Language processing, acceptability, and statistical distribution: A study of null and overt subjects in Brazilian Portuguese

• Brazilian Portuguese is shifting away from null subjects toward overt subjects.
• It nevertheless shows both repeated-name and overt-pronoun processing penalties.
• Speakers’ acceptability judgments cannot explain the overt pronoun penalty.
• Corpus language use patterns also fail to explain this effect.
• A pragmatic theory that incorporates usage statistics is proposed.

Two experiments and a corpus study tested whether Brazilian Portuguese (BP), which has been argued to be shifting from null subjects toward overt subjects indeed shows a comprehension preference for reduced over fuller anaphors for salient antecedents, and whether comprehension is better explained as an imbalance between processing cost and discourse function (pragmatics account), or simply the frequency of different constructions (usage-based account). Sentences with overt pronouns were read slower (Experiment 1) but rated more acceptable (Experiment 2) than sentences with null pronouns when the antecedents were salient. Sentences with repeated names showed weaker effects than overt pronouns. The corpus analysis confirmed that BP is changing, but that null subjects are not yet infrequent, especially in academic writings. We argue that these results reflect a “pronoun avoidance strategy” in BP related to its transitory state, and propose a new view that integrates elements from both the pragmatics and usage-based accounts.