Who are you talking about? Individual differences in pronoun comprehension

Abstract

Does individual variation in type of linguistic exposure affect processing strategies? We know that input affects language development [1] and reading proficiency [2], but less is known about how it affects spoken language processing. We ask how pronoun comprehension is shaped by the input, specifically exposure to written language. Pronouns like he and she are ambiguous, e.g. in Jill went skiing with Sue. She fell down, “she” could refer to either character. Adults follow several heuristics to interpret pronouns, including a bias toward the grammatical subject, e .g. Jill [3]. They also follow social cues such as eye gaze and pointing [5]. Children do not follow the subject bias as readily [4]. An unsolved puzzle is how these constraints are integrated, and why interpretation varies across individuals and situations. We hypothesized that exposure to written language increases one’s reliance on linguistic cues over social cues. Written texts are thematically organized, decontextualized, and complex, and may provide an ideal environment for learning the how language encodes information status. We tested three populations: 1) Adult native speakers; 2) Adult L2 speakers; 3) Children, using stimuli from [4]. Participants watched videos of a woman tell stories about two same-gender characters, e.g. Puppy is reading with Panda Bear. He wants the black book. Each puppet sat on a table in front of the speaker. At the pronoun, the speaker gazed at either 1) the subject (Puppy), 2) a neutral point (the book), or 3) the nonsubject (Panda). The next screen asked “who wants the black book?” If speakers rely on the linguistic context, they should choose the subject character; if they rely on gaze, they should choose the gazed-at character, and perform at chance in the neutral condition. For adults, we tested individual differences in print exposure using the Author Recognition Task [6], which is a proxy for exposure to written material, and correlates with measures of reading performance [7]. For children, we used the Title Recognition task [8]. Adults Native speakers (N=61) . Participants followed both subject and gaze biases, selecting the subject as pronoun referent 87% of the time in the neutral condition, 93% in gaze-to-subj. and 67% in gaze-to-nonsubj. conditions. Critically, people with higher ART scores chose the subject more (see Fig.). There was no effect of working memory. Adult L2 speakers (N=62). Proficient L2 English speakers performed the ART in both their native L1 Mandarin and L2 English. They did the same video task in English, including the gaze conditions described above, as well as pointing conditions in which the speaker both gazed and pointed. Gaze and pointing both had the expected effects. In addition, people with higher ART scores in Mandarin (but not English) chose the subject more. Thus, print exposure transfers from one language to another. Children ages 5-14 (n=54). Gaze again had the expected effect. In addition, children with higher print exposure chose the subject more. This effect eliminated the effect of age when both were in the model. These findings show that written language experience modulates spoken language processing, identifying one source of variation in pronoun comprehension. Theoretically, this supports models in which the input shapes language processing mechanisms during both development and adulthood, and in ways that transfer to other languages.