Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
An ERP study on agreement violations in Arabic: Qualitative differences between singular and plural subjects R. Muralikrishnana & Ali Idrissib a New York University Abu Dhabi, b United Arab Emirates University [email protected] Introduction Methods Verbs in Arabic generally agree with the subject in person, number and gender. However, for plural subjects, the verb agrees fully in the subject-verb (SV) order, but it only agrees in person and gender in the VS order. Thus, a partial agreement for plural subjects does not always entail ungrammaticality, whilst it does for singular subjects. In this ERP study, we investigated whether the processing system is sensitive to this idiosyncratic behaviour of plural subjects (in the VS order) even when processing verb agreement in the SV order. If that is so, qualitatively different ERP patterns must emerge at the verb for sentences with singular as opposed plural subjects. Previous studies on subject-verb agreement violations in various languages have reported a LAN in 40–74 % of cases depending upon the feature (person, number or gender) that was violated [1]. Given the paucity of ERP studies on Arabic, we used equal number of stimuli that violated one of these features, and averaged the effects for the three features to gain a first insight into the overall general effect of violating subject-verb agreement in this language. Results ERPs at the position of the verb : Singular versus Plural subject Acceptable Agreement Participants: § 28 (9 female) right-handed native speakers of Arabic § Mean age: 22.14 years; Age-range: 19 – 27 years EEG Data: § Recorded using ActiCap fixed at the scalp; 25 Ag / AgCl electrodes § Reference: Left-mastoid, re-referenced to linked mastoids offline § Ground electrode: AFZ; Offline filter: 0.3 – 20 Hz band-pass Procedure: § Rapid serial visual presentation of stimulus sentence § Tasks: Acceptability judgement followed by Probe detection Materials: § Sentences of the form: Adverb of time – Subject – Verb – PP. § Subject noun : masculine / feminine animate common noun § 3 Conditions (120 sentences in each condition per participant) • Acceptable / Violation of Agreement / Violation of Tense • No combined violation of two agreement features § 2 Subject-Types • Singular noun in sentences with a past tense adverb • Plural noun in sentence with a future tense adverb Tense Discussion Topography of effects for Agreement Violations VAG_S – ACP_S _ ` VAG_P – ACP_P Results § All violation conditions elicited a biphasic N400-P600 pattern. § Acceptable sentences with singular subjects elicited a P300 effect. § Agreement violations elicited a LAN when the subject was singular. § No difference between past-to-future and future-to-past violations. [1] Molinaro, N., Barber, H.A., & Carreiras, M. (2011). Cortex, 47(8), 908–930. [2] Osterhout, K., Mobley, L. (1995). Journal of Memory and Language, 34, 739-773. [3] Barber, H., & Carreiras, M. (2005). Journal of Cognitive Neuroscience, 17(1), 137–153. [4] Roehm, D., Bornkessel, I., Haider, H. & Schlesewsky, M. (2005). NeuroReport, 16(8), 875-878. [5] Roehm, D., Bornkessel-Schlesewsky, I., Rösler, F. & Schlesewsky, M. (2007). Journal of Cognitive Neuroscience, 19(8), 1259–1274. [6] Kretzschmar, F. (2010). Unpublished doctoral dissertation, University of Marburg, Germany. Whilst subject-verb agreement violations elicit a LAN effect in several languages [e.g., 2,3,4], a recent review has suggested that ‘a more robust expectation for a supposed-to-agree constituent in fact triggers a more reliable LAN effect’ [1, p. 919]. In light of this, the LAN elicited by singular subjects as opposed to plural subjects in our data could be interpreted as follows: there is only one possible agreement pattern for Arabic singular subjects, giving rise to a specific expectation as far as agreement is concerned; any deviation from this pattern would be a clear morphosyntactic violation, eliciting a LAN. By contrast, since Arabic verbs agree differentially to plural subjects based on argument order, a partial agreement may not always be a violation, so no LAN. Further evidence for such an expectation based interpretation stems from the P3b effect for singular subjects in acceptable sentences in our data. The P3b has been observed for linguistic stimuli whenever a prior expectation about the stimulus is fully met [5,6]. In line with this, it follows that when the expectation for a specific agreement pattern is fully met (as in the acceptable singular condition), a P3b is evoked. Such an expectation would be relatively less specific for plural subjects given their idiosyncratic behaviour, thus no P3b for acceptable sentences with plural subjects. We argue that our results show the differential nature of processing verb agreement for singular and plural subjects due to properties specific to Arabic. Poster Presented at the 26th CUNY Conference on Sentence Processing, University of South Carolina, Columbia, SC. Muralikrishnan & Idrissi (2013).