A robust index of lexical representation in the left occipito-temporal cortex as evidenced by EEG responses to fast periodic visual stimulation
Introduction
In alphabetic script, words are constituted of letters combined with specific orthographic and phonological rules and conveying a meaning. Though complex, reading proceeds rapidly and effortlessly in literate adults (85–90% of adults across the world). A skilled reader reads about five words per second in a text, so that word identification takes only about 200 ms (Rayner, 1998, Rayner et al., 2012). However, despite decades of research on this uniquely human ability, whether there are brain processes dedicated to automatic (i.e. without any explicit judgment task) discrimination of a valid visual word from a meaningless orthographically legal string of letters (i.e., a pseudoword) remains an outstanding issue. To address this issue, the present study introduces an original paradigm of EEG recording during fast periodic stimulation of visual words.
Behaviorally, a widely used task for assessing word recognition is the lexical decision task (LDT, Rubenstein et al., 1970), in which participants decide as rapidly as possible if a visually presented letter string is a word as opposed to a nonword (i.e., orthographically illegal) or a pseudoword (i.e., orthographically legal). This task has allowed to identify many variables that affect lexical decision performance, whether intrinsic to stimuli (e.g., frequency, length, orthographic regularity, semantic concreteness, number of semantic attributes, etc.), or extrinsic to stimuli (e.g., effects of the surrounding list as word-likeness at phonological or orthographic levels), constraining functional models of reading to account for these influences along the visual word recognition processes (see Balota et al. (2006) for a review; Grainger, 2008). The LDT has also helped distinguishing between subtypes of surface dyslexia since it provides access to the functional locus of a patient’s deficit without requiring a phonological output (Friedmann and Lukov, 2008). However, since the LDT requires explicit attention to, and decision about, strings of letters, it measures the outcome of several perceptual, cognitive and decisional processes. Consequently, behavioral studies using LDTs have been unable to determine whether selective representations of visual words can be activated without explicit judgments of the words, an outstanding issue in the reading literature (Shtyrov et al., 2013).
At first glance, since electro- or magneto-encephalography (EEG, MEG) can reveal differential brain responses without requiring explicit processing of the stimuli, this technique seems particularly well suited to address this issue. A visual word presented centrally triggers a negative event-related potential (ERP) peaking over left occipito-temporal sites around 170 milliseconds (ms), the word-related N170 (e.g., Bentin et al., 1999; Rossion et al., 2003; Maurer et al., 2005; Yoncheva et al., 2010) or the M170 in MEG (e.g. Tarkiainen et al., 1999). This component is larger in response to alphabetic items than to nonalphabetic strings such as pseudofonts, symbols and shapes (Bentin et al., 1999, Tarkiainen et al., 2002, Brem et al., 2009) and has been associated with a source in the left ventral occipito-temporal region (Maurer et al., 2005, Cohen et al., 2000). However, most studies report no N170 difference between words and pseudowords, or relatively late differential responses (after 200 ms) over more anterior regions (Bentin et al., 1999; Cohen et al., 2002; Wydell et al., 2003; Pammer et al., 2004; Tarkiainen et al., 1999; see also Nobre et al. (1994) for a lack of words/nonwords difference on N200 visual potentials recorded on the ventral occipito-temporal cortical surface) and a few studies have reported inconsistent effects, i.e. a small increase (Hauk and Pulvermüller, 2004, Maurer et al., 2005) or a decrease (McCandliss et al., 1997) of the N170 to words relative to pseudowords.
The lack of consistent (i.e., reliable) indexes of selective visual word representation in ERP studies is in line with the lack of systematic discriminative response in the left ventral occipito-temporal cortex in neuroimaging (Price, 2012, Seghier et al., 2012), suggesting that selective visual representation of words (i.e. different from pseudowords) do not exist. For instance, in the interactive account framework (Price and Devlin, 2011), responses to visual words merely arise as an interaction between bottom-up visual input and higher level phonological/semantic processing, and not from any selective tuning to orthographic representations. It may also be that selective responses to words are heavily task-dependent (Bentin et al., 1999, Ziegler and Goswami, 2005, Ziegler et al., 1997), so that lexicality-related N170 modulations (e.g., Hauk et al., 2012) require explicit processing of the visual stimuli. Yet, at this stage, one cannot exclude that the lack of consistent indexes of automatic selective visual word representation arises due to the difficulty of isolating selective differences between visual stimuli with a poor signal-to-noise ratio (SNR) approach such as ERPs, or an approach that accumulates neural activity over seconds such as fMRI.
Here we introduce a fast periodic visual stimulation paradigm in EEG for visual word stimulation, to potentially provide a reliable index of visual word vs. pseudoword discrimination without an explicit word-related task. Such an index would not only support the proposal of automatic activation of whole word representations (Glezer et al., 2009, Shtyrov et al., 2013), but would be valuable in assessing potential defective processes of reading in patient populations (e.g., dyslexia) and during both typical and abnormal development. Our approach is based on scalp EEG recordings during periodic visual stimulation, which results in periodic responses defined as “steady-state visual evoked potentials” (SSVEPs, Regan, 1966, Regan, 1989). Although this approach has long been confined to the study of low-level visual processes and attention, it has recently been used to measure visual discrimination responses of complex visual stimuli such as faces (Rossion and Boremanse, 2011). Most recently, Liu-Shuang et al. (2014) measured the discrimination of individual faces by presenting a sequence of identical face stimuli at a fast periodic rate (base frequency=F, 6 Hz) interspersed with different identity face stimuli (“oddball”) at a slower periodic rate (i.e., 1 novel face every 5 identical faces). In this study, a robust individual face discrimination response was recorded over the right occipito-temporal cortex, specifically at the oddball frequency rate (F/5=1.2 Hz) and its harmonics (2F=2.4 Hz, etc.). This approach, with two embedded periodic frequency rates and an analysis of the responses of interest in the frequency-domain (Braddick et al., 1986, Heinrich et al., 2009, Liu-Shuang et al., 2014), has several important advantages for the present question of interest (for review see Rossion, 2014). That is, within a few minutes of stimulation, it provides (1) high SNR visual discrimination responses that are (2) selective to the contrast between the frequent and rare stimuli, (3) objectively identifiable (i.e., at an experimentally-defined frequency rate) and directly quantifiable in the EEG spectrum and (4) obtained without any behavioral task requiring the processing of the parameter of interest, i.e. implicitly.
Considering that these advantages may prove crucial for identifying an automatic representation of visual word form, we applied this approach to the discrimination of written words from control stimuli. Specifically, human observers were presented with visual stimuli at a fast rate of 10 Hz (stimulus onset asynchrony of 100 ms), in sequences structured as follows: xxxxWxxxxWxxxxW…, where “W” refers to words, and “x”, depending on the experimental condition, to one of the following stimulus types: pseudofonts (PF), nonwords (NW, sequences of orthographically illegal letter strings), or pseudowords (PW, sequences of orthographically legal letter strings). We hypothesized that if words are discriminated from pseudofonts, nonwords, or even pseudowords, their periodic occurrence should lead to a periodic EEG response at the oddball frequency, i.e., at 2 Hz and its harmonics. Further, we hypothesized that if behavioral list context effects as reported in the literature (e.g., the more or less word-like sequence; Lupker and Pexman, 2010; Stone and Van Orden, 1993; Pugh et al., 1994) are due to visual discrimination processes rather than decisional levels, then the sequence type should determine the strength of the discrimination response. More precisely, the oddball response to words should be larger in nonword sequences (which are less word-like) than in pseudoword sequences (which are more word-like).
Section snippets
Participants
Ten right-handed healthy participants (2 males, mean age=25.3; range 20–42), all native French speakers, with normal/corrected-to-normal vision, were tested after giving written informed consent for a study that was approved by the Biomedical Ethical Committee of the University of Louvain. They received financial compensation for their participation. They were unaware of the goals of the experiment and that a change of stimulus type occurred at a periodic rate.
Stimuli
Words, pseudowords, nonwords, and
Participants
Ten right-handed healthy participants (2 males, mean age=22.7; range 20−24), all native French speakers, with normal/corrected-to-normal vision, were tested after giving written informed consent for a study that was approved by the Biomedical Ethical Committee of the University of Louvain. They received financial compensation for their participation. They were not informed about the goal of the experiment and that a change of stimulus type occurred at a periodic rate.
Stimuli
Words (W), pseudowords with
An index of automatic lexical access
The present study innovated in the domain of visual word recognition by introducing a fast periodic stimulation EEG approach that is rarely used in cognitive electrophysiology (Luck, 2014). This approach revealed automaticity and specificity of word discrimination: visual words elicited clear differential responses from letter-like control stimuli (experiment 1) and from alphabetic control stimuli (experiments 1 and 2), within a few minutes of stimulation. Remarkably, this response was recorded
Conflict of interest
None.
Acknowledgements
This work was supported by a Grant from the European Research Council (facessvep 284025) to BR. AL was supported by the ERC Grant and a PAI/UIAP Grant PAI/33. BR and GVB are supported by the Belgian National Fund for Scientific Research. The authors would like to thank Joan Liu for her advices in data analysis and Talia Retter for her careful reading of a previous version of this manuscript.
References (92)
- et al.
Visual word recognition: the journey from features to meaning (a travel update)
- et al.
Tuning of the human left fusiform gyrus to sublexical orthographic structure
NeuroImage.
(2006) - et al.
Sensitivity to orthographic familiarity in the occipito-temporal region
NeuroImage
(2008) - et al.
The what, when, where, and how of visual word recognition
Trends Cogn. Sci.
(2014) - et al.
The unique role of the visual word form area in reading
Trends Cogn. Sci.
(2011) - et al.
The neural code for written words: a proposal
Trends Cogn. Sci.
(2005) - et al.
Effects of sustained, voluntary attention on amplitude and latency of steady-state visual evoked potential: a costs and benefits analysis
Clin. Neuropsysiol.
(2002) - et al.
Steady-state responses in MEG demonstrate information integration within but not across the auditory and visual senses
Neuroimage
(2012) - et al.
Evidence for highly selective neuronal tuning to whole words in the “visual word form area”
Neuron
(2009) - et al.
The time-course of single-word reading: evidence from fast behavioral and brain responses
NeuroImage
(2012)