Saturating the Anticipatory Mechanisms of Syntactic Priming during Situated Language Understanding Eunice Fernandes, Moreno I. Coco and Holly P. Branigan Background In situated language understanding, syntactic priming (e.g., Bock, 1986) manifests in anticipatory looks to the (depicted) most probable upcoming linguistic element, according to the prime structure (e.g., Arai et al., 2007). Prime : The assassin will send the dictator the parcel. Target: The pirate will send the princess the necklace. Previous research reports apparent contradictory effects of cumulativity and surprisal: increased exposure to a structure (during an initial phase, e.g., Kaschak et al., 2006, or half an experiment, e.g., Jaeger & Snider, 2013: Study 3) can make it both more and less likely to be produced as target. Our Study We investigate repeated exposure within-trial and the interaction between this short-term priming and long-term adaptation on predictive eye-movements during situated language understanding. Fixations prop on S1 Predictor Est. S.E. t p (Intercept) 0,15 0,02 7,87 0,0001 activation 0,00 0,03 0,00 1,00 prime 0,01 0,03 0,19 0,85 Time1 -0,02 0,01 -2,31 0,02 Time2 0,00 0,01 -0,44 0,66 activation:prime 0,06 0,02 4,15 0,0001 activation:Time1 0,02 0,02 0,97 0,33 prime:Time1 0,04 0,02 2,12 0,03 activation:Time2 0,02 0,02 0,93 0,35 prime:Time2 0,01 0,02 0,58 0,56 activation:prime:Time1 -0,07 0,04 -1,63 0,10 activation:prime:Time2 -0,03 0,04 -0,73 0,47 Conclusion When a syntactic structure (e.g., LA) is repeatedly primed within a trial (Activation2), anticipatory effects get immediately saturated, and the alternative structure (e.g., HA) is instead evaluated. Long-term saturation effects also emerge through the experiment, but only on trials with less loaded short-term priming (Activation1). Overall, our results suggest an adaptive pattern of surprisal and an apparent lack of cumulative priming for dispreferred readings (HA). References Arai, M., van Gompel, R. P. G., Scheepers, C. (2007). Priming ditransitive structures in comprehension. Cognitive Psychology, 54: 218-250. Bock, J. K. (1986). Syntactic persistence in language production. Cognitive Psychology, 18: 355–387. Jagger, T. F. & Snider, N. E. (2013). Alignment as a consequence of expectation adaptation: Syntactic priming is affected by the prime’s prediction error given both prior and recent experience. Cognition, 127, 57–83. Kaschak, M., Loney, R., Borreggine, K. (2006) Recent experience affects the strength of structural priming. Cognition, 99 (3): B73-B82. Mirman, D., Dixon, J. A., Magnuson, J. S. (2008). Statistical and computational models of the visual world paradigm: Growth curves and individual differences. Journal Memory and Language, 59: 475-494. Primes Activation-1 The carrier airline will have more flight destinations next year. (Filler) The fan [sg] of the players [pl] / The fans [pl] of the player [sg] who will [sg] (…) Activation-2 The relative [sg] of the boys [pl] / The relatives [pl] of the boy [sg] who will [sg] (…) The fan [sg] of the players [pl] / The fans [pl] of the player [sg] who will [sg] (…) Target The father of the baby who will drink the beer/ the milk is tall Analyses We focus on looks to S1 (consistent with dispreferred HA interpretation) • Within-trial (short-term) priming: Time-course analysis of Proportion of Fixations on S1 from who onset up to 400 ms (8 slices, 50 ms each) after (Figure 1, Table 1). Maximal-random linear-mixed effects modelling under the Growth Curve Analysis approach is used for inferential statistics (Mirman et al., 2008), with Prime (HA, -0.5, LA, 0.5), Activation (1, -0.5; 2, 0.5) and Time (Linear, Quadratic) as predictors. Participants and Items are random effects. • Within-trial (short-term) and Across-trial (long-term adaptation) priming: Average Fixation Proportion on first 400ms along the experimental session (Figure 2, Tables 2-3). MLME modelling with predictors Prime, Activation and Adaptation (i.e., trials progression, from 0 to 48), and further analyses of ‘LA prime’ condition. Method 24 Portuguese participants read 1 or 2 (disambiguated) high (HA) or low (LA) attached relative clause primes and then heard a (temporarily ambiguous between HA and LA) auditory target (e.g., The father of the baby who will drink the beer/ the milk is tall) while their gaze to a picture was recorded. The picture depicted pronoun antecedents (S1: father, S2: baby) and verb objects (O1: beer, O2: milk) associated with HA and LA, plus 2 distractor objects. . Mean Fix. Prop. 400ms S1 Predictor Est. S.E. t p (Intercept) 0,15 0,02 8,88 0,0001 activation 0,00 0,02 -0,20 0,84 prime 0,01 0,02 0,26 0,80 adaptation 0,07 0,05 1,43 0,15 activation:prime 0,06 0,04 1,46 0,15 activation:adaptation -0,10 0,07 -1,37 0,17 prime:adaptation 0,07 0,07 0,91 0,36 activation:prime:adaptation -0,22 0,14 -1,57 0,12 Results Table 1 Table 2 Mean Fix. Prop. 400ms S1, prime LA Predictor Est. S.E. t p (Intercept) 0,15 0,02 7,27 0,0001 activation 0,03 0,03 1,00 0,32 adaptation 0,09 0,05 1,72 0,09 activation:adaptation -0,20 0,10 -1,98 0,05 Table 3 Within a trial: • experiencing a single HA prime (Activation 1) primes HA interpretation (classical priming). • but experiencing two HA primes (Activation2) saturates anticipation of HA; instead, experiencing two LA primes increases looks to HA antecedent (reversed priming). Across trials: • through the experimental session, participants increasingly show saturation of LA anticipation after experiencing a single LA prime. • but evaluation of the alternative HA structure after experiencing two LA primes occurs consistently from the beginning of the experiment. Figure 2 Figure 1 Grant SFRH/BD/72307/2010