The time course of semantic and orthographic encoding in Chinese word production: An event-related potential study

BR A IN RE S E A RCH 1 2 73 ( 20 0 9 ) 9 2 –1 05

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / b r a i n r e s

Research Report

The time course of semantic and orthographic encoding in Chinese word production: An event-related potential study Qingfang Zhang a,⁎, Markus F. Damian b a

State Key Laboratory of Brain and Cognitive Science, Institute of Psychology, Chinese Academy of Sciences, China Department of Experimental Psychology, University of Bristol, UK

b

A R T I C LE I N FO

AB S T R A C T

Article history:

Previous studies have shown that access to conceptual/semantic information precedes

Accepted 18 March 2009

phonological access in alphabetic language production such as English or Dutch. The

Available online 1 April 2009

present study investigated the time course of semantic and orthographic encoding in Chinese (a non-alphabetic language) spoken word production. Participants were shown

Keywords:

pictures and carried out a dual-choice go/nogo task based on semantic information and

Word production

orthographic information. The results of the N200 (related to response inhibition) and LRP

Dual-choice go/nogo task

(related to response preparation) indicated that semantic access preceded orthographic

N200

encoding by 176–202 ms. The different patterns of the two N200 effects suggest that they

LRP

may tap into different processes. The N200 and LRP analyses also indicate that accessing the orthographic representation in speaking is likely optional and depends on specific task requirement. © 2009 Elsevier B.V. All rights reserved.

1.

Introduction

In language production, a central issue concerns the time course of access to semantic, syntactic and phonological information. Maintaining fluent production requires retrieval of these different types of information with millisecond precision. The serial account of lexical access in spoken production (e.g., Levelt et al., 1999) postulates that semantic information is retrieved first and subsequently serves as the input for phonological encoding. By contrast, the cascading view (e.g., Peterson and Savoy, 1998) assumes that semantic information is activated before phonological information, but that phonological encoding may begin before semantic encoding has been completed. According to these two accounts, semantic information must be available before phonological information. Supporting this assumption, beha-

vioral data suggest that a word's conceptual/semantic and syntactic properties are retrieved before its phonological form is available (e.g., Dell and O'Seaghdha, 1991, 1992; Levelt et al., 1991; Peterson and Savoy, 1998; Schriefers et al., 1990). Electrophysiological studies on language production in European languages using the lateralized readiness potential (LRP) and the N200 have provided converging evidence which supports the hypothesized information retrieval sequence in more detail (Rodriguez-Fornells et al., 2002; Schmitt et al., 2000; Schmitt et al., 2001; van Turennout et al., 1997, 1998). The electrophysiological studies suggested that conceptual/semantic information is encoded about 40 to 170 ms before phonological information (Rodriguez-Fornells et al., 2002; Schmitt et al., 2000; van Turennout et al., 1997), that syntactic information is analyzed about 40 ms before phonological access (van Turennout et al., 1998), and that conceptual

⁎ Corresponding author. Institute of Psychology, Chinese Academy of Sciences, Datun Road 10A, Beijing, 100101, China. Fax: +86 10 64872070. E-mail address: [email protected] (Q. Zhang). Abbreviations: LRP, lateralized readiness potential; EEG, electroencephalogram; EOG, electro-oculogram 0006-8993/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.brainres.2009.03.049

BR A I N R ES E A RC H 1 2 7 3 ( 2 00 9 ) 9 2 –1 05

processing begins approximately 80 ms earlier than syntactic processing (Schmitt et al., 2001). Previous studies have addressed the time course of semantic vs. phonological encoding, and orthographic vs. phonological encoding in Chinese spoken production (Guo et al., 2005; Zhang et al., 2007; Zhang and Yang, 2007) (see below for details). The present study therefore addressed the time course of semantic vs. orthographic encoding in spoken production. So far, existing models of language production generally do not make any explicit claims about how orthographic representations are accessed. However, a few empirical studies have recently investigated the possibility that literate speakers may mandatorily co-activate orthographic representations.

1.1. Is orthographic information activated mandatorily in spoken production? The claim that orthography may be involved in spoken production is somewhat counterintuitive as spelling codes would prima facie appear irrelevant for phonological encoding. However, in language comprehension, rather than production, there is substantial and growing evidence for the co-activation of orthographic codes (e.g., Chéreau et al., 2007; Dijkstra et al., 1995; Donnenwerth-Nolan et al., 1981; Hallé et al., 2000; Jakimik et al., 1985; Muneaux and Ziegler, 2004; Racine and Grosjean, 2005; Seidenberg and Tanenhaus, 1979; Taft and Hambly, 1985; Ventura et al., 2004; Ziegler and Ferrand, 1998, Ziegler et al., 2004). At a general level, such results suggest that the mental lexicon may be highly interconnected, with linguistic codes in different representational formats co-activating each other on lexical access. Assuming that this claim is correct, then the possibility arises that orthographic effects previously demonstrated in speech perception may also arise in spoken production. At present, empirical evidence that would speak to the issue of co-activation between phonological and orthographic codes in this domain is limited. Lupker (1982), for example, investigated the effects of phonetic and orthographic similarity between the word and the picture's name using a pictureword interference task in English. The results showed that words which were orthographically similar to the picture name (e.g., lane–plane) facilitated naming responses by 56 ms, and phonetically similar words (e.g., brain–plane) facilitated naming by approximately 20 ms, while the orthographic plus phonetic similarity condition (e.g., year–bear) led to a 55-ms facilitation in comparison to the unrelated condition. In addition, Underwood and Briggs (1984) found no priming at all from phonetic similarity with substantial priming for the orthographic plus phonetic similarity condition. Both studies may imply that orthographic information is activated in picture naming in alphabetic scripts; however, more likely such effects arise during processing of the visually presented distractor words (e.g., Damian and Bowers, in press). Lupker and Williams (1989) asked participants to name a word (the prime) which was followed by a rhyming second word (the target), and then participants had to either name or categorize the target. They found that an orthographically and phonologically related prime produce facilitation effects of about the same magnitude on naming as on categorizing. In the

93

categorization task, participants were not required to produce the rhyming target name, thus, it seems unlikely that the rhyming relation influenced the stages involved in the actual production of the word. In summary, the available research suggests that it is mainly the orthographic relation between prime and target which produced the orthographic and phonological facilitation effect. Furthermore, there is evidence that the effects of both the orthographic and the phonological relations are probably localized at the same level of name (word form) retrieval (Starreveld and La Heij, 1995), but effects arising from input processing of prime/distractor words cannot fully be excluded. Effects of orthography have also been investigated in other experimental paradigms. Gaskell et al. (2003) targeted the fact that in English, the definite article “the” is typically pronounced as “thee” when occurring before a noun starting with a vowel, and as “thuh” otherwise. Participants were auditorily presented with nouns, and were asked to shadow them together with the determiner. It was shown that the shadowed form of the determiner depended not only on the pronunciation and stress of the target word, but crucially also on its spelling. For instance, “union” was more likely to be preceded by the article “thee” than “yellow”, despite the fact that both begin with the same phoneme. Presumably this was the case because the first letter of “union”, but not of “yellow”, corresponds to a vowel. Wheeldon and Monsell (1992) investigated long-lasting priming in language production, and specifically explored whether production of a word in a study phase (e.g., in response to a definition) can facilitate production of a form-related word in a subsequent test phase (e.g., in a picture naming task). They demonstrated longlasting facilitatory effects from the production of homographic homophones (bat), but not from heterographic homophones (sun–son). The observation that long-term facilitation of responses primed by homophones depends on the presence or absence of spelling overlap underscores the potential importance of orthographic codes in spoken production. In yet a different empirical paradigm, Damian and Bowers (2003) used the so-called “implicit priming” paradigm in which a small number of responses, typically elicited by prompt words, is produced repeatedly within an experimental block, and the presence or absence of form overlap between the responses is manipulated. They replicated previous studies (e.g., Meyer, 1990) in showing a reliable priming effect in the homogenous condition in which all response words shared initial sound and spelling (e.g., “camel”–“coffee”–“cushion”), compared to a heterogeneous condition in which this was not the case (e.g., “camel”–“gypsy”–“cushion”). No such priming effect was obtained in an inconsistent condition, however, in which all response words shared the initial sound, but differed in spelling (e.g., “camel”–“kayak”–“kidney”). Hence, although information about spelling is irrelevant to the speaking process, when retrieving the phonological codes of the response words speakers apparently co-activate orthographic codes. Other studies, however, have recently suggested that the hypothesized effects of orthography in spoken production are not automatic. Subsequent studies with the paradigm used by Damian and Bowers (2003), but conducted with Dutch (Roelofs, 2006; see also Schiller, 2007) and French (Alario et al., 2007) speakers, failed to replicate the originally reported effect.

94

BR A IN RE S E A RCH 1 2 73 ( 20 0 9 ) 9 2 –1 05

These difficulties in replicating an effect of orthography certainly warrant caution about the claim that phonological encoding in spoken production automatically entails crossactivation of orthographic representational formats. Recently the argument has been raised that automatic orthographic effects in spoken production may be stronger in languages with deep than with shallow orthographies. The argument hinges on the observation that effects from English speakers originally reported by Damian and Bowers (2003) did not replicate easily in a language such as Dutch which has a shallower orthographic transparency. Roelofs (2006) pointed out the possibility that “cross-linguistic differences in the degree to which orthography and phonology interact in speech production (are) perhaps related to differences in orthographic depth between languages”. An interesting property of the Chinese orthographic system is that it is very “deep”, i.e., the correspondences between spelling and sound are weak. Hence, studies conducted on Chinese speakers may be particularly likely to demonstrate automatic orthographic effects. The orthographic system of Chinese consists of a number of levels: strokes, radicals, characters, and words. A character is composed of one or more radicals, which, in turn, are composed of one or more strokes. Modern Chinese characters can be broadly differentiated into two categories (Li, 1993): simple and complex. Most complex characters consist of a semantic radical on the left and a phonetic radical on the right. For example, 柏 (/bai3/, cypress) is composed of a semantic radical (木, /mu1/, wood) on the left, and a phonetic radical (白, /bai2/, white) on the right. Such a Chinese character can be characterized as having a left–right structure. A radical can appear in different positions within a complex character. For example, 白 is at the top of 皂 (/zao4/, soap), at the bottom of 皆 (/jie1/, all), to the left of 皑 (/ai2/, pure white), and to the right of 柏 (adopted from Ding et al., 2004). Taft et al. (1999) proposed that the orthographic representation system of Chinese consists of three processing levels: low-level features, radicals, and characters. A visually presented word enters the orthographic system at the processing level of low-level features (i.e., strokes and stroke combinations and relationships). Then, activation passes up to the radical units level associated with the activated features (i.e., radicals and radical combinations and relationships), and this in turn passes up to the character units level. In the study reported below, we assess the time course of access to semantic information about an object name, and contrast it with the time course for access to orthographic representation. Specifically, we will assess electrophysiological markers – outlined below – for each type of code regarding their time of availability. A semantic decision task was used to tap into the stage of conceptual access during language production, whereas an orthographic decision task was used to investigate access to spelling properties of object names. In the literature on speech perception, tasks which require monitoring for phonological properties are in wide-spread use (see Connine and Titone, 1996, for review), and similar tasks are occasionally used in speech production, for instance by asking participants to categorize whether or not a specific target phoneme is part of a picture's name (i.e., /b/ or not; e.g., Özdemir et al., 2007), or to categorize the initial phoneme (vowel or not; e.g., Heim et al., 2003). Only in one previous

study an attempt has been made to adapt such a task to the monitoring of graphemic, rather than phonological, properties with electrophysiological measures in German (Hauk et al., 2001). In the present study, participants were asked to categorize picture names according to a specific orthographic property of the Chinese writing system, namely whether or not the picture's written name consisted of a left–right structure character. According to current accounts of Chinese orthography (see above), such a radical relationship judgment task requires access to mid-level representations (i.e., more complex than orthographic features, but at the same time clearly sublexical in nature). Evidence from a range of experimental paradigms suggests that reading a complex character involves the processing of its radicals and their positional relationship (Ding et al., 2004; Taft et al., 2000, 1999), hence a character's structural information is an important element of the mental orthographic representation. And indeed, participants generally find judgments concerning the structural properties of Chinese characters quite natural and straightforward (i.e., Chen et al., 1995; Perfetti and Zhang, 1991; Shen and Forster, 1999; Tsang and Chen, in press; Wong and Chen, 1999; Zhou and Marslen-Wilson, 1999; Zhou et al., 1999).

1.2. Electrophysiological markers of the time course of spoken production The relative time course of semantic vs. orthographic encoding in Chinese word production was investigated by using the N200 and LRP components, which in previous studies have been shown to be informative in constraining theoretical models (e.g., Levelt et al., 1999; Schiller, 2006). The N200 is a negative-going waveform. In a Go/noGo task, participants are asked to respond to one type of stimuli, and to withhold response for another type. Compared to the waveform on the Go trials, an ERP component, namely N200, is typically observed on the noGo trials. It is visible at a fronto-central region occurring between 100 and 300 ms after stimulus onset (Jodo and Kayama, 1992; Sasaki et al., 1993; Simson et al., 1977). It has been suggested that the amplitude of the N200 is a function of neural activity required for response inhibition (Jodo and Kayama, 1992; Sasaki and Gemba, 1993). Hence, the emergence of N200 suggests that the information which is used to determine whether or not a response is to be given must have been encoded. The peak latency of the N200 can therefore be used to determine the moment in time at which this information has become available. Note that the N200 occurs later in time when it is related to language processing (see Kutas and Schmitt, 2003). As is the case for the N200, the LRP (lateralized readiness potential) has been proved a sensitive index for estimating the timing of information processing (e.g., Rodriguez-Fornells et al., 2002; Schmitt et al., 2000; van Turennout et al., 1997). It is derived from the readiness potential, which is a slow, negative-going potential that starts to develop about 0.5 s before the execution of a voluntary hand movement and reaches its maximum just after the response is initiated. A series of studies (i.e., Kutas and Donchin, 1980) have shown that the LRP can be used as an index for exact response preparation prior to an overt Go response, or in the absence of

BR A I N R ES E A RC H 1 2 7 3 ( 2 00 9 ) 9 2 –1 05

any overt response. The LRP is computed according to the following equation: LRP = right hand ðC3
C4Þ
left hand ðC3
C4Þ: The electrode sites C3 and C4 are located above the left and the right motor cortices, respectively. The LRP yields largest amplitudes at the motor cortices contralateral to the response hand.

1.3.

Experimental paradigm

The present study used a dual-choice Go/noGo task adopted from pioneering work by van Turennout et al. (1997, 1998). Participants were asked to perform a dual decision along two dimensions, semantics and orthography. The semantic task was to determine whether the picture depicted a living or a non-living object, while the orthographic task was to decide whether or not the picture's written name consisted of a left– right structure character. In the experiment, the instruction was, for example, “Press the left button if the picture's name is a left–right structure character (i.e., 狗, /gou3/, dog), and the right button if the picture's name is not a left–right structure character (i.e., 鱼, /yu2/, fish). However, respond only if the picture depicts a living object, and not if the picture depicts a non-living object.” Thus, depending on the semantic and orthographic characteristics, a response was given with either the left or right hand, or no response at all. The experimental design consisted of two conditions. In one half of the experiment, the responding hand was contingent on semantic information (hand = semantics) and the Go/noGo decision on orthographic information (Go/ noGo = orthography). In the other condition, the response contingencies were reversed, i.e., the responding hand was contingent on orthographic information (hand = orthography) and the Go/noGo decision on semantic information (Go/ noGo = semantics). The logic of the experimental paradigm related to the N200 was as follows: in the Go/noGo = semantics condition, the peak latency of the N200 would provide an upper limit in time at which semantic information became available in order to determine whether or not to respond. In the Go/noGo = orthography condition, the peak latency of the N200 would provide an upper limit in time at which orthographic information must be retrieved for deciding to respond or not. According to the serial model or the cascading model, semantic information is encoded prior to phonological information. Furthermore, previous studies indicate that phonological and orthographic information may be co-activated simultaneously. It was then expected that the peak latency of the N200 would appear earlier in the Go/noGo = semantics condition than in the Go/noGo = orthography condition. The logic of LRP, adopted from van Turennout et al. (1997) and subsequent studies, is as follows: if information associated with the response hand dimension (e.g., semantics) is available before information associated with the Go/noGo dimension (e.g., orthography), then on noGo trials, the negative deflection typical of the LRP should be observed, which reflects an – eventually abandoned – response preparation. Crucially, given the assumptions about relative timing of the two dimensions, an LRP should not be observed on the

95

noGo trials when the response dimensions are reversed (i.e., hand = orthography; Go/noGo = semantics). An LRP should be observed on the Go trials in both conditions. The task used here required participant to monitor for semantic and orthographic characteristics during spoken production, which involves access to semantic or orthographic information plus a monitoring judgement. Although results from the dual-choice Go/noGo task could be challenged on the basis that is does not constitute a “pure” instance of language production, various aspects of language production have been examined with this technique in a number of previous studies (Rodriguez-Fornells et al., 2002; Schmitt et al., 2000, 2001; van Turennout et al., 1997). In our procedure we trained participants to name pictures in the practice session before the main session, and the results indicated that they used the intended picture names. However, the assumption that overt and covert spoken production involve very similar cognitive and cortical processes receives somewhat mixed support in the literature. For instance, in an fMRI study Palmer et al. (2001) demonstrated that regions active during overt speaking were similar to those active during covert task performance, with the addition of certain regions typically associated with motor aspects of spoken production. On the other hand, Christoffels et al. (2007) compared overt and covert picture naming in Dutch and found – among other differences – different activation patterns in the insula during overt and covert naming.

2.

Results

2.1.

Pre-experiment

The aim of this pre-test was to measure accuracy and reaction times on the target stimuli for simple semantic and orthographic decision tasks, respectively (i.e., without the dual-task paradigm). The aim was to ensure that both tasks were roughly comparable in overall difficulty, and to provide a rough indicator of how long it takes in general to make a simple semantic decision (e.g., if a picture depicted a living object, press the left button; if a non-living object, press the right button) or orthographic decision (e.g., if a picture name is a left–right structure character, press the left button; if not, press the right button). Thirteen native Chinese speakers participated in the pre-test, and they did not attend the following main experimental session. One participant was excluded due to error rates larger than 25%. Errors and reaction times deviating more than 3 standard deviations (SD) were excluded from the data analysis. For the semantic decision, the mean reaction time was 642 ms (SD = 98 ms) with a mean error rate of 6.85%. For the orthographic decision, the mean reaction time was 1312 ms (SD = 123 ms), with a mean error rate of 5.59%. A paired t test showed that the two conditions differed significantly from each other on reaction times (t(11) = 15.35, p < 0.001), but the difference in error rates was not significant (t(11) = 1.99, p = 0.072. Accuracy measures did not differ significantly across the two tasks, suggesting that they were roughly matched on overall difficulty. Response latencies were substantially slower in the orthographic than in the semantic task. On a

96

BR A IN RE S E A RCH 1 2 73 ( 20 0 9 ) 9 2 –1 05

general level, the latency pattern mirrors previous results which had compared semantic to form judgment tasks. For instance, Schmitt et al. (2000) investigated the time course of semantic and phonological encoding with a dual-choice Go/ noGo task. Mean reaction times were 617 ms for simple semantic decisions, and 841 ms for simple phonological decision; the 224-ms difference was significant. Heim et al. (2003) investigated phonological processing during language production with the fMRI method. Participants were asked to perform either a semantic decision task (SEM: natural or manmade?) or one of two phonological decision tasks on the initial phoneme of a picture name (Phon1: /b/ or not?; Phon2: vowel or not?). Mean reaction times were 646 ms for the semantic task, 812 ms for the Phon1 task, and 1015 ms for the Phon2 task, with differences significant between all tasks. Both Schmitt et al. and Heim et al. suggested that the pattern of longer reaction times for phonological relative to semantic processing during language production indicated that the simple choice based on semantics could be carried out faster than the choice based on phonology. At the same time, they argued that, based on the absence of significant differences in error rates, the tasks were matched on overall difficulty. Following this logic we likewise suggest that the semantic and orthographic decision tasks used in the present study are roughly comparable in terms of difficulty. At the same time, it

is clearly the case that simple semantic judgments can be carried out substantially faster than orthographic judgments. One should note that the mean error rates on semantic decisions were slightly higher than on orthographic decisions, indicating a potential speed-accuracy trade-off. Nevertheless, the discrepancy between the two types of tasks in the pre-test latencies is considerable. A possible inference from this pattern, which is supported by the electrophysiological results from the main experiment (see below), is that access to orthographic properties does not constitute one of the core properties of language production, and because it is “optional” and associated with additional processing, simple decision times take longer than on semantic judgments.

2.2.

Push-button reaction times

Incorrect responses and reaction times longer than 2000 ms were excluded from the behavioral and ERP data analysis. The mean reaction times for correct go responses were averaged across left and right responses for 16 participants. The mean reaction time was 1102 ms (SD = 142 ms) and 1187 ms (SD = 146 ms) for the Go/noGo = semantics condition and orthography condition, respectively. A paired t test of the reaction times was significant (t(15) = 4.35, p < 0.01) for the two contingency conditions. The error rates were not significantly

Fig. 1 – Grand average waveforms for Go and noGo trials in Go/noGo = semantics condition over nine electrode sites (F3, Fz, F4, FC3, FCz, FC4, C3, Cz, and C4).

BR A I N R ES E A RC H 1 2 7 3 ( 2 00 9 ) 9 2 –1 05

different between the two contingency conditions: hand = semantics: 10.0%, hand = orthography: 12.5%, t(15) = 2.21, p < 0.1.

2.3.

N200 analysis

One participant was excluded from all further analyses due to large electrophysiological drift. Two participants were excluded from the N200 analysis due to no significant amplitude difference between Go and noGo average waves generated in both contingency conditions. The electrophysiological signals were averaged separately for the Go and noGo trials in the two contingency conditions (Go/noGo = semantics vs. orthography), respectively. The N200 was obtained by subtracting the waveforms on the noGo trials from those on the Go trials in each Go/noGo contingency condition (semantics vs. orthography). The maximum number of trials per condition per individual was 128. On average, the number of accepted trials was 123 in the Go = semantics (hand = orthography) and 119 in the Go = orthography (hand = semantics) condition. The minimum number of accepted trials was 117 in the Go = semantics and 100 in the Go = orthography condition. Figs. 1 and 2 show grand average ERP waveforms on the Go and noGo trials in the Go/noGo = semantics and orthography conditions for 13 participants at 9 fronto-central electrode sites (F3, Fz, F4, FC3, FCz, FC4, C3, Cz, and C4). Both conditions

97

showed clear evidence of N200 effects. Waveforms on the noGo trials were more negative than on the Go trials, and this tendency occurred visibly earlier in the Go/noGo = semantics than in the Go/noGo = orthography condition. Fig. 3 shows the grand average of difference waves (noGo minus go) for the Go/noGo = semantics and orthography conditions at 9 electrode sites (the same locations as Figs. 1 and 2). It is clear that the onset and peak latencies of the two N200 effects were distinctly different between two contingency conditions. The ERP difference waves were compared by mean amplitude measures relative to the pre-stimulus baseline (− 200 to 0 ms before picture onset). The statistical comparison of the ERP difference waveforms for the two conditions has been limited to the frontal–central midline sites as the N200 is usually most clearly visible on these electrodes (Schmitt et al., 2000; Schiller, 2006). The peak latencies of the N200 effect were measured between 200 and 800 ms after picture onset at Fz, FCz and Cz for each participant. A repeated-measures ANOVA analysis was computed on N200 peak latencies and peak amplitudes with two within-participants variables: contingency condition (Go/noGo = semantics vs. orthography) and electrodes (Fz, FCz, and Cz). Greenhouse–Geisser correction was used when appropriate. For the peak latencies analyses, the main effect of contingency conditions was significant (F(1, 12) = 17.03,

Fig. 2 – Grand average waveforms for Go and noGo trials in Go/noGo = orthography condition over nine electrode sites (F3, Fz, F4, FC3, FCz, FC4, C3, Cz, and C4).

98

BR A IN RE S E A RCH 1 2 73 ( 20 0 9 ) 9 2 –1 05

Fig. 3 – Grand average difference waveforms (nogo minus go) for Go/noGo contingency on semantics and orthography conditions. Semantics and orthography condition over nine electrode sites (F3, Fz, F4, FC3, FCz, FC4, C3, Cz, and C4).

p < 0.01), reflecting a sharp difference on peak latencies (Go/ noGo = semantics: M = 373 ms, SD = 125 ms; orthography: M = 541 ms, SD = 131 ms). Neither the main effect of electrode sites nor the interaction between contingency conditions and electrode sites was significant. For the peak amplitudes analyses, the main effects of contingency conditions and electrode sites, and the interaction between contingency

conditions and electrode sites were not significant. The mean amplitude difference (across three electrode sites) of the two N200 effects was very small (0.29 μν). Fig. 4 shows the scalp distributions of the N200 effects for the Go/noGo = semantics condition (mean amplitude of the time window 330– 380 ms after picture onset) and for the Go/noGo = orthography condition (mean amplitude of the time window 510–560 ms),

Fig. 4 – Scalp distribution of the N200 effects for the Go/noGo = semantics condition (mean amplitudes of the time window 330–380 ms) and the Go/noGo = orthography condition (mean amplitudes of the time window 510–560 ms).

BR A I N R ES E A RC H 1 2 7 3 ( 2 00 9 ) 9 2 –1 05

respectively. It is observed that the N200 effects have a similar scalp distribution in both conditions, i.e. it is generated by the same neural populations.

2.4.

LRP analysis

Four stimulus-locked LRP average waveforms for each participant were calculated: (1) hand = semantics, Go = orthography, (2) hand = semantics, noGo = orthography, (3) hand = orthography, Go = semantics, (4) hand = orthography, noGo = semantics. The maximum number of trials per condition per individual was 64. On average, the number of accepted trials was 55 for Go LRP and 60 for noGo LRP in the hand = semantics condition, and 58 for Go LRP and 62 for noGo LRP in the hand = orthography condition. There was no significant difference on the number of accepted trials between Go and noGo LRPs in both conditions. The minimum number of accepted trials for Go and noGo LRPs in the hand = semantics condition was 44 and 50, respectively, and in the hand = orthography was 48 and 50, respectively. Fig. 5 shows the grand average of Go and noGo LRPs for the response contingency conditions: hand = semantics (top panel), and hand = orthography (bottom panel) at the motor cortex sites (C3 and C4). The typical LRP patterns for Go responses were obtained under both the hand = semantics and hand = orthography conditions. The LRPs were measured by mean amplitudes relative to the pre-stimulus baseline (−200 to 0 ms before picture onset). The onset latency of the LRP was measured by one-tailed serial t tests between 300 and 1000 ms after picture onset against a zero mean. The t tests were carried out stepwise with a step size of 2 ms. As in the analysis on the N200, the onset latency of the LRP was defined as the point at which four consecutive t tests yielded significant results (in the same direction). The mean onset latency for the Go LRP in the hand = semantics condition was 600 ms (from that time all t(14) < −1.807 all ps< 0.05, one-tailed), while Go LRP in the hand = orthography

Fig. 5 – Grand average of Go and noGo LRP in the hand = semantics (top panel) and the hand = orthography (bottom panel) condition.

99

condition was 592 ms (from that time all t(14) < −1.813, all ps < 0.05, one-tailed). When compared via a paired sample t test, the two Go LRPs' onset latencies were not significantly different. More important, the LRP in the hand = semantics developed for the noGo trials, and t test revealed that it was significantly divergent from baseline ranging from 514 ms to 622 ms (in this interval all ts(14)< −1.901, all ps< 0.05, one-tailed) after picture onset. When compared via a paired sample t test, the noGo LRP deviated significantly from the Go LRP starting at 716 ms after picture onset (from that time all ts(14) > 1.929, all ps < 0.05, onetailed) in the hand = semantics condition. No significant difference from baseline was obtained for the noGo LRP in the hand = orthography condition.

3.

Discussion

The present study investigated the time course of semantic and orthographic encoding in spoken Chinese word production with high temporal resolution of electrophysiological measures. The N200 and LRP were used to estimate the moment in time at which semantic and orthographic information was retrieved in order to determine whether or not a response has to be given. The results clearly indicate that semantic information was encoded prior to orthographic information in Chinese spoken production.

3.1.

N200

The N200 analysis is based on the assumption that increased negativity on the noGo trials in comparison with the Go trials reflects the moment at which relevant information is available in order to withhold a response (Gemba and Sasaki, 1989; Thorpe et al., 1996). The mean peak latency of the N200 was 176 ms earlier in the Go/noGo = semantics (362 ms) condition than in the Go/noGo = orthography (538 ms), implying that semantic information was available ahead of orthographic information by approximately that amount of time. The pre-test results showed that there was a major and significant difference on reaction times, but no significant difference on error rates, the latter result indicating similar difficulty between the two simple decision tasks. It has been suggested that the magnitude of the N200 is a function of the neural activity required for response inhibition (Jodo and Kayama, 1992), and that it is sensitive to task difficulty not only in monkeys (Gemba and Sasaki, 1989) but also in humans with a priming task (Kopp et al., 1996). Thus, the absence of a significant difference between the two N200 effects on peak amplitude in the current study also suggests comparable difficulty of semantic and orthographic decisions. Hence, the difference between the two N200 effects on peak latencies is unlikely to be attributable to semantic decisions being easier than orthographic ones; instead it suggests that semantic information was encoded prior to orthographic information in picture naming. From Figs. 1 and 2, it is observed that the waveforms of the two N200 effects were substantially different for the two response contingencies. The difference in the Go/noGo = semantics condition is most pronounced in the negativity of

100

BR A IN RE S E A RCH 1 2 73 ( 20 0 9 ) 9 2 –1 05

the signal. These characteristics are consistent with previous findings (Rodriguez-Fornells et al., 2002; Schmitt et al., 2000), hence we believe that the peak latency (362 ms) of the N200 in the Go/noGo = semantics condition provides an upper limit on the time at which semantic information became available. This time point (362 ms after picture onset) is somewhat later than the time window of lexical selection in picture naming for the “lemma” (i.e., semantic–syntactic) access stage (0– 275 ms) suggested by Indefrey and Levelt (2004). However, the peak latency of the N200 may be delayed in a dual-choice go/ nogo task due to the additional cognitive load associated with this task (Schmitt et al., 2001); furthermore it appears generally delayed in language-based, compared to nonlinguistic, tasks (see Kutas and Schmitt, 2003). In contrast, the N200 effect in the Go/noGo = orthography condition presents a different pattern (see Fig. 2): The difference is most pronounced in the later positivity of the signal. It should be noted that for the N200 effect reported in Schmitt et al. (2001), the maximum difference between Go and noGo trials similarly appeared on the positive waveforms when go/noGo decisions were based on conceptual representations. Furthermore, the difference waveforms in language-related task commonly present different patterns in different studies (i.e., Schiller et al., 2003), but these differences have to date not yet been explicitly addressed. The peak latency of the N200 may provide an upper limit on the time course of orthographic encoding. However, because Chinese orthography is not an alphabetic system, access to orthographic codes in spoken production, if automatic, is likely to be more holistic than in languages with alphabetic scripts. It is possible that the present orthographic task necessitates an additional visual judgment after accessing orthography. Participants may visualize the character (which can be interpreted as accessing orthography), followed by a spatial radical-relation judgment on the imagined orthographic code. The measured cognitive processing time hence possibly involves access to the orthographic symbol plus a spatial-relation judgment task, thus the moment in time at which the N200 emerged was delayed. Furthermore, the waveforms for Go and noGo responses are reversed in the two response contingencies at the later positive component (see Fig. 1 and Fig. 2). These differences suggest that the difference waveform in the Go/noGo = orthography condition may be a potentially different component from the one in the Go/noGo = semantics condition. On the other hand, Abdel Rahman and Sommer (2002) proposed that the N200 may indicate the termination of specific information in a dual-choice Go/noGo task, but not related to the relative timing of the beginning of these processes. If these components speak to termination, we cannot exclude the interpretation that the two types of information are actually assessed independently from each other. It is possible that semantic retrievals starts simultaneously with orthographical encoding but terminates earlier. Hence it is not clear whether the observed N200 pattern speaks to availability or rather to differences in termination of a process. Based on the above discussion – the N200 effect indicating either information availability or termination – it seems reasonable to view the peak latency of the N200 effect as providing an upper limit on the time course of semantic and

orthographic encoding during implicit naming. At minimum, our data showed that a decision based on semantic information is processed faster than the one based on orthographic information.

3.2.

LRP

The results showed that a noGo LRP developed from 514 ms to 622 ms after picture onset in the hand = semantics condition, which indicated that the noGo LRP lasted approximately 108 ms. More importantly, the Go LRP started to deviate from the noGo LRP at 716 ms after picture onset in the hand = semantics condition, which suggests that orthographic information enters the processing system at this moment in time. There was a 202-ms interval from the onset latency of the noGo LRP (514 ms) to the moment in time at which the Go and noGo LRPs deviated (716 ms). We suggest that during this interval, semantic information served to prepare one of the hands to respond, unaffected by orthographic information, which apparently was not yet available for the Go/noGo decision. In addition, the absence of LRP on the noGo trials in the hand = orthography condition indicates that orthographic information did not affect response preparation on these trials. When the Go/noGo decision was based on semantic information, orthographic information started to activate the response hand only after the Go/noGo decision had been made. The LRP data indicated that orthographic information enters the processing system at 716 ms after picture onset. This moment in time is quite late if one assumes that orthographic information is an automatic component of picture naming. As is the case for the N200 effect in the Go/ noGo = orthography condition, this finding may indicate that the present orthographic task adds an additional visual judgment after accessing orthographical representations which causes this delay. Another possibility is that orthographic information, the retrieval of which is not strictly necessary in spoken production, may be accessed only after phonological encoding in picture naming has been completed. Some recent studies (Alario et al., 2007; Roelofs, 2006; Schiller, 2007) have indicated that orthography is not mandatorily activated in speaking, hence, it is possible that requiring participants to monitor orthographic information in an implicit picture naming task is to some extent artificial. It should be noted that the onset latency of the LRP on noGo trials (514 ms) was earlier than on Go trials (600 ms) in the hand = semantics. This discrepancy replicates some previous studies (Schmitt et al., 2000, 2001; Rodriguez-Fornells et al., 2002). For instance, Rodriguez-Fornells et al. (2002) adopted a dual-choice Go/noGo task to investigate semantic and phonological encoding in speaking. They also obtained a significant difference between onset latency on the noGo (320 ms) and Go (448 ms) trials in the hand = semantics condition. According to the logic underlying the LRP approach, onset latency on the Go and noGo trials should be relatively similar. At present, we have no clear account of why onset latencies on Go and noGo trials will on some occasions deviate from each other. The N200 and LRP data revealed a different time course of semantic and orthographic encoding in the current study.

BR A I N R ES E A RC H 1 2 7 3 ( 2 00 9 ) 9 2 –1 05

The onset of the nogo-LRP (514 ms) was around 150 ms later than the peak latency of the N200 (362 ms) in the hand = semantics condition. The deviation time of the go LRP from the nogo-LRP was 716 ms after picture onset in the hand = orthography condition, which was around 170 ms later than the peak latency of the N200 (538 ms). In the literature, N200 is assumed to reflect response inhibition, while LRP is taken to reflect response preparation. N200 and LRP are considered to be independent ERP components, where components are defined in terms of their neural generators (Schmitt et al., 2000). Fuster (1997) argued that the retrieved information is evaluated by the frontal cortex, which is reflected in the N200 effect. The evaluated information is then sent to the motor cortex, which emerges in the LRP. According to Fuster's assumption, the peak latency of the N200 may appear earlier than the onset of LRP in the same contingency condition. This is what we observed here. Furthermore, the time interval between N200 and LRP was similar in both contingency conditions in the present study. However, the exact relationship between N200 and LRP is still unclear and clearly deserves further investigation (Schmitt et al., 2000). In related work, Zhang and Yang (2007) investigated the time course of semantic and phonological encoding in Chinese with the same paradigm as used in the present work, but with task dimensions along semantics and phonology. The same semantic task as the present was used, while the phonological task was to determine whether the depicted object's name was of one type (tone 1 or tone 2) or the other type (tone 3 or tone 4). Results indicated that semantic encoding occurred 152–181 ms earlier than phonological (tonal) encoding. Furthermore, Zhang et al. (2007) examined the relative time course of tonal and orthographic encoding in Chinese with the same paradigm. The tonal decision task was identical to Zhang and Yang's (2007) study, and the orthographic decision task was the same as in the present study. They found that tonal information was retrieved prior to orthographic information, and the results were taken to imply that orthographic codes are unlikely to contribute to phonological encoding in spoken word production. That is, access to orthographic information is not mandatory for speaking, even in Chinese with “deep” mapping between phonology and orthography. When participants were asked to complete a dual-choice Go/noGo task along both tonal and orthographic dimensions, and because tonal, but not orthographic, information retrieval is necessary in picture naming, tonal information was activated mandatorily and quickly when a picture was presented. By contrast, orthographic information was retrieved later according to experimental requirements, which required additional cognitive sources and took more time to access. Because semantic information retrieval takes less time than retrieval of tonal or orthographic information, the cognitive load in the dual-choice go/nogo task along either semantic and orthographic dimensions or semantic and tonal dimensions was less than in the task along tonal and orthographic dimensions. Therefore, the moment in time of orthographic retrieval in the present study may be similar to retrieval of tonal information in the task along semantic and tonal dimensions (Zhang and Yang, 2007), but later than the tonal information in the task along tonal and orthographic

101

dimensions (Zhang et al., 2007). Based on the above discussions, we suggest that orthographic information retrieval is optional and dependent on task requirements. As noted above, some recent studies failed to find orthographic activation in word production and provide some evidence for this interpretation (Alario et al., 2007; Roelofs, 2006; Schiller, 2007). Note that caution should be exerted in interpreting comparisons between two separate experiments. Because different participant groups might contribute to differences between two experiments, the comparisons outlined above cannot provide exact quantitative estimations of tonal and orthographic processing times (see also van Turennout et al., 1997). In addition, it is worth pointing out that the current line of work, together with a sizeable number of previous articles which used a similar method, crucially hinges on bimanual responses targeting a particular linguistic or conceptual aspect of the depicted object. All such manual tasks are to some extent “meta-representational” in that judgments on the target properties are not a necessary component of spoken production. Hence the measured cognitive processing times likely reflect access to semantic, phonological or orthographic representations, plus a meta-cognitive decision component. Although this fact does not invalidate the use of such methods to provide insight into the representational characteristics of the targets, the added decision component may delay the measured moment in time at which the underlying representational information became available. Specifically with regard to the orthographic judgment we used in our experiment, it could be argued that the task may to a certain extent be arbitary, and hence decision latencies may have been delayed. For studies conducted on Chinese speakers, the reported overall latencies on the form-related pre-test are certainly in line with previous work. Zhang et al. (2007) used the same orthographic task, and the preexperiment showed an average latency of 1314 ms, virtually identical to the 1312 ms reported in the present manuscript. In the earlier work, orthographic were compared to phonological judgments, and importantly, latencies on this task were again on a comparable level (1318 ms), suggesting that the orthographic judgments are not particularly slow, compared to phonological judgments in the earlier work in Chinese. As mentioned in the introduction, we are aware of only a single previous study which has used an orthographic monitoring task with speakers of an alphabetic language. Hauk et al. (2001) investigated grapheme monitoring on picture names (e.g., is the target letter “d” in “hand”?) with electrophysiological measures, and response latencies on this task were approximately 1050 ms (estimated from their Fig. 2). Damian and Singh (in preparation) recently carried out a similar experiment to investigate orthographic encoding in English adapted a similar task, and found overall grapheme monitoring latencies of 1052 ms. This compares to 1312 ms in the orthographic radical relationship monitoring task reported in our current manuscript. Based on this comparison we believe it is unlikely that there is a fundamental difference between monitoring for graphemes, and for orthographic radicals, in terms of underlying processes and mechanisms. The remaining difference in overall latencies is likely attributable to the fact that Chinese

102

BR A IN RE S E A RCH 1 2 73 ( 20 0 9 ) 9 2 –1 05

orthographic symbols are overall more complex than alphabetic ones. Is it possible that form decisions (phonological or orthographic) are relatively delayed in Chinese, compared to alphabetic languages? To this purpose, we compared latencies between phonological decision in alphabetic languages and orthographic decision in the current study. Schmitt et al. (2000) compared access to semantic to phonological properties of object names in a similar paradigm, and reported for the phonological task (is the word-initial phoneme of an object name a vowel or a consonant?) pre-test latencies of 841 ms, hence substantially faster than the 1318 ms in the tonal judgment task reported in Zhang et al. (2007). Of course, overall difficulty of the form judgment tasks may also have played a role: Schmitt et al.'s task required participants to dichotomise word-initial sounds into vowels or consonants, whereas Zhang et al.'s task required them to categorize four tones into two response categories. Heim et al. (2003) reported that latencies for a simple word-initial phoneme decision (e.g., /b/ or not) task was 812 ms, and for a simple word-initial phoneme categorization (e.g., vowel or consonant) was 1015 ms. Compared to the latter result, the average latencies of 1312 ms for the orthographic task in the current study do not appear particularly slow. Alternatively, rather than comparing the absolute time it takes to access form properties in Chinese vs. Western languages, one could compare the relative availability of semantic vs. form properties in each language group. A few studies have recently addressed the time course of semantic vs. phonological or orthographic activation in Chinese with electrophysiological methods. Zhang and Yang (2007) investigated the time course of semantic and tonal encoding in Chinese, and found that tonal information is retrieved later than semantic information by 152–181 ms. Guo et al. (2005) investigated the time course of semantic and onset encoding in Chinese, and found that semantic encoding occurred 140 ms earlier than onset encoding. In alphabetic languages, the interval from semantics to phonology varies somewhat across different studies: 80–90 ms in Schmitt et al. (2000) in English, 170 ms in Rodriguez-Fornells et al. (2002) in German, 80–120 ms in van Turennout et al. (1997) in Dutch. Because different decision tasks were used in the various studies, it is again difficult to compare these figures directly. Hence, we are at present not able to determine conclusively whether or not form decisions are relatively delayed in Chinese, compared to Western languages, and we acknowledge the urgent need for further research on these topics. Overall, the available evidence certainly does not exclude the possibility that form access in Chinese is overall slower than in Western languages. A further aspect with regard to access to orthographic properties is the possibility that the relatively long latencies in our orthographic task are partially attributable to the fact that Chinese orthography is not alphabetic, hence access to orthographic codes in spoken production, if automatic, is likely to be more holistic than in languages with alphabetic scripts. In other words, in alphabetic languages, orthographic properties of object names may become incrementally available due to a sublexical phoneme-to-grapheme conversion; by contrast, in Chinese access to the visual word form is likely all-

or-none. As a consequence, the time course indicators of orthographic access that we document are plausibly emerging at a later point in time than those that may be found in languages with alphabetic scripts. This possibility is again difficult to assess because related studies are often not entirely comparable. In conclusion, the N200 (related to response inhibition) and LRP components were used to investigate the time course of semantic and orthographic encoding in Chinese spoken production with a dual-choice go/nogo task. The results of the N200 and LRP indicate that semantic encoding occurred earlier than orthographic encoding by 176–202 ms. Due to the complexity of orthographic decision, it is not impossible that the N200 effects in two contingency conditions tap into different processes. Both the N200 and LRP analyses suggest that accessing orthographic representation is likely optional and depends on specific task requirements.

4.

Experimental procedures

4.1.

Participants

Sixteen native Mandarin Chinese speakers participated in the experiment (7 women and 9 men, at the age of 18–23 years, with the mean age of 20.4 years). Fifteen participants were right-handed, one participant was left-handed. All participants were neurologically healthy, with normal or correctedto-normal vision and normal hearing. They were paid for their participation.

4.2.

Stimuli

Seventy-two target pictures with names corresponding to monosyllabic Chinese characters were selected from a picture database in Mandarin (Zhang and Yang, 2003). The stimuli included two semantic categories: half of them depicted living objects, and half of them depicted non-living objects. Each of the two groups of pictures contained 18 pictures with left– right character names and the other 18 pictures with non left– right character names. Eight pictures were used as the practice stimuli, the remaining 64 pictures as the experimental stimuli (see Appendix A). Based on normative information provided by the Beijing Institute of Language (1986), the mean frequency of picture names in written form is 380 per million.

4.3.

Design

Each participant received eight different sets. In four sets, the responding hand was contingent on semantic information (Hand = semantics) and the Go/noGo decision was contingent on orthographic information (Go/noGo = orthography). In the other four sets, response contingencies were reversed (Hand = orthography, Go/noGo = semantics). The left- and righthand assignment, as well as the go and nogo responses were counterbalanced for each picture. Each picture was presented eight times to each participant, i.e., once in each set. The presentation order of eight sets was systematically varied across participants.

103

BR A I N R ES E A RC H 1 2 7 3 ( 2 00 9 ) 9 2 –1 05

4.4.

Procedure

Participants were tested individually in a soundproof chamber. At first, participants were asked to familiarize themselves with the pictures and to memorize their corresponding name presented below them. Then, each picture was presented without the printed name, and participants were asked to name the pictures as correctly and fast as possible. If a participant's response was wrong, the experimenter would correct the participant, and the testing would be repeated. This procedure, typical of studies using object naming tasks to investigate spoken production, guaranteed that each participant knew and used the intended names of the pictures. During the following experimental session, participants were asked to carry out a dual-choice go/nogo task without overtly naming the picture. The semantic decision was to decide whether the picture depicted a living or a non-living object, while the orthographic decision was to decide whether the picture's written name consisted of a left–right structure character (i.e., 狗, /gou3/, dog) or not (i.e., 鱼, /yu2/, fish). Depending on semantic and orthographic decisions, a response was given with either the left or right hand, or no response at all. Each set began with 16 practice trials (2 pictures in each semantic–orthographic combination category, each picture was presented two times in the practice session), followed by 64 experimental trials (16 pictures in each semantic–orthographic combination category). Each participant completed 8 sets in the experimental session, with each set consisting of 64 trials. Each trial was constructed as follows: a fixation cross appeared in the center of the computer screen for 500 ms. After a random interval of 600 to 1600 ms, the picture was presented for 2000 ms. Then a blank screen appeared for 500 ms, followed by the next trial.

4.5.

Apparatus and recordings

Electroencephalograms (EEG) were recorded with 64 electrodes secured in an elastic cap (Electro cap International). The vertical electro-oculogram (VEOG) was monitored with electrodes placed above and below the left eye, and the horizontal EOG (HEOG) was recorded with a bipolar montage using two electrodes placed on the right and left external cantus. The bilateral mastoids served as reference points and the GND electrode on the cap served as ground. Electrode impedances were kept below 5 kΩ for the EEG and eye-movement recording. The electrophysiological signals were amplified with a bandpass from 0.05 to 70 Hz and digitized at a rate of 500 Hz. Epochs of 2200 ms were obtained (−200 ms to 2000 ms) including a 200 ms pre-stimulus baseline. The EEG and EOG signals were filtered with a high-frequency cutoff point of 30 Hz. The artifact rejection criteria were from −100 μν to 100 μν. Push-button response latencies were measured from picture onset, with a time-out point set at 2500 ms, i.e., responses given after 2500 ms were registered as missing. Trials with time-outs and errors were excluded from the data analysis. Hand responses were made by pressing either the left button with the left hand, or the right button with the right hand, on a hand-held button box.

Acknowledgments This research was supported by Grants from the National Natural Science Foundation of China (30400134, 30870761) and Young Scientist Foundation of Institute of Psychology (07CX102010) to Qingfang Zhang, and an International Incoming Fellowship (IIF-2007/R1) from The Royal Society to Qingfang Zhang and Markus Damian. We thank two anonymous reviewers for helpful comments on earlier versions of this manuscript.

Appendix A Picture names used in Experiment. Non-living objects with non-left–right structure character

Non-living objects with left–right structure character

Living objects with nonleft–right structure character

Living objects with left– right structure character

斧 (/fu3/, axe) 床 (/chuang2/, bed) 书 (/shu1/, book)

球 (/qiu2/, ball) 钟 (/zhong1/, bell) 碗 (/wan3/, bowl) 弓 (/gong1/, bow) 帽 (/mao4/, cap)

熊 (/xiong2/, bear) 蚁 (/yi3/, ant) 鸟 (/niao3/, bird) 蝶 (/die2/, butterfly) 虫 (/chong2/, 猫 (/mao1/, caterpillar) cat) 牛 (niu2/, cow) 狗 (/gou3/, dog) 云 (/yun2/, cloud) 糖 (/tang2/, 蟹 (/xie4/, crab) 鸭 (/ya1/, caramel) duck) 门 (/men2/, door) 椅 (/yi3/, chair) 鹿 (/lu4/, deer) 蝇 (/ying2/, fly) 刀 (/dao1/, knife) 鼔 (/gu3/, drum) 鹰 (/ying1/, eagle) 蛙 (/wa1/, frog) 山 (/shan1/, 旗 (/qi2/, flag) 象 (/xiang4/, 豹 (/bao4/, mountain) elephant) leopard) 尺 (/chi3/, ruler) 锁 (/suo3/, lock) 鱼 (/yu2/, fish) 狮 (/shi1/, lion) 勺 (/shao2/, 枪 (/qiang1/, 花 (/hua1/, 虾 (/xia1/, spoon) machgun) flower) lobster) 凳 (/deng4/, 镜 (/jing4/, 羊 (/yang2/, 猴 (/hou2/, stool) mirror) goat) monkey) 桌 (/zhuo1/, 针 (/zhen1/, 马 (/ma3/, 猪 (/zhu1/, table) needle) horse) pig) 鸡 (/ji1/, 伞 (/san3/, 钳 (/qian2/, 鼠 (/shu3/, mouse) rooster) umbrella) pliers) 表 (/biao3/, 锯 (ju4/, saw) 兔 (/tu4/, 蛇 (/she2/, watch) rabbit) snake) 井 (/jing3/, well) 鞋 (/xie2/, shoe) 虎 (/hu3/, tiger) 鹅 (/e2/, swan) 窗 (/chuang1/, 剑 (/jian4/, 龟 (/gui1/, turtle) 树 (/shu4/, window) sword) tree)

REFERENCES

Abdel Rahman, R., Sommer, W., 2002. Does phonological encoding in speech production always follow the retrieval of semantic knowledge? Electrophysiological evidence for parallel processing. Cogn. Brain Res. 16, 372–382. Alario, F.-X., Perre, L., Castel, C., Ziegler, J.C., 2007. The role of orthography in speech production revisited. Cognition 102, 464–475.

104

BR A IN RE S E A RCH 1 2 73 ( 20 0 9 ) 9 2 –1 05

Beijing Language Institute, 1986. Modern Chinese Frequency Dictionary (in Chinese). Beijing Language Institute Publisher, Beijing. Chen, H.-C., Flores d'Arcais, G.B., Cheung, S.L., 1995. Orthographic and phonological activation in recognizing Chinese characters. Psychol. Res. 58, 144–153. Chéreau, C., Gaskell, M.G., Dumay, N., 2007. Reading spoken words: orthographic effects in auditory priming. Cognition 102, 341–360. Christoffels, I.K., Formisano, E., Schiller, N.O., 2007. Neural correlates of verbal feedback processing: an fMRI study employing overt speech. Hum. Brain Mapp. 28, 868–879. Connine, C.M., Titone, D., 1996. Phoneme monitoring. Lang. Cogn. Processes. 11, 635–645. Damian, M.F., Bowers, J.S., 2003. Effects of orthography on speech production in a form-preparation paradigm. J. Mem. Lang. 49, 119–132. Damian, M.F., Bowers, J.S., in press. Assessing the role of orthography in speech perception and production: evidence from picture–word interference tasks. Euro. J. Cogn. Psychol. Damian, M.F., Singh, A., in preparation. Phonological effects in grapheme monitoring performed on picture names. Dell, G.S., O'Seaghdha, P.G., 1991. Mediated and convergent lexical priming in language production: a comment on Levelt et al. (1991). Psychol. Rev. 98, 604–614. Dell, G.S., O'Seaghdha, P.G., 1992. Stages of lexical access in language production. Cognition 42, 287–314. Dijkstra, T., Roelofs, A., Fieuws, S., 1995. Orthographic effects on phoneme monitoring. Can. J. Exp. Psychol. 49, 264–271. Ding, G., Peng, D., Taft, M., 2004. The nature of the mental representation of radicals in Chinese: a priming study. J. Exp. Psychol. Learn. Mem. Cogn. 30, 530–539. Donnenwerth-Nolan, S., Tanenhaus, M.K., Seidenberg, M.S., 1981. Multiple code activation in word recognition: evidence from rhyme monitoring. J. Exp. Psychol. Learn. Mem. Cogn. 7, 170–180. Fuster, J.M., 1997. The Prefrontal Cortex: Anatomy, Physiology, and Neuropsychology of the Frontal Lobe, 3rd ed. Raven Press, New York. Gaskell, M.G., Cox, H., Foley, K., Grieve, H., O'Brien, R., 2003. Constraints on definite article alternation in speech production: to “thee” or not to “thee”? Mem. Cogn. 31, 715–727. Gemba, H., Sasaki, K., 1989. Potential related to no-go reaction of go/no-go hand movement task with color discrimination in human. Neurosci. Lett. 101, 262–268. Guo, T., Peng, D., Lu, C., Liu, H., 2005. The temporal course of semantic and phonological activation in Chinese word production (in Chinese). Acta Psychol. Sin. 37, 569–574. Hallé, P.A., Chéreau, C., Segui, J., 2000. Where is the /b/ in “absurde” [apsyrd]? It is in French listeners' minds. J. Mem. Lang. 43, 618–639. Hauk, O., Rockstroh, B., Euiltz, C., 2001. Grapheme monitoring in picture naming: an electrophysiological study of language production. Brain Topogr. 14, 3–13. Heim, S., Opitz, B., Müller, K., Friederici, A.D., 2003. Phonological processing during language production: fMRI evidence for a shared production-comprehension network. Cogn. Brain Res. 2003, 285–296. Indefrey, P., Levelt, W.J.M., 2004. The spatial and temporal signatures of word production components. Cognition 92, 101–144. Jakimik, J., Cole, R.A., Rudnicky, A.I., 1985. Sound and spelling in spoken word recognition. J. Mem. Lang. 24, 165–178. Jodo, E., Kayama, Y., 1992. Relation of a negative ERP component to response inhibition in a Go/noGo task. Electroencephalogr. Clin. Neurophysiol. 82, 477–482. Kopp, B., Mattler, R., Goetry, R., Rist, F., 1996. N2, P3 and the lateralized readiness potential in a noGo task involving selective response priming. Electroencephalogr. Clin. Neurophysiol. 99, 19–27.

Kutas, M., Donchin, E., 1980. Preparation to respond as manifested by movement-related brain potentials. Brain Res. 202, 95–115. Kutas, R., Schmitt, B.M., 2003. Language in microvolts. In: Banich, M.T., Mack, M.A. (Eds.), Mind, Brain, and Language: Multidisciplinary Perspectives. Lawrence Erlbaum Associates Incorporated, New York, NY, pp. 171–209. Levelt, W.JM., Schriefers, H., Vorberg, D., Meyer, A.S., Pechmann, T., Havinga, J., 1991. The time course of lexical access in speech production: a study of picture naming. Psychol. Rev. 98, 122–142. Levelt, W.J.M., Roelofs, A., Meyer, A.S., 1999. A theory of lexical access in speech production. Behav. Brain Sci. 22, 1–75. Li, D., 1993. A Study of Chinese Characters. Peking University Press, Beijing. Lupker, S.J., 1982. The role of phonetic and orthographic similarity in picture–word interference. Can. J. Psychol. 26, 349–367. Lup k er, S.J., Williams, B.A., 1989. Rhyme priming of pictures and word: a lexical activation account. J. Exp. Psychol. Learn. Mem. Cogn. 15, 1033–1046. Meyer, A.S, 1990. The time course of phonological encoding in language production: the encoding of successive syllables. J. Mem. Lang. 29, 524–545. Muneaux, M., Ziegler, J.C., 2004. Locus of orthographic effects in spoken word recognition: novel insights from the neighbor generation task. Lang. Cogn. Processes 19, 641–660. Özdemir, R., Roelofs, A., Levlet, W.J.M., 2007. Perceptual uniqueness point effects in monitoring internal speech. Cognition 105, 457–465. Palmer, E.D., Rosen, H.J., Ojemann, J.G., Buckner, R.L., Kelley, W.M., Petersen, S.E., 2001. An event-related fMRI study of overt and covert word stem completion. NeuroImage 14, 182–193. Perfetti, C.A., Zhang, S., 1991. Phonemic processes in reading Chinese words. J. Exp. Psychol. Learn. Mem. Cogn. 17, 633–643. Peterson, R.R., Savoy, P., 1998. Lexical selection and phonological encoding during language production: evidence for cascaded processing. J. Exp. Psychol. Learn. Mem. Cogn. 24, 539–557. Racine, I., Grosjean, F., 2005. The cost of deleting a schwa at the time of recognition of French words. Can. J. Exp. Psychol. 59, 240–254. Rodriguez-Fornells, A., Schmitt, B.M., Kutas, M., Münte, T.F., 2002. Electrophysiological estimates of the time course of semantic and phonological encoding during listening and naming. Neuropsychologia 40, 778–787. Roelofs, A., 2006. The influence of spelling on phonological encoding in word reading, object naming, and word generation. Psychon. Bull. Rev. 13, 33–37. Sasaki, K., Gemba, H., 1993. Prefrontal cortex in the organization and control of voluntary movement. In: Ono, T., Squire, L.R., Raiche, M.E., Perrett, D.I., Fukuda, M. (Eds.), Brain Mechanisms of Perception and Memory: From Neuron to Behavior. Oxford University Press, New York, pp. 473–496. Sasaki, K., Gemba, H., Nambu, A., Matsuzaki, R., 1993. No-go activity in the frontal association cortex of human subjects. Neurosci. Res. 18, 249–252. Schiller, N.O., 2006. Lexical stress encoding in single word production estimated by event-related brain potentials. Brain Res. 1112, 201–212. Schiller, N.O., 2007. Phonology and orthography in reading aloud. Psychon. Bull. Rev. 14, 460–465. Schiller, N.O., Bles, M., Jansma, B.M., 2003. Tracking the time course of phonological encoding in speech production: an event-related brain potential study. Cogn. Brain Res. 17, 819–831. Schmitt, B.M., Münte, T.F., Kutas, M., 2000. Electrophysiological estimates of the time course of semantic and phonological encoding during implicit naming. Psychophysiology 37, 473–484.

BR A I N R ES E A RC H 1 2 7 3 ( 2 00 9 ) 9 2 –1 05

Schmitt, B.M., Schiltz, K., Zaake, W., Kutas, M., Münte, T.F., 2001. An electrophysiological analysis of the time course of conceptual and syntactic encoding during tacit picture naming. J. Cogn. Neurosci. 13, 510–522. Schriefers, H., Meyer, A.S., Levelt, W.J.M., 1990. Exploring the time course of lexical access in language production: picture–word interference studies. J. Mem. Lang. 29, 86–102. Seidenberg, M.S., Tanenhaus, M.K., 1979. Orthographic effects on rhyme monitoring. J. Exp. Psychol. Human Learn. Mem. 5, 546–554. Shen, D., Forster, K.I., 1999. Masked phonological priming in reading Chinese words depends on the task. Lang. Cogn. Processes 14, 429–459. Simson, R., Vaughan, H.G., Ritter, W., 1977. The scalp topography of potentials in auditory and visual Go/noGo tasks. Electroencephalogr. Clin. Neurophysiol. 43, 864–875. Starreveld, P.A., La Heij, W., 1995. Semantic interference, orthographic facilitation, and their interaction in naming tasks. J. Exp. Psychol. Learn. Mem. Cogn. 21, 686–698. Taft, M., Hambly, G., 1985. The influence of orthography on phonological representations in the lexicon. J. Mem. Lang. 24, 320–335. Taft, M., Zhu, X., Peng, D., 1999. Positional specificity of radicals in Chinese character recognition. J. Mem. Lang. 40, 498–519. Taft, M., Zhu, X., Ding, G., 2000. The relationship between character and radical representation in Chinese. Acta Psychol. Sin. 32 (Suppl.), 3–12. Thorpe, S, Fize, D, Marlot, C., 1996. Speed of processing in human visual system. Nature (Lond.) 381, 520–522. Tsang, Y.-K., Chen, H.-C., in press. Do position-general radicals have a role to play in processing Chinese characters? Lang. Cogn. Proc. doi:10.1080/01690960802154615. Underwood, G., Briggs, P., 1984. The development of word recognition processes. Br. J. Psychol. 75, 243–255. van Turennout, M., Hagoort, P., Brown, C.M., 1997. Electrophysiological evidence on the time course of semantic

105

and phonological processes in speech production. J. Exp. Psychol. Learn. Mem. Cogn. 23, 787–806. van Turennout, M., Hagoort, P., Brown, C.M., 1998. Brain activity during speaking: from syntax to phonology in 40 milliseconds. Science 280, 572–574. Ventura, P., Morais, J., Pattamadilok, C., Kolinsky, R., 2004. The locus of the orthographic consistency effect in auditory word recognition. Lang. Cogn. Processes 19, 57–95. Wheeldon, L., Monsell, S., 1992. The locus of repetition priming of spoken word production. Q. J. Exp. Psychol., Human Exp. Psychol. 44 (A), 723–761. Wong, K-F-E., Chen, H.-C., 1999. Orthographic and phonological processing in reading Chinese text: evidence from eye fixations. Lang. Cogn. Processes 14, 461–480. Zhang, Q., Yang, Y., 2003. The determiners of picture naming latency (in Chinese). Acta Psychol. Sin. 35, 447–454. Zhang, Q., Yang, Y., 2007. Electrophysiological estimates of the time course of semantic and metrical encoding in Chinese speech production. Neuroscience 147, 986–995. Zhang, Q., Damian, M.F., Yang, Y., 2007. Electrophysiological estimates of the time course of tonal and orthographic encoding in Chinese speech production. Brain Res. 1184, 234–244. Zhou, X., Marslen-Wilson, W., 1999. Orthography, phonology, and semantic activation in reading Chinese. J. Mem. Lang. 41, 579–606. Zhou, X., Shu, H., Bi, Y., Shi, D., 1999. Is there phonologically mediated access to lexical semantics in reading Chinese. In: Wang, J., Inhoff, A.W., Chen, H.-C. (Eds.), Reading Chinese Script: A Cognitive Analysis. Lawrence Erlbaum Associates Incorporated,, Mahwah, NJ, pp. 135–171. Ziegler, J.C., Ferrand, L., 1998. Orthography shapes the perception of speech: the consistency effect in auditory word recognition. Psychon. Bull. Rev. 5, 683–689. Ziegler, J.C, Ferrand, L., Montant, M., 2004. Visual phonology: the effects of orthographic consistency on different auditory word recognition tasks. Mem. Cogn. 32, 732–741.