Publications and Other Articles

Conference (and other) Presentations

The Abstracts

Lexical Processing of Synonymy and Antonymy

This paper reports on three psycholinguistic experiments that explore the question of how synonyms and antonyms are linked in the mental lexicon. The experimental results confirm the expectation that synonymic and antonymic links are fundamental cognitive relations among lexical items. The results also suggest that the nature of these relations may not be identical because it was found that synonymic and antonymic pairs behaved differently in off-line and on-line experimental tasks and that they also showed differences in the way they interacted when suffixation was added as a variable. The relevance of these findings to methodological issues in psycholinguistic experimentation is discussed and the implications for further research are outlined.

Linguistic Vs. Non-Linguistic Strategies in SLA

Neuroimaging techniques such as averaging the electroencephalogram (EEG) to particular linguistic events to get Event-Related Potentials (ERPs) offer a new approach to the study of L2 processing. ERPs allow us to not only look at the end result of processing (as in RT studies), but also at how the L2 is processed in real time. This methodology is sensitive to both quantitative and qualitative differences between populations and conditions. These differences will be explored in this presentation.

The ERP experiment discussed here looked at the processing of Dutch gender-differentiated relative pronouns. In Dutch, the relative pronoun used depends on the gender of the head noun. In this presentation ERP data from German (N=14) and English (N=9) learners of Dutch as well as data from a Dutch native speaker control group (N=23) will be discussed. ERP responses to the grammatical use of the relative pronoun will be compared to an ungrammatical use of the pronoun. This study will allow us to compare native and L2 processing as well as looking at effects of the presence or absence of gender in the L1.

When native Dutch speakers are presented with a violation of the relative pronoun compared to a grammatical sentence they show an increased P600 component (a positivity at posterior electrode sites occurring between 600 and 1080 ms after the violation), indicating that they are sensitive to the grammatical violation. The P600 component is thus indicative of normal syntactic processing. Advanced German learners of Dutch also showed an increased positivity to the violations. However, their positivity was significantly delayed compared to the native speaker P600 (starting only at 810 ms), did not have a long duration (positivity lasted until 930 ms) and was more widespread. This suggests that their processing was qualitatively similar although it was somewhat more difficult (quantitative differences were found). English speakers, on the other hand, showed a significant negativity instead of a positivity (qualititative differences), occurring between 510 and 630 ms at many electrode sites.

The results demonstrate that all L2 participants were sensitive to the presented violation but that only the German participants treated it as a linguistic (or grammatical) violation. Apparently the German group, having a similar gender system in their L1, are able to process an L2 gender system similarly to native speakers. The English group, who have no grammatical gender in their L1, seem to recognize a learned pattern using a non-linguistic strategy. They may not process agreement between the pronoun and noun, but rather unconsciously recognize the mismatch between the prenominal determiner and relative pronoun (simple associative pattern recognition as opposed to rule based processing). These results suggest that the ability to process an L2 in native-like fashion is dependent upon the L1 of the learner.

SLA Research and Neuroimaging: Knowledge vs. Processing

Neuroimaging techniques have become popular techniques with which to study language processes. In particular, and of special interest for this paper, techniques for investigating electrophysiological activity while language is being processed are being used quite often in normal language processing studies. These techniques are now beginning to be used as well for L2 research. Neuroimaging is a promising new direction as it will allow us to not only look at the end result of processing (as in RT studies), but also at how the L2 is processed in real time. In this presentation on grammatical gender in Dutch L2, results from an off-line task measuring knowledge will be compared with neuroimaging results testing processing.

The Dutch gender system is a 2-gender system (common and neuter). The experiments presented here will investigate L1 effects by using 3 different L1-groups: German, Romance and English. German has a 3-gender system which is congruent with the Dutch system. Masculine and feminine gender in German are correlated with the common gender in Dutch, while the German neuter gender is correlated with the Dutch neuter gender. The Romance L1s being tested have a 2-gender system using masculine and feminine genders. There are no systematic correlations between the Romance and Dutch systems. Finally, English is a language without grammatical gender.

Experiment 1 was an off-line grammaticality judgment task. This test was to determine the knowledge the L2 participants had of the Dutch grammatical gender system. German speakers performed at a relatively high level of proficiency while Romance and English speakers were both significantly worse than the Dutch controls. The Romance speakers did perform at a higher level of accuracy than the English speakers. This suggests that having gender in L1 helps in acquiring the L2 distinctions (German and Romance speakers do better than the English speakers) but that similar gender is even more helpful.

Experiment 2 was an on-line version that compared ERPs (Event Related brain Potentials measuring electrical brain actitivity) to grammatical and ungrammatical gender combinations. While native speakers showed an increased P600 processing effect (a brain potential found for morpho-syntactic violations) to the ungrammaticalities, the L2 speakers (with German L1) showed a completely different pattern, a long-lasting negativity, to the ungrammatical version, possibly indicating that general working memory strategies, instead of linguistic processing, was occurring.

German speakers perform at a comparable proficiency level to the native speakers but by looking at the real-time processing it was shown that the brain-wave patterns are very different. The qualitatively different ERP patterns would not be expected if the L2 was being processed in the same manner as one processes their native language.

UG and the L2 Processing of Grammatical Gender

An important question for second language (L2) processing research, but one that has not received much systematic attention, is to what extent language learners use their specialized language module, and to what extent they rely on more general cognitive strategies in learning and processing a second language. The two positions have very different implications for linguistic theory.

This presentation will report on the L2 processing of the Dutch gender system. The Dutch gender system is a 2-gender system (common and neuter). The experiments presented here will investigate L1 effects by using 3 different L1-groups: German, Romance and English. German has a 3-gender system which is congruent with the Dutch system. Masculine and feminine gender in German are correlated with the common gender in Dutch, while the German neuter gender is correlated with the Dutch neuter gender. The Romance L1s being tested have a 2-gender system using masculine and feminine genders. There are no systematic correlations between the Romance and Dutch systems. Finally, English is a language without grammatical gender. If UG is fully available to the L2 learner it would be expected that advanced L2 speakers can process the Dutch gender system correctly (regardless of their L1). Results show that this is not the case.

Experiment 1 was an off-line task giving the L2 participants the time to think about the grammaticality of sentences. This test was to determine the knowledge the L2 participants had of the Dutch grammatical gender system. German and Romance speakers were both significantly worse than the Dutch controls and the German group performed significantly better than the Romance group. The English group performed significantly worse than all groups. This suggests that having gender in L1 helps in acquiring the L2 distinctions (German and Romance speakers do better than the English speakers) but that similar gender is even more helpful.

Experiment 2 was an on-line version that compared ERPs (Event Related brain Potentials measuring electrical brain actitivity) to grammatical and ungrammatical gender combinations. While native speakers showed an increased P600 response (a brain potential found for morpho-syntactic violations) to the ungrammaticalities, the L2 speakers (with German L1) showed a completely different pattern, a long-lasting negativity, to the ungrammatical version, possibly indicating that general working memory strategies, instead of linguistic processing, was occurring.

The results suggest that for at least speakers with a congruent gender system (L1 German) successful cognitive strategies are being used for off-line tasks. The different ERP patterns would not be expected if the L2 was being processed via UG.

L2 Grammatical Gender: An Event-Related Potentials (ERP) Study

This study uses EEG-technology (averaging EEGs as ERPs) to determine whether aspects of L2-processing are specifically linguistic in nature. The existence of specific ERP-responses to linguistic input for native speakers provides a method to assess L2-processing. If L2-processing uses resources specific to language, one expects ERP-responses similar to those in native speakers (although maybe delayed due to slower L2-processing). If not, patterns unrelated to normal language processing are expected.

We compare the processing of Dutch grammatical gender by native speakers and advanced L2-speakers with German L1. This is of particular interest as the Dutch and German systems are similar. The Dutch gender system consists of common and neuter. German has a gender system that is congruent with that of Dutch: German masculine and feminine correlate with the Dutch common gender and German neuter correlates with Dutch neuter. This could allow a German speaker of Dutch to have easier access to specialized linguistic knowledge regarding gender or to use more general cognitive strategies (predicting the grammatical gender of Dutch from their L1). The first method would result in similar ERP-patterns to native speakers while the second would result in different patterns.

For native speakers of Dutch, gender agreement violations result in a P600, a positive shift in the ERP, which reaches its peak around 600 ms after the presentation of the violation. The P600 is typically elicited by (morpho)syntactic anomalies detected during normal language processing. Results from an off-line task showed that Germans performed at near-native levels for detecting gender violations (98% accuracy for native speakers and 94% accuracy for German speakers). On the other hand, preliminary ERP-patterns exhibited only a negativity. This demonstrates that the Germans detect gender violations in a very different manner from native speakers.

Neuroimaging in SLA research

In the past two decades, a number of neuroimaging techniques have been developed which allow researchers to investigate electrophysiological and haemodynamic activity in a living brain while a subject reads or listens to linguistic stimuli. These methods have led to a lot of progress in our understanding of the way in which our brains process language; additionally, it has increased our insight into linguistic structure. In terms of neurolinguistics, a more detailed picture has emerged, replacing the classical picture with two language centers in the left hemisphere: additional areas in the left hemisphere, besides Broca's and Wernicke's area, are active, as are a number of areas in the right hemisphere and the cerebellum. Moreover, areas that are primarily responsible for cognitive functions other than language, for instance the primary visual cortex, play a crucial role in language processing as well. In terms of linguistic structure, these methods have contributed to the discussion between constraint-based and autonomous syntax models of parsing.

These methods may prove very useful for a number of central questions in SLA research as well; however, no systematic use has been made of them in research in this area. In this workshop, we would therefore like to present a number of the most widely used neuroimaging methods, and discuss their possibilities and limitations. Additionally, we want to raise the question how these methods can prove useful for research in SLA, focussing on the question what the added value can be of data collected using these methods, and what types of questions can more reasonably be answered using more traditional psycholinguistic methods; this is important as one should not have exaggerated expectations of the possibilities these methods present for the area of SLA, they are not a panacea. We will then present an example of a study looking at the processing of verbal agreement and finiteness processes in second language to illustrate one promising way of using one of these methods. Finally, there will be a panel discussion.

A workshop format will allow a systematic survey of the issues involved in using neuroimaging techniques in SLA research; moreover, in such a format there is ample time for questions and discussion.

The workshop proposed here, consists of three presentations, followed by a general discussion period of 15 minutes:

Neuroimaging methods: Marco Haverkort

In this presentation, a number of neuroimaging methods will be introduced (Event-Related Potentials (ERPs), Positron Emission Tomography (PET), and functional Magnetic Resonance Imaging (fMRI)) and compared in terms of possibilities and limitations. The presentation will focus on the relevance of these methods for linguistic questions.

Applications in SLA: Kees de Bot

This presentation will discuss in what respects neuroimaging methods can provide insights in SLA research and help answer questions that cannot be elucidated using more traditional psycholinguistic methods, such as elicitation techniques, self-paced reading, eye-tracking, lexical decision tasks, etc. As neuroimaging techniques are expensive and labour-intensive, it is important to realize what their added value is, and when one should use the more traditional methods.

An ERP-study: Laura Sabourin

In this presentation, processing of verbal agreement and finiteness by second language learners will be discussed, using ERPs. It will be shown how a comparison of data obtained in an off-line grammaticality judgment task and in an on-line ERP task can shed light on the distinction between grammatical knowledge and language processing in second language learning.

General panel discussion

A number of researchers (including Wander Lowie and Laurie Stowe among others) will be invited for this discussion, in addition to the presenters.

This discussion period is meant for questions that deal with several of the issues raised separately in the three presentations.

The Second Language Processing of Grammatical Gender: An ERP study

An important question for second language acquisition research is to what extent language learners use a specialized language module, and to what degree they rely on more general cognitive strategies in learning and processing a second language. Investigating second language (L2) processing is one way at getting at this issue.

The present study uses Event-Related Potentials (ERPs) to determine whether certain aspects of L2 processing are specifically linguistic in nature. The existence of specific EEG-responses to linguistic input of particular types provides an interesting method to assess second language processing. If processing in L2 takes place in the language module, one expects EEG-responses similar to those found in native speakers (although the patterns may occur later due to slower L2 processing). If not, patterns unrelated to normal language processing are expected to occur.

This paper compares the processing of Dutch gender by native speakers and advanced second language speakers of Dutch with German as their first language. This is of particular interest due to the fact that the Dutch and German systems are similar. The Dutch gender system consists of two genders: common (historically, masculine and feminine) and neuter. German comprises a gender system that is by and large congruent with that of Dutch; German masculine and feminine are correlated with the common gender and German neuter correlates with Dutch neuter. This could allow a German speaker of Dutch to either have easier access to a specialized language module as there is already a mechanism set up for the processing of grammatical gender or to use a more general cognitive strategy of using the grammatical gender from their first language in order to solve the L2 problem. The first method would result in similar brain processing patterns to native speakers while the second would result in a different pattern.

German speakers' metalinguistic knowledge about Dutch gender was assessed by using an off-line grammaticality judgment task. The results were compared with the results of an ERP experiment looking specifically at the processing of gender. For native speakers of Dutch it has been shown that gender agreement violations result in a P600, a positive shift in the EEG, which reaches its peak around 600 ms after the presentation of the violation (Hagoort & Brown, 1999). The P600 is typically elicited by (morpho)syntactic anomalies detected during normal language processing. Results from the off-line task showed that Germans performed at near-native levels for detecting gender violations (98% accuracy for native speakers and 94% accuracy for German speakers). On the other hand, preliminary EEG wave patterns exhibited only a negativity. This suggests that the Germans detect gender violations in a very different manner from native speakers.

Hagoort, P. & Brown, C.M. (1999). Gender electrified: ERP evidence on the syntactic nature of gender processing. Journal of Psycholinguistic Research. Vol. 28 (6): 715-728.

ERPs and Advanced Second Language Processing

Click here to go to the .pdf format of my poster presented at the conference.

An important question for second language acquisition research is to what extent language learners use a specialized language module, and to what degree they rely on more general cognitive strategies in learning and processing a second language. Investigating second language (L2) processing is one way at getting at this issue.

The present study uses Event-Related Potentials (ERPs) to determine whether certain aspects of L2 processing are specifically linguistic in nature. The existence of specific EEG responses to linguistic input provides an interesting method to assess second language processing. If processing in L2 takes place in the language module, one expects EEG-responses similar to those found in native speakers. If not, patterns unrelated to normal language processing are expected to occur.

This paper compares the processing of Dutch gender by native speakers and advanced second language speakers of Dutch with German as their first language. The Dutch gender system consists of two genders: common (historically, masculine and feminine) and neuter. German comprises a gender system that is congruent with that of Dutch; German masculine and feminine are correlated with the common gender and German neuter correlates with Dutch neuter.

German speakers' metalinguistic knowledge about Dutch gender was assessed by using an off-line grammaticality judgment task. The results were compared with the results of an ERP experiment looking specifically at the processing of gender. For native speakers of Dutch it has been shown that gender agreement violations result in a P600, a positive shift in the EEG, which reaches its peak around 600 ms after the presentation of the violation (Hagoort & Brown, 1999). The P600 is typically elicited by (morpho)syntactic anomalies detected during normal language processing. Results from the off-line task showed that Germans performed at near-native levels for detecting gender violations (90% accuracy). On the other hand, preliminary EEG wave patterns exhibited only a negativity. This suggests that they are doing this gender violation detection in a very different manner from native speakers.

Hagoort, P. & Brown, C.M. (1999). Gender electrified: ERP evidence on the syntactic nature of gender processing. Journal of Psycholinguistic Research. Vol. 28 (6): 715-728.

L2 Judgments on Grammatical Gender Agreement

This presentation will report on the L2 off-line processing of Dutch agreement. Non-gender agreement (finiteness and agreement) and aspects of gender agreement (within noun phrases and with the use of the relative clause) were investigated. In 3 grammaticality judgment experiments I found a hierarchy of performance on L2 gender agreement. Taken together, the 3 experiments indicate that grammatical gender is problematic independent of the general level of proficiency in L2. In terms of non-gender agreement there is no difference in performance between Dutch (native speaker control group), German and Romance groups, yet there are differences between these 3 groups on the two gender agreement experiments. German and Romance speakers were both significantly worse than the Dutch controls but the German group performed significantly better than the Romance group. An English group performed significantly worse on all experiments. This suggests that having gender in L1 helps in acquiring the L2 distinctions (German and Romance speakers do better than the English speakers) and that having similar gender is even more helpful (German speakers perform better than the Romance speakers).

Second Language Processing and the Role of the Language Module

Click here to go to the .pdf format of my poster presented at the conference.

An important question for second language acquisition (SLA) research, but one that has not received much systematic attention, is to what extent language learners use their specialized language module, and to what extent they rely on more general cognitive strategies in learning and processing a second language. Can learners make use of UG to acquire a language after the critical age or must they use more general cognitive skills? The two types of strategies have very different implications for linguistic theory: if all or certain aspects of SLA are dealt with by general problem-solving abilities, one has to reconsider the idea that UG plays a role in acquiring those aspects.

The present study uses Event-Related Potentials (ERPs) to determine whether certain aspects of second language processing are linguistic in nature. ERPs measure electro-physiological activity of the brain by means of electrodes on the scalp (EEG), while subjects perform specific (linguistic) tasks. ERPs allow the researcher to track the time course of processing at the level of the millisecond. In native speakers of a language, (morpho)syntactic anomalies affect a component of the EEG-signal known as the P600/SPS (syntactic positive shift). This component is a positive deflection in the EEG signal that occurs 600 ms after the critical word making the sentence ungrammatical. The existence of a specific EEG-response to linguistic input provides an interesting new method to assess second language processing. If processing in a second language takes place in the language module, one expects language-specific EEG-responses in second language learners that are similar to those found in native speakers, both in terms of their time course and amplitude. If not, very different patterns, unrelated to EEG-patterns that are specifically associated with language in native speakers, are expected to occur.

This paper compares the processing of Dutch gender by native speakers of Dutch and second language learners with different language backgrounds. The Dutch gender system consists of two genders: common (comprising what historically were separate masculine and feminine genders) and neuter. The second language learners came from three languages/language groups: German (which has a gender system that is congruent with that of Dutch; masculine and feminine in German are highly correlated with the so-called common gender in Dutch, and German neuter correlates highly with Dutch neuter), Romance (which has a two gender system, with a division that is different from Dutch: masculine and feminine; there is no neuter and the Romance two-way division does not correlate with the Dutch two-way division), and English (which lacks grammatical gender altogether).

Meta-linguistic knowledge about gender, off-line grammaticality judgments testing both knowledge of grammatical gender and gender agreement processes are compared with the results of an ERP experiment looking specifically at the processing of gender. For native speakers of Dutch, it has been shown that gender agreement violations result in a P600, a positive shift in the EEG, which reaches its peak around 600 ms after the presentation of the violation (Hagoort & Brown, 1999). The paper investigates whether advanced second language learners of Dutch - even if they exhibit the same knowledge as native speakers in the grammaticality judgment task - exhibit the same neuro-physiological responses to violations, thus possibly indicating that even though their knowledge is comparable to that of a native speaker, their on-line processing differs qualitatively. Comparison of data obtained with the traditional elicitation and judgement techniques with those obtained by tapping directly into electro-physiological activity in the brain associated with a specific grammatical phenomenon allows us to tease apart the contributions of the language module and those of more general cognitive strategies in the processing of a second language. Gender is specifically interesting in this respect, because it involves lexical and syntactic aspects; hence storage, computation, and their interaction can be studied simultaneously.

Hagoort, P. & Brown, C.M. (1999). Gender electrified: ERP evidence on the syntactic nature of gender processing. Journal of Psycholinguistic Research. Vol. 28 (6): 715-728.

Neurocorrelates of different types of grammatical (gender) agreement

Click here to go to the .pdf format of my poster presented at the conference.

Click here to go to the .pdf file of the abstract.

L1 Effects on the Processing of Grammatical Gender in L2

Click here to go to the .pdf format of my overheads used at the conference.

Using grammatical gender distinctions correctly in a second language poses a problem even for advanced L2-speakers, especially when the L1 and L2 differ with respect to gender distinctions. The current study tries to ascertain whether the L2 difficulty with gender is due to a lack of knowledge (either grammatical or lexical) or whether it is solely a processing problem.

This presentation will report on the L2 processing of the Dutch gender system. The Dutch gender system is a 2-gender system, derived historically from a 3-gender system in which the masculine and feminine genders have been conflated into one common gender. In Dutch, gender is expressed on the determiner, the attributive adjective and the relative pronoun. The experiment presented here will investigate facilitatory and inhibitory transfer effects by using 3 different L1-groups: German, Romance and English. German has a 3-gender system which is fairly congruent with the Dutch system in that the genders of nouns are highly correlated. Masculine and feminine gender in German are highly correlated with the common gender in Dutch while the German neuter gender is highly correlated with the Dutch neuter gender. The Romance L1s being tested (French, Italian and Spanish) have a 2-gender system like Dutch but instead of common and neuter gender, they use masculine and feminine. There is no systematic correlations between the Romance and Dutch systems. Finally, English is a language without grammatical gender distinctions.

Using a grammaticality judgment task we can see to what extent advanced adult learners of Dutch (15 subjects from each of the subject groups above) have the knowledge of the L2-gender system. The different grammatical forms of interest include the use of the correct definite determiner, use of the correct adjectival form, use of the correct relative pronoun and the form of the predicative adjective. The nouns used included both high and middle frequency nouns. The discussion will include a comparison of the different L2 groups with the native speaker control group as well as a comparison among the different L2 groups to see what effects, if any, differing L1s have on the use of the Dutch gender system. Implications of results from off-line grammaticality judgments testing knowledge for on-line processing experiments testing processing will also be considered.

How is Grammatical Gender Processed?

For second language speakers grammatical gender is the "impossible to learn" aspect of languages such as Dutch and French (and other gender languages). Is it de tafel or het tafel, is it le maison or la maison? Why do native speakers have such difficulty with gender? But, the more basic question that comes to mind is how native speakers process gender? Is grammatical gender treated as a syntactic agreement problem or as a lexical problem? What is it that must be determined in order to correctly use gender distinctions?

This study presents ERP data on the processing of the Dutch grammatical gender system by native speakers. It has previously been shown for Dutch that gender violations, in the form of a gender agreement violation between the article and the noun, result in a P600 (Hagoort & Brown, 1999). This finding suggests that gender agreement is treated as a syntactic phenomenon as the P600 component of an EEG is typically affected by syntactic anomalies. The study presented here will expand on this previous study by also looking at the electrophysiological activity to gender violations of other sorts. Theoretical consequences for what might be found in L2 processing will also be discussed.

Hagoort, P. & Brown, C.M. (1999). Gender electrified: ERP evidence on the syntactic nature of gender processing. Journal of Psycholinguistic Research. Vol. 28 (6): 715-728.


Page created July 2000
Last updated on December 14, 2005 by Laura Sabourin