Learning the statistics of pronoun reference: By word or by category?

Evidence shows that readers tend to follow recently-encountered patterns for interpreting ambiguous pronouns. If recent exposure includes frequent pronouns with prepositional object antecedents (e.g., "Matt went to the library with Ana. She took out a book"), people adapt and are more likely to assign ambiguous pronouns to prepositional antecedents than if they were exposed to pronouns with subject antecedents (Johnson & Arnold, 2022). However, it is unclear how people categorize different referential structures and how broadly they make generalizations. Johnson and Arnold (2022) found that people can learn a referential relationship specific to third-person pronouns and an antecedent's syntactic or thematic properties. The current study uses this paradigm to test how broadly people generalize categories of different types of pronouns and antecedents. Do people adapt to the behavior of categorizing "he" and "she" as individual words, or as a general class? (Experiment 1). Do people learn about likely antecedents for pronouns separately for different verb constructions (transfer vs. joint-action) and thematic roles, or broadly by grammatical function? (Experiments 2 & 3). Participants were repeatedly exposed to a referential structure of a particular type of pronouns or antecedents, and then were tested on ambiguous pronouns. All experiments showed that pronoun adaptation generalizes to new instances from the broad categorization of pronouns and antecedents.

The frequency of referential patterns guides pronoun comprehension.

There is extensive evidence that people are sensitive to the statistical patterns of linguistic elements at the phonological, lexical, and syntactic levels. However, much less is known about how people classify referential events and whether they adapt to the most frequent types of references. Reference is particularly complex because referential tokens can be multiply categorized, raising questions about what can be learned through referential exposure. We test the role of linguistic exposure to referential patterns in five experiments on pronoun comprehension, examining linguistic contexts like "X is doing something with Y" (Experiments 1a, b, and c) and transfer events like "X gave something to Y" (Experiments 2a and b). We ask whether the interpretation of ambiguous he or she pronouns is influenced by recent exposure and find that indeed it is, supporting the hypothesis that adaptation affects discourse processing. In Experiment 1, we further ask whether adaptation persists across three types of referring expressions (he or she pronouns, I/you pronouns, and names) and find that it is limited to he or she pronouns. In Experiment 2, we test whether people can learn both syntactically conditioned and semantically conditioned frequency patterns with transfer verbs. Results showed that they learned both patterns. These results provide critical new evidence that discourse processing biases are informed by exposure to referential patterns. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

The Effects of Semantic Role Predictability on the Production of Overt Pronouns in Spanish.

In order to refer in any language, speakers must choose between explicit forms of expression, such as names or descriptions, or more ambiguous forms like pronouns. Current models suggest that reference form is driven by subjecthood, where speakers in English choose pronouns for the subject, and speakers of null pronoun languages like Spanish or Italian use null pronouns. We test this generalization by examining the effect of a different factor, thematic role predictability, on reference production in Spanish. In stories about transfer events (e.g., Ana gave a ball to Liz), speakers prefer to use pronouns more for reference to goals (Liz) than sources (Rosa and Arnold, Journal of Memory and Language 94:43-60, 2017). However, this has not been examined for null pronoun languages. In two experiments, we demonstrate that Spanish speakers are also sensitive to thematic role, but it primarily affects the rate of overt pronouns (ella, el) rather than null pronouns. These results highlight the need to include semantic constraints in models of reference production for null-pronoun languages.

Individual Differences in Print Exposure Predict Use of Implicit Causality in Pronoun Comprehension and Referential Prediction.

In three experiments, we measured individual patterns of pronoun comprehension (Experiments 1 and 2) and referential prediction (Experiment 3) in implicit causality (IC) contexts and compared these with a measure of participants' print exposure (Author Recognition Task; ART). Across all three experiments, we found that ART interacted with verb bias, such that participants with higher scores demonstrated a stronger semantic bias, i.e., they tended to select the pronoun or predict the re-mention of the character that was congruent with an implicit cause interpretation. This suggests that print exposure changes the way language is processed at the discourse level, and in particular, that it is related to implicit cause sensitivity.

Semantic predictability of implicit causality can affect referential form choice.

Models of language comprehension show that predictable elements are easier to understand. Does predictability also guide production? While many models suggest it does (e.g., Arnold, 1998; Aylett & Turk, 2004; Levy & Jaeger, 2007; Jurafsky, Bell, Gregory, & Raymond, 2001; Mahowald, Fedorenko, Piantadosi, & Gibson; Orita, Vornov, Feldman & Daumé, 2015; Tily & Piantadosi, 2009), several models suggest that it does not make speakers more likely to select pronouns (Fukumura & van Gompel, 2010; Kehler et al., 2008; Kehler & Rohde, 2013; Kehler & Rohde, 2019; Stevenson et al., 1994). Claims that predictability does not affect pronoun production are based on evidence that certain semantic roles are more likely to be re-mentioned in discourse, but speakers do not prefer pronouns for those event roles, especially when predictability stems from implicit causality estimates. These findings contrast with studies of transfer verbs, where goals are more predictable than sources, and speakers do use pronouns more for goals versus sources (Arnold, 2001; Rosa & Arnold, 2017). Our study takes a closer look at the predictability of implicit causes, using a novel experimental paradigm that is more contextualized than the methods used in previous studies. In two experiments, we find that implicit causality does affect pronominalization. This suggests that predictability may play a broad role in both reference production and language production more generally.

My pronouns are they/them: Talking about pronouns changes how pronouns are understood.

The pronoun "they" can be either plural or singular, perhaps referring to an individual who identifies as nonbinary. How do listeners identify whether "they" has a singular or plural sense? We test the role of explicitly discussing pronouns (e.g., "Alex uses they/them pronouns"). In three experiments, participants read short stories, like "Alex went running with Liz. They fell down." Answers to "Who fell down" indicated whether participants interpreted they as Alex or Alex-and-Liz. We found more singular responses in discourse contexts that make Alex more available: when Alex was either the only person in the context or mentioned first. Critically, the singular interpretation was stronger when participants heard explicit instructions that Alex uses they/them pronouns, even though participants in all conditions had ample opportunity to learn this fact through observation. Results show that the social trend to talk about pronouns has a direct impact on how language is understood.

Print exposure explains individual differences in using syntactic but not semantic cues for pronoun comprehension.

Comprehenders have been shown to use both syntactic and semantic cues to understand pronouns like he and she. In Ana threw the ball to Liz. She…, there is a syntactic bias to assign "she" to the previous subject (Ana), and a semantic bias to assign it to the goal referent (Liz). How do people learn these biases? We tested how sensitivity to these cues is modulated by linguistic experience, measured with an Author Recognition Task (Stanovich & West, 1989). In two experiments, we found both the subject and goal biases overall, but higher print exposure only predicted use of the subject bias, not the goal bias. Our results suggest that the subject bias, and not the goal bias, may be learned from exposure.

Print exposure predicts pronoun comprehension strategies in children.

Language development requires children to learn how to understand ambiguous pronouns, as in Panda Bear is having lunch with Puppy. He wants a pepperoni slice. Adults tend to link he with Puppy, the prior grammatical subject, but young children either fail to exhibit this bias (Arnold, Brown-Schmidt & Trueswell, 2007) or do so more slowly than adults (Hartshorne et al., 2015a; Song & Fisher, 2005). In the current study, we test whether language exposure affects this bias in elementary-school-age children. Children listened to stories like the one above, and answered questions like "Who wants a pepperoni slice?" which reveal their pronoun interpretation. Individual variation in the rate of selecting the subject character correlated with measures of print exposure, such that children who read more are more likely to follow the subject bias. This is the first study to establish that print exposure affects spoken pronoun comprehension in children.

Talking about SOME and ALL: What determines the usage of quantity-denoting expressions?

Reference production is often studied through single dimensions of contrast (e.g., "tall glass" when there are one or two glasses of varying height). Yet real-world communication is rarely so simple, raising questions about the factors guiding more complex referents. The current study examines decisions to mention set relations (e.g., using quantity-denoting expressions like "some of the houses" to refer to 2-out-of-5 houses) versus object categories only (e.g., using bare plurals like "houses"). Two experiments used vignettes to vary discourse focus on objects (prominent vs. non-prominent) and scenes to vary the set type described (subset vs. total set). Speakers were more likely to communicate set relations of prominent objects, particularly when they elicited high name agreement in the case of total sets. Speakers' use of quantity-denoting expressions also increased listeners' sensitivity to set relations in an object-matching task. This suggests that unlike simpler forms of modification that often decrease with greater focus, quantity-denoting expressions provide additional information about the set relations of prominent referents.

The effects of utterance timing and stimulation of left prefrontal cortex on the production of referential expressions.

We examined the relationship between the timing of utterance initiation and the choice of referring expressions, e.g., pronouns (it), zeros (…and went down), or descriptive NPs (the pink pentagon). We examined language production in healthy adults, and used anodal transcranial direct current stimulation (tDCS) to test the involvement of the left prefrontal cortex (PFC) in the timing of utterance production and the selection of reference forms in a discourse context. Twenty-two subjects (11 anodal, 11sham) described fast-paced actions, e.g. The gray oval flashes, then it moves right 2 blocks. We only examined trials in contexts that supported pronoun/zero use. For sham participants, pronouns/zeros increased on trials with longer latencies to initiate the target utterance, and trials where the previous trial was short. We argue that both of these conditions enabled greater message pre-planning and greater discourse connectedness: The strongest predictor of pronoun/zero usage was the presence of a connector word like and or then, which was also tended to occur on trials with longer latencies. For the anodal participants, the latency effect disappeared. PFC stimulation appeared to enable participants to produce utterances with greater discourse connectedness, even while planning incrementally.

Explicit and Emergent Mechanisms of Information Status.

It is well established that language production and comprehension are influenced by information status, for example, whether information is given, new, topical, or predictable, and many scholars suggest that an important component of information status is keeping track of what information is in common ground (i.e., what is shared), and what is not. Information status affects both speakers' choices (e.g., word order, pronoun use, prosodic prominence) and how listeners interpret the speaker's meaning (e.g., Chafe, 1994; Prince, 1981). Although there is a wealth of scholarly work on information status (for a review, see Arnold, Kaiser, Kahn, & Kim, 2013), there is no consensus on the mechanisms by which it is used, and in fact relatively little discussion of the underlying representations and psycholinguistic mechanisms. Moreover, a major challenge to understanding information status is that its effects are notoriously variable. This study considers existing proposals about information status, focusing on two questions: (a) how is it represented; and (b) by what mechanisms is it used? I propose that it is important to consider whether representations and mechanisms can be classified as either explicit or emergent. Based on a review of existing evidence, I argue that information status representations are most likely emergent, but the mechanisms by which they are used are both explicit and emergent. This review provides one of the first considerations of information status processing across multiple domains.

Inferring Difficulty: Flexibility in the Real-time Processing of Disfluency.

Upon hearing a disfluent referring expression, listeners expect the speaker to refer to an object that is previously unmentioned, an object that does not have a straightforward label, or an object that requires a longer description. Two visual-world eye-tracking experiments examined whether listeners directly associate disfluency with these properties of objects, or whether disfluency attribution is more flexible and involves situation-specific inferences. Since in natural situations reference to objects that do not have a straightforward label or that require a longer description is correlated with both production difficulty and with disfluency, we used a mini-artificial lexicon to dissociate difficulty from these properties, building on the fact that recently learned names take longer to produce than existing words in one's mental lexicon. The results demonstrate that disfluency attribution involves situation-specific inferences; we propose that in new situations listeners spontaneously infer what may cause production difficulty. However, the results show that these situation-specific inferences are limited in scope: listeners assessed difficulty relative to their own experience with the artificial names, and did not adapt to the assumed knowledge of the speaker.

Synthesizing meaning and processing approaches to prosody: performance matters.

Words vary in acoustic prominence; for example repeated words tend to be reduced, while focused elements tend to be acoustically prominent. We discuss two approaches to this phenomenon. On the message-based view, acoustic choices signal the speaker's meaning or pragmatics, or are guided by syntactic structure. On the facilitation-based view, reduced forms reflect facilitation of production processing mechanisms. We argue that message-based constraints correlate systematically with production facilitation. Moreover, we argue that discourse effects on acoustic reduction may be at least partially mediated by processing facilitation. Thus, research needs to simultaneously consider both competence (message) and performance (processing) constraints on prosody, specifically in terms of the psychological mechanisms underlying acoustic reduction. To facilitate this goal, we present preliminary processing models of message-based and facilitation-based approaches, and outline directions for future research.

Effects of Psychological Attention on Pronoun Comprehension.

Pronoun comprehension is facilitated for referents that are focused in the discourse context. Discourse focus has been described as a function of attention, especially shared attention, but few studies have explicitly tested this idea. Two experiments used an exogenous capture cue paradigm to demonstrate that listeners' visual attention at the onset of a story influences their preferences during pronoun resolution later in the story. In both experiments trial-initial attention modulated listeners' transitory biases while considering referents for the pronoun, whether it was in response to the capture cue or not. These biases even had a small influence on listeners' final interpretation of the pronoun. These results provide independently-motivated evidence that the listener's attention influences the on-line processes of pronoun comprehension. Trial-initial attentional shifts were made on the basis of non-shared, private information, demonstrating that attentional effects on pronoun comprehension are not restricted to shared attention among interlocutors.

The effects of anodal stimulation of the left prefrontal cortex on sentence production.

BACKGROUND: Most studies in which Anodal Transcranial Direct Current Stimulation (A-tDCS) has been used to improve language production have focused on single words. Yet sentence production requires more than lexical retrieval. For example, successful suppression of the past and careful planning of the future are two critical requirements for producing a correct sentence. Can A-tDCS improves those, and by extension, production at the sentence level? OBJECTIVE/HYPOTHESIS: Given that many aspects of sentence production beyond word retrieval require frontally-mediated operations, we hypothesized that A-tDCS to the left prefrontal cortex should benefit various operation involved in producing sentences, two of which, suppression of the past and planning of the future, were targeted in this study. METHODS: We used a paradigm that elicited construction of sentences through event description, but was structured enough to allow for between-subject comparison, clear error identification, and implementation of experimental manipulations to probe certain aspects of production. RESULTS: We showed that A-tDCS to the left PFC reliably decreased the number of incomplete and errorful sentences. When the origin of this improvement was probed, we found that A-tDCS significantly decreased errors due to premature commitment to the future word (insufficient internal monitoring), and had a marginal effect on errors of perseverations (insufficient suppression of the past). CONCLUSION: We conclude that A-tDCS is a promising tool for improving production at the sentence level, and that improvement can be expected when internal monitoring and control over verbal responses is impaired, or for certain cases of perseveratory errors.

The road to understanding is paved with the speaker's intentions: cues to the speaker's attention and intentions affect pronoun comprehension.

A series of experiments explore the effects of attention-directing cues on pronoun resolution, contrasting four specific hypotheses about the interpretation of ambiguous pronouns he and she: (1) it is driven by grammatical rules, (2) it is primarily a function of social processing of the speaker's intention to communicate, (3) it is modulated by the listener's own egocentric attention, and (4) it is primarily a function of learned probabilistic cues. Experiment 1 demonstrates that pronoun interpretation is guided by the well-known N1 (first-mention) bias, which is also modulated by both the speaker's gaze and pointing gestures. Experiment 2 demonstrates that a low-level visual capture cue has no effect on pronoun interpretation, in contrast with the social cue of pointing. Experiment 3 uses a novel intentional cue: the same attention-capture flash as in Experiment 2, but with instructions that the cue is intentionally created by the speaker. This cue does modulate the N1 bias, demonstrating the importance of information about the speaker's intentions to pronoun resolution. Taken in sum, these findings demonstrate that pronoun resolution is a process best categorized as driven by an appreciation of the speaker's communicative intent, which may be subserved by a sensitivity to predictive cues in the environment.

Information structure: linguistic, cognitive, and processing approaches.

Language form varies as a result of the information being communicated. Some of the ways in which it varies include word order, referential form, morphological marking, and prosody. The relevant categories of information include the way a word or its referent have been used in context, for example, whether a particular referent has been previously mentioned, and whether it plays a topical role in the current utterance or discourse. We first provide a broad review of linguistic phenomena that are sensitive to information structure. We then discuss several theoretical approaches to explaining information structure: information status as a part of the grammar; information status as a representation of the speaker's and listener's knowledge of common ground and/or the knowledge state of other discourse participants; and the optimal systems approach. These disparate approaches reflect the fact that there is little consensus in the field about precisely which information status categories are relevant, or how they should be represented. We consider possibilities for future work to bring these lines of work together in explicit psycholinguistic models of how people encode information status and use it for language production and comprehension. WIREs Cogn Sci 2013, 4:403-413. doi: 10.1002/wcs.1234 This article is categorized under: Linguistics > Language in Mind and Brain Psychology > Language.