Brain & Language 141 (2015) 5061Contents lists available at ScienceDirect

Brain & Language

journal homepage: www.elsevier .com/locate /b&lVisuospatial complexity modulates reading in the brainhttp://dx.doi.org/10.1016/j.bandl.2014.11.0100093-934X/ 2014 Elsevier Inc. All rights reserved.

Corresponding author at: National Brain Research Centre, Nainwal Mod, NH-8,Manesar 122051, Gurgaon, Haryana, India. Fax: +91 124 2338910.

E-mail address: nandini@nbrc.ac.in (N.C. Singh).Chaitra Rao, Nandini C. Singh National Brain Research Centre, India

a r t i c l e i n f o a b s t r a c tArticle history:Accepted 17 November 2014

Keywords:Visual word recognitionReadingLanguageOrthographyVisuospatialfMRIBrainNeurocognitive processing of orthographic visuospatial complexity was examined through fMRI-basedovert naming (n = 16) of phonologically transparent, high and low frequency Hindi/Devanagari wordsthat were visually simple (, ) or complex ( , ). Participants overt behavior was mod-estly influenced by visuospatial complexity (accuracy: main effect p = .01, complexity frequency inter-action p < .07), while neuroimaging data revealed a robust effect of complexity (main effect FWE p < 104,complexity frequency interaction FWE p < 7 108). Interaction-based RoIs showed higher BOLDresponse in the VWFA to complex and left posterior temporal cortex to simple words, with greater rightlingual de-activation to complex than simple words. Subtractions confirmed additional recruitment ofVWFA, right frontal, inferior orbitofrontal, mid-temporal pole and left cerebellum by visuospatially com-plex over simple words. Finally, low frequency words activated bilateral occipital and putamen areas, leftIPL, SPL, IFG and VWFA, suggesting that effortful phonological processing in alphasyllabic Hindi/Devana-gari requires neural resources specialized for both visuospatially simple and complex orthographies.

2014 Elsevier Inc. All rights reserved.1. Introduction

Although writing systems of the world differ obviously in theirvisuospatial complexity, that is, in the number and spatial configu-ration of their visual features, research on reading has largelybypassed a direct examination of this dimension. In contrast withwriting systems based on the Roman alphabet like English, Frenchand German, alphabets of Semitic origin like Arabic, Hebrew, Per-sian and Urdu, as well as non-alphabetic systems like Chinese,Japanese and Korean have a noticeably greater visuospatialcomplexity.

The scant literature available suggests that visuospatial com-plexity does influence both cognitive processing and its neural cor-relates. An early study by Shimron and Sivan (1994) demonstratedthat HebrewEnglish bilingual readers took significantly longer toread passages in the visuospatially complex Hebrew orthographyas compared to their translation equivalents in English, despitebeing more comfortable with Hebrew than with English. A morerecent study showed that ArabicHebrew bilinguals were slowerat processing Arabic compared to Hebrew in a visual trail-makingtest (in which participants find and connect successive letters and/or numbers), a difference which the authors attributed to thegreater visual complexity of Arabic orthography compared toHebrew (Ibrahim, Eviatar, & Aharon-Peretz, 2002).

Other studies have found readers to be faster and more accurateat processing visually simple compared to complex orthographies.For example, Feldman and Turvey (1980) observed that Japanesereaders were slower at naming colours written using complex Kan-ji characters compared to the same words written in visually sim-ple Kana, despite the Kanji forms being the orthographically legalspelling of these words. Simpson and Kang (1994) reported longernaming latencies for Korean words written using complex Hanjacompared to simple Hangul. Rao, Vaid, Srinivasan, and Chen(2011) found that UrduHindi biscriptal readers were slower andmore error-prone in reading words written in their native, visuo-spatially complex Urdu than in the simpler Hindi script.

However, the visuospatially complex orthographies targeted bythese studies were also characterized by a lack of phonologicaltransparency, that is, by spelling systems which made phonology(word sound) relatively difficult to recover from their written ororthographic representation. Investigators therefore attributedthe greater cost of processing the complex orthographies in theabove studies to their relatively opaque orthography-to-phonologymapping, rather than to their complex visuospatial layout.

Comparisons in neuroimaging experiments have also adducedevidence to show that visuospatially complex orthographiesimpose an extra neurocognitive load during reading (Kuo et al.,2001; Lee, 2004; Nakamura, Dehaene, Jobert, Le Bihan, & Kouider,

http://crossmark.crossref.org/dialog/?doi=10.1016/j.bandl.2014.11.010&domain=pdfhttp://dx.doi.org/10.1016/j.bandl.2014.11.010mailto:nandini@nbrc.ac.inhttp://dx.doi.org/10.1016/j.bandl.2014.11.010http://www.sciencedirect.com/science/journal/0093934Xhttp://www.elsevier.com/locate/b&l

C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061 512005). Nevertheless, the complex orthographies investigated byneuroimaging studies have also been characterized by a lack ofphonological transparency, resulting in varied interpretations ofthe results.

1.1. Visuospatial complexity in the VWFA

With respect to the neural processing of orthographic visuospa-tial complexity, the middle section of the left fusiform gyrus(abbreviated FG) and widely termed the visual word form area orVWFA would appear to be an ideal candidate as a processing centerfor complex orthographic features. The role of this region has beenintensely debated, with one approach holding that the area is spe-cialized for reading-related orthographic processing (Dehaene &Cohen, 2011), while a competing view maintains that the FG/VWFA integrates visuospatial input with previously acquired infor-mation from other modalities, thus making it an ideal node fororthographic processing (Price & Devlin, 2011).

Evidence supporting the former view shows the FG/VWFA to besensitive towordlikeness, that is, to exhibit heightened responses tostimuli approximating real words (Binder, Medler, Westbury,Liebenthal, & Buchanan, 2006; Cohen et al., 2002; Price, Wise, &Frackowiak, 1996). The idea that the area is specialized for readingis further reinforced by reports that damage to the FG/VWFAresults in pure alexia, that is, in an inability to read despite normalvision (Binder & Mohr, 1992; Damasio & Damasio, 1983). On theother hand, the theory that the FG/VWFA carries out multisensoryintegration is strengthened by finding picture word priming inthis region (Kherif, Josse, & Price, 2011). Perhaps more convinc-ingly, this view is bolstered by recent evidence that pure alexicsexhibit not only impaired reading, but also perform poorly on othertasks requiring finer visuospatial discrimination (Roberts et al.,2013).

Notwithstanding differences in their theoretical bent, studiesacross multiple languages have highlighted the VWFA as a compo-nent of the visual word recognition network (Bolger, Perfetti, &Schneider, 2005; Jobard, Crivello, & Tzourio-Mazoyer, 2003;McCandliss, Cohen, & Dehaene, 2003; Tan, Laird, Li, & Fox, 2005).The critical importance of the VWFA to orthographic processinghas been further underlined by research tracing the developmentof FG/VWFA sensitivity to word-like stimuli in young children(Ben-Shachar, Dougherty, Deutsch, & Wandell, 2011; Brem et al.,2010) and in illiterate adults learning to read (Dehaene et al.,2010).

Indeed, research in several alphabetic languages points to a rolefor the FG/VWFA in orthography-to-phonology conversion (Fiez,Balota, Raichle, & Petersen, 1999; Kronbichler et al., 2004;Mechelli et al., 2005; Paulesu et al., 2000). Early experiments dem-onstrated a clear word frequency effect in the VWFA of readers ofalphabetic orthographies like English, German and Italian (Fiezet al., 1999; Kronbichler et al., 2004; Paulesu et al., 2000). The wordfrequency effect is a universally prevalent phenomenon per whichreaders are slower and less accurate at identifying less familiar orlow frequency compared to commonly encountered or high fre-quency words (Seidenberg, Waters, Barnes, & Tanenhaus, 1984),and is thought to arise because of the greater effort required toretrieve or assemble the phonology of a seldom encountered ornovel word. As such, the word frequency effect has been widelyexploited to evaluate the behavioral and neural mechanisms ofphonological processing.

The connection between the VWFA and phonological retrievalwas reinforced when Dehaene et al. (2004) showed that even visu-ally dissimilar prime-target pairs that were phonologically identi-cal (e.g., RADIO ) produced a significant repetitionsuppression effect, or a reduced response to repeated stimulationin this area. Mechelli et al. (2005) further demonstrated increasedfunctional connectivity during low frequency word readingbetween the FG/VWFA and the left inferior frontal gyrus, parsopercularis region (LIFG/pO), a known center of phonological pro-cessing. Recent evidence suggests that the VWFA is sensitive to lowfrequency words even in visually complex orthographies like Japa-nese Kana and Kanji (Twomey et al., 2013).

The role of the FG/VWFA in extracting phonological informationfrom orthographic input may easily be accounted for by both the-ories of VWFA function outlined above (Dehaene & Cohen, 2011;Price & Devlin, 2011). Of greater relevance to the present research,nevertheless, is the fact that the two approaches give rise to diver-gent predictions with respect to orthographic visuospatialcomplexity.

If the VWFA specializes in reading-related orthographic pro-cessing, its activation should be driven by the degree of difficultyin retrieving linguistic information from the orthographic featuresof a stimulus. For example, the frequency of lexical and sub-lexicalunits, orthographic neighborhood size and density, and spelling-sound transparency should affect VWFA response. At the sametime, the strength of VWFA activation should be independent oforthographic features that do not vary systematically in their levelof linguistic difficulty, as in the case of the visuospatially complexfeatures of Hindi/Devanagari. By contrast, the theory that the FG/VWFA integrates visuospatial with abstract multimodal informa-tion would predict that VWFA activation should correspond tothe effort of processing orthographic features, whether or not theycovary with linguistic difficulty. Indeed, the latter view wouldencourage the prediction that FG/VWFA response should reflectthe combined influence of linguistic and non-linguistic ortho-graphic difficulty of a given stimulus.

Results of a few studies have suggested that the FG/VWFA isaffected by orthographic visuospatial complexity. These studiesreported heightened neural activity in the FG/VWFA BA37 incomparisons of visually complex against simple orthographies,including Kanji versus Kana in Japanese (Ha Duy Thuy et al.,2004; Ino, Nakai, Azuma, Kimura, & Fukuyama, 2009; Nakamura,Dehaene, et al., 2005; Twomey et al., 2013) and Hanja versus Han-gul in Korean (Lee, 2004). In their meta-analytical review of theword recognition literature on Chinese and Western alphabeticorthographies, Tan et al. (2005) similarly found additional FG/VWFA response in a contrast of Chinese against alphabetic wordrecognition.

Nevertheless, two of the above studies specifically refuted thenotion that VWFA activity is correlated with the neural effort ofprocessing visually complex characters (Lee, 2004; Twomeyet al., 2013). Lee (2004) claimed that an unpublished Hanja versusHangul (visually complex versus simple) contrast which scrambledthe same characters revealed no additional FG activation by Hanja.And although Twomey et al. (2013) reported additional FG/VWFAactivation by Kanji compared to Hiragana, they found that addingthe number of strokes per character as an index of visual complex-ity in their analysis did not account for this additional activation.

Evidently, despite the recognized importance of the FG/VWFAin word recognition, its exact role in processing visuospatiallycomplex orthographies remains to be delineated. FG/VWFAinvolvement in phonological processing also requires further veri-fication, especially in non-alphabetic, visually complexorthographies.

1.2. Visuospatial complexity versus phonological processing

Apart from the FG/VWFA, several studies have implicated cer-tain regions of the brain in processing orthographic visuospatialcomplexity, including the left middle frontal gyrus (LMFG) BA9/6, the superior parietal lobule (SPL) BA7, and parts of thebilateral occipitotemporal cortex or OTC (Ha Duy Thuy et al.,

52 C. Rao, N.C. Singh / Brain & Language 141 (2015) 50612004; Ino et al., 2009; Kuo et al., 2001; Lee, 2004; Liu, Dunlap, Fiez,& Perfetti, 2007; Nakamura, Dehaene, et al., 2005; Nakamura et al.,2005; Tan et al., 2005; Twomey et al., 2013; Wu, Ho, & Chen, 2012;Yoon, Cho, Chung, & Park, 2005).

Studies of complex orthographies have attributed to the LMFGan important role in the visuospatial analysis of Chinese characters(Tan et al., 2000, 2001), as well as Japanese Kanji (Ino et al., 2009)and Korean Hanja (Yoon et al., 2005). However, other findings havealso linked this area with addressed phonological retrieval in Chi-nese (Tan et al., 2005), and even with tone processing (Bolger et al.,2005). A recent view implicated the LMFG in the integration oforthographic, phonological and semantic information (Liu et al.,2007).

Likewise, the SPL has been associated with visuospatial analysisof Japanese Kana and Kanji (Ha Duy Thuy et al., 2004), Korean Han-gul and Hanja (Lee, 2004; Yoon et al., 2005) and Chinese (Wu et al.,2012). While some studies found increased SPL activation in com-parisons of complex versus simple characters (Ino et al., 2009; Lee,2004), others suggested that this area is involved in phonologicalprocessing of complex orthographies like Chinese (Liu et al.,2007; Tan et al., 2001, 2005).

Greater bilateral OTC activation has similarly been consistentlyfound in complex compared to simple orthographies (Bolger et al.,2005; Chen, Fu, Iversen, Smith, & Matthews, 2002; Ha Duy Thuyet al., 2004; Ino et al., 2009; Ischebeck et al., 2004; Lee, 2004;Nakamura et al., 2005; Tan et al., 2005; Twomey et al., 2013;Yoon et al., 2005). Right fusiform gyrus (RFG) activity is typical inreaders of Chinese and Japanese Kanji (Bolger et al., 2005; Inoet al., 2009; Nakamura, Dehaene, et al., 2005; Tan et al., 2005),and has recently been shown to arise in English readers learningto read Chinese (Liu et al., 2007; Perfetti et al., 2007). Increasedbilateral lingual and right cuneus activation has been reported incomparisons of complex versus simple orthographiesChineseagainst Pinyin, Kanji over Kana, and Hanja more than Hangul(Chen et al., 2002; Ha Duy Thuy et al., 2004; Ino et al., 2009; Lee,2004; Nakamura, Dehaene, et al., 2005). Yet, right OTC activityhas also been linked to phonological processing in Chinese and Jap-anese (Ischebeck et al., 2004; Kuo et al., 2003; Twomey et al.,2013).

Given the close co-variance of visuospatial complexity and pho-nological transparency in the orthographies hitherto studied, it isnot surprising that diverse functions have been attributed to theobserved brain regions. A further compelling argument againstattributing a purely visuospatial function to one or more regionsidentified by previous studies lies in the fact that no study system-atically manipulated visuospatial complexity while controlling forother features of the stimuli.

1.3. Visuospatial complexity in Hindi/Devanagari

The present research examines the behavioral and neuraldynamics of processing complex visuospatial orthographic fea-tures in Hindi/Devanagari, henceforth Devanagari. Devanagari isone of the Indic group of alphasyllabaries, classified thus becauseit resembles a pure syllabary (e.g., Japanese Kana) in that eachgrapheme or akshara represents approximately one CV syllable.However, it also incorporates the alphabetic principle by distin-guishing the vowel and consonant components within an akshara(Kandhadai & Sproat, 2010; Nag & Snowling, 2012; Vaid & Gupta,2002).

Devanagari has separate aksharas for both consonants and vow-els, yet basic consonant aksharas represent CV syllables with aninherent // or schwa. Word-initial vowels are written using basicvowel aksharas, while others are typically represented by diacriticswritten to the right, left, top or bottom of the preceding consonant.Geminates and conjunct consonants are represented by ligaturinga secondary form of one consonant onto the basic akshara of theother. For example, the basic aksharas /d / and /n/ com-bine with vowel diacritics and (/u:/ and /i:/ respectively) toproduce (/du:n/, meaning valley) and (/nd i:/, river), whilethe word-initial /u:/ is written using the akshara in (/u:n/,wool). The secondary form representing is ligatured onto in (/nnd/, joy), and the secondary form (standing for , /r/) is ligatured onto (//) in (/drn/, view).

Devanagari aksharas have a one-to-one correspondence withthe phonemes they represent, resulting in high spelling-to-soundconsistency, or phonological transparency. Arguably the mostprominent feature of the orthography, however, is its visuospatial-ly complex layout, arising from the nonlinear arrangement of vow-els as well as consonant ligatures (see examples above). InDevanagari, as in other Indic alphasyllabaries, the complex visuo-spatial features systematically encode phonological information,and therefore a reasonable inference is that successful phonologi-cal retrieval in Devanagari requires accurate deciphering of visuo-spatial details.

The Devanagari orthography represents a potentially richsource of insight into the neural processing of visuospatially com-plex orthographic features. As illustrated by the above examples,the nonlinear positioning of vowel diacritics and ligatured conso-nants renders the bulk of the Devanagari lexicon visuospatiallycomplex. In addition, the presence of an inherent schwa in conso-nant aksharas, as well as the linear positioning of the vowel dia-critic (denoting the long /a:/) mean that a substantial numberof Devanagari words have a visuospatially simple orthographicconfiguration, for example (/da:n/, donation), (/na:/,intoxication) and (/nnd/, sister-in-law). Consequently,Devanagari provides the ideal platform for contrasting the pro-cessing of visuospatially simple and complex orthographicfeatures, unconfounded by differences in phonologicaltransparency.

Only a limited body of research exists on the neural processingof Devanagari, which, moreover, has not focused upon the process-ing of visuospatial complexity. The evidence so far, while suggest-ing that behavioral as well as neural costs are associated withprocessing the complex orthographic layout of Devanagari, lendsitself to diverse interpretations.

The sole behavioral study which examined the impact of com-plex visuospatial features on word reading in Devanagari foundthat visual complexity interacted with phonological processingeffort in influencing readers performance (Vaid & Gupta, 2002).Neuroimaging studies have shown that word recognition in Deva-nagari elicits activity in regions hitherto associated with visuospa-tially complex orthographies, including the FG/VWFA, LMFG, SPLand right OTC. Nevertheless, interpreting past findings from Deva-nagari in terms of visuospatial complexity is difficult, for multiplereasons.

Firstly, there is a lack of convergence in the findings from Deva-nagari. To illustrate, FG/VWFA activation while reading in Devana-gari has been documented by only one study on phrase reading(Das, Kumar, Bapi, Padakannaya, & Singh, 2009), but not by studiesof word reading (Das, Bapi, Padakannaya, & Singh, 2011; Das,Padakannaya, Pugh, & Singh, 2011; Kumar et al., 2009), althoughDas, Bapi, et al. (2011) observed activation in the right homologof the VWFA. Moreover, the role of the VWFA in processing Deva-nagari has not been investigated by any study.

Activation in the left superior parietal lobule (LSPL) BA7/40was observed during Devanagari word reading (Das, Bapi, et al.,2011), whereas phrase reading studies by Das and colleagues(Das et al., 2009; Kumar et al., 2009) reported activity in a homol-ogous part of the right superior parietal lobule (RSPL). Similarly,the right OTC was activated by Devanagari phrases (Das et al.,2009; Kumar et al., 2009), but not by single words (Das, Bapi,

C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061 53et al., 2011; Das, Padakannaya, et al., 2011). Finally, LMFG activa-tion was associated with phonological rather than visuospatialprocessing of Devanagari, when Das, Padakannaya, et al. (2011)found greater LMFG response to words of low contrasted with highfrequency; however, only monolingual Devanagari readers in thisstudy exhibited LMFG activity, whereas bilingual DevanagariEng-lish readers did not.

A second important factor in considering the Devanagari litera-ture is that none of the studies conducted so far deliberately inves-tigated visuospatial complexity. Indeed, these studies did not varythe complexity of stimuli, but limited themselves to only linearDevanagari words (e.g., /sa:t/, seven), that is, words with novowel diacritics above or below the central plane of text, and noligatured consonants. A final, methodological consideration is thatall studies contrasted word recognition in Devanagari with a pas-sive fixation condition, leaving open the possibility that activationsattributed to the visuospatial configuration of the orthographymight instead reflect lower level perceptual processing.1 The stimulus list (i = 240) was randomly divided into two sets (A and B) of 120,with 60 SIM and 60 COM words in each, and sets A (n = 24) and B (n = 19) wereadministered to different groups of raters so as to minimize rater fatigue. Ratingsobtained from these volunteers were further verified by polling a second group of 18raters who completed both sets A and B [HF versus LF, t(238) = 12.25, p < .001;SIM_HF versus SIM_LF, t(118) = 8.05, p < .001; COM_HF versus COM_LF, t(118) = 8.41,p < .001; SIM_HF versus COM_HF, t(118) = 1.63, p > .10; SIM_LF versus COM_LF t(118) = 0.95, p > .10].1.4. The current study

In the present study, therefore, we aimed to refine our under-standing of visual word recognition in Hindi/Devanagari, focusingparticularly on the neurocognitive mechanisms involved in pro-cessing visuospatial complexity. By contrasting visuospatially sim-ple with complex words within Devanagari, we thought to isolatethe extra cognitive and/or neural effort in processing visuospatiallycomplex orthographic features from any confound arising throughcontrasts in phonological transparency. Our visuospatially com-plex stimuli (COM) featured nonlinear vowel diacritics and liga-tured consonants, and represented readers typical experiencewith Devanagari text. Visuospatially simple stimuli (SIM) weretaken from Das, Bapi, et al. (2011) previous study of word recogni-tion in Devanagari, and featured words with basic CV and vowelaksharas as well as the linear diacritic .

Both COM and SIM lists included equal numbers of words ofhigh (HF) and low frequency (LF), resulting in an orthogonal varia-tion of complexity with frequency to yield four stimulus categories,namely visually simple high frequency (SIM_HF), visually simplelow frequency (SIM_LF), visually complex high frequency(COM_HF) and visually complex low frequency words (COM_LF).By comparing the processing of Devanagari words of high andlow frequency, we additionally sought to identify the mechanismsof effortful phonological retrieval, and to distinguish the influenceof visuospatial complexity on phonological processing.

We hypothesized that visuospatially complex (COM) wordswould be both behaviorally and neurally more challenging to pro-cess compared to simple (SIM) words. Based on previous findings,we further expected that cognitive task performance as well asneuroimaging data would additionally reveal a cost associatedwith processing low (LF) compared to high frequency words (HF).Finally, we thought that visuospatial complexity and word fre-quency would interact, such that, of the four stimulus categories,the greatest behavioral and neural effort would be apparent whenreading visually complex, low frequency (COM_LF) words.

With respect to the fMRI data, we predicted that one or morebrain regions hitherto associated with processing complex orthog-raphies, including the LMFG, LSPL and bilateral OTC would exhibitgreater activation in response to visuospatially complex than tosimple Devanagari words. Likewise, we anticipated that low com-pared to high frequency Devanagari words would elicit increasedactivity in (one or more) areas previously implicated in sub-lexicalphonological processing. We additionally predicated that an inter-action of visuospatial complexity with word frequency wouldreveal itself at the neural level through a systematic modulationof activity across the four stimulus categories in the abovemen-tioned brain regions.2. Materials and methods

2.1. Participants

A group of 18 neurologically normal, right-handed native read-ers of Hindi/Devanagari (4F, 14M) carried out an overt word read-ing task in the MRI scanner; data are reported here for only 16 ofthe 18 MRI participants, as two participants data were rejectedowing to excessive scanner artifacts. Participants gave informedconsent as per the requirements of the institutional human ethicscommittee.

All 16 participants whose data are reported here had receivedprimary and/or high school education through Hindi as the med-ium of instruction, and reported an average of 13.4 years(SD = 3.5) of formal Hindi instruction. In addition, participantshad received at least two years (M = 5.1, SD = 1.6) of formal educa-tion in Sanskrit, from which modern-day Hindi is descended. Fur-ther, all but one participant reported a minimum weekly readinghabit of 1 h in Hindi (M = 8.7, SD = 9.3). Finally, participants self-reported proficiency in reading Hindi on a 5-point scale averaged4.5 (SD = 0.6). Besides Hindi, English and Sanskrit, some partici-pants additionally reported familiarity with Punjabi (2), Gujarati(1), and Japanese (1).2.2. Materials

A set of 240 Hindi words was compiled to include 60 wordseach in the COM_HF, COM_LF, SIM_HF and SIM_LF categories.Visuospatially simple SIM words had no diacritics above or belowthe central plane and no ligatured consonants; these words con-tained only the schwa vowel and/or the long /a:/ whose diacritic falls within the central plane of text (e.g., /pa:lk/, spinach).Complex COM words included vowel diacritics above or below thecentral plane of text (e.g., /puls/, police), and some COMitems additionally included a ligatured consonant (e.g., /pm/, west).

Words were initially grouped into high and low frequency cat-egories based on the investigators mutual consensus, although apost hoc poll by an independent group of 441 proficient native read-ers responding to a 5-point scale revealed a significantly higher per-ceived frequency of stimuli in the high (M = 4.4, SD = 0.4) comparedto low frequency categories (M = 3.0, SD = 0.9), with t(238) = 11.57,p < .001. Separate comparisons yielded significant HF versus LF dif-ferences in both SIM [t(118) = 8.12, p < .001] and COM lists [t(118)= 9.29, p < .001]. Conversely, there were no differences betweenSIM and COM lists in the average ratings to HF [t(118) = 0.10,p > .10] or LF words [t(118) = 0.37, p > .10].

Due to orthotactic constraints, the phonemic structure of SIMand COMwords could not be completely matched, although effortswere made to include similar numbers of stimuli at each level ofphonemic length and syllable complexity in both lists. Phonemicstructure and syllable complexity of words was carefully matchedacross the HF and LF categoriesfor example, the SIM_HF item (/ba:l/, meaning hair) was matched with the SIM_LF item (/ba:

54 C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061z/, falcon), and the COM_HF word (/eb/, pocket) was matchedagainst the COM_LF (/eh/, husbands older brother). Overall,the stimuli included 117 monosyllabic, 119 disyllabic and 4 trisyl-labic words.

In addition, a set of five false-font strings was created for use inthe baseline blocks of the fMRI task, by combining componentstrokes of real Hindi/Devanagari letters in novel configurations(for example: ). The length of false-font strings was varied tomatch the range of lengths of real words used in the study. Allstimuli were typed using Yogesh bold font, size 100, and convertedinto individual bitmap images of black text against a white back-ground. The width of words spanned 2.335.16 cm on screen(approximately 1.22.6 of viewing angle), while height spanned1.242.43 cm (0.61.2).

2.3. Procedure

Participants viewed stimuli through a mirror assembly that pro-jected images displayed on a computer monitor situated outsidethe scanner. Within the birdcage head-coil, participants addition-ally wore a set of heavily foam-padded headphones which effec-tively muffled scanner noise, and also acted as a gentle restraintthat minimized head movement. Participants were instructed toname each word aloud clearly as soon as it appeared in the centerof the screen. Participants were additionally shown a sample false-font string, and asked to say, Hindi aloud to each appearance andchange in the false-font string, taking care to move their lips as lit-tle as possible throughout the experiment. Stimuli were presentedin alternating 20 s task and baseline blocks, wherein task blocksincluded 10 stimuli at 2 s/word, while baseline blocks comprised5 randomized false-font strings presented twice each per block.Baseline trials ensured discriminability of individual false-fontstrings by beginning with an initial blank gray screen (100 ms), fol-lowed by a fixation cursor + (150 ms) and then a false-font stringpresented for 1500 ms, followed finally by a 250 ms ISI.

COM words represented the full spectrum of visuospatial com-plexity typical of Devanagari, whereas SIM words were taken fromDas, Bapi, et al.s studies (2011) and Das, Padakannaya, et al.sstudies (2011) as a contrast meant to identify the additional neuraleffort involved in decoding visuospatial complexity. To maximizethis likelihood, COM stimuli were presented first to all participantsin runs separate from those of SIM stimuli. However, the order ofblocks within runs and stimuli within blocks was randomized.Stimulus presentation and data logging were controlled throughE-Prime v. 1.1 via the MRI-compatible IFIS interface (Invivo Corp,Gainesville, FL). Participants overt responses were relayed to anindependent computer and recorded by means of an MRI-compat-ible, dual-channel microphone and OptiMRI v. 2.4 software (Opto-acoustics Ltd., Yehuda, Israel).

2.3.1. Image acquisition and analysisFunctional images comprising 30 axial slices of 5 mm each

(1 mm gap) were acquired along the ACPC plane (matrix size64 64 mm, FOV 230 mm), using a 3T Philips Achieva scannerrunning a gradient echo pulse sequence (TR 2 s, TE 35 ms, flip angle90). In addition, a high-resolution T1-weighted structural scan (TR8.4 ms, TE 3.7 ms, flip angle 8) was acquired from each partici-pant. Data preprocessing and analyses were carried out usingSPM5 (http://www.fil.ion.ucl.ac.uk/spm). Functional images wererealigned for motion correction2 and then co-registered to a high-2 Following SPM5 guidelines, motion exceeding 3 mm in any of the x, y or z planes,or rotational movement (pitch, roll or yaw) in excess of 5 was set as a criterion forrejecting data with excessive head movement, but a check revealed all data in thisstudy to fall within these cutoffs. Individual motion parameters were additionallyentered as six separate motion regressors in the SPM analysis.resolution T1 image. The T1 image was normalized to MNI stereotac-tic space, and resultant parameters were used to normalize func-tional images, which were resampled to isotropic 2 2 2 mmvoxels and smoothed using an 8 mm FWHM Gaussian kernel. Infirst-level analyses, a canonical boxcar function was used in model-ing individual participants data from each of the four runs, whichwere then combined in a cumulative statistical parametric map. Ahigh-pass filter of 128 s corrected for signal drift, while the AR(1)model (Friston et al., 2002) was applied to offset autocorrelationeffects. Nine contrasts with respect to baseline activity were thencomputed per participanta global contrast of Devanagari wordreading (abbreviated ALL_WDS henceforth), individual contrasts ofthe four stimulus categories: COM_HF, COM_LF, SIM_HF and SIM_LF,as well as contrasts of all visually complex (COM), visually simple(SIM), all high frequency (HF) and low frequency words (LF).

At the second level, participants contrasts for the four individ-ual stimulus categories were first subjected to a 2 2 ANOVA toexamine the effects of complexity and frequency on neural activity.Data were then analyzed in direct subtractions (paired t-tests) ofSIM versus COM as well as HF versus LF in order to isolate brainregions specifically modulated by complexity and frequencyrespectively. Finally, an RoI analysis was carried out in order to fur-ther explore the nature of the interaction between visuospatialcomplexity and word frequency in Devanagari.3. Results

3.1. Behavioral results

Data files for two participants were not generated owing to asoftware malfunction; therefore behavioral data reported hereare summarized across the remaining 14 participants. The accu-racy of participants overt naming responses was manuallyencoded, and further processed through a custom-written softwarecode to extract vocal RT.3 Word naming accuracy and RT were thenaveraged across participants and analyzed in separate 2 2 repeatedmeasures ANOVA, with complexity and frequency as variables ofinterest. Only correct responses were included in RT analyses.Approximately 0.95% of the total data-points were further excludedowing to sub-threshold vocalizations (0.83%) or outlier responselatencies equal to or below 250 ms (0.12%).

Participants naming accuracy revealed significant main effectsof both complexity [F(1,13) = 9.08, p = .01] and frequency [F(1,13)= 9.68, p = .008], but only a marginal interaction was found [F(1,13) = 3.93, p < .07]. Post hoc paired comparisons (Bonferroni-adjusteda = .0125) further revealed this pattern to arise from signif-icantly lower accuracy in COM_LF compared to both COM_HF[MD = 1.6%, t(13) = 2.90, p = .012] and SIM_LF conditions[MD = 1.7%, t(13) = 2.90, p = .012]. On the other hand, RT datayielded a main effect of frequency [F(1,13) = 13.02, p = .003], butnot complexity (F < 1). There was no interaction of complexity andfrequency in participants word naming latency (F < 1), see Fig. 1.

3.2. Neuroimaging results

A visual inspection of group level contrasts (at FDR correctedp < .001) revealed largely overlapping patterns of activation acrossthe four stimulus categories SIM_HF, SIM_LF, COM_HF andCOM_LF, although there was an apparent increase in extent and3 The code used a moving window (developed and tuned after processing severallarge data sets recorded within the scanner) to align each participants overt namingresponses within a run with the respective stimulus onsets. Then, a BandpassButterworth filter (cutoffs: lower = 10 Hz, upper = 1000 Hz) was used to detect thefirst voice onset following stimulus presentation for each trial, and the intervalbetween stimulus and voice onset was logged as the response time for that trial.

http://www.fil.ion.ucl.ac.uk/spm

Fig. 1. Participants mean percent accuracy and reaction time (RT) on overt word naming (n = 14) in Hindi/Devanagari to visually simple (SIM) and complex (COM) words ofhigh (HF) and low frequency (LF). Significant differences indicated by double (p 6 .003) or single asterisks (p = .012), and marginal differences by a cross (p = .05). Main effectindicated by paired curly braces, paired comparisons by straight lines. Error bars indicate Standard Error (SE).

4 Talaraich coordinates were converted to MNI using the tal2mni functionprovided by http://imaging.mrc-cbu.cam.ac.uk/imaging/MniTalairach.

C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061 55intensity of activation with increasing cognitive demand, such thatcomplex, low frequency words (COM_LF) elicited most activity.The ALL_WDS contrast revealed a cumulative network for wordrecognition in Devanagari that closely resembled the universalreading network described by previous researchers (Bolger et al.,2005; Price, 2012), with activation of bilateral occipital gyri BA18/19, FG and posterior LITG BA37, left Heschls gyrus BA48, LSPL BA7/40, pars triangularis in the left inferior frontalgyrus (LIFG/pT) BA45, LIFG/pO BA44, left precentral gyrus(LPCG) and supplementary motor area (SMA) BA6, in additionto sub-cortical activation in the left putamen BA48 and bilateralcerebellum. A notable feature of the network was the significantactivation of right hemisphere regions in the superior and middletemporal gyri BA21, precentral BA48 and postcentral gyri BA4.

3.2.1. ANOVA resultsThe ANOVA of fMRI data yielded a significantmain effect of com-

plexity [F(1,45) = 44.54, corrected for multiple comparisons at FWEp < 104, cluster correction kP 100], as well as a strong interactionof complexity with frequency [F(1,45) = 98.15, FWE p < 7 108,kP 100]. There was no main effect of frequency when a stringent(FWE) correction was applied, though an effect was discernible ata relaxed threshold [F(1,45) = 17.64, FDR p < .05, kP 5]. Refer toFig. 2 for a graphical representation of ANOVA results, and to Table 1for a list of brain regionsmodulated by the interaction of complexityand frequency, main effects of each factor, as well as results of thesubtraction analyses (see below).

Thus, the neuroimaging results presented both similarities anddifferences to the behavioral data. The effect of visuospatial com-plexity was clearly evident in the fMRI results (Fig. 2A), whereasit was seen only in behavioral naming accuracy, but not in RT.On the other hand, word frequency strongly modulated bothbehavioral accuracy and RT, whereas it emerged only weakly inthe imaging data (Fig. 1). Finally, the interaction of complexityand frequency was robust in the fMRI data (Fig. 2C), but visibleonly marginally in behavioral accuracy and not at all in RT.

3.2.2. Subtraction analysesTo determine whether any brain regions were specifically acti-

vated by visually simple (SIM), visually complex (COM), high fre-quency (HF) or low frequency stimuli (LF), direct subtractionswere computed on fMRI data across stimulus categories. Thus,individual contrasts for SIM, COM, HF and LF were combined ingroup level activation maps, and four analyses were computed:SIM minus COM, HF minus LF, together with the reverse subtrac-tions. These subtractions (at FDR p < .03, kP 15) revealed brainregions additionally activated by cognitively demanding stimulustypes. Thus, COM minus SIM (Fig. 3) yielded activation in the FG BA37 and left cerebellum Crus 4/5, 6 and 8, as well as in righttemporal BA38, middle frontal BA10/46, and medial frontalregions BA25. The LF minus HF subtraction yielded bilateraloccipital activation BA18/19, a ventral parietal cluster spanningLIPL and LSMG BA2/48 and a dorsal parietal cluster in the LSPL BA7/40, in addition to a large cluster spanning LIFG/pO andLIFG/pT BA44/48 (Table 1). The reverse subtractions, that is,SIM minus COM and HF minus LF revealed no significant activity.

3.2.3. RoI analysesIn order to further delineate the role of visuospatial complexity

on neural processing in Devanagari, the reading-related BOLD sig-nal was analyzed from six regions of interest (RoIs) that have beenassociated in previous studies with either visuospatial complexityor phonological processing in Hindi/Devanagari. RoIs of radius8 mm were defined around peak coordinates in the interactioncontrast that corresponded most closely to coordinates reportedin the literature.

The coordinates for the FG/VWFA (MNI 44 48 22) wereproximal to those reported in Das et al. (2009) Devanagari studyat 42 57 11, and nearby the canonical VWFA coordinates ofMcCandliss et al. (2003):43537.4 Following the recommenda-tion of an anonymous reviewer, our VWFA RoI was further verifiedby overlaying the RoI on the Automated Anatomical Labeling orAAL anatomical map of the cerebellum (http://www.gin.cnrs.fr/spip.php?article217), thus ensuring that the VWFA RoI and cerebel-lar boundaries were distinct. One cluster in the LSPL (30 58 52)was chosen close to peaks previously reported for Devanagari byDas, Bapi, et al. (2011): 30 60 56, as well as Korean Hanja byLee (2004): 24 60 51.

Das, Bapi, et al. (2011) suggestion that the LSTGmediates alpha-bet-like phonological processing of Devanagari motivated theselection of bilateral posterior temporal RoIs, including one inthe LS/MTG (60 30 0) and another in the RS/MTG (64 4 0).Both clusters were close to Das, Bapi, et al. (2011) peak coordinatesat 60 18 4 and 58 12 4 respectively; Lee (2004) study simi-larly reported peaks in the LS/MTG: 40 31 5 and RS/MTG: 4835 2 for alphasyllabic Korean Hangul.

The LIFG/pO (52 6 24) cluster was very near Das,Padakannaya, et al. (2011) reports for both monolingual: 50 1016 and bilingual Devanagari readers: 56 12 24, and also closeby the peak reported by Ischebeck et al. (2004) for Japanese Kana:50 11 21. A final cluster in the RLin (12 76 2) was chosenbased on its proximity to the coordinates reported by Ischebecket al. (2004): 12 78 15 and Nakamura, Dehaene, et al. (2005):12 70 10 in research on Japanese.

http://www.gin.cnrs.fr/spip.php?article217http://www.gin.cnrs.fr/spip.php?article217http://imaging.mrc-cbu.cam.ac.uk/imaging/MniTalairach

Fig. 2. Main effects of visuospatial complexity (panel A) and word frequency (panel B) during overt naming of Hindi/Devanagari words, rendered at FDR p < .05 on 3D surfacesof the left and right hemispheres; color bar (panel A) represents F values. Interaction of complexity and frequency rendered at FWE p < 7 108 on 2D axial sections (panel C),with Z coordinates (section height) in yellow and color bar of F values below.

56 C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061The Marsbar toolbox was used to extract summary BOLD time-courses of individual participants at each of these locations foreach stimulus category, SIM_HF, SIM_LF, COM_HF and COM_LF,and percent signal change was calculated with respect to the globalsignal (i.e., {Task/[Task + Baseline]} * 100). The mean percent signalchange values for each RoI were then analyzed in 2 2 complexityby frequency ANOVAs.

These analyses revealed significant main effects of complexityin the RLin [F(1,15) = 5.08, p = .04] and the LS/MTG [F(1,15)= 6.69, p = .02]. A main effect of frequency emerged in the LSPL [F(1,15) = 9.47, p = .008], while the LS/MTG also exhibited a marginaleffect of frequency [F(1,15) = 3.11, p = .09]. The only RoI to showsignificant main effects of both complexity [F(1,15) = 4.54,p = .05] and frequency [F(1,15) = 5.34, p = .04] was the FG/VWFA.None of the regions revealed an interaction of the two factors,and no other difference reached statistical significance. See Fig. 4for a graphical representation of the RoI results.

4. Discussion

The present study was the first of its kind to investigate theeffect of visuospatial complexity in orthographic layout uponcognitive-behavioral as well as neural dynamics of word recogni-tion, by comparing the processing of visuospatially complex versussimple words in the phonologically transparent Devanagariorthography. Further, the study attempted to distinguish the influ-ence of orthographic visuospatial complexity upon effortful phono-logical retrieval, which was assessed through the word frequencyeffect.

Functional MR images (fMRI) were obtained while a group ofnative readers of Hindi/Devanagari overtly named high and lowfrequency Devanagari words that were either visually simple(, ) or complex ( , ). Results showed that only asubtle effect of visuospatial complexity was reflected in behavioralindices: participants naming accuracy but not RT revealed a cost ofnaming visually complex compared to simple words. In contrast,the word frequency effect was robust in both behavioral accuracyand RT (Fig. 1). The influence of visuospatial complexity wasclearly discernible in brain activation patterns, however, whilethe word frequency effect was weak (Fig. 2A versus B). The two fac-tors, complexity and frequency, also interacted only weakly inbehavioral data (p < .07), but exhibited a strong interaction in theneuroimaging results (evident at FWE corrected p < 7 108, seeFig. 2C). Overall, the data corroborated the hypothesis that visuo-

Table 1Brain areas identified in complexity frequency ANOVA and direct COM minus SIM and LF minus HF subtractions of overt word naming in Hindi/Devanagari.

Brain area (BA)b Complexity frequency [FWE p < 7 108, kP 100] Complexity [FWE p < 104, k P 100] Frequency [FDR p < .05, kP 5] COM SIM [FDR p < .03, k P 15] LF HF [FDR p < .03, kP 15]xyza k z xyz k z xyz k z xyz k z xyz k z

OccipitalLIOGc (18) 24 92 6 1236 inf 24 92 6 328 inf 22 92 6 22 4.46 24 92 6 473 4.48RIOG (18) 24 94 6 173 infLMOG (18) 26 68 34 44 4.09 26 66 30 217 4.31RSOG (19) 28 64 30 71 3.85RCal (19/18) 16 86 16 1367 inf 16 86 16 1497 inf 22 94 4 5 3.82 22 94 4 50 3.78LLin (18) 14 76 6 582 14 76 6 446 7.25RLin (18) 12 76 2 476FG (37) 44 48 22 229 inf 44 44 22 18 3.89(19) 30 56 18 20 3.98

TemporalLS/MTG (21) 60 30 0 273 infRS/MTG (48) 64 4 0 13 7.46RMTP (38) 52 14 24 18 4.51

ParietalRPoCG (43/4) 56 2 28 1003 inf 58 2 28 286 6.51 54 20 30 36 3.62LSMG (48) 52 28 28 27 3.87 48 24 26 12 3.59LIPL (2) 54 28 38 81 3.66LSPL (7/40) 30 58 52 291 inf 30 56 52 101 3.99

FrontalRIOFG (38) 42 20 16 145 4.41LMOFG (11) 24 44 4 220 infRMdOFG (11) 20 46 6 104 7.84RSMFG (25) 6 30 52 50 4.13LIFG/pO (44) 52 6 24 1098 inf 52 6 24 600 7.26 48 6 26 218 4.74 46 6 24 522 4.27LIFG/pT (48) 50 22 28 64 4.40 52 24 24 107 4.26RMFG (46) 28 46 26 28 4.00RSFG (10) 28 54 12 19 4.00LPCG (6) 52 2 38 4042 inf 52 2 38 1195 infLSMA (6) 6 6 56 234 inf

SubcorticalLPal (48) 20 2 6 126 infLPut (11) 20 18 2 86 4.21RPut (25) 14 10 8 64 3.93

CerebellumLCereb_4&5 16 34 30 201 4.48LCereb_6 26 42 36 188 4.23LCereb_8 14 68 54 42 4.01

IOG inferior occipital gyrus, MOG middle occipital gyrus, SOG superior occipital gyrus, Cal calcarine sulcus, Lin lingual gyrus, FG left fusiform gyrus, S/MTG border of superior and middle temporal gyri, MTP middletemporal pole, PoCG postcentral gyrus, SMG supramarginal gyrus, IPL inferior parietal lobule, SPL superior parietal lobule, IOFG inferior orbitofrontal gyrus, MOFG middle orbitofrontal gyrus, MdOFG medialorbitofrontal gyrus, SMFG superior medial frontal gyrus, IFG/pO inferior frontal gyrus pars opercularis, IFG/pT inferior frontal gyrus pars triangularis, MFG middle frontal gyrus, SFG superior frontal gyrus, PCG precentralgyrus, SMA supplementary motor area, Pal pallidum, Put putamen, Cereb cerebellum.

a MNI coordinates.b Brodmann Area.c Prefixes L and R stand for left and right respectively,

C.Rao,N

.C.Singh/Brain

&Language

141(2015)

5061

57

Fig. 3. 2D sagittal sections (X coordinates in yellow) representing areas additionally activated by visually complex over visually simple words (COM minus SIM subtraction)at FDR p < .03, k > 15; FG/VWFA left fusiform gyrus/visual word form area, LCereb left cerebellum (numbers denote lobules), RSMFG right superior medial frontal gyrus,RMFG right middle frontal gyrus, RSFG right superior frontal gyrus, RIOFG right inferior orbitofrontal gyrus, RMTP right middle temporal pole. Color map (bottomright) indicates Z values.

Fig. 4. Coronal sections (central panel) showing localization of RoIs in the VWFA/FG (-44 -48 -22), RLin (12 -76 -2), LS/MTG (-60 -30 0), RS/MTG (64 -4 0), LSPL (-30 -58 52)and LIFG/pO (-52 6 24). Column graphs (left and right panels) depict mean percent change in BOLD signal per RoI in response to visually simple (SIM) and complex Devanagariwords (COM) of high (HF) and low frequency (LF). Significant modulation of BOLD (p 6 .05) indicated by an asterisk over bracketed line (visual complexity) and paired curlybraces (word frequency).

58 C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061spatial complexity influences the neurocognitive processing ofDevanagari orthography, but indicated that this effect is more evi-dent at the neural than at the purely behavioral level.

The discussion below attempts to summarize the new insightsprovided by our data about the neural processing of orthographicvisuospatial complexity, as well as its influence on phonologicalprocessing in Hindi/Devanagari. Arguably the most importantresult of this study was the finding that the VWFA of Hindi/Deva-nagari readers is sensitive to the visuospatial complexity of theorthography. This interpretation was reinforced by two indepen-dent lines of evidence: firstly, the BOLD response of the VWFArevealed a significant main effect of visual complexity (Fig. 4).Secondly, visually complex words elicited additional VWFA activa-tion over visually simple words, as seen in the complex minus sim-ple (COM SIM) subtraction analysis (Fig. 3).

On one hand, the above finding adduces support for Price andDevlin (2011) theory, and corroborates the position taken byRoberts et al. (2013), that neural activity in the VWFA reflectsthe processing of higher-level visuospatial detail, irrespective ofthe linguistic difficulty of the stimulus. On the other hand, the lackof an interaction in our VWFA BOLD data between visual complex-ity and word frequency fails to support Price and Devlins claimthat the VWFA integrates visuospatial input with previouslyacquired, multimodal information. However, a weak numerical

C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061 59trend in the present results leaves open the possibility that a com-plexityfrequency interaction might emerge in the VWFA, givensufficient sample size and statistical power.

A second notable finding in the current study was that the BOLDsignal in the posterior left superior/middle temporal gyrus (LS/MTG) was significantly greater for visually simple compared tocomplex words. The posterior right S/MTG mirrored this pattern,although the differences were not statistically significant due towide individual variability in the BOLD signal. The left posteriortemporal cortex has been widely associated in the literature onreading with sub-lexical phonological assembly, that is, the retrie-val of sound from spelling (Fiebach, Friederici, Mller, & Yves vonCramon, 2002; Fiez & Petersen, 1998; Graves, Desai, Humphries,Seidenberg, & Binder, 2010; Kronbichler et al., 2004; Mechelliet al., 2005; Paulesu et al., 2000). In support of this interpretation,the LS/MTG BOLD in our results was marginally greater (p = .09) forwords of low as opposed to high frequency.

However, our finding confirms earlier reports by Das, Bapi, et al.(2011) and Lee (2004) that reading alphasyllabic orthographiesactivates the posterior superior/middle temporal cortex bilaterally.This invites a closer consideration of Das, Bapi, et al. (2011) sugges-tion that the S/MTG may be sensitive to alphabet-like features.

While the current findings are preliminary, we speculate on thepossibility of sub-specialization in the processing of complexorthographies, whereby only alphabet-like, that is, visually simple,linear features may be processed by posterior temporal regions.One study found that healthy adults who were asked to searchfor target letters in a visual array exhibited strong bilateral supe-rior temporal activation (Himmelbach, Erb, & Karnath, 2006), whileanother documented left STG activity when participants identifiedwhether a pre-specified target letter were present either at the glo-bal or at the local level in a globallocal letter detection task (Finket al., 1996). Going further, numerous cases have documented anassociation between visuospatial neglect and damage to the tempo-ral cortex (Becker & Karnath, 2007; Suchan, Rorden, & Karnath,2012).

In addition, our results revealed that reading visuospatiallycomplex Hindi/Devanagari words led to increased recruitment offrontal areas in the right hemisphere, especially the inferior orbito-frontal (RIOFG), superior (RSFG), superior mid frontal (RSMFG), andmiddle frontal gyri (RMFG), as well as left cerebellar lobules 4&5, 6and 8. Refer to Table 1, COM SIM, for coordinates and Fig. 3 for arepresentation of areas additionally activated by complex words.

The frontal activation in our data complements the differentialneural response in the posterior temporal cortices, and suggests thatthese regions may be involved in the top-down control ofvisual attention as well as visuospatial working memory whileprocessing complex orthographic features (Courtney, Petit,Maisog, Ungerleider, & Haxby, 1998; Kastner & Ungerleider, 2000;Ungerleider, Courtney, & Haxby, 1998). While activation in lobules4&5 and 8 of the left cerebellum has been linked to sensorimotorcoordination, an emerging consensus is that the left cerebellar semi-lunar or lobule 6 is integral to spatial working memory (Baier,Mller, & Dieterich, 2014; Stoodley, Valera, & Schmahmann, 2012;Tomlinson, Davis, Morgan, & Bracewell, 2014).

Besides increased right frontal activity, visually complex stimuliin our study elicited greater de-activation in the right lingual gyrus(RLin), which exhibited consistent de-activation with respect tobaseline across all stimulus types (Fig. 4). Research has previouslyimplicated the right lingual gyrus in the perception of visual con-figurations (Bellgowan, Saad, & Bandettini, 2003; Mechelli,Humphreys, Mayall, Olson, & Price, 2000), and several studies havefound right lingual activation while participants identified the glo-bal or gestalt letter in a globallocal letter detection task, but notwhen they named the local or component letter (Han et al.,2002; Fink et al., 1996, 1997; Moses et al., 2002; Weissman,Woldorff, Hazlett, & Mangun, 2002). Fink et al.s (1996,1997) studyrevealed right lingual gyrus activation during global letter detec-tion at a locus (MNI 16 72 4) very close to the coordinatesobserved in our Devanagari data.

Perhaps more convincingly, recent investigations have demon-strated that learning to read in Chinese is correlated with increasedactivation of the right lingual gyrus among native English readers(Deng, Chou, Ding, Peng, & Booth, 2011; Zhao et al., 2012), in a find-ing which the researchers impute to the necessity of discriminatingthe visuospatial configurations of Chinese characters. Thus, theright lingual de-activation in our results may be taken to reflecttop-down inhibition of activity due to the need for processing localrather than global stimulus details while reading in Hindi/Devana-gari. In this context, we note the proximity of our lateral RIOFGcluster with that reported by Fink et al. (1996, 1997) as a highercenter for regulating divided attention.

Apart from its role in processing visuospatial complexity, thecurrent study revealed the VWFA to be sensitive to phonologicaleffort, as the BOLD signal in this region was also modulated byword frequency. Other regions that exhibited a significant fre-quency effect in their BOLD response included the LIFG/pO andthe LSPL. The increased activation of these two regions by lowcompared to high frequency words was additionally confirmedby the subtraction contrast of low versus high frequency words(Table 1, LF HF). The subtraction further showed that low fre-quency words elicited greater activation in the bilateral occipitalcortex in areas BA18 and BA19, LIPL, pars triangularis (LIFG/pT),as well as the right postcentral gyrus (RPoCG) and bilateralputamen.

The VWFA and the LIFG/pO have been previously implicated inthe retrieval of phonology in alphabetic orthographies (Fiez &Petersen, 1998; Kronbichler et al., 2004; Mechelli et al., 2005),while the left inferior and superior parietal regions LIPL and LSPLhave both been linked with phonological processing in complexorthographies (Ino et al., 2009; Kuo et al., 2003; Lee, 2004; Liuet al., 2007; Nakamura, Dehaene, et al., 2005; Tan et al., 2001).The current results confirm previous reports by Das et al. (2009),Das, Bapi, et al. (2011), Das, Padakannaya, et al. (2011) of theimportance of the LIFG/pO, LIPL and LSPL to word recognition inHindi/Devanagari. Further, they suggest that phonological retrievalin an alphasyllabary might utilize the combined neural resourcesof visually simple and complex orthographies.5. Conclusion

In summary, the results of the current study substantiate theclaim that the cognitive as well as neural processing of writtenwords is influenced by the visuospatial complexity of the orthogra-phy. While, at the behavioral level our data exhibited an effect ofcomplexity only on word naming accuracy, neuroimaging resultsrevealed that visuospatial complexity diffusely modulated brainactivation across the reading network. Particularly, our resultsimplicated the FG/VWFA, bilateral superior and middle temporalcortices, right superior, middle and orbitofrontal areas and left cer-ebellar lobule 6 in processing the visuospatially complex configu-ration of Hindi/Devanagari orthography.

Further, the study highlighted the left superior parietal lobuleas well as pars opercularis of the left inferior frontal gyrus as cen-ters of effortful phonological processing. The involvement of pari-etal and inferior frontal regions in processing low frequencyHindi/Devanagari words suggests that phonological processing ofalphasyllabic scripts relies upon the neural resources recruitedby both visually simple as well as complex orthographies.

An especially interesting finding was that the FG/VWFA in read-ers of Hindi/Devanagari responded to both linguistic and non-lin-

60 C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061guistic orthographic difficulty, whereby BOLD response was mod-ulated significantly, though independently, by both word fre-quency and visuospatial complexity. This result lends somesupport to the conclusion of Moore, Durisko, Perfetti, and Fiez(2014) that the VWFA acts as a linguistic bridge, by providing opti-mal neural connectivity between visual perceptual areas and tradi-tional language processing areas.

Importantly, this study furnishes the first comprehensive blue-print of the visual word recognition network in Hindi/Devanagari,by incorporating the full range of orthographic features integral tothe orthography, including visually simple as well as complex fea-tures. The overlap between the present results and the previouslymapped neural network for Hindi/Devanagari (Das et al., 2009;Das, Bapi, et al., 2011) further strengthens their credibility.

At the same time, the present study is characterized by certaindrawbacks which limit the generalizability of these findings. In theabsence of a database containing information on key properties ofHindi/Devanagari words such as lexical frequency, neighborhoodsize, imageability and so on, stimuli could not be matched on char-acteristics known to affect neurocognitive processing, includingneighborhood size and bigram frequency (Balota, Cortese,Sergent-Marshall, Spieler, & Yap, 2004; Graves et al., 2010). Also,participants were not as well-matched in their non-Hindi languageproficiency and knowledge as they were in Hindi/Devanagari,owing to the heterogeneity of educational profiles in the targetpopulation. We acknowledge the necessity for more studies inwhich stimuli are matched on all relevant lexical and sub-lexicalvariables, and samples drawn from more linguistically homoge-neous backgrounds, in order to gain a nuanced understanding ofthe neuropsychological bases of reading in Devanagari.

Nevertheless, the current findings offer novel insights that canbe used to motivate further research on the hitherto neglecteddimension of orthographic visuospatial complexity. In particular,the role of brain areas such as the VWFA, bilateral posterior supe-rior/middle temporal cortices (S/MTG), right frontal (RIOFG, RSFG,RMFG) as well as the left parietal cortex (especially LSPL) in read-ing visuospatially complex alphasyllabaries such as Hindi/Devana-gari may be understood more clearly by analyzing the functionalconnectivity among these regions during reading. Further explora-tion of developmental changes in structural connectivity amongVWFA, S/MTG, RIOFG and LSPL may also serve to illuminate thequestion of whether there are critical differences in neurodevelop-mental markers of reading skill as well as reading disability incomplex orthographies.

Acknowledgments

We thank J. Ahlawat and Abhilash for their technical help inacquiring neuroimaging data.

References

Baier, B., Mller, N. G., & Dieterich, M. (2014). What part of the cerebellumcontributes to a visuospatial working memory task? Annals of Neurology. http://dx.doi.org/10.1002/ana.24272.

Balota, D. A., Cortese, M. J., Sergent-Marshall, S. D., Spieler, D. H., & Yap, M. J. (2004).Visual word recognition of single-syllable words. Journal of ExperimentalPsychology: General, 133, 283316.

Becker, E., & Karnath, H.-O. (2007). Incidence of visual extinction after left versusright hemisphere stroke. Stroke, 38, 31723174.

Bellgowan, P. S. F., Saad, Z. S., & Bandettini, P. A. (2003). Understanding neuralsystem dynamics through task modulation and measurement of functional MRIamplitude, latency, and width. Proceedings of the National Academy of Sciences ofthe United States of America, 100, 14151419.

Ben-Shachar, M., Dougherty, R. F., Deutsch, G. K., & Wandell, B. A. (2011). Thedevelopment of cortical sensitivity to visual word forms. Journal of CognitiveNeuroscience, 23, 113.

Binder, J. R., Medler, D. A., Westbury, C. F., Liebenthal, E., & Buchanan, L. (2006).Tuning of the human left fusiform gyrus to sublexical orthographic structure.Neuroimage, 33, 739748.Binder, J. R., & Mohr, J. P. (1992). The topography of transcallosal reading pathways:A case-control analysis. Brain, 115, 18071826.

Bolger, D. J., Perfetti, C. A., & Schneider, W. (2005). Cross-cultural effect on the brainrevisited: Universal structures plus writing system variation. Human BrainMapping, 25, 92104.

Brem, S., Bach, S., Kucian, K., Guttorm, T. K., Martin, E., Lyytinen, H., et al. (2010).Brain sensitivity to print emerges when children learn letterspeech soundcorrespondences. Proceedings of the National Academy of Sciences of the UnitedStates of America, 107, 79397944.

Chen, Y., Fu, S., Iversen, S. D., Smith, S. M., & Matthews, P. M. (2002). Testing for dualbrain processing routes in reading: A direct contrast of Chinese character andPinyin reading using fMRI. Journal of Cognitive Neuroscience, 14, 10881098.

Cohen, L., Lehericy, S., Chochon, F., Lemer, C., Rivaud, S., & Dehaene, S. (2002).Language-specific tuning of visual cortex? Functional properties of the visualword form area. Brain, 125, 10541069.

Courtney, S. M., Petit, L., Maisog, J. M., Ungerleider, L. G., & Haxby, J. V. (1998). Anarea specialized for spatial working memory in human frontal cortex. Science,279, 13471350.

Damasio, A. R., & Damasio, H. (1983). The anatomic basis of pure alexia. Neurology,33, 15731583.

Das, T., Bapi, R. S., Padakannaya, P., & Singh, N. C. (2011). Cortical network forreading linear words in an alphasyllabary. Reading and Writing, 24, 697707.

Das, T., Kumar, U., Bapi, R. S., Padakannaya, P., & Singh, N. C. (2009). Neuralrepresentation of an alphasyllabary The story of Devanagari. Current Science,97, 10331038.

Das, T., Padakannaya, P., Pugh, K. R., & Singh, N. C. (2011). Neuroimaging revealsdual routes to reading in simultaneous proficient readers of two orthographies.Neuroimage, 54, 14761487.

Dehaene, S., & Cohen, L. (2011). The unique role of the visual word form area inreading. Trends in Cognitive Sciences, 15, 254262.

Dehaene, S., Jobert, A., Naccache, L., Ciuciu, P., Poline, J.-B., Le Bihan, D., et al. (2004).Letter binding and invariant recognition of masked words. Psychological Science,15, 307313.

Dehaene, S., Pegado, F., Braga, L. W., Ventura, P., Filho, G. N., Jobert, A., et al. (2010).How learning to read changes the cortical networks for vision and language.Science, 330, 13591364.

Deng, Y., Chou, T. L., Ding, G. S., Peng, D. L., & Booth, J. R. (2011). The involvement ofoccipital and inferior frontal cortex in the phonological learning of Chinesecharacters. Journal of Cognitive Neuroscience, 23, 19982012.

Feldman, L. B., & Turvey, M. T. (1980). Words written in kana are named faster thanthe same words written in kanji. Language and Speech, 23, 141147.

Fiebach, C. J., Friederici, A. D., Mller, K., & Yves von Cramon, D. (2002). FMRIevidence for dual routes to the mental lexicon in visual word recognition.Journal of Cognitive Neuroscience, 14, 1123.

Fiez, J. A., Balota, D. A., Raichle, M. E., & Petersen, S. E. (1999). Effects of lexicality,frequency and spelling-to-sound consistency on the functional anatomy ofreading. Neuron, 24, 205218.

Fiez, J. A., & Petersen, S. E. (1998). Neuroimaging studies of word reading.Proceedings of the National Academy of Sciences of the United States of America,95, 914921.

Fink, G. R., Halligan, P. W., Marshall, J. C., Frith, C. D., Frackowiak, R. S. J., & Dolan, R. J.(1996). Where in the brain does visual attention select the forest and the trees?Nature, 382, 626628.

Fink, G. R., Halligan, P. W., Marshall, J. C., Frith, C. D., Frackowiak, R. S. J., & Dolan, R. J.(1997). Neural mechanisms involved in the processing of global and localaspects of hierarchically organized visual stimuli. Brain, 120, 17791791.

Friston, K. J., Glaser, D. E., Henson, R. N. A., Kiebel, S., Phillips, C., & Ashburner, J.(2002). Classical and Bayesian inference in neuroimaging: Applications.Neuroimage, 16, 484512.

Graves, W. W., Desai, R., Humphries, C., Seidenberg, M. S., & Binder, J. R. (2010).Neural systems for reading aloud: A multiparametric approach. Cerebral Cortex,20, 17991815.

Ha Duy Thuy, D., Matsuo, K., Nakamura, N., Toma, K., Oga, T., Nakai, T., et al. (2004).Implicit and explicit processing of kanji and kana words and non-words studiedwith fMRI. Neuroimage, 23, 878889.

Han, S., Weaver, J. A., Murray, S. O., Kang, X., Yund, E. W., & Woods, D. L. (2002).Hemispheric asymmetry in global/local processing: Effects of stimulus positionand spatial frequency. Neuroimage, 17, 12901299.

Himmelbach, M., Erb, M., & Karnath, H.-O. (2006). Exploring the visual world: Theneural substrate of spatial orienting. Neuroimage, 32, 17471759.

Ibrahim, R., Eviatar, Z., & Aharon-Peretz, J. (2002). The characteristics of arabicorthography slow its processing. Neuropsychology, 16, 322326.

Ino, T., Nakai, R., Azuma, T., Kimura, T., & Fukuyama, H. (2009). Recognition andreading aloud of kana and kanji word: An fMRI study. Brain Research Bulletin, 78,232239.

Ischebeck, A., Indefrey, P., Usui, N., Nose, I., Hellwig, F., & Taira, M. (2004). Reading ina regular orthography: An fMRI study investigating the role of visual familiarity.Journal of Cognitive Neuroscience, 16, 727741.

Jobard, G., Crivello, F., & Tzourio-Mazoyer, N. (2003). Evaluation of the dual routetheory of reading: A metanalysis of 35 neuroimaging studies. Neuroimage, 20,693712.

Kandhadai, P., & Sproat, R. (2010). Impact of spatial ordering of graphemes inalphasyllabic scripts on phonemic awareness in Indic languages. WritingSystems Research, 2, 105116.

Kastner, S., & Ungerleider, L. G. (2000). Mechanisms of visual attention in the humancortex. Annual Review of Neuroscience, 23, 315341.

http://dx.doi.org/10.1002/ana.24272http://dx.doi.org/10.1002/ana.24272http://refhub.elsevier.com/S0093-934X(14)00173-4/h0010http://refhub.elsevier.com/S0093-934X(14)00173-4/h0010http://refhub.elsevier.com/S0093-934X(14)00173-4/h0010http://refhub.elsevier.com/S0093-934X(14)00173-4/h0015http://refhub.elsevier.com/S0093-934X(14)00173-4/h0015http://refhub.elsevier.com/S0093-934X(14)00173-4/h0020http://refhub.elsevier.com/S0093-934X(14)00173-4/h0020http://refhub.elsevier.com/S0093-934X(14)00173-4/h0020http://refhub.elsevier.com/S0093-934X(14)00173-4/h0020http://refhub.elsevier.com/S0093-934X(14)00173-4/h0025http://refhub.elsevier.com/S0093-934X(14)00173-4/h0025http://refhub.elsevier.com/S0093-934X(14)00173-4/h0025http://refhub.elsevier.com/S0093-934X(14)00173-4/h0030http://refhub.elsevier.com/S0093-934X(14)00173-4/h0030http://refhub.elsevier.com/S0093-934X(14)00173-4/h0030http://refhub.elsevier.com/S0093-934X(14)00173-4/h0035http://refhub.elsevier.com/S0093-934X(14)00173-4/h0035http://refhub.elsevier.com/S0093-934X(14)00173-4/h0040http://refhub.elsevier.com/S0093-934X(14)00173-4/h0040http://refhub.elsevier.com/S0093-934X(14)00173-4/h0040http://refhub.elsevier.com/S0093-934X(14)00173-4/h0045http://refhub.elsevier.com/S0093-934X(14)00173-4/h0045http://refhub.elsevier.com/S0093-934X(14)00173-4/h0045http://refhub.elsevier.com/S0093-934X(14)00173-4/h0045http://refhub.elsevier.com/S0093-934X(14)00173-4/h0050http://refhub.elsevier.com/S0093-934X(14)00173-4/h0050http://refhub.elsevier.com/S0093-934X(14)00173-4/h0050http://refhub.elsevier.com/S0093-934X(14)00173-4/h0055http://refhub.elsevier.com/S0093-934X(14)00173-4/h0055http://refhub.elsevier.com/S0093-934X(14)00173-4/h0055http://refhub.elsevier.com/S0093-934X(14)00173-4/h0060http://refhub.elsevier.com/S0093-934X(14)00173-4/h0060http://refhub.elsevier.com/S0093-934X(14)00173-4/h0060http://refhub.elsevier.com/S0093-934X(14)00173-4/h0065http://refhub.elsevier.com/S0093-934X(14)00173-4/h0065http://refhub.elsevier.com/S0093-934X(14)00173-4/h0070http://refhub.elsevier.com/S0093-934X(14)00173-4/h0070http://refhub.elsevier.com/S0093-934X(14)00173-4/h0075http://refhub.elsevier.com/S0093-934X(14)00173-4/h0075http://refhub.elsevier.com/S0093-934X(14)00173-4/h0075http://refhub.elsevier.com/S0093-934X(14)00173-4/h0080http://refhub.elsevier.com/S0093-934X(14)00173-4/h0080http://refhub.elsevier.com/S0093-934X(14)00173-4/h0080http://refhub.elsevier.com/S0093-934X(14)00173-4/h0085http://refhub.elsevier.com/S0093-934X(14)00173-4/h0085http://refhub.elsevier.com/S0093-934X(14)00173-4/h0090http://refhub.elsevier.com/S0093-934X(14)00173-4/h0090http://refhub.elsevier.com/S0093-934X(14)00173-4/h0090http://refhub.elsevier.com/S0093-934X(14)00173-4/h0095http://refhub.elsevier.com/S0093-934X(14)00173-4/h0095http://refhub.elsevier.com/S0093-934X(14)00173-4/h0095http://refhub.elsevier.com/S0093-934X(14)00173-4/h0100http://refhub.elsevier.com/S0093-934X(14)00173-4/h0100http://refhub.elsevier.com/S0093-934X(14)00173-4/h0100http://refhub.elsevier.com/S0093-934X(14)00173-4/h0105http://refhub.elsevier.com/S0093-934X(14)00173-4/h0105http://refhub.elsevier.com/S0093-934X(14)00173-4/h0110http://refhub.elsevier.com/S0093-934X(14)00173-4/h0110http://refhub.elsevier.com/S0093-934X(14)00173-4/h0110http://refhub.elsevier.com/S0093-934X(14)00173-4/h0115http://refhub.elsevier.com/S0093-934X(14)00173-4/h0115http://refhub.elsevier.com/S0093-934X(14)00173-4/h0115http://refhub.elsevier.com/S0093-934X(14)00173-4/h0120http://refhub.elsevier.com/S0093-934X(14)00173-4/h0120http://refhub.elsevier.com/S0093-934X(14)00173-4/h0120http://refhub.elsevier.com/S0093-934X(14)00173-4/h0125http://refhub.elsevier.com/S0093-934X(14)00173-4/h0125http://refhub.elsevier.com/S0093-934X(14)00173-4/h0125http://refhub.elsevier.com/S0093-934X(14)00173-4/h0130http://refhub.elsevier.com/S0093-934X(14)00173-4/h0130http://refhub.elsevier.com/S0093-934X(14)00173-4/h0130http://refhub.elsevier.com/S0093-934X(14)00173-4/h0135http://refhub.elsevier.com/S0093-934X(14)00173-4/h0135http://refhub.elsevier.com/S0093-934X(14)00173-4/h0135http://refhub.elsevier.com/S0093-934X(14)00173-4/h0140http://refhub.elsevier.com/S0093-934X(14)00173-4/h0140http://refhub.elsevier.com/S0093-934X(14)00173-4/h0140http://refhub.elsevier.com/S0093-934X(14)00173-4/h0145http://refhub.elsevier.com/S0093-934X(14)00173-4/h0145http://refhub.elsevier.com/S0093-934X(14)00173-4/h0145http://refhub.elsevier.com/S0093-934X(14)00173-4/h0150http://refhub.elsevier.com/S0093-934X(14)00173-4/h0150http://refhub.elsevier.com/S0093-934X(14)00173-4/h0150http://refhub.elsevier.com/S0093-934X(14)00173-4/h0155http://refhub.elsevier.com/S0093-934X(14)00173-4/h0155http://refhub.elsevier.com/S0093-934X(14)00173-4/h0160http://refhub.elsevier.com/S0093-934X(14)00173-4/h0160http://refhub.elsevier.com/S0093-934X(14)00173-4/h0165http://refhub.elsevier.com/S0093-934X(14)00173-4/h0165http://refhub.elsevier.com/S0093-934X(14)00173-4/h0165http://refhub.elsevier.com/S0093-934X(14)00173-4/h0170http://refhub.elsevier.com/S0093-934X(14)00173-4/h0170http://refhub.elsevier.com/S0093-934X(14)00173-4/h0170http://refhub.elsevier.com/S0093-934X(14)00173-4/h0175http://refhub.elsevier.com/S0093-934X(14)00173-4/h0175http://refhub.elsevier.com/S0093-934X(14)00173-4/h0175http://refhub.elsevier.com/S0093-934X(14)00173-4/h0180http://refhub.elsevier.com/S0093-934X(14)00173-4/h0180http://refhub.elsevier.com/S0093-934X(14)00173-4/h0180http://refhub.elsevier.com/S0093-934X(14)00173-4/h0185http://refhub.elsevier.com/S0093-934X(14)00173-4/h0185

C. Rao, N.C. Singh / Brain & Language 141 (2015) 5061 61Kherif, F., Josse, G., & Price, C. J. (2011). Automatic top-down processing explainscommon left occipito-temporal responses to visual words and objects. CerebralCortex, 21, 103114.

Kronbichler, M., Hutzler, F., Wimmer, H., Mair, A., Staffen, W., & Ladurner, G. (2004).The visual word form area and the frequency with which words areencountered: Evidence from a parametric fMRI study. Neuroimage, 21, 946953.

Kumar, U., Das, T., Bapi, R. S., Padakannaya, P., Joshi, R. M., & Singh, N. C. (2009).Reading different orthographies: An fMRI study of phrase reading in HindiEnglish bilinguals. Reading and Writing, 23, 239255.

Kuo, W.-J., Yeh, T.-C., Duann, J.-R., Wu, Y.-T., Ho, L.-T., Hung, D., et al. (2001). A left-lateralized network for reading Chinese words: A 3T fMRI study. Neuroreport,12, 39974001.

Kuo, W.-J., Yeh, T.-C., Lee, C.-Y., Wu, Y.-T., Chou, C.-C., Ho, L.-T., et al. (2003).Frequency effects of Chinese character processing in the brain: An event-relatedfMRI study. Neuroimage, 18, 720730.

Lee, K.-M. (2004). Functional MRI comparison between reading ideographic andphonographic scripts of one language. Brain and Language, 91, 245251.

Liu, Y., Dunlap, S., Fiez, J., & Perfetti, C. (2007). Evidence for neural accommodationto a writing system following learning. Human Brain Mapping, 28, 12231234.

McCandliss, B. D., Cohen, L., & Dehaene, S. (2003). The visual word form area:Expertise for reading in the fusiform gyrus. Trends in Cognitive Sciences, 7,293299.

Mechelli, A., Crinion, J. T., Long, S., Friston, K. J., Lambon Ralph, M. A., Patterson, K.,et al. (2005). Dissociating reading processes on the basis of neuronalinteractions. Journal of Cognitive Neuroscience, 17, 17531765.

Mechelli, A., Humphreys, G. W., Mayall, K., Olson, A., & Price, C. J. (2000). Differentialeffects of word length and visual contrast in the fusiform and lingual gyri duringreading. Proceedings of the Royal Society of London, 267, 19091913.

Moore, M. W., Durisko, C., Perfetti, C. A., & Fiez, J. A. (2014). Learning to read analphabet of human faces produces left-lateralized training effects in thefusiform gyrus. Journal of Cognitive Neuroscience, 26, 896913.

Moses, P., Roe, K., Buxton, R. B., Wong, E. C., Frank, L. R., & Stiles, J. (2002). FunctionalMRI of global and local processing in children. Neuroimage, 16, 415424.

Nag, S., & Snowling, M. J. (2012). Reading in an alphasyllabary: Implications for alanguage universal theory of learning to read. Scientific Studies of Reading, 16,404423.

Nakamura, K., Dehaene, S., Jobert, A., Le Bihan, D., & Kouider, S. (2005). Subliminalconvergence of kanji and kana words: Further evidence for functionalparcellation of the posterior temporal cortex in visual word perception.Journal of Cognitive Neuroscience, 17, 954968.

Nakamura, K., Oga, T., Okada, T., Sadato, N., Takayama, Y., Wydell, T., et al. (2005).Hemispheric asymmetry emerges at distinct parts of the occipitotemporalcortex for objects, logograms and phonograms: A functional MRI study.Neuroimage, 28, 521528.

Paulesu, E., McCrory, E., Fazio, F., Menoncello, L., Brunswick, N., Cappa, S. F., et al.(2000). A cultural effect on brain function. Nature Reviews Neuroscience, 3,9196.

Perfetti, C. A., Liu, Y., Fiez, J., Nelson, J., Bolger, D. J., & Tan, L.-H. (2007). Reading intwo writing systems: Accommodation and assimilation of the brains readingnetwork. Bilingualism: Language and Cognition, 10, 131146.

Price, C. J. (2012). A review and synthesis of the first 20 years of PET and fMRIstudies of heard speech, spoken language and reading. Neuroimage, 62,816847.

Price, C. J., & Devlin, J. T. (2011). The interactive account of ventral occipitotemporalcontributions to reading. Trends in Cognitive Sciences, 15, 246253.Price, C. J., Wise, R. J. S., & Frackowiak, R. S. J. (1996). Demonstrating the implicitprocessing of visually presented words and pseudowords. Cerebral Cortex, 6,6270.

Rao, C., Vaid, J., Srinivasan, N., & Chen, H.-C. (2011). Orthographic characteristicsspeed Hindi naming but slow Urdu naming: Evidence from Hindi/Urdubiliterates. Reading and Writing, 24, 679695.

Roberts, D. J., Woollams, A. M., Kim, E., Beeson, P. M., Rapcsak, S. Z., & Lambon Ralph,M. A. (2013). Efficient visual object and word recognition relies on high spatialfrequency coding in the left posterior fusiform gyrus: Evidence from a case-series of patients with ventral occipito-temporal cortex damage. Cerebral Cortex,23, 25682580.

Seidenberg, M. S., Waters, G. S., Barnes, M. A., & Tanenhaus, M. K. (1984). When doesirregular spelling or pronunciation influence word recognition? Journal of VerbalLearning and Verbal Behavior, 23, 383404.

Shimron, J., & Sivan, T. (1994). Reading proficiency and orthography: Evidence fromHebrew and English. Language Learning, 44, 527.

Simpson, G. B., & Kang, H. (1994). The flexible use of phonological information inword recognition in Korean. Journal of Memory and Language, 33, 319331.

Stoodley, C. J., Valera, E. M., & Schmahmann, J. D. (2012). Functional topography ofthe cerebellum for motor and cognitive tasks: An fMRI study. Neuroimage, 59,15601570.

Suchan, J., Rorden, C., & Karnath, H.-O. (2012). Neglect severity after left and rightbrain damage. Neuropsychologia, 50, 11361141.

Tan, L. H., Laird, A. R., Li, K., & Fox, P. T. (2005). Neuroanatomical correlates ofphonological processing of Chinese characters and alphabetic words: A meta-analysis. Human Brain Mapping, 25, 8391.

Tan, L.-H., Liu, H.-L., Perfetti, C. A., Spinks, J. A., Fox, P. T., & Gao, J.-H. (2001). Theneural system underlying Chinese logograph reading. Neuroimage,13, 836846.

Tan, L.-H., Spinks, J. A., Gao, J.-H., Liu, H.-L., Perfetti, C. A., Xiong, J., et al. (2000). Brainactivation in the processing of Chinese characters and words: A functional MRIstudy. Human Brain Mapping, 10, 1627.

Tomlinson, S. P., Davis, N. J., Morgan, H. M., & Bracewell, R. M. (2014). Cerebellarcontributions to spatial memory. Neuroscience Letters, 578, 182186.

Twomey, T., Kawabata Duncan, K. J., Hogan, J. S., Morita, K., Umeda, K., et al. (2013).Dissociating visual form from lexical frequency using Japanese. Brain andLanguage, 125, 184193.

Ungerleider, L. G., Courtney, S. M., & Haxby, J. V. (1998). A neural system for humanvisual working memory. Proceedings of the National Academy of Sciences of theUnited States of America, 95, 883890.

Vaid, J., & Gupta, A. (2002). Exploring word recognition in a semi-alphabetic script:The case of Devanagari. Brain and Language, 81, 679690.

Weissman, D. H., Woldorff, M. G., Hazlett, C. J., & Mangun, G. R. (2002). Effects ofpractice on executive control investigated with fMRI. Cognitive Brain Research,15, 4760.

Wu, C.-Y., Ho, M.-H. R., & Chen, S.-H. A. (2012). A meta-analysis of fMRI studies ofChinese orthographic, phonological and semantic processing. Neuroimage, 63,381391.

Yoon, H. W., Cho, K.-D., Chung, J.-Y., & Park, H. W. (2005). Neural mechanisms ofKorean word reading: A functional magnetic resonance imaging study.Neuroscience Letters, 373, 206211.

Zhao, J., Li, Q.-L., Wang, J.-J., Yang, Y., Deng, Y., & Bi, H.-Y. (2012). Neural basis ofphonological processing in second language reading: An fMRI study of Chineseregularity effect. Neuroimage, 60, 419425.

http://refhub.elsevier.com/S0093-934X(14)00173-4/h0190http://refhub.elsevier.com/S0093-934X(14)00173-4/h0190http://refhub.elsevier.com/S0093-934X(14)00173-4/h0190http://refhub.elsevier.com/S0093-934X(14)00173-4/h0195http://refhub.elsevier.com/S0093-934X(14)00173-4/h0195http://refhub.elsevier.com/S0093-934X(14)00173-4/h0195http://refhub.elsevier.com/S0093-934X(14)00173-4/h0200http://refhub.elsevier.com/S0093-934X(14)00173-4/h0200http://refhub.elsevier.com/S0093-934X(14)00173-4/h0200http://refhub.elsevier.com/S0093-934X(14)00173-4/h0205http://refhub.elsevier.com/S0093-934X(14)00173-4/h0205http://refhub.elsevier.com/S0093-934X(14)00173-4/h0205http://refhub.elsevier.com/S0093-934X(14)00173-4/h0210http://refhub.elsevier.com/S0093-934X(14)00173-4/h0210http://refhub.elsevier.com/S0093-934X(14)00173-4/h0210http://refhub.elsevier.com/S0093-934X(14)00173-4/h0215http://refhub.elsevier.com/S0093-934X(14)00173-4/h0215http://refhub.elsevier.com/S0093-934X(14)00173-4/h0220http://refhub.elsevier.com/S0093-934X(14)00173-4/h0220http://refhub.elsevier.com/S0093-934X(14)00173-4/h0225http://refhub.elsevier.com/S0093-934X(14)00173-4/h0225http://refhub.elsevier.com/S0093-934X(14)00173-4/h0225http://refhub.elsevier.com/S0093-934X(14)00173-4/h0230http://refhub.elsevier.com/S0093-934X(14)00173-4/h0230http://refhub.elsevier.com/S0093-934X(14)00173-4/h0230http://refhub.elsevier.com/S0093-934X(14)00173-4/h0235http://refhub.elsevier.com/S0093-934X(14)00173-4/h0235http://refhub.elsevier.com/S0093-934X(14)00173-4/h0235http://refhub.elsevier.com/S0093-934X(14)00173-4/h0245http://refhub.elsevier.com/S0093-934X(14)00173-4/h0245http://refhub.elsevier.com/S0093-934X(14)00173-4/h0245http://refhub.elsevier.com/S0093-934X(14)00173-4/h0240http://refhub.elsevier.com/S0093-934X(14)00173-4/h0240http://refhub.elsevier.com/S0093-934X(14)00173-4/h0250http://refhub.elsevier.com/S0093-934X(14)00173-4/h0250http://refhub.elsevier.com/S0093-934X(14)00173-4/h0250http://refhub.elsevier.com/S0093-934X(14)00173-4/h0255http://refhub.elsevier.com/S0093-934X(14)00173-4/h0255http://refhub.elsevier.com/S0093-934X(14)00173-4/h0255http://refhub.elsevier.com/S0093-934X(14)00173-4/h0255http://refhub.elsevier.com/S0093-934X(14)00173-4/h0260http://refhub.elsevier.com/S0093-934X(14)00173-4/h0260http://refhub.elsevier.com/S0093-934X(14)00173-4/h0260http://refhub.elsevier.com/S0093-934X(14)00173-4/h0260http://refhub.elsevier.com/S0093-934X(14)00173-4/h0265http://refhub.elsevier.com/S0093-934X(14)00173-4/h0265http://refhub.elsevier.com/S0093-934X(14)00173-4/h0265http://refhub.elsevier.com/S0093-934X(14)00173-4/h9020http://refhub.elsevier.com/S0093-934X(14)00173-4/h9020http://refhub.elsevier.com/S0093-934X(14)00173-4/h9020http://refhub.elsevier.com/S0093-934X(14)00173-4/h0270http://refhub.elsevier.com/S0093-934X(14)00173-4/h0270http://refhub.elsevier.com/S0093-934X(14)00173-4/h0270http://refhub.elsevier.com/S0093-934X(14)00173-4/h0270http://refhub.elsevier.com/S0093-934X(14)00173-4/h0275http://refhub.elsevier.com/S0093-934X(14)00173-4/h0275http://refhub.elsevier.com/S0093-934X(14)00173-4/h0280http://refhub.elsevier.com/S0093-934X(14)00173-4/h0280http://refhub.elsevier.com/S0093-934X(14)00173-4/h0280http://refhub.elsevier.com/S0093-934X(14)00173-4/h0285http://refhub.elsevier.com/S0093-934X(14)00173-4/h0285http://refhub.elsevier.com/S0093-934X(14)00173-4/h0285http://refhub.elsevier.com/S0093-934X(14)00173-4/h0290http://refhub.elsevier.com/S0093-934X(14)00173-4/h0290http://refhub.elsevier.com/S0093-934X(14)00173-4/h0290http://refhub.elsevier.com/S0093-934X(14)00173-4/h0290http://refhub.elsevier.com/S0093-934X(14)00173-4/h0290http://refhub.elsevier.com/S0093-934X(14)00173-4/h0295http://refhub.elsevier.com/S0093-934X(14)00173-4/h0295http://refhub.elsevier.com/S0093-934X(14)00173-4/h0295http://refhub.elsevier.com/S0093-934X(14)00173-4/h0300http://refhub.elsevier.com/S0093-934X(14)00173-4/h0300http://refhub.elsevier.com/S0093-934X(14)00173-4/h0305http://refhub.elsevier.com/S0093-934X(14)00173-4/h0305http://refhub.elsevier.com/S0093-934X(14)00173-4/h0310http://refhub.elsevier.com/S0093-934X(14)00173-4/h0310http://refhub.elsevier.com/S0093-934X(14)00173-4/h0310http://refhub.elsevier.com/S0093-934X(14)00173-4/h0315http://refhub.elsevier.com/S0093-934X(14)00173-4/h0315http://refhub.elsevier.com/S0093-934X(14)00173-4/h0320http://refhub.elsevier.com/S0093-934X(14)00173-4/h0320http://refhub.elsevier.com/S0093-934X(14)00173-4/h0320http://refhub.elsevier.com/S0093-934X(14)00173-4/h0325http://refhub.elsevier.com/S0093-934X(14)00173-4/h0325http://refhub.elsevier.com/S0093-934X(14)00173-4/h0325http://refhub.elsevier.com/S0093-934X(14)00173-4/h0330http://refhub.elsevier.com/S0093-934X(14)00173-4/h0330http://refhub.elsevier.com/S0093-934X(14)00173-4/h0330http://refhub.elsevier.com/S0093-934X(14)00173-4/h0335http://refhub.elsevier.com/S0093-934X(14)00173-4/h0335http://refhub.elsevier.com/S0093-934X(14)00173-4/h0340http://refhub.elsevier.com/S0093-934X(14)00173-4/h0340http://refhub.elsevier.com/S0093-934X(14)00173-4/h0340http://refhub.elsevier.com/S0093-934X(14)00173-4/h0345http://refhub.elsevier.com/S0093-934X(14)00173-4/h0345http://refhub.elsevier.com/S0093-934X(14)00173-4/h0345http://refhub.elsevier.com/S0093-934X(14)00173-4/h0350http://refhub.elsevier.com/S0093-934X(14)00173-4/h0350http://refhub.elsevier.com/S0093-934X(14)00173-4/h0355http://refhub.elsevier.com/S0093-934X(14)00173-4/h0355http://refhub.elsevier.com/S0093-934X(14)00173-4/h0355http://refhub.elsevier.com/S0093-934X(14)00173-4/h0360http://refhub.elsevier.com/S0093-934X(14)00173-4/h0360http://refhub.elsevier.com/S0093-934X(14)00173-4/h0360http://refhub.elsevier.com/S0093-934X(14)00173-4/h0365http://refhub.elsevier.com/S0093-934X(14)00173-4/h0365http://refhub.elsevier.com/S0093-934X(14)00173-4/h0365http://refhub.elsevier.com/S0093-934X(14)00173-4/h0370http://refhub.elsevier.com/S0093-934X(14)00173-4/h0370http://refhub.elsevier.com/S0093-934X(14)00173-4/h0370

Visuospatial complexity modulates reading in 1 Introduction1.1 Visuospatial complexity in the VWFA1.2 Visuospatial complexity versus phonologic1.3 Visuospatial complexity in Hindi/Devanagari1.4 The current study

2 Materials and methods2.1 Participants2.2 Materials2.3 Procedure2.3.1 Image acquisition and analysis

3 Results3.1 Behavioral results3.2 Neuroimaging results3.2.1 ANOVA results3.2.2 Subtraction analyses3.2.3 RoI analyses

4 Discussion5 ConclusionAcknowledgmentsReferences