Canada

Complement coercion interpretation is typically operationalized in terms of verb paraphrases (she finished the book &gt; she finished reading the book), reflecting the construal of a specific event triggered by the direct object (cf. Pustejovsky &amp; Bouillon). However, when given the option, participants frequently, and selectively, refrain from providing verb paraphrases for naturally occurring sentences.

We present a corpus study of the pragmatic factors that affect the rate of refraining from giving a verb paraphrases, followed by two experiments. First, a fill-in-the-blank experiment shows that minimally different clause types containing complement coercion yield different rates of verb paraphrase. Second, we pair complement coercion sentences with possible verb paraphrases and obtain semantic (entailment) judgements from participants. Taken together, the results present a puzzle; participants’ likelihood to produce a verbal coercion interpretation is modulated by contextual factors, while their semantic comprehension of potential verb paraphrases for the same sentences is not.

CogSci 2022

Two measures for complement coercion interpretation: Interpretation vs production for complement coercion

Complement coercion interpretation is typically operationalized in terms of verb paraphrases (she finished the book > she finished reading the book), reflecting the construal of a specific event triggered by the direct object (cf. Pustejovsky & Bouillon). However, when given the option, participants frequently, and selectively, refrain from providing verb paraphrases for naturally occurring sentences.

We present a corpus study of the pragmatic factors that affect the rate of refraining from giving a verb paraphrases, followed by two experiments. First, a fill-in-the-blank experiment shows that minimally different clause types containing complement coercion yield different rates of verb paraphrase. Second, we pair complement coercion sentences with possible verb paraphrases and obtain semantic (entailment) judgements from participants. Taken together, the results present a puzzle; participants’ likelihood to produce a verbal coercion interpretation is modulated by contextual factors, while their semantic comprehension of potential verb paraphrases for the same sentences is not.

poster

Join us for CogSci2022! Our 44th Annual Conference will be centered around the theme of diversity, with an accompanying range of keynotes and invited speakers who themselves highlight the diversity of our field. Our three symposia are Cognition across Diverse Populations, Cognition within Diverse Societies, and Global Challenges for Cognitive Science. The conference is being offered in a fully hybrid environment for onsite and online attendees.
 
 [**The Program at a Glance can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/07/Cogsci-2022-PAAG-Detailed-July-12final-1.pdf)
 [**The Detailed Program can be found here**](https://cognitivesciencesociety.org/wp-content/uploads/2022/08/DeArmond-CogSci2022-program-v113-final.pdf)

This conference requires a registration to attend. Please use the link below to secure your ticket. The website will be available to all registered attendees on July 25th. 

[Register to Attend](https://cognitivesciencesociety.org/registration/)

Registration Is Required

The 44th Annual Meeting of the Cognitive Science Society highlighted research on the theme Cognitive Diversity in addition to the full breadth and diversity of research topics offered by the society’s membership. 





Syntactic adaptation to short-term exposure has been documented with both single-trial priming and cumulative priming paradigms. These studies usually involve repeated exposure to the same structure (e.g. reduced relative clauses), and therefore it remains open whether people can track context-dependent regularities through short-term exposure. In the current study, we present a self-paced-reading experiment that investigates context-dependent syntactic adaptation by manipulating the relationship between the animacy feature of the subject NP (animate vs. inanimate) and the corresponding parse of a verb following a subject NP. We analyzed the results in terms of a log-linear model for context-dependent syntactic adaptation. The results suggest that comprehenders can track and adapt to cue-based distributional regularities, but only when the short-term regularities are consistent with the long-term ones existent in their native language.

Syntactic adaptation to short-term cue-based distributional regularities

Morphological syncretism occurs in languages when one morphological category ‘merges’ with another. Cross-linguistic research on the prevalence and types of syncretic patterns has revealed that some types of syncretism are more common than others. For example, syncretism in nominal morphology is more likely to occur in non-singular categories (Baerman et al., 2005). In two artificial language learning experiments, participants were exposed to words from a miniature language with suffix markings for gender (feminine and masculine) and number (singular, dual, and plural). Participants in Experiment 1 showed no evidence of a bias for syncretism in non-singular forms. However, participants in Experiment 2 showed a general bias to infer that a suffix that marked a novel category should be identical to a known form. This bias was strongest for non-singular items, in line with the cross-linguistic typology of syncretism. Implications for learnability and typology are discussed.

Learning Biases for Syncretic Morphological Systems

Previous studies have claimed that language structures tend to minimize the linear distance between syntactic heads and their dependents, a principle known as Dependency Length Minimization (DLM). These studies, however, have largely focused on written modality. In this study we examine the role of dependency length in acceptability ratings of English and Hindi, two typologically distinct languages, using audio stimuli. With double PP constructions as a test case, our results demonstrate no effect of DLM, suggesting the preference for shorter dependencies is different in acceptability and written texts. These findings are further supported with corpus analysis of a total of 10 treebanks for the two languages, which shows additional language-specific differences in the extent of DLM. We discuss the implications of our work and call for more careful consideration of linguistic and modality-specific diversity when it comes to processing-based claims about language typology.

Does One Size Fit all in Crosslinguistic Dependency Length Minimization?

Previous research has shown that when people engage in a joint task, they tend to minimize the total required effort. This is the case even when minimizing total effort increases the amount of effort individuals would need to expend. This tendency towards co-efficiency has previously been investigated in the motor domain. The aim of the current experiments was to investigate whether these findings extend to the sharing of mental effort. Five experiments were conducted using a multiple-object tracking task and two used a memory-based task. The first two experiments confirmed that, individually, participants prefer easier tasks. The following three experiments manipulated task difficulty for a participant and their ostensible partner and measured preferences for different difficulty combinations. Our findings provide support for egoistic effort distribution strategies, with participants mostly minimizing their own effort.

Sharing of mental effort in a joint task

Knowledge diversity is widely acknowledged to be beneficial for collaborative groups engaged in problem solving. An experiment was conducted to determine whether knowledge diversity and assigned task roles for members in an online virtual collaborative group affect group task performance and individual learning and transfer, and to explore the role of explanations as a mediating variable in these effects. Two control conditions were included that involved individual work, with and without self-explanations. Results showed that the frequency of explanations in dyadic discourse were correlated with individual learning, and that groups with knowledge diversity tend to use more explanations. These findings suggest that discussing explanations is a key feature of successful group work that contributes to determining how much individual learning occurs and how well it transfers from collaborative activities to similar, novel tasks.

Learning Through Collaboration: Designing Collaborative Activities to Promote Individual Learning

We suggest that some of the mechanisms underlying network effects on cultural evolution are intuitively accessible to laypeople, and may be part of the suite of social learning strategies underlying the human capacity for cumulative culture. Interest in the psychological mechanisms underlying this capacity typically focuses on learners’ ability to identify reliable sources and capacity for high-fidelity imitation. Yet, at the population level, research suggests that network structures themselves may influence cumulative learning by changing individuals’ explore-exploit patterns. In our experiments, adults infer that more proximal or distal clusters in a fragmented network will have more similar or dissimilar technological “styles”, and prefer to seek advice from more distant experts when asked to innovate, but more proximate experts when asked to remember. Commonsense intuitions about how social networks shape our access to information and diversity- fidelity tradeoffs for memory and innovation may make us more effective social learners.

Know your network: people infer cultural dri from network structure, and expect collaborating with more distant experts to improve innovation, but collaborating with network-neighbors to improve memory

In decision-making situations that arise repeatedly, there are tradeoffs between: (i) acquiring new information to facilitate future, related decisions (exploration) and (ii) using existing information to secure expected outcomes (exploitation). Exploration choices have been well characterized in nonsocial contexts, but choices to explore (or not) in social environments are less well understood. Social environments are of particular interest because a key factor that increases exploration in nonsocial contexts is environmental uncertainty, and the social world is appreciated to be highly uncertain. Here, participants searched for rewards in a series of grids that were either described as comprising real people distributing previously-earned points (social context) or as the result of a computer algorithm or natural phenomenon (nonsocial context). Participants explored more, and earned fewer rewards, in the social versus nonsocial context, suggesting that social uncertainty prompted exploration at the cost of task-relevant goals.

Exploration is Higher in Social Contexts at the Cost of Rewards

Speakers' referential expressions often depart from communicative ideals in ways that help illuminate the nature of pragmatic language use. Patterns of overmodification, in which a speaker uses a modifier that is redundant given their communicative goal, have proven especially informative in this regard. It seems likely that these patterns are shaped by the environment a speaker is exposed to in complex ways. Unfortunately, systematically manipulating these factors during human language acquisition is impossible. In this paper, we propose to address this limitation by adopting neural networks (NN) as learning agents. By systematically varying the environments in which these agents are trained, while keeping the NN architecture constant, we show that overmodification is more likely with environmental features that are infrequent or salient. We show that these findings emerge naturally in the context of a probabilistic model of pragmatic communication.

Color Overmodification Emerges from Data-Driven Learning and Pragmatic Reasoning

In this work we introduce a structured signaling game, an extension of the classical signaling game with a similarity structure between meanings in the context, along with a variant of the Rational Speech Act (RSA) framework which we call structured-RSA (sRSA) for pragmatic reasoning in structured domains. We explore the behavior of the sRSA in the domain of color and show that pragmatic agents using sRSA on top of semantic representations, derived from the World Color Survey, attain efficiency very close to the information theoretic limit after only 1 or 2 levels of recursion. We also explore the interaction between pragmatic reasoning and learning in multi-agent reinforcement learning framework. Our results illustrate that artificial agents using sRSA develop communication closer to the information theoretic frontier compared to agents using RSA and just reinforcement learning. We also find that the ambiguity of the semantic representation increases as the pragmatic agents are allowed to perform deeper reasoning about each other during learning.

Pragmatic Reasoning in Structured Signaling Games

Empirical studies have demonstrated that when comprehenders are faced with informationally redundant utterances, they may make pragmatic inferences to accommodate the informationally redundant utterance (Kravtchenko & Demberg, 2015. Previous work has also shown that the strength of these inferences depends on prominence of the redundant utterance – if it is stressed prosodically, marked with an exclamation mark, or introduced with a discourse marker such as “Oh yeah”, atypicality inferences are stronger (Kravtchenko & Demberg, 2015; 2022; Ryzhova & Demberg, 2020).
The goal of the present paper is to demonstrate how both the atypicality inference and the effect of prominence can be modelled using the rational speech act (RSA) framework. We show that atypicality inferences can be captured by introducing joint reasoning about the habituality of events, following Degen, Tessler, and Goodman (2015); Goodman and Frank (2016). However, we find that joint reasoning models principally cannot account for the effect of differences in utterance prominence. This is because prominence markers do not contribute to the truth-conditional meaning. We then proceed to demonstrate that leveraging a noisy channel model, which has previously been used to model low-level acoustic perception (Bergen & Goodman, 2015), can successfully account for the empirically observed patterns of utterance prominence.

Premium content

Two measures for complement coercion interpretation: Interpretation vs production for complement coercion

Downloads

Next from CogSci 2022

Syntactic adaptation to short-term cue-based distributional regularities

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES