) 2.75 (.18) 2.81 (.19) 3.10 (.21) 2.69 (.17) 3.09 (.18)Orientation-based task 2.07 (.18) 2.68 (.19) 3.01 (.24) 4.61 (.17) 5.49 (.18) 5.13 (.22)

Note. Standard error of the mean (SEM) in parenthesis.


The pattern of results of first graders differedfrom that of preschoolers in three ways. First gra-ders were as accurate and fast in discriminating asin tolerating the mirrored pairs, both Fs < 1.25, yet,the association between the two tasks for these tri-als did not reach statistical significance: accuracy, r(21) = .13, p = .28; RTs, r(21) = .32, p = .07. More-over, on d scores, they were even better on orienta-tion-based than on shape-based judgments in themirrored condition, F(1, 22) = 36.57, p < .001. Sec-ond, a similar advantage for the orientation-basedover the shape-based task was observed for theother trial types, especially for rotated pairs. Onaverage, first graders were 24% more accurate indiscriminating than in tolerating these pairs, F(1,22) = 44.13, p < .001, and their average d score forthese pairs in the orientation-based task was almostthe double of that in the shape-based task, F(1,22) = 153.78, p < .001 (see Table 2). Finally, theassociation between the two tasks was significantfor rotated trials, on RTs, r(21) = .50, p = .008 (not

on accuracy, r = .09), and across trials, accuracyand RTs, rs(21) > .50.

Letters

In the analyses of variance run on accuracy andRTs, the Group 9 Trial Type 9 Letter Type interac-tion was significant, F(3, 132) = 4.79, p = .003,g2p :10, and F(3, 132) = 2.63, p = .050, g2p :056(see Figure 3). Similarly, on d scores, theGroup 9 Condition 9 Letter Type interaction wassignificant, F(2, 88) = 5.72, p = .02, g2p :115.Indeed, preschoolers were not affected by lettertype at all (neither the main effect of letter type norany interaction with other variables was significanton accuracy, d scores, and RTs, all Fs < 1.62,ps > .21). In contrast, in first graders, letter typeinteracted with trial type (accuracy: F(3, 66) = 13.81,p < .001, g2p :39, and RTs: F(3, 66) = 5.33,p = .002, g2p :19; and on d scores, with condition,F(2, 44) = 3.08, p = .05, g2p :122) and with task

Figure 3. Mean performanceaccuracy on the top, reaction times on the bottomfor reversible letters (in black) and nonreversible let-ters (in gray) by preschoolers and first graders in the experimental tasks. (A) Performance in the shape-based task. (B) Performance inthe orientation-based task. Error bars represent the standard error of the mean in each condition.


accuracy: F(1, 22) = 47.03, p < .001, g2p :68; dscores: F(1, 22) = 6.70, p = .017, g2p :233; and RTs:F(1, 22) = 5.20, p = .032, g2p :19. Actually, theimpact of letter type on first-graders accuracy wasquite specific: It was modulated by task and trialtype, F(3, 66) = 9.53, p < .001, g2p :30; LetterType 9 Task 9 Trial Type was not significant foreither d scores, F = 1.38, or RTs, F < 1. As afore-mentioned, for preschoolers, performance was notaffected by letter type, but it was modulated bytask and trial type, on accuracy, F(3, 66) = 5.61,p = .002, g2p :20, and RTs, F(3, 66) = 3.89, p = .01,g2p :15, similarly, on d scores, the Task 9 Condi-tion interaction was significant, F(2, 44) = 9.89,p < .001, g2p :31; see Table 3. Therefore, we fur-ther examined the preschoolers results in each taskacross letter type, whereas for first graders theimpact of letter type was also considered.

Shape-Based Task

In contrast to what happened for geometricshapes, for letters first graders presented an overalladvantage over preschoolers in the shape-basedtask (see Figure 3A), on accuracy, F(1, 44) = 9.63,p = .003 (d scores, F(1, 44) = 19.80, p < .001) butnot on RTs, F = 1.25 (the only significant effecton RTs was the main effect of trial type,F(3, 132) = 11.32, p < .001). This advantage wasmodulated by letter and trial type, F(3, 132) = 9.16,p < .001 (d scores: Letter 9 Condition, F(2,88) = 2.44, p = .09).

Nevertheless, both groups exhibited the samequalitative impact of trial type on their perfor-mance. As shown in Figure 3A, preschoolers pre-sented a rotation cost on shape-based judgmentsof letters: worse and slower performance on rotatedtrials (M = 59.5%, SEM = 3.8; M = 1,125 ms,

SEM = 62) than on identical trials (M = 71.5%,SEM = 3.4; M = 1,003 ms, SEM = 49), F(1,22) = 6.85 and 11.96, respectively, ps .016. Simi-larly, they had lower d scores on the rotated thanon the fully different condition (see Table 3), F(1,22) = 4.91, p = .03.

Contrary to what happened for geometricshapes, preschoolers also presented a mirror costfor letters, with worse performance on mirrored tri-als (M = 62.4%, SEM = 2.5; M = 1,082 ms,SEM = 61) than on identical trials: accuracy, F(1,22) = 6.10, p = .022; RTs, F(1, 22) = 3.39, p = .079,an effect that was also found on d scores, F(1,22) = 4.91, p = .03.

Similarly, first graders presented rotation andmirror costs on shape-based judgments of letters:For reversible letters, a rotation cost, accuracy, F(1,22) = 50.63, RTs, F(1, 22) = 9.71, both ps .005, anda mirror cost, accuracy, F(1, 22) = 31.79, p < .001,RTs, F = 1.14 (also found on d scores, with lowerperformance on the mirrored and rotated trials thanon the fully different ones, F(1, 22) = 10.82 and16.92, respectively, both ps < .005); for nonre-versible letters, the rotation and mirror costs weresignificant on RTs, F(1, 22) = 16.16 and 11.99,respectively, both ps < .001, but not on accuracy ord scores, Fs < 1 (see Table 3).

More important, first-graders shape-based judg-ments were modulated by letter and trial type onaccuracy, F(3, 66) = 17.38, p < .001 (RTs: F(3,66) = 1.55, p = .207; Letter 9 Condition, on dscores, F(2, 44) = 5.07, p = .01), because their orien-tation costs were stronger for reversible than fornonreversible letters. They were less accurate onshape-based judgments of reversible than nonre-versible letters for mirrored pairs, F(1, 22) = 13.53,p = .001 (d scores: F(1, 22) = 5.96, p = .023), androtated pairs, F(1, 22) = 62.86 (d scores: F(1,

Table 3Mean d Scores of Preschoolers and First Graders for the Two Letter Types and Across Letter Type in the Three Conditions at Test in theExperimental Tasks

Reversible letters Nonreversible letters Across letter type

Mirrored RotatedFully

different Mirrored Rotated Fully different Mirrored RotatedFully

different

PreschoolersShape-based task 2.58 (.20) 2.57 (.16) 3.03 (.28) 2.57 (.23) 2.40 (.24) 2.77 (.23) 2.57 (.17) 2.48 (.17) 2.90 (.25)Orientation-based task 1.77 (.26) 2.59 (.19) 3.40 (.27) 1.77 (.29) 2.77 (.31) 3.12 (.28) 1.77 (.24) 2.68 (.21) 3.26 (.23)

First gradersShape-based task 3.36 (.29) 3.02 (.25) 4.16 (.29) 4.27 (.30) 4.39 (.33) 4.30 (.30) 3.81 (.23) 3.70 (.25) 4.23 (.25)Orientation-based task 3.65 (.26) 4.46 (.33) 4.86 (.28) 4.63 (.28) 3.52 (.24) 4.56 (.32) 3.58 (.18) 4.51 (.28) 4.75 (.25)

Note. Standard error of the mean in parenthesis.


22) = 18.63), ps < .001, but not for identical or fullydifferent pairs, Fs < 1. Thus, first graders found itharder to classify as same the pairs that map ontodifferent letter representations, either mirrored orrotated (e.g., d-b, or d-p) than those that do not(e.g., e-).

In order to directly compare the orientation costof the two groups in the shape-based task, weadopted the orientation cost index used by Pegado,Nakamura, et al. (2014) and Kolinsky & Fernandes(2014), that is, [(x z)/(x + z)] 9 100, where x isthe proportion of correct responses on fully differ-ent trials and z is the accuracy on identical trials:The higher the orientation cost, the stronger theinterference due to an orientation transformation onshape-based judgments. This orientation cost wassignificantly modulated by group, letter type, andorientation contrast, F(1, 44) = 5.92, p = .019,g2p :12. For nonreversible letters, the orientationcost of first graders was similar to that ofpreschoolers (M = 1.34%, SEM = 3.24 vs. 8.72%,SEM = 4.35, respectively), F(1, 44) = 2.60, p = .12,and was not modulated by the orientation contrast,F(1, 44) = 2.34, p = .132. In contrast, for reversibleletters, first graders presented stronger orientationcosts than preschoolers for both mirror images(M = 19.46%, SEM = 3.56 vs. M = 5.72%,SEM = 2.26, respectively), F(1, 44) = 5.40, p = .025,and plane rotations (M = 42.77%, SEM = 6.92 vs.M = 10.79%, SEM = 5.83, respectively), F(1,44) = 12.48, p < .001.

Orientation-Based Task

Although preschoolers were somewhat sensitiveto mirror-image differences in the shape-based task,they still presented a specific difficulty in discrimi-nating mirrored letters. Their orientation-basedjudgments were the worst for the mirrored pairs(M = 50.2%, SEM = 4.1; M = 1,147 ms, SEM = 59;for d scores, see Table 3) relative to fully differentpairs (M = 78.2%, SEM = 2.8; M = 1,022 ms,SEM = 37), accuracy, d scores, and RTs, F(1,22) = 30.79, 25.72, and 8.87, respectively, allps < .01, and to rotated pairs (M = 68.7%,SEM = 3.3; M = 1,175 ms, SEM = 42), accuracy andd scores, F(1, 22) = 19.23 and 16.59, respectively,ps < .001 (on RTs, F < 1), which also differed fromeach other on accuracy, F(1, 22) = 5.95, p = .02 (dscores = 6.88, p = .016), and RTs, F(1, 22) = 23.78,p < .001 (see Figure 3B).

Contrary to what happened for shape-basedjudgments, first-graders orientation-based judg-ments were not affected by letter type, all Fs < 1.25

(either on accuracy, on d scores, or RTs). Althoughthey had an overall advantage over preschoolers inthe orientation-based task, on accuracy and dscores, F(1, 44) = 58.90 and 39.49, respectively, bothps < .001 (RTs: F < 1), discrimination of mirroredpairs was still harder than discrimination of rotatedpairs, for both reversible letters, on accuracy and dscores, F(1, 22) = 4.34 and 4.60, respectively, bothps < .05 (RTs, F < 1), and nonreversible letters, onaccuracy, d scores, and RTs, F(1, 22) = 9.65, 13.25,and 14.45, respectively, all ps .005 (see Figure 3Band Table 3).

To directly compare the two groups, we nextexamined the performance drop index previouslyconsidered for geometric shapes, and a similar pat-tern of results was found. The performance drop onmirrored trials was stronger for preschoolers thanfor first graders (M = 23.85%, SEM = 3.54 vs.M = 8.20%, SEM = 2.10, respectively), F(1,44) = 9.68, p = .003, whereas on rotated trials thetwo groups presented similar performance drops(M = 7.48%, SEM = 2.70, SEM = 2.70 vs. M = 2.06%,SEM = 1.98, respectively), F(1, 44) = 1.92, p = .17.

Comparison Between the Two Tasks and the TwoGroups

As already reported for geometric shapes,preschoolers had more difficulty in discriminating(in the orientation-based task; Figure 3B) than intolerating (in the shape-based task; Figure 3A) mir-rored letters, on accuracy, F(1, 22) = 5.81, p = .025,and on d scores, F(1, 22) = 7.87, p = .01; RTs, F < 1.Note, however, that for the other trial types (includ-ing rotated trials), preschoolers were as able to per-form orientation-based as shape-based judgments,on accuracy: Fs(1, 22) 2.94, on d scores: Fs(1,22) < 1.07, ps > .30, and on RTs: Fs 2.63, allps .10. Yet, no association was found (on accuracyor RTs) between tasks, for mirrored or rotatedpairs, or across trials, all rs(21) < .25, ps > .25.

For first graders, discrimination of mirror imagescontinued to be harder than discrimination of planerotations for both reversible letters, accuracy, F(1,22) = 4.34, p = .049 (d scores = 4.70, p = .041), RTs,F < 1, and nonreversible letters, accuracy, d scores,and RTs, F(1, 22) = 9.65, = 13.25 and 14.45, respec-tively, ps .005. Indeed, for nonreversible letters,first graders were still slower in discriminating thanin tolerating mirror images, F(1, 22) = 7.46,p = .012, on accuracy, F < 1; d scores, F(1,22) = 4.27, p = .051; this was not the case for planerotations, all Fs 1. In contrast, for reversibleletters, they were actually more accurate in


discriminating than in tolerating both the mirroredand rotated pairs, F(1, 22) = 9.76 and 62.94, respec-tively, ps .005, d scores, F = 1, and F(1,22) = 13.06, p = .001, respectively; RTs, bothFs < 1.5. Moreover, the association between taskswas significant for mirrored and rotated trials, onRTs, both rs(21) > .64, ps < .001 (accuracy, rs < .01),and across trials, accuracy and RTs, r(21) = .59 and.85, both ps .001.

As for geometric shapes, for letters the perfor-mance drop was stronger for preschoolers than forfirst graders on mirrored trials (M = 23.85%,SEM = 3.54 vs. M = 8.20%, SEM = 2.10, respec-tively), F(1, 44) = 9.68, p = .003, but not on rotatedtrials (M = 7.48%, SEM = 2.70, SEM = 2.70 vs.M = 2.06%, SEM = 1.98, respectively), F = 1.92,p = .17.

Correlation Analyses

We next examined at the individual levelwhether orientation processing was associated withliteracy-related skills (i.e., preschoolers letterknowledge and first graders reading skills, as wellas phonological awareness) rather than with visu-ospatial abilities, by considering the correlationcoefficients between these cognitive domains andthe orientation cost (in the shape-based task) andperformance drop (in the orientation-based task) formirror and rotation contrasts, separately for eachmaterial. The correlation coefficients presented in

Table 4 refer to accuracy, which was a more reliablemeasure of preschoolers orientation-based perfor-mance than RTs (but these correlation coefficientswere also checked; see Table 4, p values reportedcorrespond to one-tailed t tests; RTs indexes weremultiplied by 1 so that the correlation pattern forRTs and accuracy would be in the same direction).

Geometric Shapes

In preschoolers, sensitivity to mirror images wassignificantly associated with letter knowledge: Thebetter their letter knowledge, the stronger the mir-ror cost in shape-based judgments and the smallerthe performance drop in orientation-based judg-ments of mirrored pairs (which was also associatedwith phonological awareness; see Table 4). No asso-ciation was found between sensitivity to plane-rota-tion contrasts and any cognitive ability examined.

For first graders, mirror discrimination was asso-ciated with reading skills and phonological aware-ness (which were associated with each other, seeMethod): The better their literacy-related skills, thesmaller the performance drop (on both accuracyand RTs) for mirrored pairs in the orientation-basedtask. In contrast to what was found for preschool-ers, the performance drop for rotated pairs was alsoassociated with literacy-related skills (but onlywhen computed on RTs) and with nonverbal IQ.For the shape-based task, only one correlation wassignificant: The better the first-graders phonological

Table 4Correlation Matrix (Correlation Coefficients) Between the Ancillary Cognitive Abilities and the Orientation Cost and Performance Drop

Cognitive abilities

Geometric shapes Letters

Orientation cost(shape-based

task)

Performance drop(orientation-based

task)Orientation cost

(shape-based task)

Performance drop(orientation-based

task)

Mirror Rotation Mirror Rotation Mirror Rotation Mirror Rotation

PreschoolersNonverbal IQ (Raven) .150 .268 .016 .161 .095 .360* .149 .045aVisuospatial working memory (Corsi blocks) .068 .270 .291 .209 .015 .372* .047 .078aPhonological awareness .111 .177 .522** .153 .438* .378* .267 .104Letter knowledge .481* .128 .349 .140 .474**,a .451** .332*,a .397*

First gradersNonverbal IQ (Raven) .262 .261 .088 .325 .341* .476** .216a .308,aVisuospatial working memory (Corsi blocks) .008 .137 .053 .151 .103 .100 .065a .103aPhonological awareness .025 .339 .355*,a .241a .038a .177 .163 .170Reading index (3DM and Lobrot L3) .178 .272 .402*,a .127a .280a .043 .338,a .178

Note. Significant results (p < .05, one-tailed) are in bold, marginal results are underlined. aSignificant association (at least, |r| > .29,p < .05) for the indexes computed on reaction times.p < .10. *p < .05. **p .01.


awareness, the stronger the rotation cost. Althoughin the same direction, the association betweenphonological awareness and the mirror cost (onRTs) was unreliable, r(21) = .27, p = .10.

Letters

Preschoolers sensitivity to mirror images wassignificantly associated with letter knowledge: Thebetter their letter knowledge, the stronger the mir-ror cost in shape-based judgments and the smallerthe performance drop for mirrored pairs in orienta-tion-based judgments (see Table 4). Letter knowl-edge was also correlated with sensitivity to planerotations (but only for the indexes computed onaccuracy).

Unexpectedly, preschoolers phonological aware-ness was negatively correlated with the orientationcosts. This might be related to their adoption ofphonological labels to identify each letter shape inan orientation-invariant manner due to their limitedletter knowledge.

For first graders, mirror discrimination wasspecifically associated with reading skills: The bet-ter their reading skills, the stronger the mirror cost(on RTs) on shape-based judgments, and the lowertheir performance drop (on both accuracy and RTs)on orientation-based judgments of mirrored letters.

For both groups, the better their visuospatialabilities (i.e., nonverbal IQ and visuospatial work-ing memory), the smaller their performance drop inthe orientation-based task and the smaller their ori-entation cost in the shape-based task. This associa-tion was specific to plane rotations for preschoolersbut for first graders it was reliable for both orienta-tion contrasts.

Discussion

Learning to read leads to deep neurocognitivechanges outside the written domain, including onvisual object processing (e.g., Dehaene, Pegado,et al., 2010; Fernandes & Kolinsky, 2013; Kolinskyet al., 2011; McBride-Chang et al., 2011; Pegado,Comerlato, et al., 2014; Pegado, Nakamura, et al.,2014; Szwed et al., 2012). In this context, the pre-sent study targeted two open issues on the earlyinfluences of learning a script with mirrored sym-bols.

First, it was hitherto unknown when, duringreading development, the spillover effect of literacyon object recognition and orientation processingwould emerge. More specifically, we examined

when, in the course of literacy acquisition, mirrordiscrimination (a consequence of learning a scriptwith mirrored symbols) would become part ofvisual object recognition. To investigate this ques-tion, two groups of 5- to 7-year-old children, differ-ing on reading skillspreliterate preschoolers andfirst gradersperformed two samedifferent match-ing tasks on which orientation was either critical orirrelevant for successful performance, that is, orien-tation-based versus shape-based (orientation-inde-pendent) tasks, respectively. To our knowledge, thisis the first study to adopt a within-participantsdesign to examine in a fine-grained mannerwhether the impact of literacy would be similarwhen orientation processing was critical versusirrelevant to the task. Each task was performed ontwo categories matched in visual complexity: singleletters and geometric shapes. The latter was thenonlinguistic category used given the proximity ofits features to those of letters (cf. Hannagan et al.,2015). We thus expected that if (even incipient)changes in visual processing started to emerge earlyon, then by using this material we would be able tograsp them. We also conducted correlation analysesfor each group on each material to examine at theindividual level whether orientation processing wasassociated with literacy-related skills.

Second, it was still unclear whether the impactof learning a script with mirrored symbols wasspecific to mirror image processing or whether itwould generalize to other orientation contrasts thatare relevant for letters, like plane rotations (e.g.,d-p; Gibson et al., 1962). To study this point, thesame four trial types were used in both tasks: fullydifferent (with different shape and different orienta-tion), identical, mirrored, and rotated pairs. Wethus directly examined mirror-image versus plane-rotation processing in two tasks where orientationprocessing was critical versus irrelevant for success-ful performance.

This study represents one of the first demonstra-tions of early changes in the mirror-generalizationsystem due to literacy acquisition and providedfour original contributions on the impact that learn-ing a script with mirrored symbols has outside thewritten domain.

First, we presented the first evidence of an abso-lute and specific mirror cost on visual nonlinguisticobject recognition. For geometric shapes, a nonlin-guistic material novel for both preschoolers andfirst graders, the two groups were overall equallyable to perform shape-based judgments. Most inter-estingly, the groups differed only on mirrored pairs:Although preschoolers were immune to irrelevant


mirror-image differences, first graders exhibitedsuch strong mirror cost that they were even slowerthan preschoolers. We thus found an absolute mir-ror cost on shape-based judgments of nonlinguisticobjects, a consequence of literacy acquisition pre-dicted by the neuronal recycling hypothesis(Dehaene, 2009). In the first study to show a mirrorcost in literate adults, no other orientation contrastwas examined and the mirror cost was only rela-tive, as illiterate adults were overall slower andmore error-prone than the literate groups (Pegado,Nakamura, et al., 2014). In Kolinsky and Fernandes(2014), only literate adults (and not illiterates) wereaffected by mirror and rotation contrasts of familiarobjects, and for geometric shapes, all participants,whatever their literacy level, were sensitive to theirrelevant orientation contrasts, at least on responselatencies and mostly for plane rotations. Yet againilliterate adults were overall slower and more errorprone.

Second, by examining orientation processingwhen it was critical versus irrelevant for successfulperformance in a within-participants design, thepresent study is the first to conclusively show thatpreliterates specific difficulty with mirror discrimi-nation cannot be attributed to a general difficultywith orientation processing or because orientationis a dimension less salient than shape. On the onehand, if preschoolers had a general difficulty withorientation, then plane-rotation discriminationshould have been as hard as mirror discrimination.On the contrary, they were quite capable ofdiscriminating rotated pairs, and when comparedto first graders using the normalized index, the twogroups presented a similar performance drop forrotated pairs. On the other hand, if orientation wasa dimension less salient than shape, preschoolersshould have been worse on orientation-based thanon shape-based judgments of rotated pairs. Quitethe opposite, they presented similar rotation costsas first graders on shape-based judgments of geo-metric shapes, in line with the plane-rotation sensi-tivity of the ventral visual system (Logothetis et al.,1995; Tarr & Pinker, 1989), and more important,they were as able to explicitly discriminate plane-rotation contrasts as to tolerate them. They pre-sented the same level of interference from the irrele-vant dimension of rotated pairs in both tasks (i.e.,orientation in the shape-based task and shape inthe orientation-based task). To put it differently,preschoolers were equally sensitive to the twoincongruent dimensionsorientation and shapeasthey were as able to attend to shape as to orienta-tion of rotated pairs. Consequently, their difficulty

with mirror discrimination cannot be due to lowsensitivity to orientation in general; it seems rathergrounded on the original mirror invariance prop-erty of the ventral visual system (Dehaene, 2009;Logothetis et al., 1995). This difficulty with mirrordiscrimination is consistent with the absence of amirror cost on their shape-based judgments, denot-ing mirror invariance, and agrees with prior find-ings on illiterate adults and preliterate children(e.g., Casey, 1984; Danziger & Pederson, 1998; Fer-nandes & Kolinsky, 2013; Gibson et al., 1962; Kolin-sky et al., 2011; Nelson & Peoples, 1975; Pederson,2003), which argues for the robustness of this effect.

Third, the adoption of a within-participantsdesign also allowed us to demonstrate that theexpression of the visual consequences of literacydepends on the type of processing at stake.Whereas this impact was mirror specific when thetask did not required orientation processing andvisual object recognition was involved, it wasinstead general when explicit orientation processingwas required. The overall advantage of first gradersover preschoolers on orientation-based judgmentsdoes not seem to be due to a generic age effect, asno overall difference between groups was found onshape-based judgments of geometric shapes.Instead, it seems to be due to perceptual expertisewith the written code: Learning to read enhancesthe relevance of orientation, which then becomes acritical dimension of visual objects. In fact, the ori-entation-based task used here required both shapeand orientation processing (only exact matches hadto be considered as same), and the conjunction ofshape and orientation is essential in letter andvisual word recognition, and hence, the advantageof first graders is not surprising. Perceptual exper-tise with a visual category leads to enhancement ofthe relevant dimensions (e.g., Dehaene, Pegado,et al., 2010; Folstein et al., 2013; McCandliss et al.,2003). Thus, when learning to read, children learnto attend to critical reading-related cues, such asorientation, which were not relevant to perceptualexperience before this cultural activity took place(e.g., Casey, 1986; Gibson et al., 1962; Kolinskyet al., 2011; Nelson & Peoples, 1975). Moreover,during literacy acquisition, beginning readersbecome as able to attend to orientation as to theshape of mirrored pairs of nonlinguistic objects, asshown by their similar performance in the twoexperimental tasks for these pairs. The differentialimpact of literacy found in the shape-based versusorientation-based tasks also agrees with prior stud-ies showing that, even when the same material andprocedure is adopted, different tasks tap into


different processes underpinned by different neuralsubstrates. Although parietal regions, part of thedorsal stream, are important for explicit orientationprocessing, regions of the ventral visual stream aremainly important for processing objects shape andidentity (e.g., Gauthier et al., 2002; Harris, Benito,Ruzzoli, & Miniussi, 2008). This distinction couldalso explain why no significant association wasfound between performance in the two experimen-tal tasks for preschoolers on either geometric shapesor letters, whereas for first graders the associationwas reliable. Additionally, it was only for first gra-ders that performance in the two experimental taskswas associated with visuospatial abilities known tobe related with dorsal stream functioning (e.g., Chi-nello, Cattani, Bonfiglioli, Dehaene, & Piazza, 2013).It thus seems that literacy acquisition enhances thecross-talk between the two visual streams. In thisvein, Chinello et al. (2013) recently examined thebehavioral performance of kindergarteners (from 3to 6 years old) and adults in an extensive set offunctions related to the dorsal versus ventralstreams (e.g., visuospatial memory and grip aper-ture during grasping versus face and object recogni-tion) and found that it was only for adults, not forchildren, that visuospatial memory (assessed withthe Corsi blocks test) was associated with objectrecognition.

Fourth, the present study shows that the mirror-specific impact of literacy on visual (nonlinguistic)object recognition begins to emerge, though cru-dely, before literacy instruction, allied with letterknowledge. Although as a group preschoolers didnot present a mirror cost on their shape-based judg-ments of geometric shapes, the correlation analysesrevealed that letter knowledge was specificallyassociated with sensitivity to mirror-image differ-ences (and not to plane-rotation differences): Thehigher preschoolers letter knowledge, the strongerthe mirror cost on shape-based judgments and thesmaller the performance drop on orientation-basedjudgments of mirrored geometric shapes.

This association between letter knowledge andthe mirror cost on shape-based judgments ofpreschoolers is probably related to the nonlinguisticmaterial used here. Indeed, the degree of similarityto letters should influence the magnitude of thespillover effect of literacy on visual recognition ofother categories (Hannagan et al., 2015). This is alsoconsistent with the observation of orientation costsin identity-based judgments of illiterate adults forgeometric shapes but not for pictures of familiarobjects (Kolinsky & Fernandes, 2014). In fact, famil-iarity with letters can explain the discrepancy

between the present results (i.e., the absence of amirror cost on shape-based judgments of geometricshapes by the preschool group) and those of Kolin-sky and Fernandes (2014) with illiterate adults,given that the latter group has a long life experi-ence in a literate world (Fernandes et al., 2014).

Nevertheless, the fact that preliterate childrenalready present some sensitivity to mirror imagesagrees with prior findings suggesting an earlyimpact of the visual properties of the script to belearned on nonlinguistic visual processing(McBride-Chang et al., 2011). The present pattern ofresults adds to this evidence by showing that suchspecific impact of literacy as the one on mirror-image processing starts to emerge with letterknowledge before children are able to decode. Italso explains why the preschoolers examined herealready present a mirror cost in shape-based judg-ments of letters, which was also associated withtheir letter knowledge. This latter result is consis-tent with former observations that preschoolerswho are able to correctly write their names withoutmirrored errors are also able to discriminate mir-rored pairs of geometric shapes (Casey, 1984;Casey, 1986). It might seem at odds with the origi-nal mirror invariance of the ventral visual system(e.g., Dehaene, 2009; Logothetis et al., 1995), butprior studies have shown that the emergence of let-ter-specialized processing begins before formal liter-acy instruction in both preliterate children andilliterate adults (e.g., Cantlon et al., 2011; Fernandeset al., 2014).

Perceptual expertise with letters can explain theremarkable advantage of first graders on bothshape- and orientation-based judgments of lettersafter only ~8 months of reading instruction. On thedownside, it also explains first graders worse per-formance on shape-based judgments of mirroredand rotated pairs of reversible compared to nonre-versible letters. Experts usually show less flexibilityin selectively ignoring the dimensions relevant totheir category of expertise (e.g., Folstein et al.,2013), which explains orientation interference onletter recognition by adult readers (e.g., Corballis &Nagourney, 1978; Jolicoeur & Landau, 1984; Pegadoet al., 2011; Pegado, Nakamura, et al., 2014) andthe orientation costs found here on letters shape-based judgments by first graders. In fact, their ori-entation cost for reversible letters was so strongthat it canceled out any advantage over preschool-ers.

This pattern of results seems to agree with theresults found by Perea et al. (2011) in literateadults, using the masked priming paradigm:


Mirrored versions of reversible letters (e.g., ibea,mirrored letter underlined) significantly inhibitedthe recognition of target words (i.e., IDEA). Giventhe present observation of both mirror and rotationcosts in first-graders shape-based judgments ofreversible letters, it remains to be confirmedwhether the mirror interference reported by Pereaet al. is exclusive for mirror images. It could ratherbe due to activation of an existing letter representa-tion that is incompatible with the target word.Thus, the same interference would be expected forrotated versions of reversible letters (e.g., if ipeapreceded the target IDEA). Future research shouldexamine this prediction.

The association between explicit orientation pro-cessing of both linguistic and nonlinguistic materialand first-graders reading skills suggests that forthese beginning readers mirror discrimination is notyet fully accomplished and might continue todevelop after first grade (Cornell, 1985). For thesechildren, mirror discrimination continued to beharder than plane-rotation discrimination for bothgeometric shapes (i.e., average accuracy of 80.0%for mirrored pairs vs. 91.0% for rotated pairs, seeFigure 2) and letters (with slower performance onmirrored than rotated pairs); and this continues tobe the case in adults, even after years of readingpractice (Fernandes & Kolinsky, 2013; Gregoryet al., 2011; Kolinsky et al., 2011). Thus, mirror dis-crimination is triggered by learning a script withmirrored symbols, but it is not a dichotomous phe-nomenon fully determined by literacy. The originalmirror invariance of the visual recognition systemis not fully erased (Dehaene, 2009) and couldinstead be inhibited during recognition of visualobjects, including letters (e.g., Du~nabeitia, Molinaro,& Carreiras, 2011; Perea et al., 2011).

Neuropsychological, functional magnetic reso-nance imaging, (fMRI), and transcranial magneticstimulation studies have shown that mirror discrim-ination of linguistic and nonlinguistic objects hasdifferent neurocognitive loci. For linguistic material,mirror discrimination is underpinned by ventraloccipitotemporal regions, which are mirror invari-ant for pictures of familiar objects (e.g., Dehaene,Nakamura, et al., 2010; Nakamura, Makuuchi, &Nakajima, 2014; Pegado et al., 2011; Vinckier et al.,2006). What is, however, unclear is the temporallocus and the cognitive mechanism responsible formirror discrimination. Although beyond the scopeof the present work, this is still hotly debated, andtwo accounts have been proposed. According toone account, mirror discrimination occurs due toinhibition of mirror invariance at a late, possibly

attention-dependent, stage of visual processing(Du~nabeitia et al., 2011; Du~nabeitia et al., 2013;Perea et al., 2011). The other account proposes thatmirror discrimination becomes part of visual pro-cessing from an early time window (i.e., 100148 ms after stimulus onset; Pegado, Comerlato,et al., 2014), and evidence in favor of both has beenon the table.

These mixed results could be due to the adoptionof different paradigms, tasks, and materials, becausedifferent experimental conditions tap into differentphases of visual processing. In fact, this could alsobe the reason for the discrepancy between the mirrorcost that we found for shape-based judgments ofgeometric shapes by first graders and the mirrorinvariance found by Wakui et al. (2013) for short-term priming of familiar objects by children. Besideslow-level differences between the materials used, thelatter paradigm may tap into an earlier processingstage than the samedifferent task used in the pre-sent study (for discussions, see Kolinsky et al., 2011;Nakamura et al., 2005).

In addition to the aforementioned theoreticalimplications, the present study can also contributeto the growing interest from multiple developmen-tal perspectives on childrens print awareness andon its unique contribution to reading acquisition.Indeed, in parentchild conversations, more visualattributes are used to describe letters than pictures,and not only parents but also children emphasizeletters visual properties (Robins, Treiman, Rosales,& Otake, 2012), as if (at least implicitly) they recog-nize the importance of visual features on letterlearning and subsequent reading development. Theengagement in these conversations, especially aboutthe childs initial, was associated with better read-ing outcomes even after other factors, such asvocabulary, were controlled for (Treiman et al.,2015). More important, even before children knowwhat letters represent (i.e., the letter-sound corre-spondence), they are already sensitive to lettersvisual statistical patterns (Pollo, Kessler, & Treiman,2009; Treiman, Cohen, Mulqueeny, Kessler, &Schechtman, 2007; Treiman & Kessler, 2011).Preschoolers are better at copying and writing let-ters with the most frequent arrangement of visualfeatures in the Latin alphabet, that is, letters with ahasta on the left and a coda on the right (e.g., band F) than letters with the opposite arrangement(Pollo et al., 2009) and, hence, make more mirrorederrors on letters of the latter type (e.g., writing binstead of d; Treiman & Kessler, 2011). The presentstudy adds to this literature, showing that mirrordiscrimination, which is a necessary condition for


mastering the Latin alphabet, can be promoted byliteracy-related activities about letter forms, and thiscould happen at home during parentchildren inter-actions or at the kindergarten. Additionally, ourresults show that training orientation discriminationin general is not the best practice; preschoolers donot have difficulties with discrimination of planerotations, and this ability is not related to mirrordiscrimination. It is letter knowledge and familiaritywith letter forms that are the key. Thus, our workis part of an emergent line of research showing thatliteracy has a visual facet, crucial for learning toread, to which letter knowledge strongly con-tributes.

References

Cantlon, J. F., Pinel, P., Dehaene, S., & Pelphrey, K. A.(2011). Cortical representations of symbols, objects, andfaces are pruned back during early childhood. CerebralCortex, 21, 191199. doi:10.1093/cercor/bhq078

Casey, M. B. (1984). Individual differences in use of leftright visual cues: A reexamination of mirror-image con-fusions in preschoolers. Developmental Psychology, 20,551559. doi:10.1037/0012-1649.20.4.551

Casey, M. B. (1986). Individual differences in selectiveattention among prereaders: A key to mirror-imageconfusions. Developmental Psychology, 22, 5866.doi:10.1037/0012-1649.22.1.58

Chinello, A., Cattani, V., Bonfiglioli, C., Dehaene, S., &Piazza, M. (2013). Objects, numbers, fingers, space:Clustering of ventral and dorsal functions in youngchildren and adults. Developmental Science, 16, 377393.doi:10.1111/desc.12028

Corballis, M. C., & Nagourney, B. A. (1978). Latency tocategorize disoriented alphanumeric characters as let-ters or digits. Canadian Journal of Psychology/Revue cana-dienne de psychologie, 32, 186188. doi:10.1037/h0081685

Cornell, J. M. (1985). Spontaneous mirror-writing in chil-dren. Canadian Journal of Psychology/Revue CanadienneDe Psychologie, 39, 174179. doi:10.1037/H0080122

Danziger, E., & Pederson, E. (1998). Through the looking-glass: Literacy, writing systems and mirror-image dis-crimination. Written Language and Literacy, 1, 153164.

Dehaene, S. (2009). Reading in the brain: The new science ofhow we read. New York, NY: Penguin.

Dehaene, S., Nakamura, K., Jobert, A., Kuroki, C.,Ogawa, S., & Cohen, L. (2010). Why do children makemirror errors in reading? Neural correlates of mirrorinvariance in the visual word form area. NeuroImage,49, 18371848. doi:10.1016/j.neuroimage.2009.09.024

Dehaene, S., Pegado, F., Braga, L. W., Ventura, P., Nunes,G., Jobert, A., . . . Cohen, L. (2010). How learning toread changes the cortical networks for vision andlanguage. Science, 330, 13591364. doi:10.1126/science.1194140

Du~nabeitia, J. A., Dimitropoulou, M., Estevez, A., & Car-reiras, M. (2013). The influence of reading expertise inmirror-letter perception: Evidence from beginning andexpert readers. Mind, Brain, and Education, 7, 124135.doi:10.1111/mbe.12017

Du~nabeitia, J. A., Molinaro, N., & Carreiras, M. (2011).Through the looking-glass: Mirror reading. NeuroImage,54, 30043009. doi:10.1016/j.neuroimage.2010.10.079

Evans, M. A., Saint-Aubin, J., & Landry, N. (2009). Letternames and alphabet book reading by senior kinder-garteners: An eye movement study. Child Development,80, 18241841.

Fernandes, T., & Kolinsky, R. (2013). From hand to eye:The role of literacy, familiarity, graspability, andvision-for-action on enantiomorphy. Acta Psychologica,142, 5161. doi:10.1016/j.actpsy.2012.11.008

Fernandes, T., Vale, A. P., Martins, B., Morais, J., & Kolin-sky, R. (2014). The deficit of letter processing in devel-opmental dyslexia: Combining evidence from dyslexics,typical readers and illiterate adults. DevelopmentalScience, 17, 125141. doi:10.1111/desc.12102

Folstein, J. R., Palmeri, T. J., & Gauthier, I. (2013). Cate-gory learning increases discriminability of relevantobject dimensions in visual cortex. Cerebral Cortex, 23,814823. doi:10.1093/cercor/bhs067

Gauthier, I., Hayward, W. G., Tarr, M. J., Anderson, A.W., Skudlarski, P., & Gore, J. C. (2002). BOLD activityduring mental rotation and viewpoint-dependent objectrecognition. Neuron, 34, 161171. doi:10.1016/S0896-6273(02)00622-0

Gibson, E. J., Pick, A. D., Osser, H., & Gibson, J. J. (1962).A developmental study of discrimination of letter-likeforms. Journal of Comparative and Physiological Psychol-ogy, 55, 897906. doi:10.1037/h0043190

Gregory, E., Landau, B., & McCloskey, M. (2011). Repre-sentation of object orientation in children: Evidencefrom mirror-image confusions. Visual Cognition, 19,10351062. doi:10.1080/13506285.2011.610764

Hannagan, T., Amedi, A., Cohen, L., Dehaene-Lambertz,G., & Dehaene, S. (2015). Origins of the specializationfor letters and numbers in ventral occipitotemporal cor-tex. Trends in Cognitive Sciences, 19, 374382.doi:10.1016/j.tics.2015.05.006

Harris, I. M., Benito, C. T., Ruzzoli, M., & Miniussi, C.(2008). Effects of right parietal transcranial magneticstimulation on object identification and orientationjudgments. Journal of Cognitive Neuroscience, 20, 916926. doi:10.1162/jocn.2008.20513

Jolicoeur, P., & Landau, M. J. (1984). Effects of orientationon the identification of simple visual patterns. CanadianJournal of Psychology/Revue canadienne de psychologie, 38,8093. doi:10.1037/h0080782

Kolinsky, R., & Fernandes, T. (2014). A culturalside effect: Learning to read interferes with iden-tity processing of familiar objects. Special Issue: :The impact of learning to read on visual processing.Frontiers in Psychology, 5, 1224. doi:10.3389/fp-syg.2014.01224


http://dx.doi.org/10.1093/cercor/bhq078http://dx.doi.org/10.1037/0012-1649.20.4.551http://dx.doi.org/10.1037/0012-1649.22.1.58http://dx.doi.org/10.1111/desc.12028http://dx.doi.org/10.1037/h0081685http://dx.doi.org/10.1037/H0080122http://dx.doi.org/10.1016/j.neuroimage.2009.09.024http://dx.doi.org/10.1126/science.1194140http://dx.doi.org/10.1126/science.1194140http://dx.doi.org/10.1111/mbe.12017http://dx.doi.org/10.1016/j.neuroimage.2010.10.079http://dx.doi.org/10.1016/j.actpsy.2012.11.008http://dx.doi.org/10.1111/desc.12102http://dx.doi.org/10.1093/cercor/bhs067http://dx.doi.org/10.1016/S0896-6273(02)00622-0http://dx.doi.org/10.1016/S0896-6273(02)00622-0http://dx.doi.org/10.1037/h0043190http://dx.doi.org/10.1080/13506285.2011.610764http://dx.doi.org/10.1016/j.tics.2015.05.006http://dx.doi.org/10.1162/jocn.2008.20513http://dx.doi.org/10.1037/h0080782http://dx.doi.org/10.3389/fpsyg.2014.01224http://dx.doi.org/10.3389/fpsyg.2014.01224

Kolinsky, R., Verhaeghe, A., Fernandes, T., Mengarda, E.J., Grimm-Cabral, L., & Morais, J. (2011). Enantiomor-phy through the looking glass: Literacy effects on mir-ror-image discrimination. Journal of ExperimentalPsychology: General, 140, 210238. doi:10.1037/A0022168

Logothetis, N. K., Pauls, J., & Poggio, T. (1995). Shaperepresentation in the inferior temporal cortex of mon-keys. Current Biology, 5, 552563. Available at http://www.ncbi.nlm.nih.gov/pubmed/7583105

Macmillan, N. A., & Creelman, C. D. (Eds.). (2005). Detec-tion theory: A users guide (2nd ed.). Mahwah, NJ: Erl-baum.

McBride-Chang, C., Zhou, Y. L., Cho, J. R., Aram, D.,Levin, I., & Tolchinsky, L. (2011). Visual spatial skill: Aconsequence of learning to read? Journal of ExperimentalChild Psychology, 109, 256262. doi:DOI 10.1016/j.-jecp.2010.12.003

McCandliss, B. D., Cohen, L., & Dehaene, S. (2003). Thevisual word form area: Expertise for reading in thefusiform gyrus. Trends in Cognitive Sciences, 7, 293299.doi:10.1016/s1364-6613(03)00134-7

Nakamura, K., Dehaene, S., Jobert, A., Le Bihan, D., &Kouider, S. (2005). Subliminal convergence of Kanji andKana words: Further evidence for functional parcella-tion of the posterior temporal cortex in visual wordperception. Journal of Cognitive Neuroscience, 17, 954968. doi:10.1162/0898929054021166

Nakamura, K., Makuuchi, M., & Nakajima, Y. (2014).Mirror-image discrimination in the literate brain: Acausal role for the left occpitotemporal cortex. SpecialIssue: Research topic: The impact of learning to read onvisual processing. Frontiers in Psychology, 5, 478.doi:10.3389/fpsyg.2014.00478

Nelson, R. O., & Peoples, A. (1975). A stimulus-responseanalysis of letter reversals. Journal of Literacy Research, 7,329340. doi:10.1080/10862967509547152

Pederson, E. (2003). Mirror-image discrimination amongnonliterate, monoliterate, and biliterate Tamil subjects.Written Language and Literacy, 6, 7191. doi:10.1075/wll.6.1.04ped

Pegado, F., Comerlato, E., Ventura, F., Jobert, A., Naka-mura, K., Buiatti, M., . . . Dehaene, S. (2014). Timing theimpact of literacy on visual processing. Proceedings of theNational Academy of Sciences of the United States of America,111, E5233E5242. doi:10.1073/pnas.1417347111

Pegado, F., Nakamura, K., Braga, L. W., Ventura, P.,Filho, G. N., Pallier, C., . . . Dehaene, S. (2014). Liter-acy breaks mirror invariance for visual stimuli: Abehavioral study with adult illiterates. Journal of Exper-imental Psychology: General, 143, 887894. doi:10.1037/a0033198

Pegado, F., Nakamura, K., Cohen, L., & Dehaene, S.(2011). Breaking the symmetry: Mirror discriminationfor single letters but not for pictures in the visual wordform area. NeuroImage, 55, 742749. doi:10.1016/j.neu-roimage.2010.11.043

Perea, M., Moret-Tatay, C., & Panadero, V. (2011). Sup-pression of mirror generalization for reversible letters:Evidence from masked priming. Journal of Memory andLanguage, 65, 237246. doi:10.1016/j.jml.2011.04.005

Pollo, T. C., Kessler, B., & Treiman, R. (2009). Statisticalpatterns in childrens early writing. Journal of Experi-mental Child Psychology, 104, 410426. doi:10.1016/j.-jecp.2009.07.003

Reis, A., Faisca, L., Castro, S. L., & Petersson, K. M.(2013). Reading predictors across schooling. In L. M.Morgado & M. L. Vale-Dias (Eds.), Desenvolvimento eEducac~ao [Development and education] (pp. 31173132)Coimbra, Portugal: Almedina.

Robins, S., Treiman, R., Rosales, N., & Otake, S. (2012).Parentchild conversations about letters and pictures.Reading and Writing, 25, 20392059. doi:10.1007/s11145-011-9344-5

Sim~oes, M. (2000). Investigac~oes no ambito da aferic~ao nacio-nal do Teste das Matrizes Progressivas Coloridas de Raven(M.P.C.R.) [Research for national measurement of theColored Progressive Matrices of Raven (C.P.M.R.)]. Lis-boa, Portugal: Fundac~ao Calouste Gulbenkian.

Sucena, A., & Castro, S. L. (2008). Aprender a ler e Avaliara Leitura [Learning how to read and the assessment ofreading]. Coimbra, Portugal: Almedina.

Szwed, M., Ventura, P., Querido, L., Cohen, L., &Dehaene, S. (2012). Reading acquisition enhances anearly visual process of contour integration. Developmen-tal Science, 15, 139149. doi:10.1111/j.1467-7687.2011.01102.x

Tarr, M. J., & Pinker, S. (1989). Mental rotation and orien-tation-dependence in shape recognition. Cognitive Psy-chology, 21, 233282. doi:10.1016/0010-0285(89)90009-1

Treiman, R., Cohen, J., Mulqueeny, K., Kessler, B., &Schechtman, S. (2007). Young childrens knowledgeabout printed names. Child Development, 78, 14581471.doi:10.1111/j.1467-8624.2007.01077.x

Treiman, R., & Kessler, B. (2011). Similarities among theshapes of writing and their effects on learning. WrittenLanguage and Literacy, 14, 3957.

Treiman, R., Schmidt, J., Decker, K., Robins, S., Levine, S.C., & Demir, O. E. (2015). Parents talk about letterswith their young children. Child Development, 86, 14061418. doi:10.1111/cdev.12385

Vinckier, F., Naccache, L., Papeix, C., Forget, J., Hahn-Barma, V., Dehaene, S., & Cohen, L. (2006). Whatand where in word reading: Ventral coding of writ-ten words revealed by parietal atrophy. Journal of Cog-nitive Neuroscience, 18, 19982012. doi:10.1162/jocn.2006.18.12.1998

Wakui, E., Juttner, M., Petters, D., Kaur, S., Hummel, J.E., & Davidoff, J. (2013). Earlier development of analyt-ical than holistic object recognition in adolescence. PLoSONE, 8, e61041. doi:10.1371/journal.pone.0061041

Wechsler, D. (1997). The Wechsler memory scale (3rd ed.).San Antonio, TX: The Psychological Corporation.

http://dx.doi.org/10.1037/A0022168http://www.ncbi.nlm.nih.gov/pubmed/7583105http://www.ncbi.nlm.nih.gov/pubmed/7583105http://dx.doi.org/DOI 10.1016/j.jecp.2010.12.003http://dx.doi.org/DOI 10.1016/j.jecp.2010.12.003http://dx.doi.org/10.1016/s1364-6613(03)00134-7http://dx.doi.org/10.1162/0898929054021166http://dx.doi.org/10.3389/fpsyg.2014.00478http://dx.doi.org/10.1080/10862967509547152http://dx.doi.org/10.1075/wll.6.1.04pedhttp://dx.doi.org/10.1075/wll.6.1.04pedhttp://dx.doi.org/10.1073/pnas.1417347111http://dx.doi.org/10.1037/a0033198http://dx.doi.org/10.1037/a0033198http://dx.doi.org/10.1016/j.neuroimage.2010.11.043http://dx.doi.org/10.1016/j.neuroimage.2010.11.043http://dx.doi.org/10.1016/j.jml.2011.04.005http://dx.doi.org/10.1016/j.jecp.2009.07.003http://dx.doi.org/10.1016/j.jecp.2009.07.003http://dx.doi.org/10.1007/s11145-011-9344-5http://dx.doi.org/10.1007/s11145-011-9344-5http://dx.doi.org/10.1111/j.1467-7687.2011.01102.xhttp://dx.doi.org/10.1111/j.1467-7687.2011.01102.xhttp://dx.doi.org/10.1016/0010-0285(89)90009-1http://dx.doi.org/10.1111/j.1467-8624.2007.01077.xhttp://dx.doi.org/10.1111/cdev.12385http://dx.doi.org/10.1162/jocn.2006.18.12.1998http://dx.doi.org/10.1162/jocn.2006.18.12.1998http://dx.doi.org/10.1371/journal.pone.0061041

Into the Looking Glass: Literacy Acquisition and Mirror ... · Into the Looking Glass: Literacy...

Documents

Transcript of Into the Looking Glass: Literacy Acquisition and Mirror ... · Into the Looking Glass: Literacy...