Adele E. Goldberg
Biodata
Adele Goldberg is Professor of Linguistics at Princeton University. After getting her Ph.D. from UC Berkeley in 1992, she taught at UCSD and at the University of Illinois, Urbana-Champaign before moving to Princeton in 2004. She has also taught at several previous LSA Linguistic Institutes (1999, 2001, 2005, 2007) and at the LOT winter school in Amsterdam (1995, 2006). She is author of Constructions: a Construction Grammar approach to argument structure (1995) and Constructions at Work: the nature of generalization in language (2006). Her research interest is the psychology of language, particularly the way form and meaning are related, and the way these pairings are learned and represented.
Abstract
“Learning what not to say: the nature of statistical preemption in a-adjective distribution”
How do learners constrain constructional overgeneralizations? A number of theorists have suggested that statistical preemption plays a role (Foraker et al., 2007; Goldberg, 1995; 2006; Pinker, 1981). This hypothesis has not received a great deal of attention in the experimental literature, except in notable work by Brooks and colleagues. Brooks has demonstrated that seeing novel intransitive verbs in periphrastic causative constructions significantly preempts children’s use of them in simple transitives (Brooks & Tomasello, 1999; Brooks & Zizak, 2002). The present work extends this finding by demonstrating that the existence of a category enhances preemption effects, and by demonstrating that speakers use preemptive contexts rationally (discounting pseudo-preemptive contexts).
A-adjectives are multisyllabic adjectives that begin with schwa, can be morphologically segmented into a- plus a semantically related stem (e.g., a-live, a-sleep, cf. adult ≠ a-dult), and crucially, disprefer appearing prenominally (??the asleep boy). We examined adult naturalistic production in four experiments, all of which required participants to describe scenes in which one of two animals with different adjective labels moved to a star. This procedure resulted in either a relative clause (RC) or prenominal use of the targeted adjective (e.g., 1 or 2).
(1) Prenominal: The sleepy/??asleep/?adax fox moved to the star.
(2) Relative Clause (RC): The fox that’s sleepy/asleep/adax moved to the star.
Experiment 1 established that novel a-adjectives (e.g., adax) disprefer prenominal use relative to non-a adjectives (e.g., chammy) to a significant extent. This indicates that participants tentatively assimilate never-before-seen a-adjectives to the category of familiar a-adjectives. Experiment 2 tested to see whether learners were sensitive to statistical preemption, and found that witnessing a novel a-adjective used in a preemptive (RC) context just three times dramatically decreased incorrect prenominal uses so that all novel a-adjectives behaved indistinguishably from familiar a-adjectives in avoiding prenominal distributions. Experiment 3 showed that learners disregard pseudo-preemptive input. That is, speakers did not display an increased avoidance of prenominal uses when exposed to pseudo-preemptive contexts like (3), presumably because they rationally attributed adax¹s appearance in the RC to the complex adjective (cf. 4), rather than to adax.
(3) The hamster, adax and proud of itself, moved to the star.
(4) *The proud of itself hamster moved to the star.
Fillers were used to obscure the goal of the experiment and debriefing confirmed that speakers were unaware of our manipulations.
To our knowledge this set of experiments is the first to systematically explore various factors involved in statistical preemption. They shed light on the vexed question of how speakers learn what not to say, a critical issue within the constructionist approach to language.
