2018 Friday Session C 1100

Friday, November 2, 2018 | Session C, Terrace Lounge | 11am

Look at THAT: Using deixis to evaluate prediction-via-simulation with L1 children, L1 adults, and L2 adults
T. Reuter, C. Lew-Williams

A number of recent theories propose that prediction facilitates efficient language processing.1,2 Supporting this idea are findings that listeners can use verb semantics3 and number markings4 to predict upcoming referents. However, precisely how prediction occurs during language processing remains uncertain. One prominent theory is prediction-via-simulation: Listeners use language production mechanisms to simulate the speaker’s production2, which is contingent, at least in part, on perspective-taking and on experience.

To evaluate this theory, we assessed whether listeners can use spatial deixis (this, that, these, and those) to predict the plurality and proximity of a speaker’s referent. In an eye-tracking task, English L1 adults (n=28), English L1 children (n=28), and English L2 adults (n=22) viewed scenes that included one speaker and four referents. Referents were arranged such that two were proximal to the speaker and two were distal; two were plural and two were singular. Participants listened to deictic sentences (e.g., Look at that wonderful apple) and neutral sentences (e.g., Look at the wonderful apple). In this design, successful use of deixis to predict the proximity of the target required perspective-taking, because deictic words are typically anchored on the speaker’s perspective.5 We evaluated this ability by comparing listeners with varying language production experience, thus providing an important empirical test of prediction-via-simulation.

We first measured whether listeners could predict the target referent’s plurality. We did so by analyzing looks to plural referents for singular and plural deictic sentences (this/that vs. these/those) and for singular and plural neutral sentences. For all three groups, we found that the plurality effect began earlier for deictic sentences (L1 adults: -600–1000 ms, ps<0.05, L1 children: -100–1000, ps<0.01, L2 adults: -600–1000, ps<0.05) than for neutral sentences (L1 adults: 300–1000, ps<0.001, L1 children: 500–1000, ps<0.001, L2 adults: 300–1000, ps<0.01). The earlier divergence for deictic sentences suggests listeners used deixis number marking to predict the referent’s plurality (Figure 1).

To evaluate whether listeners predicted the referent’s proximity to the speaker – the main test for prediction-via-simulation – we analyzed looks to proximal referents for proximal and distal deictic sentences (this/these vs. that/those) and for proximal and distal neutral sentences. For L1 adults, the proximity effect began earlier for deictic sentences (200–1000 ms, ps<0.001) as compared to neutral sentences (400–1000, ps<0.001). In contrast, for L1 children and L2 adults, the proximity effect did not begin earlier for deictic sentences (L1 children: 400–1000, ps<0.01, L2 adults: 400–1000, ps<0.001) than for neutral sentences (L1 children: 300–1000, ps<0.05, L2 adults: 400–1000, ps<0.01). Thus, only L1 adults predicted the referent’s proximity (Figure 2).

We argue that prediction-via-simulation supports processing in the mature, native speaker, but that extensive experience with cues in a language may be required before perspective-taking can support prediction. This three-group investigation goes beyond the goal of assessing whether prediction occurs and evaluates how prediction occurs – a crucial goal for defining prediction’s role in language processing and learning.