Title: Using acceptability judgments to study language change: Age differences reflect differences in speakers' linguistic systems Abstract: Apparent-time studies, which track change in a linguistic variable by comparing speakers of different ages within a single community, are a common approach to study variation and change. However, comparing older and younger participants presents a potential confound for linguists using psycholinguistic measures which correlate with processing ability because processing ability decreases with age. Waters & Kaplan (2001) found that older participants were less able to tolerate the processing load which accompanies syntactic complexity. Thus, differences between age-groups could be attributable to this general cognitive pattern, as opposed to a change in linguistic constraints. This project examines ongoing contact between Malayalam (Dravidian) and English in India. I present results from an acceptability judgment task which show that age correlates positively with flexibility in constituent order. I conclude that the results are more likely to reflect differences in speakers’ linguistic systems as opposed to a general decrease in tolerance for syntactic complexity. Canonical order in Malayalam is SOV; all 6 logical orders are grammatical and attested. Non- canonical orders, which are more complex at the level of syntax and pragmatics, result in increased processing difficulty and lowered acceptability ratings (e.g., Weskott & Fanselow 2011, Kaiser & Trueswell 2004, Miyamoto & Takahashi 2001). Given that older individuals have been shown to be less tolerant of increased processing load, age should correlate with higher ratings for canonical orders and lower ratings for non-canonical orders: decreased flexibility. 44 native-speakers of Malayalam residing in the same area of India participated in this study. Language background surveys were conducted with each participant, who were aged from 18-82 and represented a variety of socio-economic statuses. Malayalam has no subject-verb agreement and differential object marking, and each experimental stimulus had three constituents: an animate subject, an inanimate object, and a verb. This ensured that the semantic role of each argument would be unambiguous. Participants rated 5 tokens of each of the 6 logical variants of transitive sentences on a 7-point scale, along with 40 filler items of varying acceptability. Stimuli were distributed among six lists pseuorandomly using a Latin Square. Results (as z-scores computed across fillers and experimental items) are presented in Figure 1 (error bars=SE). The data is grouped by “Younger” and “Older” based on a median split in age. The difference between canonical SOV and noncanonical orders was larger for the younger partic- ipants than for the older participants, reflecting less flexibility. The difference between the groups is driven by a relative preference for SOV for younger speakers: treating age as a continuous variable, the difference between SOV and OSV is negatively correlated with age (Fig. 2), though this correlation is small (Pearson’s r=-.27) and just marginally significant (p<0.064). Age does not correspond to higher acceptability across the board, as shown in Figure 3. Higher ratings for more syntactically complex sentences is the opposite of what is predicted for older speakers, which suggests that the differences in acceptability are not due to decreased tolerance for processing load. Age correlates positively with several measures of contact with English in this population (medium of school instruction, engagement with English media, attitude towards code-switching), suggesting that language experience could be a source of variation in constituent order. In fact, languages with flexible orders have been shown to become more rigid due to contact (Heine 2008), and, as Malayalam has undergone contact-induced changes in other domains, this is a plausible explanation of the results. Though apparent-time studies are not ideal (they downplay language change across the lifespan), using this approach with acceptability tasks can reveal relatively subtle inter-speaker variation.