Conference Program (subject to minor changes). Click on the names for titles & abstracts and use the 'up' link to come back to the schedule.

Albert Costa (University of Barcelona)

Some answers to the problem of controlling two languages in bilingual speech production

Bilingual speakers have to learn to control their two languages in order to achieve successful communication. In any given conversation, these speakers have to focus on one language while ignoring the representations of their other language that may act as possible competitors. Thus, a main issue in bilingual language production refers to the attentional mechanisms that allow bilingual speakers to control their lexicalization process. In this talk, I will discuss the most relevant proposals on this issue and I will present some experiments aiming at adjudicating between these proposals.
Up

Jonathan Grainger (University of Aix-en-Provence)

A set of component processes is described within a functional architecture for word recognition. The hypothesized architecture and processing mechanisms are justified on the basis of behavioural results obtained with the masked priming paradigm, and allow precise predictions concerning the relative timing of the different component processes. The combination of masked priming and EEG recordings is used to put these predictions to test. Event-related potentials (ERPs) show a series of priming-sensitive components (with distinct temporal and topographic signatures) starting from around 100 ms post-target onset, that are interpreted as reflecting feature-level, orthographic, phonological, and semantic processing during visual word recognition.
Up

Session 1: language Production

Helena Trompelt (University of Leipzig)

H. Trompelt, D. Bordag, & T. Pechmann

The production of verbs in German - evidence of picture-word-interference-experiments

In three experiments we used a picture-word distractor paradigm to explore German verbs in speech production. In Experiment 1 and 2 we focussed on the retrieval of word class information as a syntactic property. Subjects used verbs in the 3rd pers.sg. to name pictures of actions under four conditions: identical, neutral (xxxxx), same class (verb) or different class (closed class word) distractors. In the same class condition we employed five abstract verbs in infinitive. Additionally, we manipulated the factor SOA. The analyses showed a significant word class effect at SOA -100 ms. The production of verbs was delayed by distractors of the same word class. This is consistent with similar findings of Pechmann & Zerbst (2002) concerning the production of nouns. We assume for verbs, too, that the same activated features compete for the same syntactic slot. In Experiment 3 we used the same paradigm to investigate the production of regular and irregular verbs. We again constructed four conditions (identical, neutral, congruent, incongruent). In the congruent condition the picture and the distractor were either both regular or irregular, in the incongruent condition they differed in regularity. The (ir)regularity of verbs manifests itself in the past tense. Past tense regular forms are constructed with the suffix –te (lachen-lachte) (laugh-laughed), irregular forms have a special past tense stem (schlafen-schlief) (sleep-slept). Subjects named the pictures in 3rd pers.sg. in the present tense where the (ir)regularity is not overtly visible. We again observed a significant effect at SOA -100 ms: RTs were longer in congruent than incongruent condition. Possible interpretations of the observed effect, especially with respect to competition between same vs. different verbal frames, are discussed.
Up

Lies Notebaert (Ghent University)

L. Notebaert & R. Hartsuiker

Do speech planning troubles cause disfluencies?

Disfluencies, like self-corrections, repetitions, and silent or filled pauses, occur about 6 times per 100 words. Many accounts relate disfluencies to speech planning problems. Thus, disfluencies might result from an overload on conceptualization processes (Henderson, Goldman-Eisler, & Skarbek, 1966), from (covert) repair of speech planning errors (Postma & Kolk, 1993), or they might be signals of upcoming delays in speech delivery (Clark & Fox Tree, 2002). Two picture naming experiments in Dutch tested whether, and at which production stages, planning difficulties cause disfluencies. Experiment 1 manipulated speech planning difficulty at the early stage of lexical selection (by varying picture name agreement) and Experiment 2 at the late stage of form encoding (by varying word length in syllables). Both variables affect picture naming latencies (Severens, Van Lommel, Ratinckx, & Hartsuiker, 2005).

Speakers described networks of line drawings (Figure 1) in synchrony with a red dot that moved through the network, pacing speech rate. This task elicits many disfluencies and allows experimental control over speech content (e.g., Schnadt & Corley, 2006). Utterances referring to the pictures (e.g., "to the mailbox") were analyzed, and disfluencies were categorized as self-corrections (e.g., overt repairs, false starts), repetitions, and "pauses" (silent and filled pauses, prolongations). Experiment 1 varied name agreement, using Severens et al.'s (2005) Dutch picture norms. The low- and high-agreement conditions were matched for frequency, age of acquisition (AoA), and word length. Disfluencies occurred more often in the low-agreement condition (321) than in the high-agreement condition (250). This was particularly the case for pauses (low agreement: 250; high agreement: 178). Experiment 2 varied the number of syllables of the picture names. The conditions were matched for f requency, AoA, and name agreement. Disfluencies occurred more often with short words (349) than long words (269). This counterintuitive effect was mainly driven by pauses (short words: 268; long words: 200).

Schnadt and Corley (2006) observed increased disfluency rates in a condition that confounded low name agreement with low word frequency. Our study shows a frequency-independent effect of name agreement and an effect of word length. Because these variables affect linguistic planning stages, we conclude that disfluencies are not solely caused by high conceptualization demands. In contrast to current views, Experiment 2 showed that disfluency rates sometimes increase when speech planning is easier (the short-word condition). We suggest that speakers attempt to deliver short and long words at a similar pace, and use pauses to synchronize this timing.
Up

Sarah Bernolet (Ghent University)

S. Bernolet, S. Schoonbaert, & R. Hartsuiker

Long-term syntactic priming in spoken and written sentence production

Speakers tend to reuse the syntactic structures of recently processed sentences (syntactic priming). Lexicalist accounts explain this as the result of residual activation of syntactic representations, which are connected to lexical representations. Implicit learning accounts explain it as a result of continuous learning of syntactic procedures. Lexicalist accounts are supported by a “lexical boost”: Priming is stronger when prime and target sentences use identical rather than different verbs (Pickering & Branigan, 1998). Implicit learning accounts are supported by the longevity of priming in spoken production: Priming remains even when 10 fillers separate prime and target (Bock & Griffin, 2000). However, it is unclear whether there is long-term priming in writing. Branigan et al. (1999), testing only same-verb conditions, found rapid decay in a written completion task; Speybroeck et al. (2005) replicated this, but observed the strongest decay in same-verb conditions .

To test whether priming decays at different rates in speaking and writing, our experiment directly compared long-term priming and lexical boost in these modalities. Forty-eight naïve participants took part twice (1 week between sessions), once in each modality. Order of modality was counterbalanced. Naïve participants interacted with a pseudo-participant, who produced (scripted) prime sentences. The participants described each other pictures; in one condition they spoke, and in another condition they typed, using a (simulated) computer chat system. Prime sentences were Dutch Prepositional Object (PO) and Double Object (DO) datives (1-2 provide English translations). We varied prime structure, verb-repetition (same vs. different), and number of fillers separating prime and target (Lag: 0, 2, or 6).The dependent variable was the proportion of PO responses.

(1; PO) The nun sells a book to the dancer
(2; DO) The nun sells the dancer a book

There was a main effect of prime, a prime x lag interaction (suggesting some decay of priming), but no prime x lag x modality interaction (suggesting similar decay in written and spoken production). Importantly, there were significant priming effects at each lag separately, both in written and spoken production. The lexical boost was only significant at lag 0 in spoken production. Unexpectedly, priming effects and the lexical boost were smaller in session 2, perhaps suggesting that episodic memory can affect syntactic choice. Consistent with predictions of Chang et al.’s (in press) implicit learning account, syntactic priming is long-lived, and the lexical boost is short-lived, in both speaking and writing.
Up

Session 2: Bilingualism

Kristof Strijkers (Ghent University)

K. Strijkers, A. Costa, M. Santesteban, & R. Hartsuiker

Lexical access in bilingual speech production: ERP correlates from language switching in highly proficient bilinguals

An important ability for bilinguals is that of keeping their two languages separated during speech production. Several researchers have argued that the attentional mechanisms responsible for this ability rely on inhibitory processes (Green, 1986, 1998), while others argue that they rely on language-specific selection processes (Costa, et al., 1999). The asymmetrical switching cost found when bilinguals switch between their L1 and L2 (e.g., Meuter & Allport, 1999), is in favour of the inhibition hypothesis. However, recently Costa and Santesteban (2004) showed that the switch cost for highly proficient bilinguals is symmetrical. On top of this symmetrical switch cost, the authors found a rather remarkable result: Naming was always faster in L2. An ERP-experiment of picture naming in a language switching task is reported, in which the brain correlates of the symmetrical switch cost and language effect were explored. In the behavioural results, the symmetrical switching cost and language effect were replicated. In the electrophysiological measures, three main switch-related ERP results were found: (1) no modulation of the ‘inhibitory’ N2 component, supporting the notion that highly proficient bilinguals do not make use of inhibition for lexical selection. (2) A late positive complex (LPC) over frontal and parietal cortices associated with the general switch cost due to stimulus-response reconfiguration, and (3) a novel late negative complex (LNC) over parietal and temporal cortices that may be associated with a language specific switch cost. Finally, two language-related ERP-components, who may be responsible for the language effect, were visible: (1) An early positive component (P2) mainly over frontal brain region s; and (2) a late negative component (LNC) over parietal and temporal brain areas. In summary, we conclude that highly proficient bilinguals do not rely on inhibitory control processes, and that they have shifted from inhibitory processes to language-selective processes.
Up

Robert Maier (Ghent University, University of Edinburgh)

Translating Dutch passives into English: the influence of structural properties on spontaneous sentence translation

Studies of cross-linguistic structural priming such as Hartsuiker, Pickering and Veltkamp (2004), Loebell and Bock (2003), or Meijer and Fox Tree (2003) support the assumption that at least some structural-syntactic information is shared between the systems of different languages. To study the sharing of structures, combinations of languages A and B are useful where language A offers one or more alternatives to the syntactic structure in question, whereas language B does not. This relation holds true between passives in Dutch and English: while both languages allow Pat Aux Vpart Ag structures as in (1) and (2), only Dutch allows a "wide" passive with Pat Aux O Vpart as in (3).

(1) The cow is pulled by the farmer.
(2) De koe wordt getrokken door de boer.
(3) De koe wordt door de boer getrokken.

Dutch-English bilinguals translated active sentences and passives of type (2) and (3) from Dutch into English. Method and results of this experiment will be discussed - early data analysis finds that reaction times to active sentences are shorter than to the class of passives as a whole, while no significant differences in reaction times between translation of "wide" and "narrow" Dutch passives have been found.
Up

Erica Smits (University of Antwerp)

E. Smits, D. Sandra, H. Martensen & T. Dijkstra

The activation of non-target language spelling-to-sound rules in word naming: The between-language inconsistency effect in pure and mixed-language stimulus lists

Dutch-English bilinguals named English words that had a between-language inconsistent rime, (BOOST), a consistent rime, (SKIP), a typically English rime (TAKE) and nonwords had an inconsistent rime (SPEEF), or a consistent rime (SLEM). In Experiment 1, the stimulus list contained only English words, whereas Experiments 2 and 3 also contained Dutch words. In Experiment 2, participants were asked to name the nonwords in English, whereas they could choose the language of pronunciation in Experiment 3. The results show that the three types English words were named equally fast in Experiment 1. The responses to typically English words were slowed down significantly and to the same extent in the two mixed language experiments in comparison to the pure list (27 ms and 36 ms). The English words with an inconsistent rime were slowed down significantly more so (43 ms in Experiment 2 and 74 ms in Experiment 3). The difference between both mixed conditions was not significant. The same pattern was observed for the number of erroneous Dutch pronunciation responses. Nonwords showed the same pattern of results as the English words, and between-language inconsistent Dutch words (GOOR), were responded to more slowly in both mixed experiments than typically Dutch words (SNAAR). In conclusion, a mixed stimulus list has two separate effects: 1. Response insecurity (alternate English and Dutch pronunciation) slows down responding. 2. A response conflict arises between both possible pronunciations for between-language consistent items. This indicates that both sets of grapheme-to-phoneme conversion rules (English and Dutch) are activated bottom-up. It is, however, difficult to explain the null-effect of inconsistency in the pure list. The absence of the effect can either indicate a lack of activation of the Dutch GPC-rules, or a response mechanism that favors the English response. We will address the problems that arise with both explanations.
Up

Eva Van Assche (Ghent University)

E. Van Assche, W. Duyck, & R. Hartsuiker

Processing of interlingual homographs is sentences: Evidence from reaction times and ERPs

This study investigates the activation of semantic information in the first language (L1) when reading words or sentences in the second language (L2). The experiments make use of interlingual homographs (words with the same orthographic form but different meanings across language, e.g. stage is an English-Dutch homograph) as critical stimuli. A first pilot study generated words that are semantically related to the English reading and words that are semantically related to the Dutch reading of the homographs (e.g. the English related word pair stage-audience and the Dutch related word pair stage-work). For each related target word, an unrelated control word was selected (e.g. the English unrelated word pair stage-marriage and the Dutch unrelated word pair stage-feel). In a first experiment, we tested the priming pattern generated by these word pairs. Dutch-English bilinguals had to perform an English semantic relatedness judgment task. The results showed faster reaction times (RTs) for the English related pairs but slower RTs and more errors for the Dutch related pairs. This means that there is an influence of the L1-semantic representations of the homographs when performing a task in the L2. In the following two experiments, the homographs will be presented in a high (e.g. The public is waiting for the band to get on stage) or low constraint sentence context (e.g. Lisa is waiting for her three friends to build the stage) to investigate the role of sentence constraint on the activation of semantic representations. Participants will perform the same semantic relatedness judgment task in a reaction time and ERP- experiment.
Up

Session 3: Sentence Comprehension 1

Jana Haeussler (University of Konstanz)

Interference effects during subject retrieval

The process of computing syntactic representations during language comprehension includes the checking of various dependencies between items which can be separated by several other items. In order to do so the first element of a dependency relation has to be retrieved from the syntactic structure built so far on encountering the second element. In this presentation, I will focus on subject-verb agreement in German verb-final sentences. Checking subject-verb agreement requires subject retrieval after encountering the clause-final verb. The subject’s number specification has then to be compared with the verb’s number specification. An interfering element (a ‘distractor’) sharing some properties with the subject might disrupt this checking process. In instead of retrieving the number specification of the subject, the number specification of the distractor might be retrieved, leading to a seeming agreement error if subject and distractor differ in number. I will present results from four experiments using the method of speeded-grammaticality judgments. All experiments investigated embedded subject-object sentences with the verb in clause-final position. Subject and object either matched or mismatched in number (cf. (1)).

(1)
a.        dass die Lehrerin das Kind natürlich getröstet hat
            that the teacher-SG the child-SG of-cause comforted has
b.         dass die Lehrerin die Kinder natürlich getröstet hat
            that the teacher-SG the children-PL of-cause comforted has

Participants made more judgment errors for sentences like (1b) (and corresponding sentences with a plural subject and a singular object). Experiment 1 shows, that an adverbial between subject and object reduces the number of judgment errors. Experiment 2 and 3 show that unambiguous case-marking reduces interference. An ongoing fourth experiment investigates the role of animacy. Preliminary results show that inanimate objects cause less interference than animate objects. I will discuss how these results can be used to gain insights on the checking mechanism and the actual retrieval cues.
Up

Els Severens (Ghent University)

E. Severens, & R. Hartsuiker

Subject-verb agreement investigated with Event-Related potentials

The aim of the present study was to investigate if morphophonological factors influence the comprehension of sentences containing attraction errors. In Dutch it has been found in a production study that more errors are made when participants have to orally finish sentence fragments like De straat bij de kerken (The street [neuter gender] near the churches) than when they have to finish sentence fragments like Het plein bij de kerken (The square [common gender] near the churches) (Hartsuiker, Schriefers, Bock, & Kikstra, 2003). This can be explained by a biased self-monitor for which it is harder to detect errors when the article is ambiguous in number; as in the sentence fragments starting with “de”. In the present study participants had to read sentences like De straat bij de kerk/kerken is/zijn schoon (The street [neuter gender] near the church/churches is/are clean)and Het plein bij de kerk/kerken is/zijn schoon (The square [common gender] near the church/church es is/are clean) while their EEG was measured. It has been found that when subject and object match in number there is a clear P600 effect in both the “de” and the “het” condition. When subject and object mismatch in number there was a small P600 with an anterior distribution in the “het” condition. In the “de” condition there were no significant differences between the correct and the incorrect sentences. The results will be discussed in light of a biased self-monitor.
Up

Ingrid Christoffels (Maastricht University)

I. Christoffels, V. van de Ven, E. Formisano, & N. Schiller

Speaking but not hearing: The sensory consequences of verbal feedback in the brain

Speakers usually hear themselves while speaking and use this external auditory feedback to monitor their speech. In effect this means that speaking always introduces speech input. Recently, we showed that the cortical response in the bilateral superior temporal gyrus (STG) was attenuated when speaking with normal feedback. This finding corresponds to a feedback mechanism which compares the expected sensory consequence of an action (motor-to-sensory discharge) with actual sensory feedback, dampening the sensory response. In this framework the amount of overlap between the expected and actual auditory feedback is expected to determine the amount of attenuation. In the present study, this prediction was tested by parametrically manipulating the quality of verbal feedback during speaking. Preliminary findings suggest that the quality of feedback indeed affects the amount of attenuation in the STG. This supports the idea of a precise prediction of sensory input that is the consequence of speech. However, STG activity also appears to be reduced in general during speaking compared to listening. The attenuation of the cortical response to speech may be an important mechanism by which self-generated and external speech can be distinguished.
Up

Session 4: Phonology and Auditory Word Recognition

Alain Content (Free University Brussels)

The role of syllable onsets in spoken word recognition

As the classic saying illustrates, <i>it's hard to recognize speech</i>, partly because of the absence of absolute cues to word boundaries. Previous studies in French using either the word spotting task or the cross-modal repetition priming paradigm have provided converging evidence that the alignment between the beginning of a word and the beginning of a syllable (as for «robe», <i>dress</i>, in the nonsense utterance /nal.ROb.gan/ vs. /na.fROb.gan/) is an influential factor in spoken word recognition. Other cues such as rhythm and lexical knowledge are however known to contribute to word segmentation. In this talk, I'll present a series of follow-up cross-modal priming experiments in which we examined the interplay between syllable onset segmentation and lexical knowledge. As in previous experiments, a priming advantage was obtained when the target was aligned with the beginning of a syllable (e.g. «rose» with either /dul.Roz/, aligned or /vu.bRoz/, misaligned auditory primes). However, the onset alignment effect disappeared when the preceding context was meaningful, suggesting that the onset alignment effect was cancelled out by the segmentation cue provided by the identification of the preceding context word. A further experiment manipulated the phonetic realization of the prime sequences. The results indicate that the influence of lexical information is modulated by fine phonetic or prosodic features that index the structure of the utterance.
Up

Michael Stevens (Ghent University)

M. Stevens, J.M. McQueen, & R. Hartsuiker

Perceptual adaptation to accent characteristics

When we listen to someone, we have to deal with the talker's idiosyncrasies in their realization of speech sounds. Norris, McQueen and Cutler (2003) showed that part of the adaptation to a specific talker is achieved by a lexically-driven retuning of phoneme boundaries. In their exposure-test paradigm, listeners first heard a talker who produced an ambiguous fricative that was equally similar to /f/ and /s/. One group of listeners was lexically biased to interpret the ambiguous sound as /f/ (they heard e.g. /wItlo?/, from witlof, 'chicory'). The other group was biased to interpret it as /s/ (they heard e.g. /na:ldbo?/; from naaldbos, 'pine forest'). In the subsequent test phase, listeners from the /f/-biased group categorized more sounds on an /f/-/s/ continuum as /f/ than listeners with /s/-biased exposure. We can thus compensate for the idiosyncratic production of a speech sound by shifting our phoneme boundaries after exposure to that sound in appropriate lexic al contexts.

The current study investigates whether this perceptual adaptation is also used in adaptation to dialectically marked speech. Dutch and Flemish listeners were exposed to words produced by a Flemish talker whose realization of either /x/ or /h/ was ambiguous (the ambiguous /x/ reflects an actual property of the dialect of West-Flanders). Contrary to our expectations, no lexically-driven retuning was found. Both /x/- and /h/-biased listeners learned to interpret more sounds as /x/ after exposure. This is attributed to the fact that /h/ is strongly influenced by coarticulation. As the realization of this sound is not stable across contexts, it is futile to adapt its representation every time a different realization is heard.
Up

Marjolein Reinaerts (Radboud University Nijmegen)

M. Reinaerts, P. Gulpen, H. Diepstra, & B. Maassen

The phonological development in dutch children at genetic risk of dyslexia between ages of 23 and 35 months

To date, dyslexia can be reliably diagnosed after a child has received a few years of reading and writing education. By that time, however, many children have already experienced negative consequences of the disorder. The ‘Dutch Longitudinal Study of Dyslexia’ aims to find early precursors of dyslexia, to be used for early intervention. Several studies examined the early language skills in Dutch children at-risk of dyslexia1,2. Until now, no research has been published on the early phonological development of these children. Aim of the current study was to look for deviancies in the phonological development of young Dutch children at-risk of dyslexia.

Spontaneous speech samples were collected from thirteen at-risk children (based on confirmed dyslexia in first-degree relatives) and eight control children at ages of 23, 29 and 35 months in a semi-structured setting of book reading and playing. Verbal responses were phonetically transcribed, such that speech sounds and syllable structures could be determined and compared to the targets.

The collected speech measures show consistent age-related changes, demonstrating the sensitivity of these measures to speech development. Importantly, the at-risk group shows a tendency to perform poorer than the control group. In particular, a subgroup of the at-risk children produce more atypical phonological processes and less syllable initial clusters correct. These subtle speech production deficits of at-risk children may be used as early indicators of dyslexia.
Up

Session 5: Visual Word Recognition

Dirk Koester (Maastricht University)

D. Koester, T. Gunter, & H. Holle

Neurophysiological correlates of lexical-semantic composition

Most languages have morphologically complex words that consist of two or more words, so-called constituents (e.g. bath + towel). The number of compound constituents is linguistically not limited (bath + towel + rack…). In German, Dutch, or English, compounding is done on a daily basis and many novel compounds are created. In order to understand a novel compound, decomposition into its constituents is necessary. The present experiment explores how these constituents are semantically integrated. Is integration done incrementally or does the parser await the end of the compound where integration is done? In order to explore this issue, subjects were auditorily presented with novel compounds that consisted of 3 noun constituents. Semantic expectancy between the first and the second constituent was varied (high vs. low). Similarly and independently, the semantic expectancy of the third constituent compared to the combination of constituent 1 and 2 was manipulated. ERPs o n the third constituent showed a clear N400 effect, indicating that semantic integration took place at the last constituent of the compound. The second constituent showed a small negativity with a left central scalp distribution followed by a parietally distributed positivity for low expected items. Although the pattern of results at the second constituent was significantly different from that of the third constituent, it suggests that the ERPs showed sensitivity towards the lexical-semantic relation between the first and the second constituent. This result suggests that semantic processes, possibly integration, are not restricted to the final, ie. crucial constituent of novel compounds in German.
Up

Emmanuel Keuleers (University of Antwerp)

Leanlex: A library for selecting and processing information from lexical databases in Python
http://www.cpl.ua.ac.be/Members/emmanuel/leanlex/

For many psycholinguists, lexical information sources such as CELEX (Baayen, Piepenbrock & Gulikers, 1995) are an invaluable resource for constructing experimental stimuli, obtaining material for computational modeling, and a variety of other tasks. However, selecting and processing data from these sources is often met with apprehension. The need to make this task accessible is apparent in the proliferation of specialized tools, graphical interfaces, and pre-processed files in spreadsheet format. While the value of these solutions is not to be underestimated, they do not escape the trade-off between ease of use and flexibility, and problems can quickly fall beyond their scope. The only current alternative is to process the data files containing the “raw” lexical information in high-level programming languages, a solution that trades off ease of use for flexiblity: it is time-intensive, difficult to master, and because the code is not always easy to read, it is unap pealing to unexperienced programmers. Leanlex is a general purpose solution for selecting and processing lexical data that alleviates some of the problems associated with the latter approach while maintaining all the flexibility inherent in the use of a high-level programming language. Leanlex offers immediate and exhaustive access to lexical information in the Python language, which is known for its readability, conciseness, and ease of use. The aim of Leanlex is to allow psycholinguists to resolve simple and complex problems involving lexical information in easy to read self-documenting code, without sacrificing either ease of use or flexibility.
Up

Session 6: Language Comprehension & Self-Monitoring

Jieun Kiaer (King's College London)

Processing Sentece-initial Dative NP in Korean: via case and prosody

This paper argues that in Korean, a sentence-initial dative NP as in (1) can be resolved incrementally in on-line processing by case-prosody interaction. Based on Dynamic Syntax(Cann et al 2005), which assumes that structure building is incremental from left-to-right, I argue that incremental/local parsing via constructive case is default in Korean (See Miyamoto2002 for Japanese case).And non-local parsing is cued by prosody. I focus on Intonational Phrase boundary(IP). In(1), the dative particle –hanthey provides an instruction to build a third argument, anticipating a ditransitive verb. However, it may not provide the information that determines whether the dative-NP node is to be resolved locally(1a) or non-locally(1b). IP boundary at this point helps the parser’s decision. When IP boundary comes after dative-NP, parsers resolve it in the non-local, embedded clause(1b). When the IP boundary appears after the topic-NP, parsers resolve it in the local, matrix clause(1a). I present two novel psycholinguistic tests to probe case and prosody interaction. In off-line completion and phrasing, subjects were asked to phrase the fragment (underlined in (1)) and complete it naturally. Results show strong correlation between phrasing and completion, supporting Implicit Prosody Hypothesis (Fodor 1998). In on-line cross-modal self-paced reading, the auditory input is given in the first region (underlined in (2)) and the other regions were presented visually. Slow-down was observed in region 3 when IP boundary appears after the first dative NP.

Examples

(1)        Pomi-hanthey Jina-nun Mina-ka sakwa-lul cwuessta-ko malhaysse.
            Pomi-DAT Jina-TOP Mina-NOM apple-ACC gave-COMP said

            a. Jina said to Pomi that Mina gave an apple to her
            (IP boundary after Jina-nun: Dative NP(=Pomi-hanthey) Matrix Clause reading)
            b. Jina said that Mina gave an apple to Pomi.
            (IP boundary after Pomi-hanthey: Dative NP(=Pomi-hanthey) Embedded Clause reading)

(2)        Pomi-hanthey% Jina-nun Mina-ka(1) sakwa-lul(2) Kiho-hanthey(3) cwuessta-ko(4) malhaysse.(5)
            Pomi-DAT Jina-TOP Mina-NOM apple-ACC Kiho-DAT gave-COMP said
Up

Nina Versteeg (Utrecht Institute of Linguistics OTS)

N. Versteeg, T. Sanders, & F. Wijnen

Knowing what’s going on. The influence of contextual knowledge on the on-line resolution of structural ambiguity

NP/S-coordination ambiguities such as those in (1) were investigated (Versteeg, Sanders & Wijnen, submitted).

1a. Pete was very annoyed by the unfounded opinion of Mary and Ellen, a remark that they both found very rude. [NP-coordination]

1b. Pete was very annoyed by the unfounded opinion of Mary and Ellen listened attentively to his words. [S-coordination]

Previous research has shown that readers are inclined to initially interpret the ambiguous NP Ellen as part of a complex object NP, as in sentence (1a) (Frazier, 1987; Hoeks, Vonk & Schriefers, 2002). However, when you knów that Pete is annoyed by Mary but not by Ellen, does that affect your inclination to treat the last two proper names in the string “Pete was very annoyed by (…) Mary and Ellen…” as a coordinated noun phrase? To investigate this question, a completion study, a judgment study and a moving window self-paced reading study were conducted. Temporarily ambiguous S-coordinations were embedded in contexts that were either supportive or non-supportive of S-coordination. In brief, the results of all three experiments show that parsing “… X and Y …” as an S-coordination is easier when the discourse context makes it clear that X and Y do not belong together. This supports the notion of an interactive parser, in which contextual knowledge has an immediate impact on structural attachment decisions.
Up

Katrien Colman (Groningen State University)

K. Colman, J. Koerts, M. van Beilen, R. Bastiaanse , K.L. Leenders

The role of cognitive mechanisms is sentence comprehension in Dutch speaking Parkinson's disease patients: preliminary data

Conclusion
The deficient performance of PD patients on processing the incorrect long passive sentences cannot be explained by deficits in verbal working memory, but rather by deficits in cognitive set-switching and complex sequencing abilities.
Up

All questions, registration, etc please e-mail pif2006@ugent.be