Elsevier

Brain and Language

Volume 127, Issue 3, December 2013, Pages 497-509
Brain and Language

Word-specific repetition effects revealed by MEG and the implications for lexical access

https://doi.org/10.1016/j.bandl.2013.09.013Get rights and content

Highlights

Abstract

This magnetoencephalography (MEG) study investigated the early stages of lexical access in reading, with the goal of establishing when initial contact with lexical information takes place. We identified two candidate evoked responses that could reflect this processing stage: the occipitotemporal N170/M170 and the frontocentral P2. Using a repetition priming paradigm in which long and variable lags were used to reduce the predictability of each repetition, we found that (i) repetition of words, but not pseudowords, evoked a differential bilateral frontal response in the 150–250 ms window, (ii) a differential repetition N400m effect was observed between words and pseudowords. We argue that this frontal response, an MEG correlate of the P2 identified in ERP studies, reflects early access to long-term memory representations, which we tentatively characterize as being modality-specific.

Introduction

The mental lexicon stores different kinds of information necessary for language comprehension, such as perceptual codes that map (visual or auditory or tactile) input onto stored lexical representations, motor codes, and syntactic and semantic information. As such, the lexicon acts as an interface between perceptual processes and higher-level linguistic processes. Understanding the mapping from orthographic visual input onto lexical representations is an important step in understanding language comprehension, but there is considerable disagreement about the neurobiological mechanisms and the time course of lexical access from print. We address these questions using magnetoencephalography (MEG), capitalizing on the well-established repetition priming effect.

Research on reading has used a variety of empirical methods, from classic behavioral psycholinguistic paradigms to eye-tracking and electrophysiological approaches (EEG and MEG). Different time frames have been implicated in the early stages of lexical access, often by different methodologies. Therefore, to derive an increasingly accurate and granular time-course of lexical access, it is necessary to investigate how their results fit together.

Lexical access research has largely focused on behavioral tasks with single words (e.g., lexical decision, speeded or delayed naming, and semantic categorization) to probe the variables that affect lexical retrieval. Reaction time (RT) is typically used as a dependent variable. However, the late and highly variable time frame associated with task execution (about 500–900 ms), coupled with the decisional nature of such tasks, raises questions about whether RT constitutes an accurate index of lexical access (see Balota and Chumbley, 1984, Balota and Chumbley, 1990 for review).

Research using eye-tracking has implicated a much earlier time-frame. Because the experimental paradigms typically do not require any task beyond reading, fixation latencies are thought to constitute more faithful indices of the underlying cognitive processes involved in reading than button presses or vocal responses (Radach & Kennedy, 2004). The average fixation time during reading is about 250 ms, but eye movement preparation takes about 150 ms. Therefore, in order for fixation time to reflect lexical processing, access to lexical information must already be underway by 100 ms, much earlier than the time frame implicated by experiments using single word reading and self-paced sentence reading (Sereno, Rayner, & Posner, 1998).

The most commonly studied electrophysiological response in the context of lexical access is the N400 and its MEG counterpart (alternatively called N400m or M400; see Kutas and Federmeier, 2000, Kutas and Federmeier, 2011, Lau et al., 2008 for review). The N400 displays many of the characteristics of a lexically-induced ERP. First, it is reliably elicited by words or word-like stimuli, either in sentential contexts (Van Petten & Kutas, 1990) or in isolation or pairs (Holcomb & Neville, 1990), but not for nonword-like written stimuli, like non-meaningful, orthographically illegal strings (e.g., NPMO, TBUA, LSIA; Rugg & Nagy, 1987, but see Laszlo & Federmeier, 2008). Second, it is found across paradigms – such as passive reading (Van Petten & Kutas, 1990), lexical decision (Holcomb & Neville, 1990), semantic discrimination (Rugg, Doyle, & Wells, 1995), and repetition judgment (Curran & Dien, 2003) – suggesting a high degree of task-independence. Third, the N400 seems to reflect meaning retrieval (see Kutas and Federmeier, 2000, Kutas and Federmeier, 2011). Finally, the N400 is found in both the visual and auditory modalities (Holcomb & Neville, 1990) as well as in signed languages (Kutas, Neville, & Holcomb, 1987), suggesting modality independence.

One of the earlier ERP components reliably implicated in reading is the occipitotemporal N170 and its MEG counterpart (alternatively called N170m or M170). ERP and MEG research show this response to be sensitive to orthographic processing (Cornelissen et al., 2003, Maurer et al., 2005, Simon et al., 2004, Tarkiainen et al., 2002, Tarkiainen et al., 1999), but the findings are divided on whether it reflects access to lexical information. Some studies report lexicality effects at the N170, which distinguishes words vs pseudowords and nonwords (Compton et al., 1991, McCandliss et al., 1997), whereas other studies report no sensitivity of the N/M170 to lexicality (Cornelissen et al., 2003, Fiorentino and Poeppel, 2007, Maurer et al., 2005, Pylkkänen et al., 2002, Simon et al., 2004, Stockall et al., 2004). Similarly, the N170 appears to be sensitive to lexical or morpheme frequency (Sereno et al., 1998, Zweig and Pylkkänen, 2009) – but not always (Embick, Hackl, Schaeffer, Kelepir, & Marantz, 2001). Finally, Sekiguchi, Koyama, and Kakigi (2001) report a Japanese MEG repetition priming experiment in which a component peaking around 200 ms was sensitive to the repetition of words but not pseudowords, speculating that the source of this effect lies in inferior occipitotemporal cortex, the same source implicated in the occipitotemporal N/M170 effects for words (McCandliss et al., 2003, McCandliss et al., 1997, Tarkiainen et al., 2003, Tarkiainen et al., 1999).

Another ERP component occasionally implicated in lexical processing in electrophysiological research is the P2, a response generally showing a frontocentral distribution. To our knowledge, this component has been only identified in ERP work, and lacks a clearly identified MEG counterpart. One of the clearest characterizations of this component was reported by Van Petten, Kutas, Kluender, Mitchiner, and McIsaac (1991), who studied lexical repetitions within naturalistic texts and described an early effect of repetition between 180 and 300 ms, with a more anterior scalp distribution than the subsequent negativity (N400). Early reports for P2 effects for orthographic string repetition (Nagy and Rugg, 1989, Rugg, 1987) were found for immediate repetition only, as opposed to the long lags used by Van Petten et al. (1991; see also Curran & Dien, 2003). There is at least some evidence that the P2 effect for repetition of strings is word specific, suggesting lexical processing is being reflected. For instance, an ERP study by Rugg et al. (1995), found that word repetition elicited early frontal effects starting at around 240 ms, while repetition of pseudowords elicited no such effects. Further evidence comes from masked priming experiments which reported P2 enhancement effects for immediate word repetition (Misra and Holcomb, 2003, Woollams et al., 2008). Moreover, in a sentence processing experiment, Dambacher, Kliegl, Hofmann, and Jacobs (2006) reported strong effects of lexical frequency, but not contextual predictability, in the P2 component at frontocentral electrodes, leading the authors to link the P2 to word recognition procedures (see also Barnea and Breznitz, 1998, Curran et al., 1993). The ERP findings of Hauk, Davis, Ford, Pulvermüller, and Marslen-Wilson (2006) are also compatible with a frontocentral P2 that reflects contact with lexical entries. These authors performed a novel multiple regression analysis of single-word reading ERP data and reported a lexicality effect (words vs pseudowords) starting at 160 ms, with a frontocentral scalp distribution (Hauk et al., 2006, Fig. 5). Moreover, the early bilateral evoked MEG response found by Sekiguchi et al. (2001) is also compatible with this component, although the authors themselves linked it to the M170. Finally, it is important to note that at least one study (Joyce & Rossion, 2005) has proposed that the N170 response for human faces can also surface as a central positivity (called the Vertex Positive Potential, VPP) depending on the choice of reference electrode sites. Although it should be pointed out that there are anatomical and functional differences between the N170 obtained in response to visually presented words and the N170 obtained in response to human faces, the results of Joyce and Rossion (2005)’s study still raise an important question as to whether the frontocentral P2 reported in several reading studies is a functionally independent evoked response, or simply a different scalp projection of the N170 obtained in response to visually presented words. Incidentally, MEG can help adjudicate this question, as MEG data does not depend on reference sites and is less spatially smeared than ERP data, due to the lack of volume conduction distortion (Hämäläinen et al., 1993, Nunez, 1986).

Given that there are robust effects at the N/M170 and the P2 related to lexical processing, it seems that the N400 is too late to reflect (at least) the early stages of lexical access. Some eye-tracking data suggest that lexical access in the context of normal reading is under way even earlier, around 100 ms. However, most eye-tracking studies investigate the natural reading of sentences, which might recruit top-down processes to a larger extent than single word reading, and therefore may not reflect the identical collection of processes. The most plausible remaining candidates for early lexical access are the two early electrophysiological evoked responses, the occipitotemporal N/M170 and the frontocentral P2.

To test the proposal that lexical access is reflected in physiological responses roughly 200 ms after word-onset, it is important to identify a task that can index lexicality. Two approaches can be pursued. First, one can manipulate well-known properties of lexical items (phonological, morphological, semantic) and seek neurophysiological (or behavioral) correlates that co-vary with the manipulated features (e.g., Hauk et al., 2006). Second, one can aim for a signal of ‘global’ lexical access. Here we pursue the latter strategy, using an experimental paradigm seemingly sensitive to overall lexicality independently of local (sublexical) manipulations.

By definition, words have a representation in long-term memory, whereas pseudowords do not. However, distinguishing the processing of pseudowords and existing words is more difficult than this would suggest. For instance, it has long been known that repeated presentation of words in a lexical decision task improves performance (e.g., Scarborough, Cortese, & Scarborough, 1977). Such repetition priming effects are generally interpreted as reflecting facilitation in retrieving information from long-term memory (i.e., the lexicon). Repetition effects are also investigated in electrophysiological studies, where an N400 modulation occurs for word repetition (Rugg, 1987, Rugg et al., 1995, Van Petten et al., 1991).

However, it is also known that not only words, but phonologically legal nonwords (pseudowords) and even illegal nonwords (nonwords) can, in some circumstances, elicit behavioral repetition effects (Bowers, 1994, Logan, 1988), particularly in lexical decision experiments (Wagenmakers, Zeelenberg, Steyvers, Shiffrin, & Raaijmakers, 2004). The repetition effect of pseudowords has also been observed in ERP research, both in immediate priming paradigms (Deacon, Dynowska, Ritter, & Grose-Fifer, 2004) and short-term priming contexts (Rugg et al., 1995, Fiebach et al., 2005), showing very similar neural signatures to the repetition of words. Within immediate priming contexts, repetition of words and pseudowords alike elicit smaller evoked potentials within the N400 time window (300–500 ms). Immediate repetition effects of both words and pseudowords have also been replicated in MEG for the same time window (Pylkkänen, Stringfellow, Flagg, & Marantz, 2000).

Presumably, the repetition effects for pseudowords and nonwords should not have their origin in access to or modification of long-term memory representations (which do not exist). Following this line of reasoning, several models accommodate repetition effects of pseudowords by linking facilitation to retrieval of the episodic memory trace formed in the first encounter with the item (Logan, 1988, Wagenmakers et al., 2004). This raises the possibility that episodic encoding could also be at play in the repetition of words. Indeed, some researchers argue that there is little evidence or theoretical necessity for positing that modification of access conditions for long-term memory representations has anything to do with the repetition effect for words (see Tenpenny, 1995 for review).

Alternatively, even though pseudowords are not stored in long-term memory, it is conceivable that they nonetheless engage the lexical system to some extent, perhaps by facilitating access to similar words (e.g., Bowers, 2000). A number of experiments provide evidence compatible with this hypothesis. For example, Deacon et al. (2004) reported a semantic priming experiment in which a prime-target pair of pseudowords (plynt–tlee) showed N400 priming effects that could only be explained by positing that the pseudowords accessed the semantic representations of the words they resembled in form (plant–tree). In the same vein, Holcomb et al. (2002) reported that words and pseudowords alike can show N400 modulation effects due to the size of their lexical neighborhoods (i.e., lexical entries that resemble the experimental stimuli in form) – another result that is hard to explain without allowing that pseudowords at least partially activate pre-existing lexical representations. Therefore, on the basis of these results, a simple comparison of words vs pseudowords might not provide the relevant information about whether access to lexical representations happens.

However, other researchers claim that, despite their sometimes similar behavioral and electrophysiological profiles (as reflected by N400 modulations), there are indeed two different mechanisms underlying the repetition effects of words and pseudowords (e.g., Bowers, 1996, Wagenmakers et al., 2004), and that these mechanisms are dissociable experimentally by varying tasks or task demands (e.g., Wagenmakers et al., 2004; see Bowers, 2000 for review). Dual-process accounts generally posit that two different and potentially opposing processes underlie the triggering of a lexical decision: a fast-acting familiarity-based process, and a slow-acting, episodic memory-based facilitatory process. Moreover, these processes can interact with strategic processes (for e.g., words being more familiar than pseudowords could produce a word bias in repeated presentations) in tasks such as lexical decision to either produce or eliminate repetition priming of pseudowords (Orfanidou et al., 2011, Wagenmakers et al., 2004). Dual-process models receive some support from the MEG study conducted by Sekiguchi et al. (2001). Using a medium term visual repetition priming paradigm, these authors reported word-specific reduction in the amplitude of the MEG evoked response over left hemisphere sensors in both the 200–300 and 300–500 ms post-stimulus onset window. They concluded that the neural sources of these effects subserve the mental lexicon (i.e., long term memory storage of words), but not general visual episodic memory traces.

We designed an MEG study to identify the time window in which early contact with the mental lexicon occurs. Our design featured variable lag medium term priming in a simple lexical decision task (e.g., Fiebach et al., 2005). The lag between repetition of words and pseudowords was varied to reduce the predictability of repetitions, since predictability has been shown to affect ERP and MEG responses very early in the processing stream, as well as in the N400 time window (Dambacher et al., 2009, Dikker et al., 2010). Given the arguments above that the N400 response does not reflect the earliest stages of lexical access, we expected that either the occipitotemporal M170 or an MEG correlate of the frontocentral P2 (henceforth, P2m) should be more sensitive to the repetition of words than to the repetition of pseudowords, even though we might expect to find behavioral repetition priming effects in both cases, due to the more slow-acting episodic-memory retrieval process that often dominates the triggering of a lexical decision (Wagenmakers et al., 2004).

Therefore, our hypothesis predicts a statistical interaction between Lexicality (word, pseudoword) and Order of Presentation (1st and 2nd) in the temporal window compatible with the M170 or P2m (150–250 ms), in which larger repetition effects should be observed for words compared to pseudowords. In addition, based on prior research, we also expected to observe a simple repetition effect for words and pseudowords alike in the N400 time window (i.e., a simple main effect of Order of Presentation).

However, if a lack of interaction between Lexicality and Order of Presentation in the 150–250 ms window were to be followed by a significant interaction in the 300–500 ms window, then this would have been interpreted as evidence that, contrary to our expectations, it is the N400, and not the earlier responses, that reflects the earliest stages of lexical retrieval.

Section snippets

Participants

Twenty-two native English speakers (11 women) participated in the experiment (mean age: 20, age range: 18–24). Fifteen data sets were included in the final analyses. All were right-handed, had normal or corrected-to-normal vision, and reported no history of hearing problems, language disorders or mental illness. All participants gave their written informed consent to take part in the study, which was approved by the University of Maryland Institutional Review Board. Subjects were paid for their

Results

We defined a repetition effect as an average difference between first and second presentations that is significantly different from zero. We had a directional prediction for the behavioral results: Second presentations should be faster than first presentations. We also had a directional prediction for the electrophysiological data: Repetition of words, because they involve modifications in the conditions of access to specific items in long term memory, should yield, in the 150–250 ms window,

General discussion

Behavioral repetition effects were found for both words and pseudowords, a result that is sometimes reported in written word lexical decision studies (e.g., Fiebach et al., 2005, Pylkkänen et al., 2000, Scarborough et al., 1977, Wagenmakers et al., 2004,), which might indicate that the lexical decision response in our repeated pseudowords was dominated by facilitatory slow-acting episodic memory retrieval processes.

Contrary to the behavioral findings, however, the MEG results demonstrated clear

Conclusion

In summary, the behavioral and MEG results discussed here are compatible with existing findings on word and pseudoword repetition in the MEG and ERP literatures, and suggest that (i) bottom-up, automatic contact with putative modality-specific (orthographic) representations of lexical items in long term memory happens around 200 ms post-stimulus onset in reading, and (ii) the N400 repetition effect can be related to post-lexical processing of words and the attempt to retrieve pseudowords from

Acknowledgments

This work was supported by NIH Grant 2R01DC05660 to D.P. The authors would like to thank Jeff Walker for invaluable technical assistance and Katie Yoshida for extensive editorial help in a previous version of this manuscript.

References (79)

  • E.F. Lau et al.

    A lexical basis for N400 context effects: Evidence from MEG

    Brain and Language

    (2009)
  • K. Marinkovic et al.

    Spatiotemporal dynamics of modality-specific and supramodal word processing

    Neuron

    (2003)
  • B.D. McCandliss et al.

    The visual word form area: Expertise for reading in the fusiform gyrus

    Trends in Cognitive Science

    (2003)
  • B.D. McCandliss et al.

    Brain plasticity in learning visual words

    Cognitive Psychology

    (1997)
  • P. Nunez

    The brain’s magnetic field: Some effects of multiple sources on localization methods

    Electroencephalography and clinical Neurophysiology

    (1986)
  • L. Pylkkänen et al.

    Neuromagnetic evidence for the timing of lexical activation: An MEG component sensitive to phonotactic probability but not to neighborhood density

    Brain and Language

    (2002)
  • T. Sekiguchi et al.

    The effect of stimulus repetition on cortical magnetic responses evoked by words and nonwords

    Neuroimage

    (2001)
  • L. Stockall et al.

    The precise time course of lexical activation: MEG measurements of the effects of frequency, probability, and density in lexical decision

    Brain and Language

    (2004)
  • A. Tarkiainen et al.

    Category-specific occipitotemporal activation during face perception in dyslexic individuals: An MEG study

    Neuroimage

    (2003)
  • R.H. Baayen et al.

    The CELEX lexical database [CD-ROM]. Linguistic data consortium

    (1993)
  • D.A. Balota et al.

    Are lexical decisions a good measure of lexical access? The role of word frequency in the neglected decision stage

    Journal of Experimental Psychology: Human Perception and Performance

    (1984)
  • D.A. Balota et al.

    Where are the effects of frequency in visual word recognition tasks? Right where we said they were! Comment on Monsell, Doyle, and Haggard (1989)

    Journal of Experimental Psychology: General

    (1990)
  • A. Barnea et al.

    Phonological and orthographic processing of Hebrew words: Electrophysiological aspects

    Journal of Genetic Psychology

    (1998)
  • J.S. Bowers

    Does implicit memory extend to legal and illegal nonwords?

    Journal of Experimental Psychology: Learning, Memory and Cognition

    (1994)
  • J.S. Bowers

    Different perceptual codes support priming for words and pseudowords: Was Morton right all along?

    Journal of Experimental Psychology: Learning, Memory and Cognition

    (1996)
  • J.S. Bowers

    In defense of abstractionist theories of repetition priming and word identification

    Psychonomic Bulletin Review

    (2000)
  • M. Brysbaert et al.

    Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English

    Behavior Research Methods

    (2009)
  • A.D. Cheveigné et al.

    Denoising based on time-shift PCA

    Journal of Neuroscience Methods

    (2007)
  • P.E. Compton et al.

    A cognitive-anatomical approach to attention in lexical access

    Journal of Cognitive Neuroscience

    (1991)
  • P. Cornelissen et al.

    Cortical effects of shifting letter position in letter strings of varying length

    Journal of Cognitive Neuroscience

    (2003)
  • T. Curran et al.

    Differentiating amodal familiarity from modality-specific memory processes: An ERP study

    Psychophysiology

    (2003)
  • M. Dambacher et al.

    Event-related potentials reveal rapid verification of predicted visual input

    PLoS ONE

    (2009)
  • D. Deacon et al.

    Repetition and semantic priming of nonwords: Implications for theories of N400 and word recognition

    Psychophysiology

    (2004)
  • R.P. Dhond et al.

    Spatiotemporal maps of brain activity underlying word generation and their modification during repetition priming

    Journal of Neuroscience

    (2001)
  • S. Dikker et al.

    Early occipital sensitivity to syntactic category is based on form typicality

    Psychological Science

    (2010)
  • C.J. Fiebach et al.

    Neuronal mechanisms of repetition priming in occipitotemporal cortex: spatiotemporal evidence from functional magnetic resonance imaging and electroencephalography

    The Journal of neuroscience

    (2005)
  • R. Fiorentino et al.

    Compound words and structure in the lexicon

    Language and Cognitive Processes

    (2007)
  • R.E. Greenblatt et al.

    Randomization-based hypothesis testing from event-related data

    Brain Topography

    (2004)
  • M. Hämäläinen et al.

    Magnetoencephalography – Theory, instrumentation, and applications to noninvasive studies of the working human brain

    Reviews of Modern Physics

    (1993)
  • Cited by (0)

    View full text