Brain and Cognition 88 (2014) 83–89

Contents lists available at ScienceDirect

Brain and Cognition

journal homepage: www.elsevier .com/ locate /b&c

Investigating letter recognition in the brain by varying typeface:An event-related potential study

http://dx.doi.org/10.1016/j.bandc.2014.05.0010278-2626/� 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author. Address: School of Psychology, Social Work and SocialPolicy, GPO Box 2741, Adelaide 5001, Australia.

E-mail address: Hannah.Keage@unisa.edu.au (H.A.D. Keage).

Hannah A.D. Keage a,⇑, Scott Coussens a, Mark Kohler a, Myra Thiessen b, Owen F. Churches c

a Cognitive Neuroscience Laboratory, School of Psychology, Social Work and Social Policy, University of South Australia, Australiab School of Art, Architecture and Design, University of South Australia, Australiac Brain and Mind Laboratory, School of Psychology, Flinders University, Australia

a r t i c l e i n f o a b s t r a c t

Article history:Accepted 2 May 2014Available online 28 May 2014

Keywords:Letter recognitionEvent-related potentialTypeface

We aimed to investigate the contributions of visual letter form and abstract letter identity to the timecourse of letter recognition, by manipulating the typeface (i.e. font) in which letters were presented.Twenty-six adult participants completed a modified one-back task, where letters where presented ineasy-to-read typefaces (‘‘fluent’’ letter stimuli) or difficult-to-read typefaces (‘‘disfluent’’ letter stimuli).Task instructions necessitated that participant’s focus on letter identity not visual letter form. Electro-encephalography was collected and event-related potentials (ERPs) were calculated relative to letterstimuli. It was found that typeface affected both early-mid (N1 amplitude and P2–N2 amplitude andlatency) and late processing (450–600 ms), thereby including time points whereby it is theorised thatabstract identity is extracted from visual letter form (that is, 300 ms post-stimulus). Visual featuresof the letter therefore affect its processing well beyond the currently theorised point at which abstractinformation is extracted; which could be explained by a feedback loop between abstract letter represen-tations and lower-level visual form processing units, which is not included in current cognitive readingmodels.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

Reading words, by recognising groupings of letter forms aswords with meaning, is an extremely complex task. There are anumber of influential cognitive models of reading, such as theInteractive Activation Model (McClelland & Rumelhart, 1981),and refinements such as the Dual Route Cascaded Model(Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001) and the MultipleReadout Model (Grainger & Jacobs, 1996). All models include asimilar hierarchical system, where basic feature perception feedsforward onto a letter-identity recognition processing level. Thislevel incorporates a stage where letter meaning is abstracted fromits visual features, referred to as an abstract letter unit (Coltheartet al., 2001) or abstract letter identity (Grainger & Jacobs, 1996).These early processing stages and the concept of the abstract letterunit have gone largely untested however (Chang, Furber, &Welbourne, 2012), with the vast majority of research focusing onwhole-word reading (Finkbeiner & Coltheart, 2009). Our abilityto accurately identify individual letters underpins our capacity to

read words, yet we know very little about how the brain recognisesletters.

Letter recognition involves a highly constrained visual patternrecognition system: constructing and identifying an abstract con-cept of a letter from basic visual features (Grainger, Rey, & Dufau,2008). There appear to be abstract letter specific representationswithin the brain, predominantly within the left occipito-temporalcortex (Dehaene et al., 2004; Grainger et al., 2008). After a letteris presented, low-level visual feature processing is seen withinthe first 100 ms (ms) post-stimulus (i.e. the letter) which is unre-lated to letter meaning (Madec, Rey, Dufau, Klein, & Grainger,2012). High-level visual representations are activated around120–180 ms post-stimulus, and this is the likely time where visualfeatures are mapped on to letter representations (Carreiras, Perea,Gil-López, Abu Mallouh, & Salillas, 2013; Cornelissen, Tarkiainen,Helenius, & Salmelin, 2003; Madec et al., 2012; Petit, Midgley,Holcomb, & Grainger, 2006; Rey, Dufau, Massol, & Grainger,2009; Wong, Gauthier, Woroch, Debuse, & Curran, 2005). Thisstage appears to be case-specific (i.e. upper or lower case), butinsensitive to variation in type size (Carreiras et al., 2013; Petitet al., 2006). Processing stages around 220–300 ms appear to bemore abstract and case-independent (Petit et al., 2006). This timewindow includes recurrent processing within occipital regions

84 H.A.D. Keage et al. / Brain and Cognition 88 (2014) 83–89

which contributes to the conscious identification of the letter byamplifying or sustaining activity (Madec et al., 2012). Activity after300 ms post-stimulus has been reported to be task/responserelated (Madec et al., 2012) and therefore unrelated to visual letterform. However, findings from a recent priming experiment havesuggested that the P3 component which peaks after 300 ms ismodulated, to a small degree, by the visual similarity of letters(i.e. prime and target similarity) (Carreiras et al., 2013). Notably,although letter recognition is lateralised to the left hemisphere,there is evidence to suggest that the processing of letter visualform (typeface) is lateralised to the right hemisphere (Bartonet al., 2010).

An interesting manipulation to test our current understandingof letter recognition in the brain is to manipulate the typeface inwhich letters are presented (the font, e.g. Arial, Freestyle Script,or Comic Sans), which would enable processing of visual letterform and letter meaning to be teased apart. According to theresearch above, typeface should not affect activity 300 ms post let-ter stimulus. There only appears one study manipulating typeface(of roman letters) in a letter recognition task (Wamain, Tallet,Zanone, & Longcamp, 2012). They reported that typeface influ-enced processing of letter stimuli around 100 ms and 500–600 ms but not at the N1/170 ERP component (Wamain et al.,2012). However, the task instructions had participants direct theirattention to the typeface of the letters which would have shiftedattention toward visual letter form not abstract letter meaning.

In a whole word study, Chauncey, Holcomb, and Grainger(2008) assessed masked repetition priming of whole words (i.e.rather than single letters) by varying typeface (Arial and Gigi fonts)and size. They found that typeface (but not size) affected the pos-terior N150, but not later components. The authors suggested thatthe N/P150 component reflected feature-level processing of wordinformation; the timing corresponding with investigations of letterrecognition, where visual features are mapped on to letter repre-sentations around 120 ms (Madec et al., 2012; Petit et al., 2006;Rey et al., 2009). However, it is unclear how applicable this findingis to single letter recognition.

Behavioural studies have demonstrated that typeface influ-ences the feature detection of letters (Pelli et al., 2009; Sanocki,1988) and effects perceptual speed, memory, perceived effort, atti-tude and the detection of deception from linguistic information(Bar & Neta, 2006; Diemand-Yauman, Oppenheimer, & Vaughan,2011; Dreisbach & Fischer, 2011; Gauthier, Wong, Hayward, &Cheung, 2006; Jolicoeur, Snow, & Murry, 1987; Song & Schwarz,2008a, 2008b). For example, Diemand-Yauman et al. (2011)reported that memory was better for descriptions of fictional alienspecies when they were presented in a disfluent (e.g. hard-to-readitalicised Bodoni) as compared to a fluent (e.g. easy-to-read Arial)typeface.

We aimed to investigate letter recognition processes within thebrain by varying the typeface in which letters were presented. Bydoing so, we sought to disentangle the processing of visual letterfeatures and abstract letter meaning. There is no cognitive modelwhich explicitly details how the same letter is recognised as thesame letter if presented in a different typeface (Finkbeiner &Coltheart, 2009; Grainger et al., 2008). However, we are all awarethat an ‘a’ is an ‘a’, regardless of the font it is presented in. Weemployed a one-back paradigm and based our selection of typefacefrom the typographic literature, where typefaces are placed on afluency spectrum, and harder-to-read fonts are referred to as dis-fluent typefaces and easy-to-read fonts are referred to as fluenttypefaces. We hypothesised that typeface fluency would affectERP componentry prior to 300 ms. Further, we predicted that activ-ity after 300 ms (corresponding to the letter being abstracted, withinformation unrelated to the typeface in which it was presented)would not differ between fluent and disfluent letter stimuli.

2. Methods

2.1. Participants

There were a total of 26 right-handed participants (5 male and21 female), with a mean age of 24.01 years (SD = 9.71; range19–59 years). Exclusion criteria were a history of psychologicaland psychiatric disorders (including those associated withlanguage such as Specific Language Impairment and Dyslexia),unconsciousness, drug abuse, and visual disturbance not able tobe corrected.

2.2. Experimental paradigm

The task was a modified one-back task in which participantswere instructed to press a response button with one index fingerwhen they saw the same letter twice in a row (i.e. a target stimu-lus), and another button with the other finger for non-repeats (i.e.an update stimulus). The buttons used to indicate stimuli identitywere counterbalanced across participants. The target probabilitywas 38%. Stimuli were presented for 500 ms and the inter-stimulusinterval varied between 1250 and 1750 ms. The task was presentedin two blocks, each lasting around 8 min (n = 240 stimuli perblock), with a short break provided to participants between blocks.The stimulus train is illustrated in Fig. 1.

Different sets of letters were used for each trial block, one hadthe letters o, a, e and c, and the other f, l, t and I. These sets mini-mised inter-letter visual form differences (Grainger et al., 2008).There were four pseudo-randomised versions of each block, andblock order was counterbalanced. Each block had half the letterspresented in a fluent typeface (Arial or Times New Roman) and halfin a disfluent typeface (Lucida Blackletter or Edwardian Script). Thesame font did not appear more than three times in a row, and tar-get stimuli had an equal probability of being presented in a disflu-ent or fluent font. Letter aspect ratio (i.e. the internal ratio of eachletter) was controlled over typefaces, as it is known that expertreaders tune to aspect ratio but not to the other font changes(Gauthier et al., 2006). Letter aspect ratio is illustrated in Fig. 2,along with the four fonts employed in this experiment to varytypeface.

2.3. Psychophysiological recording and measures

Participants were seated in a darkened, sound attenuated roomapproximately 60 cm from the monitor that presented the stimuli.Electroencephalography (EEG) was recorded using a modifiedQuickcap (Compumedics Neuroscan, Charlotte, North Carolina,USA). Sixty-four silver/silver-chloride electrodes were arrangedaccording to the 10–20 system (American Electroencephalo-graphic Society, 1994). Reference was at the tip of the nose andground at FPZ. Vertical and horizontal eye movements wererecorded in bipolar channels with electrodes 1 cm above andbelow the left eye and from the outer canthus of each eye. Contin-uous EEG was recorded using a Synamps II amplifier (Compumed-ics Neuroscan) that sampled the analogue signal at 1000 Hz withan analogue bandpass filter between 0.1 and 100 Hz. Impedanceat each electrode was reduced to below 5 KX at the start of theexperiment. EEG data was edited using SCAN4.5 (CompumedicsNeuroscan); data contaminated by excessive motor movementwas removed, eye-blink artefact was removed using the ocularartefact reduction tool within SCAN4.5 and only trials involving acorrect response were included in ERP averages. On average, 77%of trials were kept for ERP analysis.

Peak amplitude and latency measures were taken from the fol-lowing ERP components of interest – posterior N1, occipital P2,

Fig. 1. Illustration of the stimulus train.

Fig. 2. (A) An illustration of typographic letter aspect ratio, i.e. the relative proportions of the x-height and ascenders/descenders and (B) how letter aspect ratios differbetween fonts of the same point size. In this experiment we controlled letter aspect ratios for the four fonts employed (C).

H.A.D. Keage et al. / Brain and Cognition 88 (2014) 83–89 85

occipital N2, and P3. The occipital P2 and N2 components weretreated as a complex, and peak-to-peak amplitude and latency dif-ferences were calculated. The mean amplitude 450–600 ms post-stimulus (i.e. after the P3 component) was also calculated, whichwere termed the sustained late activity. The time windows forpeak detection, and electrodes included in analyses, are outlinedin Table 1.

2.4. Statistical analyses

Repeated measures ANOVAs were conducted for each compo-nent/complex (peak amplitude and latency), and for late-sustainedactivity. The repeated measures were typeface (fluent/disfluent),

Table 1Summary of ERP components and areas measured in the study.

Window used (ms post-stimulus) Meas

Posterior N1 100–200 PeakOccipital N2 200–230 PeakOccipital P2 220–260 PeakCentral P3 300–450 PeakSustained late activity 450–600 Mean

response-relevance (target/non-target), and hemisphere (left/right). Paired t-tests were used for planned comparisons (withBonferroni correction).

3. Results

Group averaged ERPs for electrodes of interest (PO7, PO8, CP1,CP2, CB1 and CB2) are displayed in Fig. 3.

3.1. Behavioural

Participants completed the task well with an overall error rateof 10% (notably all stimuli required a response). Reaction times

ure taken Electrodes included in analyses

Left hemisphere Right hemisphere

amplitude and latency PO7 PO8amplitude and latency CB1 CB2amplitude and latency CB1 CB2amplitude and latency CP1 CP2amplitude CP1 CP2

Fig. 3. Group-averaged ERPs from electrodes – CP1, CP2, PO7, PO8, CB1 and CB2 – where components of interest were maximal.

86 H.A.D. Keage et al. / Brain and Cognition 88 (2014) 83–89

were quicker to target stimuli (M = 550.32, SD = 83.57) than updatestimuli (M = 598.44, SD = 103.22) [t(26) = 3.55, p = .002]. Therewere no significant differences (p > .05) in reaction time and errorrates following fluent as compared to disfluent stimuli (error rates:fluent stimuli 10.50% and disfluent stimuli 10.06%; reaction time:fluent stimuli 568.60 ms and disfluent stimuli 576.33 ms).

3.2. Psychophysiological

3.2.1. Posterior N1 ERP componentThere was a main effect of typeface on the posterior N1

amplitude [F(1,25) = 17.67, P < .001], where disfluent stimuli

(M = �8.42 lV, SD = 5.93) produced larger N1 components thanfluent stimuli (M = �6.45 lV, SD = 5.41). This was qualified by athree way interaction between response-relevance, typeface andhemisphere [F(1,25) = 4.80, p = .038], however after planned com-parisons were run, it was found that this interaction was drivenby effects irrelevant to the hypotheses. Posterior N1 latency didnot vary as a function of stimuli, typeface and hemisphere, orinteractions between these factors.

3.2.2. Occipital P2–N2 ERP complexThere was a main effect of typeface for the P2–N2 ERP complex

[F(1,25) = 28.02, p < .001], where disfluent stimuli generated a

H.A.D. Keage et al. / Brain and Cognition 88 (2014) 83–89 87

larger complex (M = 2.68 lV, SD = 2.90; measured from the P2 tothe N2 peak) than fluent stimuli (M = 0.26 lV, SD = 2.35). Therewas also a main effect of response-relevance [F(1,25) = 4.72,p = .039] where targets generated a larger complex (M = 1.80 lV,SD = 2.54) than updates (M = 1.14 lV, SD = 2.44). This main effectwas furthered by a two-way interaction with hemisphere[F = 4.97, p = .035], which represented target stimuli eliciting largercomplexes over the left hemisphere (M = 1.71 lV, SD = 2.77) ascompared to update stimuli (M = 0.80 lV, SD = 2.67) but no differ-ences between target and update stimuli over the righthemisphere.

There was a main effect of response-relevance for the P2–N2complex latency (i.e. the time from the P2 to the N2 component)[F(1,25) = 5.55, p = .027], with target stimuli eliciting longer com-plexes (M = 37.04 ms, SD = 12.24) than update stimuli (M = 33.79,SD = 11.04). There were no other main effects or interactions forP2–N2 complex latency.

3.2.3. P3 ERP componentP3 components were larger following target stimuli

(M = 20.09 lV, SD = 6.28), as compared to update stimuli(M = 16.15 lV, SD = 5.67) [F(1,25) = 41.93, p < .001]. The maineffect for typeface on P3 amplitude did not reach conventional sig-nificance levels [F(1,25) = 4.19, p = .051], which represented P3components being larger following disfluent stimuli (M = 18.48,SD = 5.85) than fluent stimuli (M = 17.76 lV, SD = 5.84). P3 compo-nents following target stimuli (M = 400 ms, SD = 40.18) were alsoearlier than following update stimuli (M = 417.19, SD = 47.51)[F(1,25) = 6.25, p = .019]. P3 latency did not vary relative to fluency.

3.2.4. Sustained-late activityThere was a main effect of typeface [F(1,26) = 14.72, p = .001]

where disfluent stimuli (M = 14.24 lV, SD = 5.51) generated moresustained late activity than fluent stimuli (M = 12.73 lV,SD = 6.25); and a main effect of response-relevance[F(1,25) = 9.38, p = .005) where sustained-late activity was greaterfollowing targets (M = 14.38 lV, SD = 5.98) than updates(M = 12.60 lV, SD = 6.00).

3.2.5. Sensitivity analysesGiven update ERPs were computed on twice as many trials as

target ERPs, we conducted sensitivity analyses to ensure resultswere not driven by unequal number of epochs forming ERPs foreach condition. We identified the smallest number of epochs forany condition for each individual (was always a target condition,either target fluent or target disfluent) and then used this numberto calculate ERPs using a random selection of epochs for each con-dition, with the same number of epochs across conditions for eachparticipant. Results were robust, with the only difference betweenresults concerning P3 latency, where the updated results revealeda significant interaction (that was not significant in original analy-ses) between response-relevance and typeface [F(1,25) = 6.72,p = .016], where update disfluent stimuli elicited a later P3 compo-nent than target disfluent stimuli. This effect may be driven by theaddition of noise into the ERP calculations.

4. General discussion

This experimental paradigm illustrates the complexity of theletter recognition process, where visual features of the letter affectits processing well beyond the currently theorised point at whichabstract information is extracted. The typeface of a letter thereforeinfluences processes during letter recognition and the use of lettermeaning by higher-order cognitive processes. Results regardingthe N1 conflict with another study of letter recognition via the

use of typeface (Wamain et al., 2012), however this is likely dueto experimental demand differences – with ours necessitating afocus on letter meaning and theirs on visual letter form. Interest-ingly, despite these psychophysiological differences in the process-ing of letters presented in different typefaces there were noperformance differences (i.e. reaction time and error). This maybe due to the task being relatively simple, or that participantsbecame quickly accustomed with the employed fonts (Sanocki,1988).

The posterior N1 (peaking around 160 ms) was larger followingdisfluent letter stimuli, as compared to fluent. This was accordingto hypothesis, as it has been previously reported that this compo-nent is sensitive to visual characteristics of letters (Carreiras et al.,2013) as well as words (Chauncey et al., 2008). The N1 is thoughtto reflect the activation of the higher-order visual representation ofletters (Cornelissen et al., 2003; Madec et al., 2012; Petit et al.,2006; Rey et al., 2009; Wamain et al., 2012) and attention to suchobject representations (Churches, Wheelwright, Baron-Cohen, &Ring, 2010). Our results therefore indicate that the initial abstrac-tion of letter meaning is more difficult when the letter is presentedin a disfluent typeface, and further, that such a presentation cap-tures more attention than material written in a fluent typeface.Thus, previous behavioural findings that memory is greater forinformation written in disfluent typefaces (e.g. Diemand-Yaumanet al., 2011) may be due to the increased attention captured bythe disfluent typefaces at the time of encoding (Desimone &Duncan, 1995).

Interestingly, experience with letters has been found to increasethe amplitude of the N1, specifically over the left hemisphere, afinding similar to pseudoletters evoking a smaller posterior N1than normal letters (Wong et al., 2005). Participants in the currentstudy were likely all very experienced with English letters (Maureret al., 2006). However, it is also likely that they were more experi-enced with the fonts of Arial and Times New Roman than Edwar-dian Script and Blackletter. Hence, these data suggest thatexperience with a typeface operates counter to experience withletters, with more disfluent/unfamiliar typefaces increasing earlyprocessing.

Marked differences relative to typeface were also seen for theoccipital P2 and N2 components, around 220 and 260 ms respec-tively. This effect is very clear looking at the ERPs, with this com-plex nearly undetectable following fluent letter stimuli butpronounced following disfluent letter stimuli (the reason for ana-lysing as a complex rather than separate components). Accordingto this time-window, these components reflect activation ofabstract letter representations (Petit et al., 2006), including recur-rent processing within occipital regions contributing to the con-scious identification of the letter by amplifying or sustainingactivity (Madec et al., 2012). In the broader object recognitionliterature, this time window has been associated with target detec-tion and salience evaluation (Potts, Liotti, Tucker, & Posner, 1996;Potts & Tucker, 2001), a product of interaction between frontaland occipito-temporal processing (Bar, 2003). Bar (2003) suggeststhat partially-analysed course versions of the presented stimulusare rapidly projected to the prefrontal cortex from early-visualareas, to create expectations, which are later fed-back to occip-ito-temporal areas to guide fine-grained visual processing. Thisprocess facilitates object recognition by limiting the number ofvisual representations that need to be considered.

Therefore, the finding that letters presented in disfluent type-faces generate larger P2–N2 complexes suggests that the cognitivedemands of abstracting letter meaning from visual letter form aregreater for disfluent as compared to fluent typefaces (Petit et al.,2006). These cognitive demands we hypothesise may be drivenby (1) greater recurrent occipital processing required for disfluentas compared to fluent stimuli, to strengthen bottom-up signals for

88 H.A.D. Keage et al. / Brain and Cognition 88 (2014) 83–89

later conscious recognition (Madec et al., 2012; Wyatte, Curran, &O’Reilly, 2012); and/or (2), a more complicated interplay betweentop-down and bottom-up processes (Bar, 2003) as prefrontalexpectatory representations are likely to be less accurate for disflu-ent as compared to fluent letter stimuli. Further experimentalmanipulations would be required to investigate these possibilities.

In this experiment, target stimuli also generated larger P2–N2complexes (regardless of typeface) as compared to update stimuli,specifically over the left hemisphere. This has been reported previ-ously during a digit target-identification task (Potts et al., 1996)and likely reflects the interplay between frontal and occipito-temporal processing (Bar, 2003), which drives salience evaluation(Potts & Tucker, 2001). The degree of saliency will also be affectedby target probability (which was less than non-target probabilityin this experiment) that has been long established to affect ERPcomponentry (Duncan-Johnson & Donchin, 1977). The effect beingover the left, rather than the right, hemisphere is the product of thelateralization of letter processing (Dehaene et al., 2004; Graingeret al., 2008; Polk et al., 2009). This was the only significant effectfor hemisphere (i.e. no hemispheric effect for other components/sustained-late activity), despite letter processing being lateralisedto left hemisphere (Dehaene et al., 2004; Grainger et al., 2008;Polk et al., 2009), which is likely due to the typeface of letters beinglateralised to the right (Barton et al., 2010) and therefore counter-ing the left-letter processing predominance.

Activity after 300 ms post-stimulus should be task/responserelated (Madec et al., 2012), rather than related to the typefacein which the letter was presented in. The effect of typeface onthe P3 did not reach conventional significance levels (p = .051),however the effect was in the expected direction; with disfluentletters generating a larger P3 than fluent letters. This effect is sim-ilar to a recent experiment comparing Roman and Arabic letters(Carreiras et al., 2013). This would suggest that letters presentedin disfluent typefaces are more difficult to consciously recognise(Koenig-Robert & VanRullen, 2013). Supporting this effect butagainst our hypothesis, sustained activity 450–550 ms post-stimu-lus also showed an effect of fluency, whereby mean activity wasgreater for disfluent than fluent stimuli. This suggests that is takesmore energy to maintain disfluent letter stimuli in working mem-ory than fluent letter stimuli (even though typeface was unrelatedto task instructions) (Luck & Vogel, 2013), that is, typeface affectsthe maintenance of working memory. This sustained-late activitytypeface effect is similar to a study by Wamain et al. (2012) whoreported that the motor information embodied within lettersaffects their processing between 500 and 600 ms.

This study is not without limitations. We only used two sets offour letter stimuli, so to reduce inter-letter visual form differences.It would be essential to ensure effects remained when all letters ofalphabet were assessed. Further, it will be essential to comparethese letter findings to whole-word reading to investigate readingand the abstraction of information via the use of typeface, to teaseapart processes related to visual features. Although letters areinherently language-based, and our experimental paradigm reliedon language comparison (i.e. was the letter the same as the oneprior?), we cannot ensure that these effects are driven by languagerather than object recognition processes. However, it is of note thata recent study reported a rhyme-effect indexed by the N450 fol-lowing single letter presentations, indicating automatic mappingof letters to phonological processes (Stevens, McIlraith, Rusk,Niermeyer, & Waller, 2013). Further, these findings bring intoquestion the role of perceived effort and familiarity (Song &Schwarz, 2008b). Our disfluent typefaces are inarguably less famil-iar than our fluent typefaces. This would be the case even for type-face experts, typographers.

In conclusion, findings showed how the visual form of a letterinfluences its processing through to late processing stages, which

could suggest a feedback loop between abstract letter representa-tions and lower-level visual processing areas. Although no currentcognitive reading models include such a feedback loop, thissuggestion would be supported by neurobiological models of lan-guage (Poeppel, Emmorey, Hickok, & Pylkkänen, 2012) as well asmodels of object recognition (Bar, 2003). These neural processingdifferences related to typeface existed without any performancedifferences, which highlights the sensitivity of ERP data. Thisexperiment has shown the utility of typeface to tease apart cogni-tive processes involved in letter recognition.

References

American Electroencephalographic Society (1994). Guideline thirteen: Guidelinesfor standard electrode position nomenclature. AmericanElectroencephalographic Society. Journal of Clinical Neurophysiology: OfficialPublication of the American Electroencephalographic Society, 11(1), 111–113.

Bar, M. (2003). A cortical mechanism for triggering top-down facilitation in visualobject recognition. Journal of Cognitive Neuroscience, 15(4), 600–609.

Bar, M., & Neta, M. (2006). Humans prefer curved visual objects. PsychologicalScience, 17(8), 645–648.

Barton, J. J. S., Sekunova, A., Sheldon, C., Johnston, S., Iaria, G., & Scheel, M. (2010).Reading words, seeing style: The neuropsychology of word, font andhandwriting perception. Neuropsychologia, 48(13), 3868–3877.

Carreiras, M., Perea, M., Gil-López, C., Abu Mallouh, R., & Salillas, E. (2013). Neuralcorrelates of visual versus abstract letter processing in Roman and Arabicscripts. Journal of Cognitive Neuroscience, 1–11.

Chang, Y.-N., Furber, S., & Welbourne, S. (2012). Modelling normal and impairedletter recognition: Implications for understanding pure alexic reading.Neuropsychologia, 50(12), 2773–2788.

Chauncey, K., Holcomb, P. J., & Grainger, J. (2008). Effects of stimulus font and sizeon masked repetition priming. Language and Cognitive Process, 23(1), 183–200.

Churches, O., Wheelwright, S., Baron-Cohen, S., & Ring, H. (2010). The N170 is notmodulated by attention in autism spectrum conditions. Neuroreport, 21(6),399–403.

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

Cornelissen, P., Tarkiainen, A., Helenius, P., & Salmelin, R. (2003). Cortical effects ofshifting letter position in letter strings of varying length. Journal of CognitiveNeuroscience, 15(5), 731–746.

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: Behavioral andneuroimaging evidence. Psychological Science, 15(5), 307–313.

Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual-attention.Annual Review of Neuroscience, 18, 193–222.

Diemand-Yauman, C., Oppenheimer, D. M., & Vaughan, E. B. (2011). Fortune favorsthe bold (and the Italicized): Effects of disfluency on educational outcomes.Cognition, 118(1), 111–115.

Dreisbach, G., & Fischer, R. (2011). If it’s hard to read. . . try harder! Processingfluency as signal for effort adjustments. Psychological Research, 75(5). 376–373/383.

Duncan-Johnson, C. C., & Donchin, E. (1977). On quantifying surprise: The variationof event-related potentials with subjective probability. Psychophysiology, 14(5),456–467.

Finkbeiner, M., & Coltheart, M. (2009). Letter recognition: From perception torepresentation. Cognitive Neuropsychology, 26(1), 1–6.

Gauthier, I., Wong, A. C., Hayward, W. G., & Cheung, O. S. (2006). Font tuningassociated with expertise in letter perception. Perception, 35(4), 241–259.

Grainger, J., & Jacobs, A. (1996). Orthographic processing in visual word recognition:A multiple readout model. Psychological Review, 103, 518–565.

Grainger, J., Rey, A., & Dufau, S. (2008). Letter perception: From pixels topandemonium. Trends in Cognitive Sciences, 12(10), 381–387.

Jolicoeur, P., Snow, D., & Murry, J. (1987). The time to identify disorientated letters:Effects of practice and font. Canadian Journal of Psychology, 41(3), 303–316.

Koenig-Robert, R., & VanRullen, R. (2013). SWIFT: A novel method to track theneural correlates of recognition. NeuroImage, 81, 273–282.

Luck, S. J., & Vogel, E. K. (2013). Visual working memory capacity: Frompsychophysics and neurobiology to individual differences. Trends in CognitiveSciences, 17(8), 391–400.

Madec, S., Rey, A., Dufau, S., Klein, M., & Grainger, J. (2012). The time course of visualletter perception. Journal of Cognitive Neuroscience, 24(7), 1645–1655.

Maurer, U., Brem, S., Kranz, F., Bucher, K., Benz, R., Halder, P., et al. (2006). Coarseneural tuning for print peaks when children learn to read. NeuroImage, 33(2),749–758.

McClelland, J., & Rumelhart, D. (1981). An interactive activation model of contexteffects in letter perception: Part 1. Psychological Review, 88, 275–407.

Pelli, D. G., Majaj, N. J., Raizman, N., Christian, C. J., Kim, E., & Palomares, M. C.(2009). Grouping in object recognition: The role of a Gestalt law in letteridentification. Cognitive Neuropsychology, 26(1), 36–49.

H.A.D. Keage et al. / Brain and Cognition 88 (2014) 83–89 89

Petit, J.-P., Midgley, K., Holcomb, P. J., & Grainger, J. (2006). On the time course ofletter perception: A masked priming ERP investigation. Psychonomic Bulletin &Review, 13(4), 674–681.

Poeppel, D., Emmorey, K., Hickok, G., & Pylkkänen, L. (2012). Towards a newneurobiology of language. The Journal of Neuroscience, 32(41), 14125–14131.

Polk, T. A., Lacey, H. P., Nelson, J. K., Demiralp, E., Newman, L. I., Krauss, D. A., et al.(2009). The development of abstract letter representations for reading:Evidence for the role of context. Cognitive Neuropsychology, 26(1), 70–90.

Potts, G. F., Liotti, M., Tucker, D. M., & Posner, M. I. (1996). Frontal and inferiortemporal cortical activity in visual target detection: Evidence from highspatially sampled event-related potentials. Brain Topography, 9(1), 4–14.

Potts, G. F., & Tucker, D. M. (2001). Frontal evaluation and posterior representationin target detection. Cognitive Brain Research, 11(1), 147–156.

Rey, A., Dufau, S., Massol, S., & Grainger, J. (2009). Testing computational models ofletter perception with item-level event-related potentials. CognitiveNeuropsychology, 26(1), 7–22.

Sanocki, T. (1988). Font regularity constraints on the process of letter recognition.Journal of Experimental Psychology: Human Perception and Performance, 14(3),472–480.

Song, H., & Schwarz, N. (2008a). Fluency and the detection of misleading questions:Low processing fluency attenuates the Moses illusion. Social Cognition, 26(6),791–799.

Song, H., & Schwarz, N. (2008b). If it’s hard to read, it’s hard to do: Processingfluency affects effort prediction and motivation. Psychological Science, 19(10),986–988.

Stevens, C., McIlraith, A., Rusk, N., Niermeyer, M., & Waller, H. (2013). Relativelaterality of the N170 to single letter stimuli is predicted by a concurrentneural index of implicit processing of letternames. Neuropsychologia, 51(4),667–674.

Wamain, Y., Tallet, J., Zanone, P.-G., & Longcamp, M. (2012). Brain responses tohandwritten and printed letters differentially depend on the activation state ofthe primary motor cortex. NeuroImage, 63(3), 1766–1773.

Wong, A. N., Gauthier, I., Woroch, B., Debuse, C., & Curran, T. (2005). An earlyelectrophysiological response associated with expertise in letter perception.Cognitive, Affective, & Behavioral Neuroscience, 5(3), 306–318.

Wyatte, D., Curran, T., & O’Reilly, R. (2012). The limits of feedforward vision:Recurrent processing promotes robust object recognition when objects aredegraded. Journal of Cognitive Neuroscience, 24(11), 2248–2261.