Word co‐occurrence patterns in language corpora contain a surprising amount of conceptual knowledge. Large language models (LLMs), trained to predict words in context, leverage these patterns to achieve impressive performance on diverse semantic tasks requiring world knowledge. An important but understudied question about LLMs’ semantic abilities is whether they acquire generalized knowledge of common events. Here, we test whether five pretrained LLMs (from 2018's BERT to 2023's MPT) assign a higher likelihood to plausible descriptions of agent−patient interactions than to minimally (...) different implausible versions of the same event. Using three curated sets of minimal sentence pairs (total n = 1215), we found that pretrained LLMs possess substantial event knowledge, outperforming other distributional language models. In particular, they almost always assign a higher likelihood to possible versus impossible events (The teacher bought the laptop vs. The laptop bought the teacher). However, LLMs show less consistent preferences for likely versus unlikely events (The nanny tutored the boy vs. The boy tutored the nanny). In follow‐up analyses, we show that (i) LLM scores are driven by both plausibility and surface‐level sentence features, (ii) LLM scores generalize well across syntactic variants (active vs. passive constructions) but less well across semantic variants (synonymous sentences), (iii) some LLM errors mirror human judgment ambiguity, and (iv) sentence plausibility serves as an organizing dimension in internal LLM representations. Overall, our results show that important aspects of event knowledge naturally emerge from distributional linguistic patterns, but also highlight a gap between representations of possible/impossible and likely/unlikely events. (shrink)

Spoken communication occurs in a “noisy channel” characterized by high levels of environmental noise, variability within and between speakers, and lexical and syntactic ambiguity. Given these properties of the received linguistic input, robust spoken word recognition—and language processing more generally—relies heavily on listeners' prior knowledge to evaluate whether candidate interpretations of that input are more or less likely. Here we compare several broad-coverage probabilistic generative language models in their ability to capture human linguistic expectations. Serial reproduction, an experimental paradigm where (...) spoken utterances are reproduced by successive participants similar to the children's game of “Telephone,” is used to elicit a sample that reflects the linguistic expectations of English-speaking adults. When we evaluate a suite of probabilistic generative language models against the yielded chains of utterances, we find that those models that make use of abstract representations of preceding linguistic context (i.e., phrase structure) best predict the changes made by people in the course of serial reproduction. A logistic regression model predicting which words in an utterance are most likely to be lost or changed in the course of spoken transmission corroborates this result. We interpret these findings in light of research highlighting the interaction of memory-based constraints and representations in language processing. (shrink)

I shall attempt something rash in this paper. I shall draw your attention to some past and current work on perception by psychologists and others. I shall concentrate on work in vision and hearing. This outline will occupy the first part of my lecture. I shall then go on, in the second part, to suggest that this scientific work has certain philosophical implications. This whole attempt is a bit rash for obvious reasons. It is not easy to outline fairly and (...) accurately past and current work in any branch of science. I am very liable, therefore, to do an injustice to the efforts of psychologists and others in this field. What makes matters more difficult for me is that I also have to show that this work is of philosophical interest. What has led me to embark on this perilous enterprise is a hunch I have developed in recent years. I have the hunch that philosophers who are interested in perception would do well to pay rather more attention than they have been wont to do in the past to the work and discourse coming out of the scientific laboratory and similar places. (shrink)

Frequency effects are pervasive in studies of language, with higher frequency words being recognized faster than lower frequency words. However, the exact nature of frequency effects has recently been questioned, with some studies finding that contextual information provides a better fit to lexical decision and naming data than word frequency ( Adelman et al., 2006 ). Recent work has cemented the importance of these results by demonstrating that a measure of the semantic diversity of the contexts that a word occurs (...) in provides a powerful measure to account for variability in word recognition latency ( Johns et al., 2012, 2015 ; Jones et al., 2012 ). The goal of the current study is to extend this measure to examine bilingualism and aging, where multiple theories use frequency of occurrence of linguistic constructs as central to accounting for empirical results ( Gollan et al., 2008 ; Ramscar et al., 2014 ). A lexical decision experiment was conducted with four groups of subjects: younger and older monolinguals and bilinguals. Consistent with past results, a semantic diversity variable accounted for the greatest amount of variance in the latency data. In addition, the pattern of fits of semantic diversity across multiple corpora suggests that bilinguals and older adults are more sensitive to semantic diversity information than younger monolinguals. (shrink)

I shall attempt something rash in this paper. I shall draw your attention to some past and current work on perception by psychologists and others. I shall concentrate on work in vision and hearing. This outline will occupy the first part of my lecture. I shall then go on, in the second part, to suggest that this scientific work has certain philosophical implications. This whole attempt is a bit rash for obvious reasons. It is not easy to outline fairly and (...) accurately past and current work in any branch of science. I am very liable, therefore, to do an injustice to the efforts of psychologists and others in this field. What makes matters more difficult for me is that I also have to show that this work is of philosophical interest. What has led me to embark on this perilous enterprise is a hunch I have developed in recent years. I have the hunch that philosophers who are interested in perception would do well to pay rather more attention than they have been wont to do in the past to the work and discourse coming out of the scientific laboratory and similar places. (shrink)

This paper puts forward a general framework for thought about human information processing. It is intended to avoid some of the problems of pipeline or stage models of function. At the same time it avoids the snare of supposing a welter of indefinitely many separate processes. The approach is not particularly original, but rather represents the common elements or presuppositions in a number of modern theories. These presuppositions are not usually explicit, however, and making them so reduces the danger of (...) slipping back into earlier modes of thought.The key point is to distinguish between persisting representations and the processes that translate one representation into another. Various classes or groups of persisting representations can be distinguished by the experimental treatments that interfere with them. In particular, there now seem to be several kinds of short-term or temporary storage, different from each other as well as from longterm memory; the translating processes also have several different modes or kinds. A particularly important aspect of the current position is that a model of this general type no longer requires some external agent to direct and control long sequences of behaviour. (shrink)

This paper puts forward a general framework for thought about human information processing. It is intended to avoid some of the problems of pipeline or stage models of function. At the same time it avoids the snare of supposing a welter of indefinitely many separate processes. The approach is not particularly original, but rather represents the common elements or presuppositions in a number of modern theories. These presuppositions are not usually explicit, however, and making them so reduces the danger of (...) slipping back into earlier modes of thought.The key point is to distinguish between persisting representations and the processes that translate one representation into another. Various classes or groups of persisting representations can be distinguished by the experimental treatments that interfere with them. In particular, there now seem to be several kinds of short-term or temporary storage, different from each other as well as from longterm memory; the translating processes also have several different modes or kinds. A particularly important aspect of the current position is that a model of this general type no longer requires some external agent to direct and control long sequences of behaviour. (shrink)