Abstract

This paper shows that the nature of letters—consonant versus vowel—modulates the process of letter position assignment during visual word recognition. We recorded Event Related Potentials while participants read words in a masked priming semantic categorization task. Half of the words included a vowel as initial, third, and fifth letters (e.g., acero [steel]). The other half included a consonant as initial, third, and fifth (e.g., farol [lantern]). Targets could be preceded 1) by the initial, third, and fifth letters (relative position; e.g., aeo—acero and frl—farol), 2) by 3 consonants or vowels that did not appear in the target word (control; e.g., iui—acero and tsb—farol), or 3) by the same words (identity: acero–acero, farol–farol). The results showed modulation in 2 time windows (175–250 and 350–450 ms). Relative position primes composed of consonants produced similar effects to the identity condition. These 2 differed from the unrelated control condition, which showed a larger negativity. In contrast, relative position primes composed of vowels produced similar effects to the unrelated control condition, and these 2 showed larger negativities as compared with the identity condition. This finding has important consequences for cracking the orthographic code and developing computational models of visual word recognition.

Introduction

Printed word recognition is a key process for reading. Whether or not words are recognized visually as a whole via their constituent letters has been debated for more than a century. Although early research suggested that words could be identified by the use of word shape (see Cattell 1886) and some investigators still argue that supraletter features such as word shape play a role in visual word recognition (e.g., Healy et al. 1987; Healy and Cunningham 1992; Allen et al. 1995), most theorists currently support the idea that words are initially formed from component letters (analytical models; e.g., the search model, Forster 1976; the multiple read-out model, Grainger and Jacobs 1996; the interactive–activation model, McClelland and Rumelhart 1981; and the activation–verification model, Paap et al. 1982). Recent research has shown that words in alphabetic languages are processed via their constituent letters (see Pelli et al. 2003). However, the process of coding for letters (i.e., how letters are assigned within words, including their identity and position) remains poorly understood. To recognize a printed word, we need to process the identity and position of its letters, hence distinguishing between trail and trial, but not between tABLe and TabLE or chair and CHAIR. Thus, 1 critical issue for understanding visual word recognition is the nature and functioning of the orthographic input code to the system. Recently, a neurobiological model has been proposed (the local combination detector or LCD model; Dehaene et al. 2005; see also Vinckier et al. 2007) as a new attempt to crack the orthographic code. One important aspect of the LCD model is that it distinguishes between letter identity and letter order processes, taking into account neurobiological constraints and the development of new coding schemes for computational models of visual word recognition. According to the Dehaene et al. proposal, the brain decodes words through a hierarchy of LCDs in the occipito–infero–temporal pathway, which are sensitive to increasingly larger fragments of words. In particular, they tentatively propose detectors for letter shapes in V4, abstract letter detectors in V8, which represent letters denoting their identities but abstracted from their visual appearance (e.g., CaSe, font, and size), and detectors for letter strings in the left fusiform gyrus. However, as with current computational models, this proposal assumes that the process of assigning letters within words is similar for both consonant and vowels. Here, we present evidence from event related potentials (ERPs) that challenges this assumption.

Most current computational models of visual word recognition (e.g., McClelland and Rumelhart 1981; Paap et al. 1982; Grainger and Jacobs 1996; Coltheart et al. 2001; Perry et al. 2007) have assumed that letter position coding was “channel specific.” That is, letters are assumed to be tagged to their positions in the orthographic representation of the word presented when the identities of the letters have been encoded. However, 2 recent findings have challenged this assumption: the transposed-letter effect and the relative position priming effect.

The transposed-letter priming effect refers to the fact that nonword primes created by transposing 2 letters from a real word produce form-priming effects, relative to the orthographic control creating by replacing letters (e.g., jugde–JUDGE vs. jupte–JUDGE; Perea and Lupker 2003, 2004; see also Perea and Carreiras 2006a, 2006b, 2006c, 2008; Duñabeitia et al. 2007; Perea et al. 2008). The relative position priming effect has been defined as a “variety of orthographic priming that involves a change in length across prime and target such that shared letters can have the same order without being matched in terms of absolute, length-dependent position” (Grainger 2008, p. 16). Target words like BALCON preceded by nonword primes created by replacing some letters with hyphen marks (e.g., B-LC-N; the absolute position condition) or with the same letters without the hyphens (e.g., BLCN; the relative position condition) are recognized faster than when preceded by nonsense symbol strings (e.g., %%%%%%), or than when preceded by unrelated control strings (e.g., CRTR) (see also Humphreys et al. 1990). Importantly, the size of the effect is the same for the absolute and the relative position conditions. In addition, the effects are canceled when relative position is violated (e.g., BCLN instead of BLCN) (Peressotti and Grainger 1999; Grainger, Granier, et al. 2006), that is, when the same letters as those used as relative position primes are presented in a partially scrambled order. This condition produces similar reaction times as those to the unrelated primes. The relative position priming effect also vanishes at long prime durations (ca. 80 ms) (Grainger, Granier, et al. 2006). Thus, because these effects are short lived and similar for absolute and relative position primes, the findings suggest that the orthographic information extracted at early stages of processing is letter identity and the relative ordering of letters in the input string, with no specific encoding of the absolute position of each letter.

The transposed-letter effect and the relative position priming effect generate serious problems for standard computational models (e.g., McClelland and Rumelhart 1981; Grainger and Jacobs 1996; Coltheart et al. 2001) because none of them can account for either of these 2 effects: 1) transposed-letter neighbors are more similar to target words than replaced-letter neighbors and 2) 2 strings that differ in length and therefore in the absolute position of the letters but that do not differ in the relative order/position of the letters activate each other to a great extent. However, a number of input “coding schemes” have recently been proposed that successfully capture the existence of these effects (e.g., SERIOL model, Whitney 2001; SOLAR model, Davis 1999; open-bigram model, Grainger and van Heuven 2003; Overlap Open-Bigram model Grainger, Granier, et al. 2006; and overlap model, Gomez et al. 2008), although the basic mechanisms of how letter position is encoded differ across these models (e.g., via the activation of open bigrams in the SERIOL and open-bigram models, via a spatial coding in the SOLAR model, or via a noisy perceptual input in the overlap model). Interestingly, these input schemes are favored in the LCD model. There is 1 caveat, however: These models assume that consonants and vowels are processed in exactly the same way, but this does not seem to be the case.

Recent transposed-letter experiments (e.g., Perea and Lupker 2004; see also Perea and Carreiras 2006a; Carreiras et al. 2007, 2009) have found a differential role of consonants and vowels in transposed-letter similarity effects. For instance, Perea and Lupker (2004) obtained a masked priming effect for consonant transpositions (relovución–REVOLUCIÓN vs. the replaced-letter control condition (retosución–REVOLUCIÓN), but not for vowel transpositions (reluvoción–REVOLUCIÓN vs. relavición–REVOLUCIÓN) (see Duñabeitia et al. 2009 for the suitability of this control). There are specific hypotheses suggesting that the consonant–vowel structure is determined and coded very early in word processing (e.g., Berent and Perfetti 1995). Furthermore, recent research with other paradigms and methodologies (e.g., reaction time ERPs, functional magnetic resonance imaging) in auditory and visual word recognition suggests that the processing of vowels and consonants may be different. For instance, adults are more likely to replace vowels than consonants when instructed to change 1 phoneme from a nonword to make it a word (“kebra” becomes “cobra” rather than “zebra”) (Van Ooijen 1996; Cutler et al. 2000). Moreover, participants find it very difficult to use transitional probabilities between successive vowels when presented with an auditory string of artificial speech and later on are asked to identify possible words, whereas this process is much easier for successive consonants, demonstrating the greater importance of consonants for word identification than of vowels (Peña et al. 2002; Bonatti et al. 2005). Vowels and consonants produce different effects on regional brain activation when participants are engaged in a lexical decision task with pseudowords created by changing 2 consonants or 2 vowels (Carreiras and Price 2008). Readers find more difficulty in recognizing a word when 2 consonants as compared with 2 vowels are slightly delayed during the presentation of the word (Lee et al. 2001, 2002; Carreiras et al. 2009). Infants fail to discriminate vowels in a lexical acquisition task, whereas they do discriminate between consonants (Nazzi 2005). Finally, a number of neuropsychological studies have found a double dissociation, so that some patients find vowels more difficult to produce than consonants, whereas others show the reverse pattern (e.g., Caramazza et al. 2000).

One potential target area for identifying differences between consonants and vowels is the letter assignment process in visual word recognition. If differences are found here, the effect will have very important theoretical consequences for the development of computational models of word recognition and letter identification, as well as for the LCD model, which assumes that some of the new coding schemes have neurobiological plausibility, but does not make any distinction between consonants and vowels in the letter assignment process. To investigate this possibility, we will measure ERPs on the relative position priming effect. Previous behavioral research (Duñabeitia and Carreiras submitted) has already shown that the relative position priming effect vanishes when the letters manipulated are vowels, whereas it is preserved when the letters are consonants (e.g., csn–CASINO vs. aia–ANIMAL). In addition, these authors showed that the relative position that only holds for consonants cannot be accounted for just by the frequency of letters: When they presented primes that contained only vowels, only high frequency consonants or only low-frequency consonants, they found a significant relative position priming effect of the same size for the 2 consonant conditions but no effects for vowel primes. Thus, despite the fact that vowels are typically of higher frequency than consonants, Duñabeitia and Carreiras showed that the frequency of the letters by itself is not the factor responsible for the vanishing of the relative position priming effect (note that, otherwise, high frequency consonants should have produced reduced priming effects as compared with low-frequency consonants). Finally, they also showed that the inclusion of repeated letters in the primes did not have an impact on the relative position priming and type of letter (consonant, vowel) interaction (e.g., lbl–GLOBAL and cbr–ICEBERG vs. iia–DIGITAL and iea–MINERAL). However, to claim that the consonant–vowel difference is influencing the letter position assignment process, it is critical to show that it is affecting the very early moments of visual word recognition. So far, some previous attempts to show early consonant–vowel effects using transposed-letter stimuli either showed late effects more related with response bias than with perceptual encoding (e.g., Carreiras et al. 2007) or only early effects for vowels (Carreiras et al. 2009). This paper investigates in detail the time course of the processing of vowels and consonants during letter position assignment in visual word recognition, using the relative position priming paradigm combined with the recording of ERPs.

ERPs are functionally decomposable to a greater extent than behavioral data, thus enabling conclusions not only about the existence of processing differences between vowels and consonants but also about the level of processing at which these differences occur. ERPs are voltage changes recorded from the scalp and extracted from the background electroencephalogram by averaging time-locked responses to stimuli onset. Of specific interest for our study are 3 components: N/P150, N250, and N400. The N/P150 component has a posterior scalp distribution focused over right occipital scalp sites and is larger (i.e., more positive) to target words that are repetitions of a prior masked prime word compared with targets that are unrelated to their corresponding masked prime (Holcomb and Grainger 2006). In addition, the N/P150 component has been found with the masked priming paradigm using single letters as primes and targets (Petit et al. 2006). They found that the P150 was significantly larger to mismatches between prime and target letter case, but more so when the features of the lower and upper case versions of the letters were physically different compared with when they were physically similar (e.g. A–A compared with a–A vs. C–C compared with c–C). Thus, this component seems to be sensitive to processing at the level of visual features. Recent research has shown that this early ERP component can take the form of a bipolar effect across frontal and occipital electrodes (e.g., Chauncey et al. 2008), so that although it is more positive over occipital areas, it is more negative over frontal areas of the scalp. The N250 component has been associated with the degree of prime–target orthographic overlap and phonological overlap in masked priming, suggesting that it is sensitive to processing sublexical representations (Grainger, Kiyonaga, and Holcomb 2006; Holcomb and Grainger 2006; Carreiras et al. 2009; Carreiras, Perea, et al. 2009), at least when the prime strings do not constitute real words (Duñabeitia et al. 2009). The N250 has a more widespread scalp distribution than the N/P150. Two windows can be isolated in the N250, the first window being more sensitive to orthographic processing and the second to phonological processing (see Grainger, Kiyonaga, and Holcomb 2006; Carreiras et al. 2009; Carreiras, Perea, et al. 2009). The N400 component is a negative deflection occurring around 400 ms after word presentation that has been associated with lexical–semantic processing (see Kutas and Federmeier 2000; Holcomb et al. 2002; Muller, Duñabeitia, and Carreiras, unpublished data). Specifically, the amplitude of this negativity is an inverse function of lexical frequency, of lexicality (e.g, Neville et al. 1992; see also Carreiras et al. 2005; Barber and Kutas 2007). In addition, items from small orthographic or syllabic neighborhoods produce an N400 of smaller amplitude than items from a large orthographic or syllabic neighborhood (Holcomb et al. 2002; Barber et al. 2004).

In particular, we asked: 1) whether consonants and vowels have a differential influence on the letter assignment process in visual word recognition, for which the relative position priming was manipulated for vowels and consonants, and 2) whether the time course for consonants and vowels in the letter assignment process differs, because it has been shown to be different in letter identification (Carreiras et al. 2009). To this end, we sought differences in the relative position priming effects between strings that share only consonants (e.g., “frl” priming “farol,” the Spanish for “bluff” and “lantern”), or that share only vowels (e.g., “aeo” priming “acero,” the Spanish for “steel”). We also included an identity condition (e.g., “farol” priming “farol” and “acero” priming “acero”) and an unrelated condition (e.g., “tsb” and “iui”) as control conditions.

If the contribution of consonants and vowels to the letter assignment process is different and has a different time course, we expect our manipulation to have a differential impact on the ERP components described above. More specifically, for both consonant and vowels, we expect a difference in early perceptual components between the identity and the other 2 conditions, reflected in the N/P150 component, the reason being that there is a lower perceptual change in the identity condition.

Differences between relative position priming for consonants and vowels are expected in the N250 component (i.e., a component that has been posited to be sensitive to orthographic overlap), assuming that there is differential orthographic processing of these 2 types of letters in the initial stages of visual word recognition. The reason for this expectancy relies on the tight link between consonants, orthographic processing and lexical selection processes, and the importance of vowels in orthographic encoding, because the role of vowels is more related to rhythmic and syntactic patterns, as proposed by Nespor et al. (2003) (see also Carreiras and Price 2008). Finally, late differences between the manipulation of consonants and vowels should also be noticeable in the N400 component, according to the lexical constraint hypothesis: Vowel primes and consonant primes may trigger different patterns of activation in the lexicon, and more critically, different numbers of lexical candidates. Consonant primes are much more constraining than vowel primes, because there are far fewer words consistent with the consonant primes than with the vowel primes. Hence, vowel primes activate many lexical units and therefore produce more dispersion of the activation and probably more competition during the lexical selection processes.

Method

Participants

Twenty-seven undergraduate students (16 women) from the University of La Laguna participated in the experiment in exchange for course credit. All of them were native Spanish speakers, with no history of neurological or psychiatric impairment and with normal or corrected-to-normal vision. All participants were right handed, as assessed with an abridged Spanish version of the Edinburgh Handedness Inventory (Oldfield 1971).

Materials

A total of 258 words were selected from the Spanish LEXESP database (Sebastián-Gallés et al. 2000) and analyzed with the B-PAL software (Davis and Perea 2005). All of the words were 5 or 6 letters long (mean: 5.47). Half of these words (namely, 129) included a vowel as first, third, and fifth letters (e.g., acero [steel]). The mean frequency of these words was 14.23 appearances per million (range: 0.18–204.46); the mean length was 5.57 letters; and the mean number of orthographic neighbors was 1.51 (range: 0–8). The other half (the remaining 129 words) was composed of words that included a consonant as first, third, and fifth letters (e.g., farol [lantern]). These words were matched to the previous ones in frequency (mean: 17.97; range 0.18–201.07), length (mean: 5.36), and number of orthographic neighbors (mean: 2.20; range: 0–11) in a pairwise manner using the Match software (van Casteren and Davis 2007). See Table 1 for a summary of the characteristics of the materials. These words could be preceded 1) by the first, third, and fifth letters (relative position priming condition; e.g., aeo–acero and frl–farol), 2) by 3 consonants or vowels that did not appear in the target word (control priming condition; e.g., iui–acero and tsb–farol), and 3) by the same word (identity priming condition, e.g., acero–acero and farol–farol) as an extra control condition. Three lists of materials were created so that each target appeared once in each, but each time in a different priming condition. Different participants were assigned to each of the lists.

Table 1

Characteristics of the materials used in the experiment

 Frequency
 
Length
 
N
 
 Mean (SD) Range Mean (SD) Range Mean (SD) Range 
Consonants (farol) 17.97 (28.84) 0–201 5.36 (0.48) 5–6 2.27 (2.15) 0–8 
Vowels (acero) 14.23 (29.24) 0–204 5.57 (0.50) 5–6 1.51 (1.66) 0–11 
 Frequency
 
Length
 
N
 
 Mean (SD) Range Mean (SD) Range Mean (SD) Range 
Consonants (farol) 17.97 (28.84) 0–201 5.36 (0.48) 5–6 2.27 (2.15) 0–8 
Vowels (acero) 14.23 (29.24) 0–204 5.57 (0.50) 5–6 1.51 (1.66) 0–11 

Mean (M) frequency, length, and N values are provided with the standard deviation (SD, within parentheses) and ranges.

In order to make the go/no-go semantic categorization possible, we included a set of 40 animal names in the item set of each list (e.g., erizo [hedgehog], lince [linx]) of similar length, frequency, and structure to the critical words (targets: mean length 5.60, mean frequency 5.54, and number of orthographic neighbors 2.05; primes: mean length 5.55, mean frequency 9.78, and number of orthographic neighbors 2.07). These words were primed by a new set of 40 unrelated nonanimal prime words. A prime visibility test was also included in order to check for conscious identification of the masked primes. To this end, the 40 animal names were presented as primes and followed by 40 unrelated nonanimal target words. Thus, each list contained 338 trials of which 258 were experimental trials, 40 were trials with animal names for the go/no-go semantic categorization task, 12% of the total amount of trials, and finally 40 were trials for the prime visibility test, 12% of the total amount of trials.

Procedure

Participants were individually tested in a well-lit soundproof room. The presentation of the stimuli and recording of the responses was carried out using Presentation software on a computer associated to a CRT monitor. All stimuli were presented on a high-resolution monitor that was positioned at eye level 80 cm in front of the participant. Each trial consisted in the presentation of a forward mask created by hash mark symbols for 500 ms, followed by the displaying of the prime for 50 ms, and immediately followed by the presentation of the target. Primes and targets were presented in lowercase following previous work on the relative position priming effect (e.g., Peressotti and Grainger 1999; Grainger, Granier et al. 2006) in which primes and targets were also presented in the same case. Primes and targets were presented in Courier New font. In order to minimize physical overlap between primes and targets, different font sizes were used for these strings. Each character of the prime strings had a width of 0.12 in, whereas each character of the targets had a width of 0.16 in (note that Courier New font is a nonproportional font in which all letters occupy the same amount of space). Under these conditions, no saccades were required during reading of each stimulus, because the strings filled less than 1.5° of the visual field. Target items remained on the screen for 500 ms. The intertrial interval varied randomly between 700 and 900 ms. After this interval, an asterisk was presented for 1000 ms in order to allow participants’ blinks (see Fig. 1 for schematic representation of each trial). All items were presented in a different random order for each participant. Participants performed a go/no-go semantic categorization task: They were instructed to press the spacebar on the keyboard only when the letter string displayed referred to an animal name. Twenty warm-up trials, containing different stimuli from those used in the experimental trials, were provided at the beginning of the session. Participants were asked to avoid eye movements and blinks during the interval when the row of hash marks or the asterisk was not present. Each session lasted approximately 1 h and 15 min.

Figure 1.

Schematic representation of each experimental trial.

Figure 1.

Schematic representation of each experimental trial.

Electroencephalogram Recording and Analyses

Scalp voltages were collected using a BrainAmp recording system from 32 Ag/AgCl electrodes that were mounted in an elastic cap (ElectroCap International, Eaton, USA, 10-10 system). Figure 2 shows the schematic distribution of the recording sites. Linked earlobes were used as reference. Eye movements and blinks were monitored with 4 further electrodes providing bipolar recordings of the horizontal (Heog−, Heog+) and vertical (Veog−, Veog+) electrooculogram. Interelectrode impedances were kept below 10 K. EEG was filtered with an analogue band-pass filter of 0.01–100 Hz, and a digital 30-Hz low-pass filter was applied before analysis. The signals were sampled continuously throughout the experiment with a sampling rate of 256 Hz.

Figure 2.

Schematic flat representation of the 32 electrode positions from which EEG activity was recorded (front of head is at the top). Scheme of the distributed analysis is expressed here by different styled lines.

Figure 2.

Schematic flat representation of the 32 electrode positions from which EEG activity was recorded (front of head is at the top). Scheme of the distributed analysis is expressed here by different styled lines.

As in prior research with the masked priming technique, the focus was on the word targets. Only trials free of ocular artifacts (blinks and eye movements) and muscular artifacts were averaged and analyzed (more than 95% of the trials). Epochs of the EEG up to 550 ms after the onset of the target word (i.e., 50 ms after the target word disappeared from the monitor) were the primary data. The baseline correction was performed using the average EEG activity in the 200 ms preceding the onset of the target as a reference signal value—we also visually examined ERPs using baselines calculated during the 100 ms before the target and the 100 ms immediately preceding prime and the outcome was the same as the one presented here. Separate ERPs were formed for each of the experimental conditions, each of the subjects, and each of the electrode sites (see electrode numbers in Fig. 2).

We planned 2 analysis strategies: a “local” one in the occipital electrodes and a “distributed” one to evaluate effects evident all over the scalp.

As demonstrated in recent studies by Holcomb and Grainger (2006) masked priming studies show the earliest effects around 100–150 ms in the occipital electrodes. We evaluated these effects through the local analysis considering only the occipital (O1, O2) and temporal–occipital electrodes (PO7/PO8). For each time window of interest, we ran a 4-way analysis of variance (ANOVA) with Electrode (O1/O2, PO7/PO8), Hemisphere (Left, Right), Letter (Vowel, Consonant), and Condition (Identity, Relative, and Control) as factors.

The distributed analysis consisted initially in separate ANOVAs on 4 groups of electrodes (Fig. 2, see also Neville et al. 1992; Holcomb and Grainger 2006 for similar analysis patterns) considering all the 6 conditions. In the Midline (electrodes linked by thick solid lines), we ran a 3-way ANOVA considering the Electrode factor (5 levels: Fpz, Fz, Cz, Pz, and Oz), the type of Letter (2 levels: Vowels, Consonants), and the Condition (3 levels: Identity, Relative, and Control). The effects at the electrodes of the Column 1 group (electrodes linked by thin solid lines) were initially evaluated by means of a 4-way ANOVA crossing the Letter and Condition factors with the Electrode factor (3 levels: FC1/FC2, C3/C4, and CP1/CP2) and with the Hemisphere (2 levels: Left, Right). The Column 2 electrodes (electrodes linked by thin dotted lines) were analyzed crossing the same Letter, Condition, and Hemisphere factors with a 4-level Electrode factor (F3/F4, FC5/FC6, CP5/CP6, and P3/P4). Finally, the Column 3 electrodes (electrodes linked by thin dashed lines) were analyzed crossing Letter, Condition, and Hemisphere with a 6-level Electrode factor (Fp1/Fp2, F7/F8, T3/T4, T5/T6, PO7/PO8, and O1/O2). In order to evaluate the contrasts between conditions, we ran pairwise comparisons for each group of electrodes with the same spatial factors (Electrode and Hemisphere) with a 2-level Condition factor (Identity vs. Relative, Identity vs. Control, and Relative vs. Control). When appropriate, P values are reported using the Greenhouse and Geisser (1959) correction.

Mean amplitudes were obtained for different time windows. A first local analysis was performed on the 100–250 ms time window for the N/P150 and the N250 effects that could emerge in the occipital electrodes. The N250 was analyzed using the distributed analysis in the 150–250 ms time window. This time window has been shown to elicit orthographic effects in previous ERP studies (Grainger, Kiyonaga, and Holcomb 2006). A further distributed analysis was performed for the N400 (350–450 ms) component.

Results

Behavioral Measures

Participants correctly categorized 97.3% (standard deviation, SD = 0.8) of the animal names when these words were presented as targets. None of the participants made more than 5% of errors in the categorization task.

When the animal names were presented as masked primes, participants only identified them in 0.5% (SD = 0.5) of the trials. In fact, none of the participants identified the masked animal names in more than 5% of the cases. This result confirms that the participants were mostly unaware of the existence of the masked primes, and even where they did perceive something, they were not able to consciously recognize the briefly presented word.

Electrophysiological Measures

ERP grand averages time locked to the onset of the target words are represented in Figure 3. The upper panel of Figure 3 shows the contrasts between the identity, unrelated and relative position priming conditions for consonants, whereas the lower panel of Figure 3 shows the same conditions for vowels. Visual inspection of Figure 3 reveals differences in amplitude for the experimental conditions. In this first time window, which corresponds to the deflection starting around 100 ms, and lasting until 250 ms, there is a clear positive deflection in the anterior areas of the scalp and a corresponding negative deflection in the posterior areas. This first component appears to be bimodal in the posterior areas with an initial negative peak around 150 ms and a second peak of similar amplitude around 200 ms. The grand average showed different initial trends between Consonants and Vowels. Considering the 3 Vowel conditions, both in the anterior and posterior electrodes, the Relative position and Control conditions are more negative compared with the Identity condition. On the other hand, comparing the 3 Consonant conditions, the Control condition is more negative compared with the Identity and Relative position conditions. Interestingly, in the posterior electrodes (O1, O2, Oz, PO7, and PO8), the above-described dissociation for the N250 among consonants and vowels is preceded by an earlier negativity for the Identity condition. We consider this negativity as reflecting the N/P150 elicited by the prime–target physical overlap. For both consonants and vowels, the peak is more negative for the Identity compared with the other 2 conditions.

Figure 3.

ERP waves corresponding to the identity, relative position, and unrelated conditions for consonants (upper panel) and vowels (lower panel). Negative voltages are plotted up.

Figure 3.

ERP waves corresponding to the identity, relative position, and unrelated conditions for consonants (upper panel) and vowels (lower panel). Negative voltages are plotted up.

100–250 ms Window (N/P150 and Occipital N250)

The effects identified by visual inspection are supported by the overall ANOVA for each group in the 100–250 ms time window. We ran a local ANOVA on the occipital electrodes to analyze the bimodal waveform evident in the grand average (see electrode O2 in Fig. 4).

Figure 4.

ERP effects in the electrode O2 for the consonant (upper panel) and vowel (lower panel) manipulation. The N/P150 and the N250 modulations visible in the occipital electrodes are evident here.

Figure 4.

ERP effects in the electrode O2 for the consonant (upper panel) and vowel (lower panel) manipulation. The N/P150 and the N250 modulations visible in the occipital electrodes are evident here.

For both the Vowel and the Consonant manipulation, the early negative peak (around 150 ms) showed a more negative trend for the Identity condition compared with the other 2 conditions. The second negative peak (around 200 ms) showed an opposite trend: The Control condition is more negative compared with the Identity, whereas the Relative position condition is similar to the Control for the Vowel manipulation, and more similar to the Identity for the Consonant manipulation. These occipital effects were evaluated splitting the whole time window into 2 parts. In fact, we divided the 100–250 ms time window into 2 consecutive time windows of 75 ms each. Our expectation was to replicate the Letter by Condition interaction in the late time window, but not in the early window. This is exactly what happened: In the early time window only the main effect of Condition emerged (F(2,52) = 11.531, P < 0.001); in the late time window, we found both a main effect of Condition (F(2,52) = 4.144, P < 0.05) and an interaction Letter with Condition (F(2,52) = 3.227, P < 0.05). When we included the time Window (Early, Late) as a factor in the ANOVA, a strong interaction of Window by Condition emerged (F(2,52) = 14.599, P < 0.001), thus confirming the trend evident in the grand average. To examine this effect, we ran 2-way ANOVAs with Electrode (5 levels: PO7, O1, Oz, O2, and PO8) and Condition (2 levels) as factors in each time window, comparing in a pairwise manner the conditions for the Consonant and Vowel manipulation (Table 2).

Table 2

F values for the condition factor (2 levels, see comparison column) and the electrode (5 levels: PO7, O1, Oz, O2, and PO8) by condition interaction for the ANOVA performed in the occipital electrodes in the early (100–175 ms) and late (175–250 ms) time windows corresponding to the negative deflection evident on the posterior electrodes in the grand average

Manipulation Time window (ms) Comparison Condition, df (1,26) Electrode × condition, df (4,104) 
Vowel 100–175 Identity versus relative 6.164* 5.409** 
Identity versus control 12.485** 2.499# 
Relative versus control 1.548ns 1.626ns 
175–250 Identity versus relative 4.004* 0.454ns 
Identity versus control 3.573* 1.217ns 
Relative versus control 0.100ns 0.383ns 
Consonant 100–175 Identity versus relative 14.045** 5.487* 
Identity versus control 5.568* 6.717** 
Relative versus control 1.096ns 0.527ns 
175–250 Identity versus relative 0.218ns 0.348ns 
Identity versus control 4.919* 0.522ns 
Relative versus control 9.071** 0.333ns 
Manipulation Time window (ms) Comparison Condition, df (1,26) Electrode × condition, df (4,104) 
Vowel 100–175 Identity versus relative 6.164* 5.409** 
Identity versus control 12.485** 2.499# 
Relative versus control 1.548ns 1.626ns 
175–250 Identity versus relative 4.004* 0.454ns 
Identity versus control 3.573* 1.217ns 
Relative versus control 0.100ns 0.383ns 
Consonant 100–175 Identity versus relative 14.045** 5.487* 
Identity versus control 5.568* 6.717** 
Relative versus control 1.096ns 0.527ns 
175–250 Identity versus relative 0.218ns 0.348ns 
Identity versus control 4.919* 0.522ns 
Relative versus control 9.071** 0.333ns 

Ns: not significant; #: P < 0.1; *: P < 0.05; **: P < 0.01.

150–250 ms Window (N250)

The N250 effects identified by visual inspection are supported by the distributed ANOVA considering all the 6 conditions in the 150–250 ms time window. A main effect of Condition was found (Midline: F(2,52) = 5.020, P < 0.05; Column 1: F(2,52) = 4.411, P < 0.05; Column 2: F(2,52) = 3.867, P < 0.05). Moreover, a significant interaction of Electrode and Condition emerged (Midline: F(8,208) = 5.939, P < 0.01; Column 1: F(4,104) = 6.263, P < 0.01; Column 2: F(6,156) = 8.94, P < 0.01; and Column 3: F(10,260) = 7.368, P < 0.01). Finally, more crucial for our manipulation, Letter and Condition interacted significantly in all groups (Midline: F(2,52) = 3.764, P < 0.05; Column 1: F(2,52) = 3.360, P < 0.05; Column 2: F(2,52) = 5.038, P < 0.05; and Column 3: F(2,52) = 5.249, P < 0.01).

Separate planned comparisons for vowels and consonants confirmed the visual inspection evaluation. The Vowel conditions revealed statistical differences between the Identity and the Relative position condition: An effect of Condition was significant in the central areas (Midline: F(1,26) = 7.478, P < 0.05; Column 1: F(1,26) = 7.714, P < 0.05; Column 2: F(1,26) = 7.217, P < 0.05) and the interaction Electrode with Condition emerged for the more lateral electrodes (Column 2: F(3,78) = 5,138, P < 0.05; Column 3: F(5,130) = 4.131, P < 0.05). This last interaction is probably due to the more posterior effect for the Identity condition compared with the Relative position condition (see topography of the ERP differences in Fig. 5, upper panel).

Figure 5.

Topographical maps of both the N250 (upper panel) and the N400 (lower panel) effects elicited by the identity (control–identity) and relative position (control–relative) conditions compared with the control condition for the consonants and vowels manipulations. Differential effects are calculated between 200 and 230 ms for the N250 and between 350 and 450 for the N400, that is, at the peak of the effect.

Figure 5.

Topographical maps of both the N250 (upper panel) and the N400 (lower panel) effects elicited by the identity (control–identity) and relative position (control–relative) conditions compared with the control condition for the consonants and vowels manipulations. Differential effects are calculated between 200 and 230 ms for the N250 and between 350 and 450 for the N400, that is, at the peak of the effect.

Also the Identity and the Control conditions differed, as shown by main effect of Condition (Midline: F(1,26) = 5.452, P < 0.05; Column 1: F(1,26) = 4.889, P < 0.05) and the interaction Electrode by Condition (Midline: F(4,104) = 5.484, P < 0.01; Column 1: F(2,52) = 5.834, P < 0.01; Column 2: F(3,78) = 7.596, P < 0.01; and Column 3: F(5,130) = 5.359, P < 0.05) that confirms the more anterior effect between these 2 conditions evident in the grand average. The Relative position and the Control conditions were not statistically different for the Vowel manipulation (see Fig. 5, upper panel).

The Consonant conditions showed statistical differences between the Identity and the Control conditions: A main effect of Condition emerged in the central areas (Midline: F(1,26) = 6.376, P < 0.05; Column 1: F(1,26) = 5.29, P < 0.05; Column 2: F(1,26) = 5.256, P < 0.05), but the interaction Electrode by Condition showed that the effect is mainly posterior (Midline: F(4,104) = 5.134, P < 0.05; Column 1: F(2,52) = 5.891, P < 0.05; Column 2: F(3,78) = 5.467, P < 0.01; and Column 3: F(5,130) = 7.665, P < 0.01) (maps in Fig. 5). Comparing the Relative position and the Control conditions, a main effect of Condition emerged (Midline: F(1,26) = 9.761, P < 0.01; Column 1: F(1,26) = 6.516, P < 0.01; Column 2: F(1,26) = 11.386, P < 0.01; and Column 3: F(1,26) = 11.717, P < 0.01). The Identity and the Relative position conditions did not statistically differ for the Consonant manipulation.

350–450 Window

Finally, the 350–450 ms time window was selected to evaluate the N400 modulation. The grand average concerning the Vowel manipulation shows a more pronounced negativity both for the Relative position and the Control condition. Interestingly, the visual inspection of the grand average for the Consonant manipulation shows that the N400 for the Relative position condition is markedly reduced, being similar to the Identity condition, particularly at the anterior electrodes (see Fig. 3, upper panel, and Fig. 5, lower panel). The Control condition remains more negative than the Identity.

The overall ANOVAs showed a main effect of Condition (Midline: F(2,52) = 5.205, P < 0.05; Column 1: F(2,52) = 7.856, P < 0.01; Column 2: F(2,52) = 7.101, P < 0.01; and Column 3: F(2,52) = 3.892, P < 0.05), an interaction Electrode by Condition (Midline: F(8,208) = 2.972, P < 0.05; Column 2: F(6,156) = 3.495, P < 0.05; and Column 3: F(10,260) = 3.466, P < 0.05). Moreover, the interaction Letter by Condition (Midline: F(2,52) = 4.374, P < 0.05; Column 1: F(2,52) = 4.551, P < 0.05; Column 2: F(2,52) = 4.455, P < 0.05; and Column 3: F(2,52) = 4.201, P < 0.05) confirmed the trend evident in the previous time windows.

Planned comparisons for the Vowel manipulation showed a main effect of Condition for the comparison between the Identity and the Relative position conditions (Midline: F(1,26) = 6.580, P < 0.05; Column 1: F(1,26) = 9.482, P < 0.01; Column 2: F(1,26) = 10.296, P < 0.01; and Column 3: F(1,26) = 5.922, P < 0.05). The comparison between the Identity and Control conditions showed the following interactions: Electrode by Condition (Midline: F(4,104) = 3.84, P < 0.05; Column 2: F(3,78) = 5.79, P < 0.05; and Column 3: F(5,130) = 5.36, P < 0.05), Hemisphere by Condition (Column 2: F(1,26) = 4.841, P < 0.05) and Electrode by Hemisphere by Condition (Column 3: F(5,130) = 5.097, P < 0.01) in the more lateral electrodes. These effects are due to the slightly left-anterior distribution of the N400 (see Identity effect for vowels in Fig. 5, lower panel). No reliable effects emerged from the comparison between the Relative position and Control conditions.

The Consonant manipulation in the N400 time window showed a main effect of Condition between the Identity and the Control for all groups (Midline: F(1,26) = 9.273, P < 0.01; Column 1: F(1,26) = 12.266, P < 0.01; Column 2: F(1,26) = 11.090, P < 0.01; and Column 3: F(1,26) = 6.986, P < 0.05). Similarly, the analyses comparing the Relative position and Control conditions also showed a main effect of Condition for the more central electrodes (Midline: F(1,26) = 5.311, P < 0.05; Column 1: F(1,26) = 7.263, P < 0.05; Column 2: F(1,26) = 4.589, P < 0.05) (see topography of the consonant effects in Fig. 5, lower panel). The analyses comparing the Identity and the Relative position conditions did not show any statistically reliable effects.

Discussion

The present experiment has demonstrated clear evidence of differential processing of consonants and vowels during visual word recognition by showing that the consonant–vowel status of shared letters critically modulates the relative position priming. Overall, we found a very consistent pattern in the data. In the first window (100–175), we found that the identity condition differed from the other 2 conditions—relative position and unrelated priming conditions—both for consonants and vowels. This result is consistent with previous findings showing that the N/P150 is sensitive to processing at the level of visual features, probably reflecting the mapping of visual features onto letter-level representations (e.g., Holcomb and Grainger 2006; Petit et al. 2006; Chauncey et al. 2008; Carreiras et al. 2009). In addition, we found that the N250 (175–250) was sensitive to the consonant–vowel distinction. In this window, the amplitude of the relative position priming conditions was similar to the identity condition for consonants and to the unrelated condition for vowels. The 2 identity conditions also differed from the 2 unrelated conditions. (The identity and the control conditions did not contain the same number of letters in the present experiment, because identity was designed as a control condition. This fact, together with important differences in the experimental procedures [e.g., the use/absence of a backward mask, different size of prime and target stimuli] may account for the slightly different results obtained in Holcomb and Grainger [2006] and in the present experiment when contrasting identity and control conditions. The effects are more limited in spatial extent and arise slightly earlier in the present experiment.) It seems that from a very early moment of word processing, the brain establishes differences between consonants and vowels. Previous results have related the N250 component to orthographic processing (Grainger, Kiyonaga, and Holcomb 2006; Carreiras et al. 2009; Carreiras, Perea, et al. 2009). Thus, these results also provide converging evidence that the N250 reflects processing at the level of form representations. Strikingly, the relative position priming condition for consonants parallels the effects of the identity condition, which seems to suggest that consonants are the main driving force in the process of visual word recognition. Finally, similar effects were again obtained in the N400 component. For consonants, the control condition is more negative than the relative position and the identity conditions; for vowels, the control and the relative position conditions are more negative than the identity condition. Later effects on the N400 are likely to reflect lexical competition processes and/or the mapping of whole-word representations onto semantics (see Barber et al. 2004; Carreiras et al. 2005; Holcomb and Grainger 2006; Muller, Duñabeitia, and Carreiras, unpublished data).

Although the N/P150 effect was restricted to the occipital areas, the N250 effect was a more widely spread distributed ERP negativity. Differences in the topographical distribution add more evidence to the idea that these 2 components are tapping into 2 different processes of visual word recognition. Although effects in the N/P150 seem to be related to early perceptual processes, effects in the N250 seem to be related to orthographic processes. In fact, Grainger, Kiyonaga, and Holcomb (2006) found that transposed-letter primes modulated the ERP signal in a window between 150 and 250 ms after stimuli presentation. More specifically, they presented targets preceded by transposed-letter primes (e.g., barin–BRAIN) and their controls (e.g., bosin–BRAIN). Replaced-letter controls produced larger amplitude in the 150–250 ms window than transposed-letter primes (see also Carreiras et al. 2009; Duñabeitia et al. 2009). Furthermore, 1 recent study has found evidence for a dissociation in the processing of vowels and consonants in the N250 (Carreiras et al. 2009). In the Carreiras et al. study, an anterior N250 effect was found when manipulating consonants (in a delayed letter paradigm), whereas a posterior effect was found when manipulating vowels. Thus, the differences in the 175–250 ms window may reflect a genuine differential computation of consonants and vowels as different objects with different linguistic properties. In fact, some theoretical proposals acknowledge that although the role of consonants is specifically linked to lexical selection, the role of vowels is more related to rhythmic and syntactic patterns (Nespor et al. 2003; see also, Bonati et al. 2005, 2007). In addition, there is a computational model, the CDP+ model (Perry et al. 2007), which makes an early distinction between consonants and vowels in the input representation to align graphemes to a graphosyllabic template with onset, vowel and coda constituents. Thus, the N250 could be tapping this early processing of consonants and vowels as 2 different entities, which will have differential influence on later computations during the stream of word processing.

On the other hand, the pattern of data obtained in the N400 window may be a consequence of the differential lexical activation produced by consonants and vowels. Vowel and consonant relative position primes create different lexical activation patterns according to the view that consonant primes are much more constraining than vowel primes. That is, there are far fewer words consistent with the consonant relative position skeleton than with the vowel relative position skeleton, as a consequence of the fact that most languages have a higher number of consonants than vowels. Thus, the observed effect in the N400 component may well be reflecting the lexical constraint provided by the relative position primes. More concretely, consonant relative position primes activate few lexical units, because that consonantal pattern does not match with too many word forms. Importantly, the target word will be one of these candidates, and therefore, it is expected that the consonant relative position priming condition activates the target word to a similar extent to the identity priming condition, due to the reduced number of competing lexical representations. For instance, consider the example farol used in our experiment; in the Spanish corpus, only 4 different words coincide with the consonantal pattern -f-r-l- (the target farol, and the competing lexical representations “farola,” “foral,” and férula). In contrast, vowel relative position primes activate many lexical units, and once those lexical units start competing with one another, there is little priming effect to be had on the actual target recognition. For instance, taking the word acero as an example, the pattern -a-e-o- is found to be present in more than 150 Spanish words (the target acero, but also “asesor,” “cajero,” “alero,” “casero,” “granero,” etc.), thus yielding higher values of global lexical activation and competition than the consonantal patterns. Therefore, the results of the low constraint and/or high competition of the relative position priming for vowels do not favor the recognition of the target, as happens with the vowel unrelated condition.

Thus, from a very early moment in the process of visual word recognition, the brain uses a system that allows for coding the relative position of letters, which is mainly based on consonants, because this code is good enough to trigger the whole-word representation. It seems to be as good for consonants as the complete string of letters. However, this is not the case for vowels. The information contained in vowels is not as valuable as that contained in consonants in accessing words, considering the high lexical activation dispersion that vowel-based skeletons lead to. These results challenge 1 important assumption of most models of visual word recognition in alphabetic orthographies about how the human processing system encodes letter positions when creating internal orthographic representations, that is, that all letters are encoded in a similar fashion at early stages of orthographic processing, independently of their consonant–vowel status. The basic conclusion from this study is that not only are the identity and position of a character in a string of letters different perceptual dimensions, but consonants and vowels are also 2 different objects with differentiated roles in the processes of visual word recognition. Present models of orthographic encoding are unable to explain these findings in their current form. None of the “classical” computational models of visual word recognition or the “recent” ones that incorporate a difference between letter position and letter identity (e.g., McClelland and Rumelhart 1981; Grainger and Jacobs 1996; Davis 1999; Coltheart et al. 2001; Whitney 2001; Grainger and van Heuven 2003; Dehaene et al. 2005; Gomez et al. 2008) can account for differential processing of consonants and vowels. One noticeable exception, however, is the recent CDP+ model (Perry et al. 2007), which makes an early distinction between consonants and vowel in the input representation to align graphemes to a graphosyllabic template with onset, vowel and coda constituents. Hence, in the CDP+ model, the consonant–vowel differentiation is a basic assumption that underlies the ortho-phonological encoding of the input units. However, it has to be mentioned that this model in its current version has not been tested with a large pool of masked priming lexical decision data, because its major focus is on reading aloud. Future research should be aimed at exploring the validity of this model in terms of the most influential orthographic effects (e.g., form priming, transposed-letter priming, or relative position priming, among others).

It could be argued, however, that the differences reported between vowels and consonants in relative position priming effects do not pose a problem for the computational models, because the reason for the difference could simply be letter frequency: Vowels are more frequent than consonants. Much evidence supports the idea that perceptual systems become selectively efficient at processing input that is encountered frequently. Thus, if the consonant–vowels processing difference could be accounted for by just letter frequency, any of the current models could easily account for the present results. However, Duñabeitia and Carreiras (unpublished data) tested whether the consonant–vowel differentiation could be a consequence of the frequencies of the letters and ruled out a purely frequency account. When they presented primes that contained only vowels, only high frequency consonants, or only low-frequency consonants, they found a significant relative position priming effect of the same size for the 2 consonant conditions but no effects for vowel primes. Thus, it is quite unlikely that the differences observed in the present experiment, which mirrors the pattern obtained by Duñabeitia and Carreiras (unpublished data), could be reduced to a letter frequency effect. However, it remains to be seen whether other frequency accounts such as the co-occurrence of letters can account for the observed differences between consonants and vowels. In any case, we should also acknowledge that a complete reduction of processing differences between consonants and vowels to computation of statistical regularities has been advocated as quite unlikely in speech segmentation (Bonatti et al. 2007). Although consonants have a prominent role in word identification, vowels have a prominent role in rule extraction. Using artificial speech streams in which words and rules were implemented over either consonants or vowels, Toro et al. (2008) showed that participants identified words by computing statistical dependencies among consonants, but failed to do so when vowels carried the same dependencies. In contrast, participants extracted simple structural generalizations when vowels instantiate them, but failed to do so when they were instantiated by consonants.

Finally, 1 important finding that deserves attention is the null difference found between the consonant relative position priming condition and the identity priming condition (namely, the similar ERP effects in the N250 and N400 components for frl–farol and farol–farol). Consonant relative position primes elicited an N250 and N400 of similar amplitude to the identity condition, whereas vowel relative position primes produced N250 and N400 effects of similar amplitude to the unrelated condition. In spite of the undeniable importance of the consonant–vowel distinction in the relative position priming effect for current models of word recognition and orthographic encoding, the negligible difference between the consonant relative position and identity conditions provides interesting insights about the way in which the visual word recognition system extracts the key information of words (i.e., letter identities). The most plausible interpretation of these findings derives from the lexical constraint imposed by consonants. Considering that consonants are much more constraining than vowels for word recognition (due to the higher number of consonants as compared with the typically limited number of vowels), and according to the view that consonants have a more marked role in orthographic processing that vowels, it could be assumed that when a string is briefly presented (for around 50 ms), the visual word recognition system only extracts basic features of this string to be processed, mainly focusing on consonants. If this is so, it is feasible to assume that for both identical (farol) and relative position consonant primes (frl), the orthographic processor might be fired, extracting the same amount of information (f-r-l and frl, respectively). In the case of identity primes in the vowel condition (acero), the same mechanism would only lead to the identification of the consonants from the identity condition (-c-r-). However, in the case of the vowel relative position priming, there is no consonantal information to be extracted from the input. Therefore, a much larger number of possible candidates compatible with the 3 vowels (aeo) input will be activated. Accordingly, a reasonable number of possible lexical candidates that fit with these extracted patterns would then be activated only in the case of the 2 identity conditions and the consonant relative position condition and the appearance of the target words would fire a matching procedure, based on classical candidate suppression mechanisms (e.g., lateral inhibition processes), that would finally lead to the correct lexical candidate selection. Such a mechanism of orthographic processing based on consonant detection would account both for the null difference between absolute and relative position priming conditions (f-r-l and frl from farol) that has previously been reported (Grainger, Granier, et al. 2006) and for the effects observed in the present study. This mechanism predicts no differences between the consonant relative position and identity priming conditions, whereas clear differences are to be expected between vowel relative position and identity priming conditions, both at early orthographic encoding stages (reflected in the N250 component) and in later lexico-semantic processing stages (as shown in the N400 component).

Conclusion

The empirical data on 2 phenomena—transposed-letter and relative position priming—have given rise to the general consensus that there is a level of orthographic processing where some form of approximate, flexible coding of letter positions operates. How is letter position information coded during the earliest phases of visual word recognition? It is assumed that individual letters are the key elements for orthographic processing and that the mechanism used to code for the positions of these letters is what critically determines the nature of the orthographic code. Here we show that letter position assignment is modulated by whether the letter is a consonant or a vowel. The results reported here for ERPs, which are very sensitive to the time course of early processing in the visual word recognition system, suggest that consonants seem to trigger the coding but vowels do not, or at least not to the same extent. Consonants and vowels seem to be distinctive elements in the initial stages of visual word recognition, and they seem to trigger different patterns of activation in the lexicon, as shown by their differential effects on a series of ERP components, which reflect a cascade of processes triggered on presentation of a printed word and which are sensitive to form representations (N250) and lexico-semantic representations (N400). These results challenge 1 important assumption of most models of visual word recognition in alphabetic orthographies about how the human processing system encodes letter positions when creating internal orthographic representations. These consonant–vowel differences should be taken into consideration when developing computational models of visual word recognition.

Funding

Spanish Ministry of Education and Science (SEJ2006-09238) and CONSOLIDER-INGENIO 2010 (CSD2008-00048).

Conflict of Interest: None declared.

References

Allen
PA
Wallace
B
Weber
TA
Influence of case type, word frequency, and exposure duration on visual word recognition
J Exp Psychol Hum Percept Perform
 , 
1995
, vol. 
21
 (pg. 
914
-
934
)
Barber
H
Kutas
M
Interplay between computational models and cognitive electrophysiology in visual word recognition
Brain Res Rev
 , 
2007
, vol. 
53
 (pg. 
98
-
123
)
Barber
H
Vergara
M
Carreiras
M
Syllable-frequency effects in visual word recognition: evidence from ERPs
Neuroreport
 , 
2004
, vol. 
15
 (pg. 
545
-
548
)
Berent
I
Perfetti
CA
A rose is a REEZ: the two cycles model of phonology assembly in reading English
Psychol Rev
 , 
1995
, vol. 
102
 (pg. 
146
-
184
)
Bonatti
LL
Peña
M
Nespor
M
Mehler
J
Linguistic constraints on statistical computations: the role of consonants and vowels in continuous speech processing
Psychol Sci
 , 
2005
, vol. 
16
 (pg. 
451
-
459
)
Bonatti
LL
Peña
M
Nespor
M
Mehler
J
On consonants, vowels, chickens, and eggs
Psychol Sci
 , 
2007
, vol. 
18
 (pg. 
924
-
925
)
Caramazza
A
Chialant
D
Capasso
D
Miceli
G
Separable processing of consonants and vowels
Nature
 , 
2000
, vol. 
403
 (pg. 
428
-
430
)
Carreiras
M
Gillon-Dowens
M
Vergara
M
Perea
M
Are vowels and consonants processed differently? ERP evidence with a delayed letter paradigm
J Cogn Neurosci
 , 
2009
, vol. 
21
 (pg. 
275
-
288
)
Carreiras
M
Perea
M
Vergara
M
Pollatsek
A
The time course of orthography and phonology: ERP correlates of masked priming effects in Spanish
 , 
2009
 
Psychophysiology. Forthcoming
Carreiras
M
Price
CJ
Brain activation for consonants and vowels
Cereb Cortex
 , 
2008
, vol. 
18
 (pg. 
1727
-
1735
)
Carreiras
M
Vergara
M
Barber
H
Early ERP effects of syllabic processing during visual word recognition
J Cogn Neurosci
 , 
2005
, vol. 
17
 (pg. 
1803
-
1817
)
Carreiras
M
Vergara
M
Perea
M
ERP correlates of transposed-letter similarity effects: are consonants processed differently from vowels?
Neurosci Lett
 , 
2007
, vol. 
419
 (pg. 
219
-
224
)
Carreiras
M
Vergara
M
Perea
M
ERP correlates of transposed-letter priming effects: the role of vowels vs. consonants
Psychophysiology
 , 
2009
, vol. 
46
 (pg. 
34
-
42
doi: 10.1111/j.1469-8986.2008.00725.x
Cattell
JM
The time taken by cerebral operations
Mind
 , 
1886
, vol. 
11
 (pg. 
220
-
242
)
Chauncey
K
Holcomb
PJ
Grainger
J
Effects of stimulus font and size on masked repetition priming: an ERP investigation
Lang Cogn Process
 , 
2008
, vol. 
23
 (pg. 
183
-
200
)
Coltheart
M
Rastle
K
Perry
C
Ziegler
J
Langdon
R
DRC: a dual-route cascaded model of visual word recognition and reading aloud
Psychol Rev
 , 
2001
, vol. 
108
 (pg. 
204
-
256
)
Cutler
A
Sebastián-Gallés
N
Soler-Vilageliu
O
van Ooijen
B
Constraints of vowels and consonants on lexical selection: cross-linguistic comparisons
Mem Cognit
 , 
2000
, vol. 
28
 (pg. 
746
-
755
)
Davis
CJ
The self-organising lexical acquisition and recognition (SOLAR) model of visual word recognition
 , 
1999
University of New South Wales
 
Unpublished doctoral dissertation
Davis
CJ
Perea
M
BuscaPalabras: a program for deriving orthographic and phonological neighborhood statistics and other psycholinguistic indices in Spanish
Behav Res Methods
 , 
2005
, vol. 
37
 (pg. 
665
-
671
)
Dehaene
S
Cohen
L
Sigman
M
Vinckier
F
The neural code for written words: a proposal
Trends Cogn Sci
 , 
2005
, vol. 
9
 (pg. 
335
-
341
)
Duñabeitia
JA
Molinaro
N
Laka
I
Estévez
A
Carreiras
M
N250 effects for letter transpositions depend on lexicality: casual or causal?
NeuroReport
 , 
2009
 
DOI: 10.1097/WNR.0b013e3283249b1c
Duñabeitia
JA
Perea
M
Carreiras
M
Do transposed-letter similarity effects occur at a morpheme level? Evidence for morpho-orthographic decomposition
Cognition
 , 
2007
, vol. 
105
 
3
(pg. 
691
-
703
)
Forster
KI
Wales
RJ
Walker
EW
Accessing the mental lexicon
New approaches to language mechanisms
 , 
1976
Amsterdam
North Holland
(pg. 
257
-
287
)
Gomez
P
Ratcliff
R
Perea
M
The overlap model: a model of letter position coding
Psychol Rev
 , 
2008
, vol. 
115
 (pg. 
577
-
601
)
Grainger
J
Cracking the orthographic code: an introduction
Lang Cogn Process
 , 
2008
, vol. 
23
 
1
(pg. 
1
-
35
)
Grainger
J
Granier
JP
Farioli
F
Van Assche
E
van Heuven
W
Letter position information and printed word perception: the relative-position priming constraint
J Exp Psychol Hum Percept Perform
 , 
2006
, vol. 
32
 (pg. 
865
-
884
)
Grainger
J
Jacobs
AM
Orthographic processing in visual word recognition: a multiple read-out model
Psychol Rev
 , 
1996
, vol. 
103
 (pg. 
518
-
565
)
Grainger
J
Kiyonaga
K
Holcomb
PJ
The time-course of orthographic and phonological code activation
Psychol Sci
 , 
2006
, vol. 
17
 (pg. 
1021
-
1026
)
Grainger
J
van Heuven
WJB
Bonin
P
Modeling setter position coding in printed word perception
The mental lexicon
 , 
2003
New York
Nova Science
(pg. 
1
-
23
)
Greenhouse
S
Geisser
S
On methods in the analysis of profile data
Psychometrica
 , 
1959
, vol. 
24
 (pg. 
94
-
112
)
Healy
AF
Cunningham
TF
A developmental evaluation of the role of word shape in word recognition
Mem Cognit
 , 
1992
, vol. 
20
 (pg. 
141
-
150
)
Healy
AF
Oliver
WL
McNamara
TP
Detecting letters in continuous text: effects of display size
J Exp Psychol Hum Percept Perform
 , 
1987
, vol. 
13
 (pg. 
279
-
290
)
Holcomb
PJ
Grainger
J
On the time course of visual word recognition: an event-related potential investigation using masked repetition priming
J Cogn Neurosci
 , 
2006
, vol. 
18
 (pg. 
1631
-
1643
)
Holcomb
PJ
Grainger
J
O'Rourke
T
An electrophysiological study of the effects of orthographic neighborhood size on printed word perception
J Cogn Neurosci
 , 
2002
, vol. 
14
 (pg. 
938
-
950
)
Humphreys
GW
Evett
LJ
Quinlan
PT
Orthographic processing in visual word identification
Cogn Psychol
 , 
1990
, vol. 
22
 (pg. 
517
-
560
)
Kutas
M
Federmeier
K
Electrophysiology reveals semantic memory use in language comprehension
Trends Cogn Sci
 , 
2000
, vol. 
4
 (pg. 
463
-
470
)
Lee
H-W
Rayner
K
Pollatsek
A
The relative contribution of consonants and vowels to word identification during reading
Journal of Memory & Language
 , 
2001
, vol. 
44
 (pg. 
189
-
205
)
Lee
H-W
Rayner
K
Pollatsek
A
The processing of consonants and vowels in reading: evidence from the fast priming paradigm
Psych Bull Rev
 , 
2002
, vol. 
9
 (pg. 
766
-
772
)
McClelland
JL
Rumelhart
DE
An interactive activation model of context effects in letter perception: part 1. An account of basic findings
Psychol Rev
 , 
1981
, vol. 
88
 (pg. 
375
-
407
)
Nazzi
T
Use of phonetic specificity during the acquisition of new words: differences between consonants and vowels
Cognition
 , 
2005
, vol. 
98
 (pg. 
13
-
30
)
Nespor
M
Peña
M
Mehler
J
On the different roles of vowels and consonants in speech processing and language acquisition
Lingue e Linguaggio
 , 
2003
, vol. 
2
 (pg. 
221
-
247
)
Neville
HJ
Mills
DL
Lawson
DS
Fractionating language: different neural subsystems with different sensitive periods
Cereb Cortex
 , 
1992
, vol. 
2
 (pg. 
244
-
258
)
Oldfield
RC
The assessment and analysis of handedness: the Edinburgh inventory
Neuropsychologia
 , 
1971
, vol. 
9
 (pg. 
97
-
113
)
Paap
KR
Newsome
SL
McDonald
JE
Schvaneveldt
RW
An activation–verification model for letter and word recognition: the word superiority effect
Psychol Rev
 , 
1982
, vol. 
89
 (pg. 
573
-
594
)
Pelli
DG
Farell
B
Moore
DC
The remarkable inefficiency of word recognition
Nature
 , 
2003
, vol. 
423
 (pg. 
752
-
756
)
Peña
M
Bonatti
LL
Nespor
M
Mehler
J
Signal-driven computations in speech processing
Science
 , 
2002
, vol. 
298
 (pg. 
604
-
607
)
Perea
M
Carreiras
M
Do orthotactics and phonology constrain the transposed-letter effect?
Lang Cogn Process
 , 
2008
, vol. 
23
 (pg. 
69
-
92
)
Perea
M
Carreiras
M
Do transposed-letter similarity effects occur at a syllable level?
Exp Psychol
 , 
2006
, vol. 
53
 (pg. 
308
-
315
)
Perea
M
Carreiras
M
Do transposed-letter effects occur across lexeme boundaries?
Psych Bull Rev
 , 
2006
, vol. 
13
 (pg. 
418
-
422
)
Perea
M
Carreiras
M
Do transposed-letter similarity effects occur at a prelexical phonological level?
Quart J Exp Psychol
 , 
2006
, vol. 
59
 (pg. 
1600
-
1613
)
Perea
M
Duñabeitia
JA
Carreiras
M
Transposed-letter priming effects for close versus distant transpositions
Exp Psychol
 , 
2008
, vol. 
55
 (pg. 
397
-
406
)
Perea
M
Lupker
SJ
Can CANISO activate CASINO? Transposed-letter similarity effects with nonadjacent letter positions
J Mem Lang
 , 
2004
, vol. 
51
 (pg. 
231
-
246
)
Perea
M
Lupker
SJ
Does jugde activate COURT? Transposed-letter similarity effects in masked associative priming
Mem Cognit
 , 
2003
, vol. 
31
 (pg. 
829
-
841
)
Peressotti
F
Grainger
J
Letter position coding in random consonant arrays
Percept Psychophys
 , 
1999
, vol. 
57
 (pg. 
875
-
890
)
Perry
C
Ziegler
JC
Zorzi
M
Nested incremental modeling in the development of computational theories: the CDP+ model of reading aloud
Psychol Rev
 , 
2007
, vol. 
114
 (pg. 
273
-
315
)
Petit
JP
Midgley
KJ
Holcomb
PJ
Grainger
J
On the time course of letter perception: a masked priming ERP investigation
Psychon Bull Rev
 , 
2006
, vol. 
13
 (pg. 
674
-
681
)
Sebastián-Gallés
N
Martí
MA
Carreiras
M
Cuetos
F
Lexesp: una base de datos informatizada del español
 , 
2000
Spain
Universitat de Barcelona
Toro
JM
Nepor
M
Mehler
J
Bonatti
LL
Finding words and rules in a speech stream
Psychol Sci
 , 
2008
, vol. 
19
 (pg. 
137
-
144
)
van Casteren
M
Davis
MH
Match: a program to assist in matching the conditions of factorial experiments
Behav Res Methods
 , 
2007
, vol. 
39
 
4
(pg. 
973
-
978
)
van Ooijen
B
Vowel mutability and lexical selection in English: evidence from a word reconstruction task
Mem Cognit
 , 
1996
, vol. 
24
 (pg. 
573
-
583
)
Vinckier
F
Dehaene
S
Jobert
A
Dubus
JP
Sigman
M
Cohen
L
Hierarchical coding of letter strings in the ventral stream: dissecting the inner organization of the visual word-form system
Neuron
 , 
2007
, vol. 
55
 (pg. 
143
-
156
)
Whitney
C
How the brain encodes the order of letters in a printed word: the SERIOL model and selective literature review
Psychon Bull Rev
 , 
2001
, vol. 
8
 (pg. 
221
-
243
)