In typical statistical learning studies, researchers define sequences in terms of the probability of the next item in the sequence given the current item (or items), and they show that high probability sequences are treated as more familiar than low probability sequences. Existing accounts of these phenomena all assume that participants represent statistical regularities more or less as they are defined by the experimenters—as sequential probabilities of symbols in a string. Here we offer an alternative, or possibly supplementary, hypothesis. Specifically, (...) rather than identifying or labeling individual stimuli discretely in order to predict the next item in a sequence, we need only assume that the participant is able to represent the stimuli as evincing particular similarity relations to one another, with sequences represented as trajectories through this similarity space. We present experiments in which this hypothesis makes sharply different predictions from hypotheses based on the assumption that sequences are learned over discrete, labeled stimuli. We also present a series of simulation models that encode stimuli as positions in a continuous two‐dimensional space, and predict the next location from the current location. Although no model captures all of the data presented here, the results of three critical experiments are more consistent with the view that participants represent trajectories through similarity space rather than sequences of discrete labels under particular conditions. (shrink)

Statistical learning refers to the ability to identify structure in the input based on its statistical properties. For many linguistic structures, the relevant statistical features are distributional: They are related to the frequency and variability of exemplars in the input. These distributional regularities have been suggested to play a role in many different aspects of language learning, including phonetic categories, using phonemic distinctions in word learning, and discovering non-adjacent relations. On the surface, these different aspects share few commonalities. Despite this, (...) we demonstrate that the same computational framework can account for learning in all of these tasks. These results support two conclusions. The first is that much, and perhaps all, of distributional statistical learning can be explained by the same underlying set of processes. The second is that some aspects of language can be learned due to domain-general characteristics of memory. (shrink)

From a theoretical perspective, most discussions of statistical learning have focused on the possible “statistical” properties that are the object of learning. Much less attention has been given to defining what “learning” is in the context of “statistical learning.” One major difficulty is that SL research has been monitoring participants’ performance in laboratory settings with a strikingly narrow set of tasks, where learning is typically assessed offline, through a set of two-alternative-forced-choice questions, which follow a brief visual or auditory familiarization (...) stream. Is that all there is to characterizing SL abilities? Here we adopt a novel perspective for investigating the processing of regularities in the visual modality. By tracking online performance in a self-paced SL paradigm, we focus on the trajectory of learning. In a set of three experiments we show that this paradigm provides a reliable and valid signature of SL performance, and it offers important insights for understanding how statistical regularities are perceived and assimilated in the visual modality. This demonstrates the promise of integrating different operational measures to our theory of SL. (shrink)

One of the themes that has been widely addressed in both the implicit learning and statistical learning literatures is that of rule learning. While it is widely agreed that the extraction of regularities from the environment is a fundamental facet of cognition, there is still debate about the nature of rule learning. Rey and colleagues show that the comparison between human and non‐human primates can contribute important insights to this debate.

Cognitive Science, Volume 46, Issue 4, April 2022.

In 2006, Perruchet and Pacton (2006) asked whether implicit learning and statistical learning represent two approaches to the same phenomenon. This article represents an important follow‐up to their seminal review article. As in the previous paper, the focus is on the formation of elementary cognitive units. Both approaches favor different explanations on what these units consist of and how they are formed. Perruchet weighs up the evidence for different explanations and concludes with a helpful agenda for future research.

Cognitive Science, Volume 46, Issue 2, February 2022.

Humans have the ability to learn surprisingly complicated statistical information in a variety of modalities and situations, often based on relatively little input. These statistical learning (SL) skills appear to underlie many kinds of learning, but despite their ubiquity, we still do not fully understand precisely what SL is and what individual differences on SL tasks reflect. Here, we present experimental work suggesting that at least some individual differences arise from stimulus-specific variation in perceptual fluency: the ability to rapidly or (...) efficiently code and remember the stimuli that SL occurs over. Experiment 1 demonstrates that participants show improved SL when the stimuli are simple and familiar; Experiment 2 shows that this improvement is not evident for simple but unfamiliar stimuli; and Experiment 3 shows that for the same stimuli (Chinese characters), SL is higher for people who are familiar with them (Chinese speakers) than those who are not (English speakers matched on age and education level). Overall, our findings indicate that performance on a standard SL task varies substantially within the same (visual) modality as a function of whether the stimuli involved are familiar or not, independent of stimulus complexity. Moreover, test–retest correlations of performance in an SL task using stimuli of the same level of familiarity (but distinct items) are stronger than correlations across the same task with stimuli of different levels of familiarity. Finally, we demonstrate that SL performance is predicted by an independent measure of stimulus-specific perceptual fluency that contains no SL component at all. Our results suggest that a key component of SL performance may be related to stimulus-specific processing and familiarity. (shrink)

Identifying wordlike units in language is typically done by applying a battery of criteria, though how to weight these criteria with respect to one another is currently unknown. We address this question by investigating whether certain criteria are also used as cues for learning an artificial language—if they are, then perhaps they can be relied on more as trustworthy top‐down diagnostics. The two criteria for grammatical wordhood that we consider are a unit's free mobility and its internal immutability. These criteria (...) also map to two cognitive mechanisms that could underlie successful statistical learning: learners might orient themselves around the low transitional probabilities at unit boundaries, or they might seek chunks with high internal transitional probabilities. We find that each criterion has its own facilitatory effect, and learning is best where they both align. This supports the battery‐of‐criteria approach to diagnosing wordhood, and also suggests that the mechanism behind statistical learning may not be a question of either/or; perhaps the two mechanisms do not compete, but mutually reinforce one another. (shrink)

Extracting the regularities of our environment is one of our core cognitive abilities. To study the fine-grained dynamics of the extraction of embedded regularities, a method combining the advantages of the artificial language paradigm and the serial response time task was used with a group of Guinea baboons in a new automatic experimental device. After a series of random trials, monkeys were exposed to language-like patterns. We found that the extraction of embedded patterns positioned at the end of larger patterns (...) was faster than the extraction of initial embedded patterns. This result suggests that there is a learning advantage for the final element of a sequence that benefits from the contextual information provided by previous elements. (shrink)

We introduce a set of biologically and computationally motivated design choices for modeling the learning of language, or of other types of sequential, hierarchically structured experience and behavior, and describe an implemented system that conforms to these choices and is capable of unsupervised learning from raw natural-language corpora. Given a stream of linguistic input, our model incrementally learns a grammar that captures its statistical patterns, which can then be used to parse or generate new data. The grammar constructed in this (...) manner takes the form of a directed weighted graph, whose nodes are recursively (hierarchically) defined patterns over the elements of the input stream. We evaluated the model in seventeen experiments, grouped into five studies, which examined, respectively, (a) the generative ability of grammar learned from a corpus of natural language, (b) the characteristics of the learned representation, (c) sequence segmentation and chunking, (d) artificial grammar learning, and (e) certain types of structure dependence. The model's performance largely vindicates our design choices, suggesting that progress in modeling language acquisition can be made on a broad front—ranging from issues of generativity to the replication of human experimental findings—by bringing biological and computational considerations, as well as lessons from prior efforts, to bear on the modeling approach. (shrink)

We introduce a set of biologically and computationally motivated design choices for modeling the learning of language, or of other types of sequential, hierarchically structured experience and behavior, and describe an implemented system that conforms to these choices and is capable of unsupervised learning from raw natural‐language corpora. Given a stream of linguistic input, our model incrementally learns a grammar that captures its statistical patterns, which can then be used to parse or generate new data. The grammar constructed in this (...) manner takes the form of a directed weighted graph, whose nodes are recursively (hierarchically) defined patterns over the elements of the input stream. We evaluated the model in seventeen experiments, grouped into five studies, which examined, respectively, (a) the generative ability of grammar learned from a corpus of natural language, (b) the characteristics of the learned representation, (c) sequence segmentation and chunking, (d) artificial grammar learning, and (e) certain types of structure dependence. The model's performance largely vindicates our design choices, suggesting that progress in modeling language acquisition can be made on a broad front—ranging from issues of generativity to the replication of human experimental findings—by bringing biological and computational considerations, as well as lessons from prior efforts, to bear on the modeling approach. (shrink)

Cognitive Science, Volume 46, Issue 9, September 2022.

Results from two self-paced reading experiments in English are reported in which subject- and object-extracted relative clauses (SRCs and ORCs, respectively) were presented in contexts that support both types of relative clauses (RCs). Object-extracted versions were read more slowly than subject-extracted versions across both experiments. These results are not consistent with a decay-based working memory account of dependency formation where the amount of decay is a function of the number of new discourse referents that intervene between the dependents (Gibson, 1998; (...) Warren & Gibson, 2002). Rather, these results support interference-based accounts and decay-based accounts where the amount of decay depends on the number of words or on the type of noun phrases that intervene between the dependents. In Experiment 2, presentation in supportive contexts was directly contrasted with presentation in null contexts. Whereas in the null context the extraction effect was only observed during the RC region, in a supportive context the extraction effect was numerically larger and persisted into the following region, thus showing that extraction effects are enhanced in supportive contexts. A sentence completion study demonstrated that the rate of SRCs versus ORCs was similar across null and supportive contexts (with most completions being subject-extractions), ruling out the possibility that an enhanced extraction effect in supportive contexts is due to ORCs being less expected in such contexts. However, the content of the RCs differed between contexts in the completions, such that the RCs produced in supportive contexts were more constrained, reflecting the lexical and semantic content of the preceding context. This effect, which we discuss in terms of expectations/lexico-syntactic priming, suggests that the enhancement of the extraction effect in supportive contexts is due to the facilitation of the subject-extracted condition. (shrink)

Are similar, or even identical, mechanisms used in the computational modeling of speech segmentation, serial image processing, and music processing? We address this question by exploring how TRACX2, a recognition‐based, recursive connectionist autoencoder model of chunking and sequence segmentation, which has successfully simulated speech and serial‐image processing, might be applied to elementary melody perception. The model, a three‐layer autoencoder that recognizes “chunks” of short sequences of intervals that have been frequently encountered on input, is trained on the tone intervals of (...) melodically simple French children's songs. It dynamically incorporates the internal representations of these chunks into new input. Its internal representations cluster in a manner that is consistent with “human‐recognizable” melodic categories. TRACX2 is sensitive to both contour and proximity information in the musical chunks that it encounters in its input. It shows the “end‐of‐word” superiority effect demonstrated by Saffran et al. (1999) for short musical phrases. The overall findings suggest that the recursive autoassociative chunking mechanism, as implemented in TRACX2, may be a general segmentation and chunking mechanism, underlying not only word‐ and image‐chunking, but also elementary melody processing. (shrink)