Content uploaded by Bruce D Mccandliss
Author content
All content in this area was uploaded by Bruce D Mccandliss on Dec 04, 2014
Content may be subject to copyright.
1
The development of visual expertise for words: the contribution of
electrophysiology
Urs Maurer and Bruce D. McCandliss
Skilled readers are able to process written text at a remarkable speed that surpasses the rate of typical
speech. A significant part of this fluent processing of connected writing involves computations applied to
individual words. Individual words are processed in order to activate corresponding information about
word meaning and pronunciation in the reader’s mental lexicon. The current chapter in this book on single
word processes focuses on the contribution of electrophysiology for understanding single word processes,
especially processes associated with accessing the visual forms of written words. Although some
processes applied to single words in isolation have been demonstrated to interact with other processes
related to text (for review see Balota, 1994), investigations of mental processes at the single word level
represent a critical component process within reading, and also provides scientifically pragmatic
paradigms for examining sub-processes involved, from processing the visual forms of words to accessing
linguistic representations of phonology and semantics.
The process of rapidly processing visual word forms to access linguistic representations may be
understood as a form of perceptual expertise (Gauthier & Nelson, 2001) that develops with reading
experience (McCandliss, Cohen, & Dehaene, 2003). Aspects of this fast perceptual process for visual
words that is inherent in our ability to rapidly process text have been isolated by several cognitive
paradigms (reviewed in this book, chapter x), and such processes have been localized to a network of
brain regions (also reviewed in this book, chapter y). This chapter investigates the contribution of
electrophysiology to understand the neural basis of this skill, as well as its development. Specifically, we
examine early perceptual responses related to visual words as well as responses to speech sounds that may
2
be critical to the foundation of reading skills, and have been implicated in both typical and atypical
development of reading ability.
Behavioral evidence for rapid perceptual processing of visual words
Naturally, when focusing on fast visual perceptual expertise for reading, issues of time-course of
perception of visual words become critical. Eye-tracking studies of reading provide insights into the time-
course of information processing for visual words, and have shown that readers typically fixate a word for
a short period of time (typically between about 200 and 300 ms) before moving to fixate the next (Rayner,
1998). During this brief fixation, information about the currently fixated word, such as its lexical
frequency, influences the amount of time the eye remains on that word, thus providing a lower limit for
the estimation of the ‘eye-mind’ lag in lexical access, suggesting that some information regarding the
word being viewed is accessed within the first 200 milliseconds (Reichle, Rayner, & Pollatsek, 2003).
Such research, examining single word processing in the context of connected text, converges nicely with a
large body of cognitive research conducted using eye tracking, naming, and lexical decision tasks applied
in paradigms that present single words in isolation. For example, converging evidence from studies of
brief, masked presentation of visual words suggests that the rapid perceptual processes we apply to letter
strings are facilitated by common patterns of combining letters into word forms – the word superiority
effect (Reicher, 1969) demonstrates that subjects are more accurate in detecting brief exposures to letters
presented within words than letters presented alone or within random consonant strings. Such perceptual
facilitation even occurs for letters embedded in pronouncable nonwords (pseudoword superiority effect,
for a recent study see Grainger, Bouttevin, Truc, Bastien, & Ziegler, 2003). Such studies provide
additional information into the nature of processes that occur within early perceptual processes applied to
visual words. While these behavioral studies allow inference about fast cognitive processes during
reading, such evidence is open to questions about the time-course of processing, as they potentially reflect
post-perceptual strategies. Electrophysiology research provides direct means of examining early
3
components of visual word perception through the analysis of electrical signals of brain activity recorded
on the human scalp.
A brief introduction to event-related potentials in electrophysiology
Basic concepts
The electroencephalogram (EEG) is the recording of fluctuating voltage differences between electrodes
placed on the scalp, and is measured with millisecond precision. The event-related potential (ERP)
represents the electrophysiological response in the EEG that is time- and phase-locked to a particular
event. An ERP is extracted by averaging many time-locked EEG epochs suppressing the background
activity in the EEG that is unrelated to the particular event.
Traditionally, ERPs are depicted as waveforms at particular electrodes, and the peaks and troughs in the
waveforms are thought to reflect components, which are typically labeled according to their polarity and
timing. For example components may be labeled as P1, N1, N170, P300, N400, with the letter depicting
whether the component was a positive or negative deviation from baseline, and the component number
representing the timing, either in cardinal order (as in P1, N1, P2, N2) or the millisecond latency of the
peak (e.g. P300: a positive-going component peaking at about 300 ms after stimulus onset). Additional
labels, such as “visual” or “occipital”, are often added to the component name, because polarity and
timing can vary with stimulus modality and electrode site.
To account for varying head sizes and head forms in interindividual comparisons, electrodes are placed in
relation to landmarks (inion, nasion, left and right preauricular points) that can be found on each head. By
dividing the distance between these landmarks into equidistant parts of 10% each, one of the commonly
used electrode placement system - the 10-10 system - creates a grid on the scalp to place the electrodes
(Chatrian, Lettich, & Nelson, 1985). The labels of the grid points (e.g. Oz, P1, C2, F3, Fpz) indicate their
anterior-posterior locations on the scalp (Fp: frontopolar, F: frontal, C: central, P: parietal, O: occipital)
and their relations to the midline (z: central, odd numbers: left; even numbers: right) with increasing
numbers indicating increasing excentricity.
4
ERP mapping approach
In traditional ERP analyses, topographic effects such as laterality are often limited by examination of a
few channels on the scalp, but modern EEG systems allow recordings from many channels (commonly as
many as 128) providing additional topographic information. One approach to topographic analysis of
multi-channel ERP data, termed a “mapping” approach, has been developed that looks at the data
primarily as a sequence of topographic ERP maps changing in topography and global strength over time
(Lehmann & Skrandies, 1980). Within this approach, analysis methods have been developed that take full
advantage of the additional topographic information while preserving the high time-resolution benefits of
ERP data (e.g. Brandeis & Lehmann, 1986; Lehmann & Skrandies, 1980; Michel et al., 2004; Pascual-
Marqui, Michel, & Lehmann, 1995).
ERP map strength can be described by Global Field Power (GFP; Lehmann & Skrandies, 1980), which is
computed as the root mean square of the values at each electrode in an average-referenced map. ERP map
topography can be described by map features, such as the locations of the positive and negative maxima
(Brandeis & Lehmann, 1986) or the locations of the positive and negative centroids (centers of gravity;
Brandeis, Vitacco, & Steinhausen, 1994).
The use of topographic information for ERP analysis is important, because it can resolve ambiguities that
result from the fact that amplitude differences between two conditions recorded at a single electrode can
result from identical topographies which are stronger in one condition compared to the other or they can
result from different topographies which may or may not differ in global strength. This distinction is
important because topographic information allows a characterization of the underlying neural processes:
Different topographies are produced by different neural networks, and identical topographies are likely to
reflect the same neural networks.
In an additional step, ERP topographies can be used to estimate location and orientation of the underlying
cortical sources, provided that a number of assumptions can be validly adopted. Assumptions are
necessary to mathematically solve the “the inverse problem” which captures the fact that one can model
5
the scalp topography given a set of neural sources of known location, orientation, and strength, but given a
known scalp topography, many potential combinations of number, location, orientation, and strength of
neural sources are equally plausible as solutions. Thus, different assumptions about the sources of
electrical activity and its propagation to the scalp, as implemented in different source estimation models,
have an influence on the nature of the results (for a recent review see Michel 2004).
Topographic 3D centroid analysis
The large number of electrodes used in modern EEG systems results in a vast amount of topographic
information. Selecting only a subset of these channels for analysis can lead to results that are biased by the
pre-selection of the channels. The use of map centroids offers an un-biased means for topographic ERP
analysis (see Figure 1). The positive and negative 3D centroids are the centers of gravity of the positive
and negative fields in 3D space (e.g. Talairach coordinate system, Talairach & Tournoux, 1988) and are
computed as the voltage-weighted locations of all electrodes showing positive or negative values,
respectively. Accordingly, an ERP map consisting of 129 electrodes can be reduced to 2 centroids, each
defined by 3 values representing the x-, y-, and z-coordinates of the 3D space. Subsequent statistical
analyses can be computed for the x-, y-, and z-coordinates of the centroids resulting in topographic
differences in 3 spatial dimensions “left-right”, “posterior-anterior”, and “inferior-superior”.
Importantly, although the centroids are located within the head space - which is typical for centers of
gravity of scalp measures, they are by no means estimations of the underlying sources. The advantage of
the centroid measure vs. source estimation is that the centroids are features of the measured topography,
whereas source estimations depend on additional assumptions that may or may not apply.
Overview of electrophysiology of visual word processing
Visual word processing has been extensively investigated with ERP measurements, and various aspects of
psychological processes involved in reading have been linked to several different ERP components.
Perhaps the most studied language component in response to visual words is the N400, a component
6
linked to semantic processes. The N400 component is a negative deflection which is significantly
enhanced when a word is presented as semantically anomalous within a sentence (Friederici, Gunter,
Hahne, & Mauth, 2004; Kutas & Federmeier, 2000). Studies on syntactic violations in sentence processing
have revealed a rather different component, the P600. Differences in topography of the N400 and P600
suggest that distinct processes are involved in processing semantic and syntactic violations, even in cases
when manipulations of syntactic expectations produce ERP responses to syntactic violations within 400
ms (Friederici et al., 2004). N400-like effects have also been used to investigate phonological processes in
visual word tasks (Rugg, 1984), although some aspects of phonological processing may occur between
about 200 and 350 ms, earlier than semantic N400 effects (Bentin, Mouchetant-Rostaing, Giard, Echallier,
& Pernier, 1999). For the purposes of the current chapter, each of these effects appear to implicate
processes that occur much later than the time-course of specialized word perception, which other cognitive
research suggests occurs within the first 200 ms of word perception. Reading-related effects, however,
have also been reported in earlier components, especially in the N170 component.
Perceptual expertise N170 effects
The visual N170 (or N1) component of the ERP peaks between 150 and 200 ms and shows a topography
with posterior negativity and anterior positivity. Although elicited by visual stimuli in general, the N170 is
strongly elicited by certain classes of visual stimuli, such as faces (Bentin et al., 1999; Rossion, Joyce,
Cottrell, & Tarr, 2003), compared to control stimuli.
The psychological principles that lead to a larger N170 for one stimulus category compared to another
may lie in perceptual expertise with the stimulus category at hand. Increased N170 responses were elicited
in bird-experts looking at birds (Tanaka & Curran, 2001), and in car-experts looking at cars (Gauthier,
Curran, Curby, & Collins, 2003). An increase of the N170 could even be induced by expertise-training
with novel objects (e.g. “greebles”, Rossion, Gauthier, Goffaux, Tarr, & Crommelinck, 2002). These
results suggest that extensive visual experience with an object category leads to fast, specialized
processing within the first 200 ms.
7
This perceptual expertise framework for evaluating N170 effects may also account for increased N170
responses that skilled readers show for visual words. A robust reading-related N170 specialization in
electric fields (as well as a similar component in magnetic fields of magnetoencephalography) is found for
contrasts between categories of stimulus classes including words versus other low-level visual control
stimuli such as strings of meaningless symbols, forms, shapes, or dots (Bentin et al., 1999; Brem et al.,
2005; Eulitz et al., 2000; Maurer, Brandeis, & McCandliss, 2005; Maurer, Brem, Bucher, & Brandeis, in
press; Schendan, Ganis, & Kutas, 1998; Tarkiainen, Helenius, Hansen, Cornelissen, & Salmelin, 1999;
Zhang, Begleiter, Porjesz, & Litke, 1997). Such overall reading-related N170 specialization appears to be
automatic, as it is also detected in tasks that do not require the words to be read (Bentin et al., 1999; Brem
et al., 2005; Eulitz et al., 2000; Maurer et al., 2005; Maurer et al., in press; Schendan et al., 1998;
Tarkiainen et al., 1999).
Examination of the pattern of stimuli that elicit an N170 response provides support for a form of similarity
gradient in these implicit tasks, such that the more the stimuli resemble letter-strings, the larger their N170
component, as found e.g. in a larger N170 for word-like pseudofonts compared to control stimuli (Eulitz et
al., 2000; Schendan et al., 1998; Tarkiainen et al., 1999). On the other hand, words, pseudowords, and
even consonant strings have been shown to produce similar N170 responses, which differed from that
elicited by symbol strings and other visual forms (Bentin et al., 1999; Maurer et al., in press).
Although specialization for words appears to be one example out of a broader class of perceptual expertise
stimuli that affect the N170, there is also evidence that visual words represent a special case of perceptual
expertise, as N170 responses to words are typically left-lateralized (for review see Maurer et al., 2005).
Left-lateralization of the N170
Several studies have shown that overall reading-related N170 specialization is left-lateralized (Bentin et
al., 1999; Maurer et al., 2005; Maurer et al., in press; Tarkiainen et al., 1999), with larger amplitudes over
the left hemisphere for words than for low-level visual control stimuli. This left-lateralized N170
topography elicited by visual words stands in contrast to N170 responses for other forms of perceptual
8
expertise related to faces or objects of expertise, which are typically bilateral or right-lateralized (Rossion
et al., 2003; Tanaka & Curran, 2001). The degree of left-lateralization in the N170, however, varies across
studies and seems to depend on additional factors, which are not yet fully understood. In pursuit of a
suitable explanation we propose below factors that may help explain this variability.
Source localization estimates for the N170 elicited by words found left-lateralized sources in inferior
occipitotemporal cortex (Maurer et al., in press; Michel et al., 2001; Rossion et al., 2003). This is in
agreement with intracranial recordings finding a negative component around 200 ms in basal
occipitotemporal cortex (Nobre, Allison, & McCarthy, 1998) and with source localization of the word-
specific M170, the N170 analogue recorded with MEG (Tarkiainen et al., 1999).
The characteristic trend towards a left-lateralized N170 topography for words might be linked to similarly
left-lateralized hemodynamic activation during visual word tasks. Functional neuroimaging studies
reported reading-related activation in many areas of the extrastriate visual cortex, especially in the left
hemisphere (Petersen, Fox, Posner, Mintun, & Raichle, 1988; Price, Wise, & Frackowiak, 1996;
Tagamets, Novick, Chalmers, & Friedman, 2000). In particular, an area in the left fusiform gyrus, located
in the inferior part of the occipito-temporal cortex, may constitute a Visual Word Form Area, because it
shows sensitivity for visual word forms at a highly abstract level (for a review see McCandliss et al.,
2003).
Similar sensitivity for abstract properties of visual words may also already occur during the N170
component.
Sensitivity of the N170 for higher language functions
Several studies have also examined the nature of cognitive processing indexed by the N170 word effect by
investigating additional stimulus category contrasts. Comparing consonant strings with pseudowords or
words serves to control letter expertise while contrasting information on the structure of a word form. One
set of results demonstrated that consonant strings have larger N170 amplitudes than words (Compton,
Grossenbacher, Posner, & Tucker, 1991; McCandliss, Posner, & Givon, 1997), and orthographically
9
irregular pseudowords were in between (McCandliss et al., 1997). Such results provide support for the
notion that perceptual expertise for words indexed by the N170 reflects not just expertise for letter
recognition, but also expertise associated with recognition of familiar patterns of letters within visual word
forms. Other studies, however, failed to show N170 sensitivity to the contrast between words and
pseudowords (Bentin et al., 1999; Wydell, Vuorinen, Helenius, & Salmelin, 2003), and differences
between consonant strings and words were found only during explicit lexical and semantic tasks, not
during implicit reading (Bentin et al., 1999).
In contrast to such studies that make inferences based on comparisons of different stimulus categories (i.e.
words, pseudowords, consonant string), studies that compared different levels of word frequency provide
more consistent results regarding the role of lexical information in modulating the N170. Low frequency
words typically elicit larger N170 amplitudes than high frequency words in lexical or semantic decision
tasks (Assadollahi & Pulvermuller, 2003; Hauk & Pulvermuller, 2004; Neville, Mills, & Lawson, 1992;
Sereno, Brewer, & O'Donnell, 2003; Sereno, Rayner, & Posner, 1998, but see also Proverbio, Vecchi, &
Zani, 2004), providing evidence of perceptual expertise at the level of accessing specific words. Thus,
different approaches to the question of whether N170 responses are sensitive to specific word
representations, such as categorical distinctions between words and pseudowords vs. within category
parametric manipulations of word frequency) provide contrasting answers, and raise new potential
questions regarding the nature of processing applied to pseudowords.
As these studies based their analyses mostly on a few channels, an ERP mapping approach, which takes
full advantage of the topographic information available, may be able to better characterize the processes
involved and to resolve some ambiguities.
N170 ERP mapping studies in German and English
Reading-related N170 specialization was investigated in two separate studies with the same paradigm in
Zurich, Switzerland (Maurer et al., in press), and subsequently at the Sackler Institute in New York
(Maurer et al., 2005).
10
In both studies, literate adult subjects looked at blocks of serially presented stimuli that contained runs of
either words, pseudowords, symbol strings, or pictures. For each class of stimuli the subjects remained
motionless in response to the majority of stimuli, but pressed a button whenever they detected an
occasional immediate stimulus repetition. This ‘one-back’ paradigm allowed ERPs to be calculated for
each word on its initial presentation without any reaction on the part of the subject, and behavioral
responses to be collected on occasional repeated stimuli, thus ensuring that subjects engaged in the task.
Moreover, as repeated words could be detected even without reading, this implicit reading task could
potentially be applied with illiterate children, thus allowing the investigation of changes due to learning to
read.
In the Zurich study, German-speaking adults viewed German stimuli, and EEG was recorded from 43
electrodes. Data were analyzed with an ERP mapping strategy, i.e. differences in N170 maps were
measured according to global strength (GFP) and topography (3D centroids).
Statistical t-maps showed that among ERP responses to stimuli that required no manual response, larger
N170 amplitudes were found for words than symbols particularly over left occipitotemporal electrodes
consistent with earlier studies analyzing waveforms at selected channels (Bentin et al., 1999). The
mapping analyses showed that GFP was stronger for words than for symbols and that N170 topography
differed between words and symbols, implicating different neural networks involved in word and symbol
processing within the first 200 milliseconds. The most prominent topographic feature that differed
between word and symbol maps was found in the distribution of the centroids along the inferior-superior z
coordinate axis. These centroid differences reflected that the negative fields over the posterior part of the
scalp were most pronounced at inferior sites for words and at superior sites for symbols, whereas over the
anterior part the positive fields were most pronounced at inferior sites for symbols and at superior sites for
words. The centroid differences also reflected that the largest negative differences occurred at left inferior
occipito-temporal electrodes at the edge of the electrode montage.
11
For the word-pseudoword contrast, however, there were no N170 differences in response to German
words and pseudowords, suggesting that reading-related N170 specialization generalized from words to
pseudowords, and thus may reflect perceptual expertise for letters or well-ordered letter strings.
For the pseudoword-symbol contrast, similar prominent topographic differences in the centroid
distribution were found as for the word-symbol contrast. In addition the centroids were more left-
lateralized for pseudowords than for symbols. This reflected that the negative fields of the N170 were
more left-lateralized for pseudowords than for symbols, which was also apparent in the word-symbol
comparison, where it reached significance in the last two thirds of the N170.
The topographic analysis of the Zurich study extended earlier studies by showing that reading-related
N170 specialization is characterized not only by more left-lateralized fields but by more inferior negative
fields over the posterior part of the head and more superior positive fields over the anterior part of the
head. This suggests that the differences arose because different neural networks were activated in an early
phase of word processing compared to symbol processing, rather than the same network being more
strongly activated.
In the New York study we investigated whether the effects from the Zurich study could be replicated with
the same paradigm (after adaptations for language) in a sample of English speaking participants (Maurer
et al., 2005). An additional aim of the study was to apply a high-density EEG recording system, because
this system samples more densely and extends the coverage on the scalp to more inferior locations than
traditional EEG systems do. The traditional 10-20 electrode placement system covers only regions as
much inferior as the Oz and Fpz electrodes, thus – roughly speaking – electrodes are only placed about
above the ears. For signals that presumably originate in inferior brain regions, a more inferior sampling
may provide a better characterization of the resulting scalp topography. For this reason the 10-20 system
has been extended to more inferior regions in some studies (e.g. Bentin et al., 1999; Maurer et al., in
press). However, high-density recordings sample from sites that are located even more inferior than those
in these previous studies (Luu & Ferree, 2000). As the maximal effect in the Zurich study was found at the
12
edge of the montage at posterior inferior channels, more inferior sampling may provide better
characterization of the effects.
The central contrast for the replication study was the most robust contrast of words versus unfamiliar
symbol strings in the N170. The statistical t-maps in the New York study showed that words elicited
larger N170 amplitudes than symbols at left inferior occipitotemporal electrodes similar to the Zurich
results. The mapping analyses revealed that GFP was not larger for words than for symbols, but as in the
Zurich study the N170 topographies differed between words and symbols, confirming that specialized
neural networks are active within the first 200ms of word presentation. These topographic differences
showed very similar characteristics to the results of the Zurich study with a different centroid distribution
along the inferior-superior z coordinate axis between word and symbol responses, suggesting that similar
neural networks are specialized for reading across languages.
In addition, a topographic effect was also found in the left-right axis, suggesting a larger involvement of
the left hemisphere in word processing and the right hemisphere in symbol processing. A similar
difference in left-lateralization was also present in the Zurich data, where it reached significance in the last
two thirds of the N170.
These results suggested that overall reading-related N170 specialization can be detected across different
languages and EEG systems and that the maximal effect is inferior to the coverage of traditional EEG
montages. Because the topographic effects where consistent while the GFP effects varied across studies, it
can be inferred that similar neural networks are activated across languages, whereas the relative strength
of the engagement of these networks may depend on additional factors. Finally, contrasts between results
in German and English may provide additional insights into how differences in these writing systems may
lead to different forms of perceptual expertise for reading.
One difference that emerged between the English and German studies involved the responses to
pseudowords. While the Zurich study revealed comparable N170 effects for words and pseudowords, in
English, N170 topographic effects for pseudowords were not identical to words. Words and pseudowords,
when compared to symbol strings, both demonstrated similar topographic effects in the inferior-superior
13
z-axis in both English and German, yet words were more strongly left-lateralized than pseudowords in
English.
These findings may suggest that inferior-superior topographic effects may be indexing some form of
processing which is constant across these two languages, but that the left-lateralized topographic effect
reflects a form of processing that is more language-specific. We suggest that the inferior-superior N170
modulation indicates visual expertise for letters or well-ordered letter strings, and may reflect more
general visual perceptual expertise effects that are also found with stimuli outside of reading.
Considering that pseudowords elicit a left-hemisphere modulation of the N170 in German, but not in
English, may reflect differences between these writing systems that impact the processing of novel visual
word forms. In fact a prominent difference between the two languages involves the degree of consistency
with which letters map onto word sounds. As a result, pseudowords are more ambiguous for English
speakers to pronounce than for German speakers. Thus the left-lateralized subtype of perceptual expertise
may specifically relate to processes involved in mapping letters onto word sounds. The lack of such a left-
lateralization for English pseudowords may suggest that such processes are less automatic in English
(Zevin & Balota, 2000), and are engaged to a lesser degree while detecting pseudoword repetitions,
because repetition detection does not require explicit pronunciation of the stimuli.
Although direct comparisons of these studies with English and German subjects may be limited by the use
different words, pseudowords, and EEG systems, the results suggest that left-lateralization may be related
to spelling-to-sound mapping, which leads to formulation of a more general hypothesis about learning to
read and left-lateralized specialization of the N170 word effect — the phonological mapping hypothesis.
The phonological mapping hypothesis
Converging evidence from electrophysiological and hemodynamic studies suggests that left-lateralized
activation is a characteristic of visual word processing in the brain. As the left hemisphere has long been
known to be dominantly involved in speech perception and speech production, one straightforward
14
hypothesis is that during reading acquisition the left-lateralized characteristic of the visual language
system is induced by the pre-existing left-lateralization of the auditory language system.
More specifically, the left-lateralization might be driven by a particular aspect of auditory language
processing, namely phonological processing, which leads to the phonological mapping hypothesis
(McCandliss & Noble, 2003). The essence of this hypothesis is that given that phonological processes are
typically left-lateralized (Price, Moore, Humphreys, & Wise, 1997; Rumsey et al., 1997), specialized
processing of visual words in visual brain areas also becomes left-lateralized.
The results of the Zurich and the New York ERP mapping studies suggest that the phonological mapping
hypothesis also accounts for fast reading-related expertise in the N170 component. Accordingly, the
characteristic left-hemispheric modulation of the N170 may be specifically related to the influence of
grapheme-phoneme conversion established during learning to read. This left-hemispheric modulation may
add up to the inferior-superior modulation thought to reflect visual expertise for letters or well-ordered
letter strings, and which may also develop during learning to read. This inferior-superior modulation of the
N170 might be more generally related to visual discrimination learning and thus might be less language-
specific. However, it cannot be ruled out that this modulation could nonetheless be shaped by grapheme-
phoneme conversion or other language-specific aspects during learning to read.
In its simplest form the phonological mapping hypothesis for the left-lateralized N170 component has
several implications for reading-related N170 specialization:
1) The left-lateralization of the N170 responses to visual words should be more pronounced in
scripts using grapheme-phoneme conversion rules, but less pronounced in logographic scripts
which are based on lexical morphemes. Furthermore, the phonological mapping hypothesis is
specific to the left-lateralized modulation of the N170, thus the inferior-superior N170 modulation
should not be influenced by scripts that differ in their phonological properties.
2) Reading-related N170 specialization, with inferior-superior and left-lateralized modulations,
should develop in children when they learn to read, as well as in laboratory experiments that
simulate this process.
15
3) Reading-related N170 specialization in dyslexic readers should show a smaller degree of left-
lateralization, because of the phonological core deficit that has been associated with dyslexia,
although such disorders could also affect the inferior-superior modulation.
4) Early phonological ability should predict the degree of N170 specialization with reading
acquisition, especially with respect to its left-lateralization. Remediation of the phonological core
deficit through intervention should specifically increase left-lateralization of the N170
specialization.
We consider the implications of these facets of the phonological mapping hypothesis in a broader
consideration of the literature on reading-related N170 specialization.
N170 specialization in scripts of different language systems
Comparisons between fundamentally different writing systems may allow conclusions about processes
involved during early visual word processing. For example a study with Koreans who were educated in
both Chinese characters and written English, reported a left-lateralized N170 for both English words and
Korean words, but a bilateral N170 for Chinese characters and pictures (Kim, Yoon, & Park, 2004). Both
English and Korean writing systems map characters onto phonemes, whereas Chinese uses a logographic
script, in which graphic symbols represent lexical morphemes. Thus, left-lateralization in the N170 was
confined to language systems that use spelling-to-sound mapping that can be described at the grapheme-
phoneme level, which suggests that the left-lateralization, developed during reading acquisition, is
mediated by phonological processing related to grapheme-phoneme conversion.
Such cross-cultural and cross-linguistic differences, although confounded by many challenges of between-
group, between-lab, and between-culture factors, nonetheless provide support for the phonological
mapping hypothesis for the left-lateralization of the N170.
One such confound in cross-linguistic studies is the possibility of a difference in lateralization between
first and second languages. For example, Proverbio et al. (2002) reported N170 responses for bilinguals
that suggested a left-lateralization for the first language (Slovenian), but not for the second language
16
(Italian; Proverbio et al., 2002), a result which was also found in Italian-English interpreters (Proverbio et
al., 2004), suggesting that reading skill for languages acquired later in life may be organized somewhat
differently than for languages acquired early in life. Contrasts between first and second languages,
however, may be complicated by differences in perceptual expertise across the two languages, and thus
further studies are needed to clarify whether this lateralization effect is related to differences in spelling-
to-sound mapping between first and second languages, differences in the strength of specialization, or
whether it represents an additional factor driving left-lateralized N170 specialization.
Learning to read and development of N170 specialization
The predictions of the phonological mapping hypothesis on reading acquisition can be tested with
developmental studies in children learning to read, as well as within laboratory experiments with adults in
which aspects of reading acquisition are simulated.
The most direct evidence for specialization due to learning to read can be obtained by studying the same
children before and after the start of reading acquisition. Within the context of the Zurich study, described
above, children’s N170 responses to words and symbol strings were recorded before and after learning to
read. EEG was recorded from pre-literate children in kindergarten, in the same paradigm as described
above, as they detected occasional immediate repetitions of words, and pseudowords, (a task which they
could perform even without reading skills), as well as immediate repetitions of symbols and pictures. As
reported above, adult readers in this paradigm had shown a large difference in the N170 between words
and symbol strings, with a bilateral occipito-temporal topography, which was slightly stronger on the left.
In contrast, the kindergarten children did not show a reliable difference in the N170 elicited by words and
symbols (Maurer et al., in press). This result demonstrates that reading-related N170 specialization
develops between kindergarten and adulthood. It also suggests that rudimentary levels of visual familiarity
with print and letter knowledge are not sufficient to produce the typical reading-related N170 response, as
the children could tell whether a stimulus string consisted of letters or other symbols, and could name
about half of the letters in the alphabet. A further analysis which examined subsets of children with high
17
and low letter knowledge confirmed that the children with low letter knowledge did not show a N170
specialization at all, but revealed that the children with high letter knowledge showed a weak
specialization. The topography of this difference, however, was right-lateralized and strikingly different
from the adult N170 effect, which suggested that the children’s difference indicated a precursor to the
mature fast specialization in adults. This precursor in non-reading children is presumably due to their
letter knowledge or visual familiarity with print (Maurer et al., in press).
The absence of reading-related N170 specialization in pre-literate Swiss kindergartners, and especially the
lack of a left-lateralized N170 modulation, lends some support to the phonological mapping hypothesis,
which suggests that N170 responses develop as a result of increased mapping from letters to sounds. In
addition, these same children have recently participated in the same paradigm during the middle of the 2
nd
grade, after they had mastered initial reading training. Forthcoming analyses of the longitudinal data will
reveal whether reading-related N170 specialization emerges quickly with the initial reading training or
whether it develops rather slowly, and may also provide additional subset analyses between children who
differ in their initial phonological skills (Maurer et al., submitted).
One cross-sectional developmental study also provides a similar developmental account of N170
specialization. Posner and McCandliss (2000) reported a study looking at four-, seven-, and ten-year-old
children, using a contrast between words and consonant strings previously reported to demonstrate visual
N170 effects in adults (Compton et al., 1991; McCandliss et al., 1997), modified to ensure that the words
used were familiar to both the seven- and ten-year-olds in the study. Using an implicit reading task, they
reported that no N170 specialization for words over consonant strings emerged with initial learning to read
(between age 4 and 7), but showed that 10-year olds began to demonstrate some evidence of differential
N170 responses to words versus consonant strings (Posner & McCandliss, 2000). These results suggest
that familiarity with words alone is unlikely to account for such N170 expertise effects, as the 7 year olds
demonstrate familiarity but no N170 effect, and suggest instead that such effects may arise gradually over
development of extensive expertise with fluent visual word recognition, a process that is emerging in 10-
year-olds.
18
Developmental studies with children provide crucial data on learning to read that is ecologically
valid, and provide insights into the nature of the processes that create the adult specialization, yet such
studies raise questions about whether observed changes are specifically linked to learning-based increases
in reading expertise or to maturation processes that play out over this same age range. Developmental
studies can be usefully complemented by training studies with skilled adult readers to address these very
issues. McCandliss and colleagues (McCandliss et al., 1997) introduced a novel way of investigating the
impact of visual familiarity and lexical status of visual stimuli, by holding the exact stimulus set constant
across a series of repeated testing sessions, but manipulating subjects’ experience with a subset of the
stimuli. With repeated measures this study investigated potential changes in the N170 induced over the
course of 5 weeks as students spent 50 hours learning to read 60 words of a 120-word artificial
pseudoword-language called “Keki” (the other 60 were reserved for testing purposes only). All
pseudowords were comprised of familiar Roman letters, yet followed a visual word form structure
generated by an algorithm that was designed to deviate from English in subtle but identifiable ways (e.g.
all words ended in a vowel). N170 responses to several classes of stimuli were collected over the course of
the five week learning experiment, and overall results demonstrated a significant effect of stimulus class,
such that consonant strings elicited larger N170 responses than words, and responses to the Keki
pseudowords fell in between. The central finding from the training study was that fifty hours of training,
which increased both the visual familiarity and the associated meanings for the trained Keki words, did
not change the stimulus-class effects on the N170 (i.e. no training-by-stimulus-type interaction was
present for the N170 component). Even after training, the N170 for trained and untrained Keki words were
not significantly different, and responses to the entire class of Keki items were still significantly more
negative than for words, and significantly less negative than for consonant strings. In contrast, a
component subsequent to the N170 demonstrated a significant and systematic training effect for the
trained Keki items in relation to the other stimuli, from approximately 280 to 360 ms, revealing the
sensitivity of the electrophysiological technique to training effects. From these results, the authors
concluded that the N170 likely reflects orthographic structure, as letter familiarity was held constant
19
across stimuli, and lexical familiarity was manipulated over 50 hours to no effect, and yet robust
differences persisted across the three classes of stimuli (well structured English words, slightly atypically
structured Keki words, and strongly atypically structured consonant strings). Furthermore, they suggested
that, since the N170 was unresponsive to 50 hours of studying the novel structure of Keki word forms
relative to consonant strings, such processes may change very slowly over time. Considering this pattern
of results in the context of the phonological mapping hypothesis draws focus to the fact that the Keki
words could be decoded via grapheme-phoneme associations related to reading English both before
training and throughout training, and thus a lack of N170 training-related changes for the specifically
trained Keki items might be predicted. In order to address this implication, future research in training
studies should employ novel graphical features that lie outside the ability to generalize based on already
existing grapheme-phoneme decoding abilities, and to directly contrast training methods that encourage
learning via associations between graphic features and phonemes versus training that encourages
associations between entire visual characters and auditory words.
N170 specialization in dyslexia
The phonological mapping hypothesis of the left-lateralized N170 expertise effect has important
implications for dyslexia because it provides a developmental pathway account for how well-documented
core phonological deficits present in early childhood and other precursors of dyslexia (for a review see
Shaywitz, 2004) impact the developing neural mechanisms underlying fluent visual word recognition.
Furthermore, individual differences in phonological mapping ability may also relate to the degree of
reading-related N170 specialization in dyslexic children and adults, especially with respect to its left-
lateralization.
Evidence for reduced reading-related N170 specialization in dyslexia has come from
magnetoencephalographic studies. Helenius and colleagues (1999) presented words and symbol strings to
dyslexic adults who attended to the stimuli and were prepared to report them if prompted. In normally
reading adults, sources in the inferior occipito-temporal cortex, predominantely in the left hemisphere,
20
differed between words and symbols around 150ms (Tarkiainen et al., 1999). In dyslexic subjects,
however, such word-specific sources were undetectable in the same time range (Helenius et al., 1999).
This pattern of results is corroborated by another MEG study which found that words and pseudowords
activate sources in the left occipito-temporal cortex in normal readers between 100 and 200 ms, but less so
in dyslexic readers (Salmelin, Service, Kiesila, Uutela, & Salonen, 1996). Such results are at least
consistent with the phonological mapping hypothesis, in that they present further evidence on the link
between adult expertise in reading and the left-lateralized N170, and are consistent with the notion that
phonological core deficit in dyslexia may impact the process of progressively increasing left-lateralized
recruitment of visual regions that are the hallmark of reading-related expertise in the form of the N170.
However, such developmental claims about the late emergence of left-lateralized N170 responses for
skilled readers, and not dyslexics, require developmental data. Interestingly, two studies that directly
compared dyslexic children to age-matched controls did not find group differences in visual word
processing in the N170 time range (Brandeis et al., 1994; Simos, Breier, Fletcher, Bergman, &
Papanicolaou, 2000). This contrast between the adult and child literature may suggest that differences
between normal and dyslexic reading develop only late during childhood and become manifest in
adulthood only with the emergence of visual expertise in skilled adult readers.
Future developmental work on the cognitive and neural basis of the N170 effect in dyslexia will need to
include longitudinal designs that examine early manifestations of phonological deficits, and relate such
deficits directly to the emergence of behavioral and neurophysiological indexes of perceptual expertise in
reading. Such developmental work on early phonological deficits may also be enhanced by the inclusion
of electrophysiological measures of phonological processes, as behavioral assays of phonological deficits
may reflect not only deficiencies in phonological processing, but also deficiencies in other processes such
as executive attention functions (for a review see Noble, McCandliss, & Farah, submitted). Such
electrophysiological studies of phonological processing have tried to directly measure brain processes
related to the phonological core deficit, thus aiming to improve prediction of dyslexia.
21
One candidate for a neurophysiological measure of phonological processing deficits in dyslexia is the
mismatch negativity (MMN), a component of the auditory ERP. The MMN is regarded as a measure of
the auditory memory or the central sound representation (Naatanen, Tervaniemi, Sussman, Paavilainen, &
Winkler, 2001). MMN responses are also elicited by deviant phonemes and thus may represent a measure
for phoneme representations in the brain (Naatanen, 2001). The MMN is also very suited to be used with
children, as it measures automatic discrimination, i.e. the participants are given a distracting task, such as
reading a book or watching a silent video. The MMN is also regarded as developmentally stable as it has
been elicited in young children and even in infants (Cheour, Leppanen, & Kraus, 2000), although in
children it can change its topography under certain conditions (Maurer, Bucher, Brem, & Brandeis,
2003a).
Currently, several longitudinal studies with children from families with risk for dyslexia are being
conducted that have obtained MMN measures before the start of reading acquisition.
In the Zurich study (described above) that looked at development of reading-related N170 specialization
in children before and after learning to read, a subgroup of children came from families with one or more
parents demonstrating symptoms of dyslexia. The kindergarten children were tested with two auditory
oddball paradigms containing tone stimuli (standard: 1000Hz, deviants: 1030 Hz, 1060 Hz) and phoneme
stimuli (standard: ba, deviants: da, ta). Between approximately 300 and 700 ms the children showed a
frontally negative mismatch response to the deviant stimulus compared to the standard. This late-MMN
differed between children at risk and control children (Maurer, Bucher, Brem, & Brandeis, 2003b).
Children at risk for dyslexia demonstrated an attenuated late-MMN response following deviant tone
stimuli, and demonstrated an atypical topography of the late-MMN in response to deviant phoneme
stimuli. This topographic difference following deviant phonemes was potentially informative, as the
control children showed one major positive pole, which was strongly left-lateralized, indicating left-
lateralized mismatch processing, whereas the children at risk showed two positive poles of the MMN,
indicating bilateral mismatch processing.
22
These results suggest deviant automatic phoneme perception in children at risk for dyslexia. The
attenuated MMN to tones may suggest that the deviant phoneme processing is related to a more low-level
auditory processing deficit. Pending longitudinal results may reveal whether such effects are early
predictive markers of specific dyslexia-risk for these individual children, or merely markers of familial
risk. Moreover, such longitudinal designs provide the framework to test whether these measures of speech
perception can predict the degree of reading-related N170 specialization and its left-hemispheric
modulation.
Evidence for predictive values of ERP measures of early speech perception for later reading ability comes
from studies that have followed children from early perception of speech through development of early
reading skills. The Jyvaskyla longitudinal study in Finland, followed the development of children at
familial risk for dyslexia in contrast to typically developing children. Testing, including ERP recordings,
started in the first days after birth, and will continue intermittently until the 3
rd
grade. The ERP data,
assessing basic speech processing and automatic mismatch response to speech stimuli, showed that the at-
risk infants already differed from the control group during their first days and months of infancy (Lyytinen
et al., 2004; Lyytinen et al., 2004). Comparison of the ERP data from the first days of life with later
language development showed correlations with receptive language skills and verbal memory (Guttorm et
al., 2005). Preliminary data also indicates correlations with initial reading and spelling skills (Lyytinen et
al., 2004). The results from this longitudinal study are generally consistent with earlier reports that ERP
responses to speech sounds recorded within hours after birth are strongly correlated with reading ability at
8 years of age (Molfese, 2000). In this study, some selected indexes derived from the ERP results
collected in infancy were able to support discrimination among children into three different groups of
reading and IQ impairments with an overall accuracy of 81%.
Such longitudinal studies provide evidence for the role of early speech processing in later language
development and reading acquisition. These studies, however, did not investigate reading-related N170
specialization and thus do not allow for a test of the phonological mapping hypothesis with regard to the
role of phonological processing for specialized visual word recognition. Based on our review of current
23
findings, such developmental studies would need to include children beyond age seven to characterize the
rise of perceptual expertise and N170 responses to visual words.
Conclusions.
Behavioral studies have indicated that word-specific information is processed within the first 200ms of
stimulus presentation. Such fast visual word processing ability in skilled adults may rely on left-lateralized
visual expertise effects linked to the N170 component. Converging evidence shows larger N170
amplitudes, especially over the left hemisphere, for words compared to visual control stimuli such as
symbol strings, but results regarding specialization among different types of letter strings and the degree
of the left-lateralization of the word N170 suggest large variation due to additional factors involved. An
ERP mapping approach that takes advantage of modern multi-channel EEG recordings in participants with
different language backgrounds suggested two overlapping processes in the N170, leading to the
formulation of the phonological mapping hypothesis for the development of reading-related N170
specialization. A left-lateralized modulation may develop under the influence of grapheme-phoneme
conversion during learning to read and reflects the involvement of spelling-to-sound mapping during
visual word processing. Furthermore, a more domain-general inferior-superior modulation may develop
through visual discrimination learning during reading acquisition. This hypothesis frames a set of specific
predictions regarding reading-related N170 specialization in language systems using different scripts, in
learning to read, and in dyslexia, that can be tested in specific studies. Results that allow such tests are just
emerging and seem to support the predictions of the phonological mapping hypothesis for N170
specialization. Emerging and future results that directly examine the developmental and learning changes
that link phonological processes to the emergence of expertise in fluent visual word recognition via
development and training studies will provide more direct evidence that bear on such predictions, and will
likely provide further neural-circuitry-level insights into the developmental and learning pathways that
give rise to fluent visual word recognition.
24
References
Assadollahi, R., & Pulvermuller, F. (2003). Early influences of word length and frequency: a group study
using MEG. Neuroreport, 14(8), 1183-1187.
Balota, D. A. (1994). Visual word recognition: the journey from features to meaning. In M. A.
Gernsbacher (Ed.), Handbook of Psycholinguistics (pp. 303-358). New York: Academic Press.
Bentin, S., Mouchetant-Rostaing, Y., Giard, M. H., Echallier, J. F., & Pernier, J. (1999). ERP
manifestations of processing printed words at different psycholinguistic levels: time course and scalp
distribution. J Cogn Neurosci, 11(3), 235-260.
Brandeis, D., & Lehmann, D. (1986). Event-related potentials of the brain and cognitive processes:
approaches and applications. Neuropsychologia, 24(1), 151-168.
Brandeis, D., Vitacco, D., & Steinhausen, H. C. (1994). Mapping brain electric micro-states in dyslexic
children during reading. Acta Paedopsychiatr, 56(3), 239-247.
Brem, S., Lang-Dullenkopf, A., Maurer, U., Halder, P., Bucher, K., & Brandeis, D. (2005).
Neurophysiological signs of rapidly emerging visual expertise for symbol strings. Neuroreport, 16(1),
45-48.
Chatrian, G. E., Lettich, E., & Nelson, P. L. (1985). Ten percent electrode system for topographic studies
of spontaneous and evoked EEG activity. Am J EEG Technol, 25, 83-92.
Cheour, M., Leppanen, P. H., & Kraus, N. (2000). Mismatch negativity (MMN) as a tool for investigating
auditory discrimination and sensory memory in infants and children. Clin Neurophysiol, 111(1), 4-16.
Compton, P. E., Grossenbacher, P., Posner, M. I., & Tucker, D. M. (1991). A cognitive-anatomical
approach to attention in lexical access. J Cogn Neurosci, 3(4), 304-312.
Eulitz, C., Eulitz, H., Maess, B., Cohen, R., Pantev, C., & Elbert, T. (2000). Magnetic brain activity
evoked and induced by visually presented words and nonverbal stimuli. Psychophysiology, 37(4),
447-455.
Friederici, A. D., Gunter, T. C., Hahne, A., & Mauth, K. (2004). The relative timing of syntactic and
semantic processes in sentence comprehension. Neuroreport, 15(1), 165-169.
25
Gauthier, I., Curran, T., Curby, K. M., & Collins, D. (2003). Perceptual interference supports a non-
modular account of face processing. Nat Neurosci, 6(4), 428-432.
Gauthier, I., & Nelson, C. A. (2001). The development of face expertise. Curr Opin Neurobiol, 11(2),
219-224.
Grainger, J., Bouttevin, S., Truc, C., Bastien, M., & Ziegler, J. (2003). Word superiority, pseudoword
superiority, and learning to read: a comparison of dyslexic and normal readers. Brain Lang, 87(3),
432-440.
Guttorm, T. K., Leppanen, P. H., Poikkeus, A. M., Eklund, K. M., Lyytinen, P., & Lyytinen, H. (2005).
Brain event-related potentials (ERPs) measured at birth predict later language development in children
with and without familial risk for dyslexia. Cortex, 41(3), 291-303.
Hauk, O., & Pulvermuller, F. (2004). Effects of word length and frequency on the human event-related
potential. Clin Neurophysiol, 115(5), 1090-1103.
Helenius, P., Tarkiainen, A., Cornelissen, P., Hansen, P. C., & Salmelin, R. (1999). Dissociation of normal
feature analysis and deficient processing of letter-strings in dyslexic adults. Cereb Cortex, 9(5), 476-
483.
Kim, K. H., Yoon, H. W., & Park, H. W. (2004). Spatiotemporal brain activation pattern during
word/picture perception by native Koreans. Neuroreport, 15(7), 1099-1103.
Kutas, M., & Federmeier, K. D. (2000). Electrophysiology reveals semantic memory use in language
comprehension. Trends Cogn Sci, 4(12), 463-470.
Lehmann, D., & Skrandies, W. (1980). Reference-free identification of components of checkerboard-
evoked multichannel potential fields. Electroencephalogr Clin Neurophysiol, 48(6), 609-621.
Luu, P., & Ferree, T. (2000). Determination of the Geodesic Sensor Nets' Average Electrode Positions and
their 10-10 International Equivalents.Eugene, Oregon: Electrical Geodesics, Inc.
Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T., Kulju, P., Laakso, M. L., Leiwo, M., Leppanen, P.,
Lyytinen, P., Poikkeus, A. M., Richardson, U., Torppa, M., & Viholainen, H. (2004). Early
26
development of children at familial risk for dyslexia--follow-up from birth to school age. Dyslexia,
10(3), 146-178.
Lyytinen, H., Aro, M., Eklund, K., Erskine, J., Guttorm, T., Laakso, M. L., Leppanen, P. H., Lyytinen, P.,
Poikkeus, A. M., & Torppa, M. (2004). The development of children at familial risk for dyslexia: birth
to early school age. Ann Dyslexia, 54(2), 184-220.
Maurer, U., Brandeis, D., & McCandliss, B. (2005). Fast, visual specialization for reading in English
revealed by the topography of the N170 ERP response. Behav Brain Func, 1(1), 13.
Maurer, U., Brem, S., Bucher, K., & Brandeis, D. (in press). Emerging neurophysiological specialization
for letter strings. J Cogn Neurosci.
Maurer, U., Brem, S., Kranz, F., Bucher, K., Benz, R., Halder, P., Steinhausen, H.-C., & Brandeis, D.
(submitted). Fast visual brain activation is tuned for print when children learn to read.
Maurer, U., Bucher, K., Brem, S., & Brandeis, D. (2003a). Development of the automatic mismatch
response: from frontal positivity in kindergarten children to the mismatch negativity. Clin
Neurophysiol, 114(5), 808-817.
Maurer, U., Bucher, K., Brem, S., & Brandeis, D. (2003b). Altered responses to tone and phoneme
mismatch in kindergartners at familial dyslexia risk. Neuroreport, 14(17), 2245-2250.
McCandliss, B. D., Cohen, L., & Dehaene, S. (2003). The visual word form area: expertise for reading in
the fusiform gyrus. Trends Cogn Sci, 7(7), 293-299.
McCandliss, B. D., & Noble, K. G. (2003). The development of reading impairment: a cognitive
neuroscience model. Ment Retard Dev Disabil Res Rev, 9(3), 196-204.
McCandliss, B. D., Posner, M. I., & Givon, T. (1997). Brain plasticity in learning visual words. Cogn
Psychol, 33(1), 88-110.
Michel, C. M., Murray, M. M., Lantz, G., Gonzalez, S., Spinelli, L., & Grave de Peralta, R. (2004). EEG
source imaging. Clin Neurophysiol, 115(10), 2195-2222.
27
Michel, C. M., Thut, G., Morand, S., Khateb, A., Pegna, A. J., Grave de Peralta, R., Gonzalez, S., Seeck,
M., & Landis, T. (2001). Electric source imaging of human brain functions. Brain Res Brain Res Rev,
36(2-3), 108-118.
Molfese, D. L. (2000). Predicting dyslexia at 8 years of age using neonatal brain responses. Brain Lang,
72(3), 238-245.
Naatanen, R. (2001). The perception of speech sounds by the human brain as reflected by the mismatch
negativity (MMN) and its magnetic equivalent (MMNm). Psychophysiology, 38(1), 1-21.
Naatanen, R., Tervaniemi, M., Sussman, E., Paavilainen, P., & Winkler, I. (2001). "Primitive intelligence"
in the auditory cortex. Trends Neurosci, 24(5), 283-288.
Neville, H. J., Mills, D. L., & Lawson, D. S. (1992). Fractionating language: Different neural subsystems
with different sensitive periods. Cereb Cortex, 2, 244-258.
Noble, K. G., McCandliss, B. D., & Farah, M. J. (submitted). Socioeconomic background predicts
individual differences in neurocognitive abilities.
Nobre, A. C., Allison, T., & McCarthy, G. (1998). Modulation of human extrastriate visual processing by
selective attention to colours and words. Brain, 121 (Pt 7), 1357-1368.
Pascual-Marqui, R. D., Michel, C. M., & Lehmann, D. (1995). Segmentation of brain electrical activity
into microstates: model estimation and validation. IEEE Trans Biomed Eng, 42(7), 658-665.
Petersen, S. E., Fox, P. T., Posner, M. I., Mintun, M., & Raichle, M. E. (1988). Positron emission
tomographic studies of the cortical anatomy of single- word processing. Nature, 331(6157), 585-589.
Posner, M., & McCandliss, B. D. (2000). Brain circuitry during reading. In R. Klein & P. McMullen
(Eds.), Converging methods for understanding reading and dyslexia (pp. 305-337). Cambridge: MIT
Press.
Price, C. J., Moore, C. J., Humphreys, G. W., & Wise, R. J. S. (1997). Segregating semantic from
phonological processes during reading. J Cogn Neurosci, 9(6), 727-733.
Price, C. J., Wise, R. J., & Frackowiak, R. S. (1996). Demonstrating the implicit processing of visually
presented words and pseudowords. Cereb Cortex, 6(1), 62-70.
28
Proverbio, A. M., Cok, B., & Zani, A. (2002). Electrophysiological measures of language processing in
bilinguals. J Cogn Neurosci, 14(7), 994-1017.
Proverbio, A. M., Vecchi, L., & Zani, A. (2004). From orthography to phonetics: ERP measures of
grapheme-to-phoneme conversion mechanisms in reading. J Cogn Neurosci, 16(2), 301-317.
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychol
Bull, 124(3), 372-422.
Reicher, G. M. (1969). Perceptual recognition as a function of meaninfulness of stimulus material. J Exp
Psychol, 81(2), 275-280.
Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-Z reader model of eye-movement control in
reading: comparisons to other models. Behav Brain Sci, 26(4), 445-476; discussion 477-526.
Rossion, B., Gauthier, I., Goffaux, V., Tarr, M. J., & Crommelinck, M. (2002). Expertise training with
novel objects leads to left-lateralized facelike electrophysiological responses. Psychol Sci, 13(3), 250-
257.
Rossion, B., Joyce, C. A., Cottrell, G. W., & Tarr, M. J. (2003). Early lateralization and orientation tuning
for face, word, and object processing in the visual cortex. Neuroimage.
Rugg, M. D. (1984). Event-related potentials and the phonological processing of words and non-words.
Neuropsychologia, 22(4), 435-443.
Rumsey, J. M., Horwitz, B., Donohue, B. C., Nace, K., Maisog, J. M., & Andreason, P. (1997).
Phonological and orthographic components of word recognition. A PET-rCBF study. Brain, 120 (Pt
5), 739-759.
Salmelin, R., Service, E., Kiesila, P., Uutela, K., & Salonen, O. (1996). Impaired visual word processing
in dyslexia revealed with magnetoencephalography. Ann Neurol, 40(2), 157-162.
Schendan, H. E., Ganis, G., & Kutas, M. (1998). Neurophysiological evidence for visual perceptual
categorization of words and faces within 150 ms. Psychophysiology, 35(3), 240-251.
Sereno, S. C., Brewer, C. C., & O'Donnell, P. J. (2003). Context effects in word recognition: evidence for
early interactive processing. Psychol Sci, 14(4), 328-333.
29
Sereno, S. C., Rayner, K., & Posner, M. I. (1998). Establishing a time-line of word recognition: evidence
from eye movements and event-related potentials. Neuroreport, 9(10), 2195-2200.
Shaywitz, S. E. (2004). Overcoming dyslexia: a new and complete science-based program for overcoming
reading problems at any level.New York: Alfred A. Knopf.
Simos, P. G., Breier, J. I., Fletcher, J. M., Bergman, E., & Papanicolaou, A. C. (2000). Cerebral
mechanisms involved in word reading in dyslexic children: a magnetic source imaging approach.
Cereb Cortex, 10(8), 809-816.
Tagamets, M. A., Novick, J. M., Chalmers, M. L., & Friedman, R. B. (2000). A parametric approach to
orthographic processing in the brain: an fMRI study. J Cogn Neurosci, 12(2), 281-297.
Talairach, J., & Tournoux, P. (1988). Co-planar Stereotaxic Atlas of the Human Brain.Stuttgart: Thieme.
Tanaka, J. W., & Curran, T. (2001). A neural basis for expert object recognition. Psychol Sci, 12(1), 43-
47.
Tarkiainen, A., Helenius, P., Hansen, P. C., Cornelissen, P. L., & Salmelin, R. (1999). Dynamics of letter
string perception in the human occipitotemporal cortex. Brain, 122(Pt 11), 2119-2132.
Wydell, T. N., Vuorinen, T., Helenius, P., & Salmelin, R. (2003). Neural correlates of letter-string length
and lexicality during reading in a regular orthography. J Cogn Neurosci, 15(7), 1052-1062.
Zevin, J. D., & Balota, D. A. (2000). Priming and attentional control of lexical and sublexical pathways
during naming. J Exp Psychol Learn Mem Cogn, 26(1), 121-135.
Zhang, X. L., Begleiter, H., Porjesz, B., & Litke, A. (1997). Visual object priming differs from visual
word priming: an ERP study. Electroencephalogr Clin Neurophysiol, 102(3), 200-215.
30
Figure legends
Figure 1. ERP maps and centroids.
The different distribution of the positive (indicated by a “+”) and negative (indicated by a “-”) fields in the
N170 between words and symbols is illustrated in maps seen from the front and from the back (upper
figure, left and middle). The difference maps show that the maximal negative difference appears at the left
occipito-temporal electrodes at the edge of the montage (upper figure, right). The different distribution of
the positive and negative fields between word and symbol N170 in the 129-channel maps are summarized
by the centroid locations in Talairach space (lower figure), which can be used for further statistics. The
centroids are shown from the back with their x- and z-coordinates. The additional posterior-anterior y-axis
is not depicted here.
31
151050-5-10-15
50
40
30
20
10
0
_
+
_
+
l
e
f
t
inferior
superior
r
i
g
h
t
words
symbols
mm
mm
N170 centroids
word N170 symbol N170 difference
+2
-2
0_V
+1.5
-1.5
0_V
b
a
c
k
f
r
o
n
t
