Relative clause processing in Tongan: an effect of syntactic ergativity on the object preference

In the psycholinguistic literature, it has been established that the processing cost for the subject relative clauses (SRC) is lower than that for the object relative clauses (ORC) in many languages (Kwon, et al., 2010, for review). However, most of the previous studies rely on the data from a small set of languages, which typically have nominative-accusative case alignment with SVO or SOV word order. This study examines the relative clause processing in Tongan, a Polynesian language with VSO word order and ergative-absolutive alignment to address the question of whether the widely observed SRC preference is indeed a universal processing bias.

Previous studies on relative clause processing in ergative languages are sparse and show mixed results. While Chamorro, a verb-initial language showed the subject advantage (Wagers, et al., 2018), Avar (Polinsky, et al., 2012) and Basque (Carreiras, et al., 2010), with SOV order, showed the ABS advantage, preferring ORC. Tongan differs from these languages in that it exhibits syntactic ergativity: transitive SRC requires a pre-verbal resumptive pronoun (RP), ne, while ORC doesn’t, as shown in (1). This extra grammatical complexity with SRC is expected to give ABS a processing advantage in Tongan.

A self-paced reading study was conducted using the software Linger (written by Douglas Rohde), with which the reading time for each phrase was measured. Sixteen sets of SRC target sentence pairs and sixteen sets of ORC target sentence pairs were prepared, along with 40 filler sentences. The stimulus sentences were followed by a yes-no comprehension question. A sample set of SRC target sentence pairs is shown in (2).

The critical region (underlined) contains an absolutive noun phrase (ABS NP) in both conditions. If native speakers of Tongan have the ORC preference, they would posit a gap in the object position upon seeing the tense and verb that mark the beginning of the relative clause. Under the SRC condition (2b), the ORC preference leads the reader to predict an object gap corresponding to e loea ‘the lawyer’ and consequently, takes the 3rd person singular pronoun, ne to be the subject rather than an RP. The appearance of the ABS NP ‘the pilot’ in (2b) will then reject such an ORC analysis and cause a filled-gap effect (Stowe, 1986), compared to the same ABS NP under the no-gap condition (2a), where there is no need to posit a gap in the ABS object position and hence the existence of the ABS NP should cause no trouble.

Sixty-one native speakers of Tongan participated in the study. The mean accuracy rate for the comprehension questions for the SRC target sentence pairs was 69%. One participant was excluded whose accuracy rate was exceedingly low; similarly, responses were excluded that were beyond 4,000ms and below 100ms. The rest of the data was modeled using LME, with random intercepts for participants and items. The mean reading time at the critical region showed that there was a slowdown in the SRC condition (2b), compared to its control no-gap condition (2a) (b= 162.55, t = 3.63, p < 0.01). On the other hand, no region showed a significant difference in the ORC target sentence pairs.

The reading time pattern suggested that there was a filled-gap effect in the SRC condition. Upon seeing the verb, the parser wrongly posited a gap in the ABS object position, assuming that it is reading an ORC, but the structure turned out to be an SRC. This ORC preference suggests that there is an ABS advantage in Tongan, like Avar and Basque. Unmarkedness of ABS in ergative languages influences on the processing cost of the relative clauses, although the ORC structure actually creates a longer filler-gap dependency (both structurally and linearly). This suggests that the structural distance in RCs (Hawkins, 2004; O’Grady, 1997) or the linear distance (Gibson, 2000) does not play a major role in Tongan relative clause processing. Subject advantage was not observed, unlike Avar and Chamorro. The subject advantage in Avar, an ergative language (Polinsky, et al. 2012), suggests that ergative-absolutive case system is in principle compatible with the subject advantage. We suggest that the syntactic ergativity is one key factor that suppresses the subject advantage in Tongan.

To conclude, Tongan exhibits an ORC preference unlike many other languages examined in the literature. We argue that syntactic ergativity contributes to such a preference, further suggesting that the subject advantage in RC is not universal, but the grammar of a language like syntactic ergativity may have a strong influence for the bias.

Acknowledgement: This paper is co-authored with Hajime Ono (Tsuda University), Koichi Otaki (Kanazawa Gakuin University), Manami Sato (Okinawa International University), ‘Ana Heti Veikune (University of the South Pacific), Peseti Vea (University of the South Pacific), and Masatoshi Koizumi (Tohoku University), and was presented at the 26th meeting of the Austronesian Formal Linguistics Association. The study is funded by JSPS-KAKENHI Grant Number JP 15H02603.

Selected References: Carreiras, M., Duñabeitia, J.A., Vergara, M., de la Cruz-Pavía, I., & Laka, I. 2010. Subject relative clauses are not universally easier to process: Evidence from Basque. Cognition 115, 79–92. //Polinsky, M., Gallo, C.G., Graff, P., & Kravtchenko, E. 2012. Subject preference and ergativity. Lingua 122, 267–277. //Wagers, M. Borja, M.F., & Chung, S. 2018. Grammatical licensing and relative clause parsing in a flexible word-order language. Cognition 178, 207–221.