scholarly journals Statistical learning reflects inferences about unique predictive relations

2021 ◽  
Author(s):  
Anna Leshinskaya ◽  
Sharon L. Thompson-Schill
Keyword(s):  
Statistical Learning ◽  
Conditional Probability ◽  
Learning Paradigm ◽  
Trade Off ◽  
Event Pair ◽  
Statistical Property ◽  
Statistical Regularities ◽  
Predictive Relations ◽  
High Base ◽  
The Mind

The mind adeptly registers statistical regularities in experience, often incidentally and implicitly. We used a visual statistical learning paradigm to study what kinds of statistics it spontaneously computes in such conditions. We found that participants’ learning of pairwise predictive relations was best explained by an inferentially sophisticated quantity, deltaP, which reflects whether a high conditional probability between an event pair is unique. We showed that uniqueness can be reduced by either a strong competing predictor or an overall high base rate of the outcome. Both can result from normalization: if predictors of the same effect trade off, a predictor must raise the probability of the effect more than the others to be effective. Adding normalization to the Rescorla-Wagner learning model captures these results. We argue that the uniqueness of a relation is an intrinsically important statistical property that governs learning without incentive or deliberation.

Prior real-world experience influences non-linguistic statistical learning

2020 ◽  
Author(s):  
Stephen Charles Van Hedger ◽  
Ingrid Johnsrude ◽  
Laura Batterink
Keyword(s):  
Statistical Learning ◽  
Real World ◽  
Language Experience ◽  
Pitch Range ◽  
Test Items ◽  
Instrument Measuring ◽  
Statistical Regularities ◽  
Context Free ◽  
Specific Learning

Listeners are adept at extracting regularities from the environment, a process known as statistical learning (SL). SL has been generally assumed to be a form of “context-free” learning that occurs independently of prior knowledge, and SL experiments typically involve exposing participants to presumed novel regularities, such as repeating nonsense words. However, recent work has called this assumption into question, demonstrating that learners’ previous language experience can considerably influence SL performance. In the present experiment, we tested whether previous knowledge also shapes SL in a non-linguistic domain, using a paradigm that involves extracting regularities over tone sequences. Participants learned novel tone sequences, which consisted of pitch intervals not typically found in Western music. For one group of participants, the tone sequences used artificial, computerized instrument sounds. For the other group, the same tone sequences used familiar instrument sounds (piano or violin). Knowledge of the statistical regularities was assessed using both trained sounds (measuring specific learning) and sounds that differed in pitch range and/or instrument (measuring transfer learning). In a follow-up experiment, two additional testing sessions were administered to gauge retention of learning (one day and approximately one-week post-training). Compared to artificial instruments, training on sequences played by familiar instruments resulted in reduced correlations among test items, reflecting more idiosyncratic performance. Across all three testing sessions, learning of novel regularities presented with familiar instruments was worse compared to unfamiliar instruments, suggesting that prior exposure to music produced by familiar instruments interfered with new sequence learning. Overall, these results demonstrate that real-world experience influences SL in a non-linguistic domain, supporting the view that SL involves the continuous updating of existing representations, rather than the establishment of entirely novel ones.

scholarly journals The Mind In Its Own Place: The Difficulties and Benefits of Thinking for Pleasure

2019 ◽  
Author(s):  
Timothy D. Wilson ◽  
Erin Corwin Westgate ◽  
Nick Buttrick ◽  
Daniel Gilbert
Keyword(s):  
Positive Affect ◽  
Intervention Study ◽  
Present Evidence ◽  
Trade Off ◽  
Everyday Lives ◽  
The Right ◽  
The Mind ◽  
Is Value

This chapter is concerned with a type of thinking that has received little attention, namely intentional “thinking for pleasure”—the case in which people deliberately focus solely on their thoughts with the goal of generating positive affect. We present a model that describes why it is difficult to enjoy one's thoughts, how it can be done successfully, and when there is value in doing so. We review 36 studies we have conducted on this topic with just over 10,000 participants. We found that thinking for pleasure does not come easily to most people, but can be enjoyable and beneficial under the right conditions. Specifically, we found evidence that thinking for pleasure requires both motivation and the ability to concentrate. For example, several studies show that people enjoy thinking more when it is made easier with the use of “thinking aids.” We present evidence for a trade-off model that holds that people are most likely to enjoy their thoughts if they find those thoughts to be personally meaningful, but that such thinking involves concentration, which lowers enjoyment. Lastly, we review evidence for the benefits of thinking for pleasure, including an intervention study in which participants found thinking for pleasure enjoyable and meaningful in their everyday lives.

scholarly journals Complementary learning systems within the hippocampus: A neural network modeling approach to reconciling episodic memory with statistical learning

2016 ◽  
Author(s):  
Anna C. Schapiro ◽  
Nicholas B. Turk-Browne ◽  
Matthew M. Botvinick ◽  
Kenneth A. Norman
Keyword(s):  
Neural Network ◽  
Statistical Learning ◽  
Entorhinal Cortex ◽  
Network Modeling ◽  
Learning Systems ◽  
Neural Network Modeling ◽  
Trade Off ◽  
Rapid Learning ◽  
Rapid Extraction

AbstractA growing literature suggests that the hippocampus is critical for the rapid extraction of regularities from the environment. Although this fits with the known role of the hippocampus in rapid learning, it seems at odds with the idea that the hippocampus specializes in memorizing individual episodes. In particular, the Complementary Learning Systems theory argues that there is a computational trade-off between learning the specifics of individual experiences and regularities that hold across those experiences. We asked whether it is possible for the hippocampus to handle both statistical learning and memorization of individual episodes. We exposed a neural network model that instantiates known properties of hippocampal projections and subfields to sequences of items with temporal regularities. We found that the monosynaptic pathway — the pathway connecting entorhinal cortex directly to region CA1 — was able to support statistical learning, while the trisynaptic pathway — connecting entorhinal cortex to CA1 through dentate gyrus and CA3 — learned only individual episodes, with apparent representations of regularities resulting from associative reactivation through recurrence. Thus, in paradigms involving rapid learning, the computational trade-off between learning episodes and regularities may be handled by separate anatomical pathways within the hippocampus itself.

scholarly journals Prediction suppression in monkey inferotemporal cortex depends on the conditional probability between images

2016 ◽  
Vol 115 (1) ◽  
pp. 355-362 ◽  
Author(s):  
Suchitra Ramachandran ◽  
Travis Meyer ◽  
Carl R. Olson
Keyword(s):  
Conditional Probability ◽  
Joint Probability ◽  
Inferotemporal Cortex ◽  
Fixed Sequence ◽  
Weak Response ◽  
Predictive Relations ◽  
Area Te

When monkeys view two images in fixed sequence repeatedly over days and weeks, neurons in area TE of the inferotemporal cortex come to exhibit prediction suppression. The trailing image elicits only a weak response when presented following the leading image that preceded it during training. Induction of prediction suppression might depend either on the contiguity of the images, as determined by their co-occurrence and captured in the measure of joint probability P( A, B), or on their contingency, as determined by their correlation and as captured in the measures of conditional probability P( A| B) and P( B| A). To distinguish between these possibilities, we measured prediction suppression after imposing training regimens that held P( A, B) constant but varied P( A| B) and P( B| A). We found that reducing either P( A| B) or P( B| A) during training attenuated prediction suppression as measured during subsequent testing. We conclude that prediction suppression depends on contingency, as embodied in the predictive relations between the images, and not just on contiguity, as embodied in their co-occurrence.

scholarly journals Reading as Statistical Learning

2018 ◽  
Vol 49 (3S) ◽  
pp. 634-643 ◽  
Author(s):  
Joanne Arciuli
Keyword(s):  
Implicit Learning ◽  
Statistical Learning ◽  
Literacy Instruction ◽  
Word Reading ◽  
Implicit Methods ◽  
New Techniques ◽  
Linguistic Research ◽  
Statistical Regularities ◽  
Basic And Applied Research ◽  
Deep Orthography

Purpose The purpose of this tutorial is to explain how learning to read can be thought of as learning statistical regularities and to demonstrate why this is relevant for theory, modeling, and practice. This tutorial also shows how triangulation of methods and cross-linguistic research can be used to gain insight. Method The impossibility of conveying explicitly all of the regularities that children need to acquire in a deep orthography, such as English, can be demonstrated by examining lesser-known probabilistic orthographic cues to lexical stress. Detection of these kinds of cues likely occurs via a type of implicit learning known as statistical learning (SL). The first part of the tutorial focuses on these points. Next, studies exploring how individual differences in the capacity for SL relate to variability in word reading accuracy in the general population are discussed. A brief overview of research linking impaired SL and dyslexia is also provided. The final part of the tutorial focuses on how we might supplement explicit literacy instruction with implicit learning methods and emphasizes the value of testing the efficacy of new techniques in the classroom. The basic and applied research reviewed here includes corpus analyses, behavioral testing, computational modeling, and classroom-based research. Although some of these methods are not commonly used in clinical research, the depth and breadth of this body of work provide a compelling case for why reading can be thought of as SL and how this view can inform practice. Conclusion Implicit methods that draw on the principles of SL can supplement the much-needed explicit instruction that helps children learn to read. This synergy of methods has the potential to spark innovative practices in literacy instruction and remediation provided by educators and clinicians to support typical learners and those with developmental disabilities.

scholarly journals Non-native phonemes in adult word learning: evidence from the N400m

2009 ◽  
Vol 364 (1536) ◽  
pp. 3697-3709 ◽  
Author(s):  
Christian Dobel ◽  
Lothar Lagemann ◽  
Pienie Zwitserlood
Keyword(s):  
Statistical Learning ◽  
Word Learning ◽  
Native Language ◽  
The Novel ◽  
Learning Paradigm ◽  
Minimal Pairs ◽  
Lexical Activation ◽  
Semantic Access ◽  
Novel Words

Newborns are equipped with a large phonemic inventory that becomes tuned to one's native language early in life. We review and add new data about how learning of a non-native phoneme can be accomplished in adults and how the efficiency of word learning can be assessed by neurophysiological measures. For this purpose, we studied the acquisition of the voiceless, bilabial fricative /Φ/ via a statistical-learning paradigm. Phonemes were embedded in minimal pairs of pseudowords, differing only with respect to the fricative (/aΦo/ versus /afo/). During learning, pseudowords were combined with pictures of objects with some combinations of pseudowords and pictures occurring more frequently than others. Behavioural data and the N400m component, as an index of lexical activation/semantic access, showed that participants had learned to associate the pseudowords with the pictures. However, they could not discriminate within the minimal pairs. Importantly, before learning, the novel words with the sound /Φ/ showed smaller N400 amplitudes than those with native phonemes, evidencing their non-word status. Learning abolished this difference indicating that /Φ/ had become integrated into the native category /f/, instead of establishing a novel category. Our data and review demonstrate that native phonemic categories are powerful attractors hampering the mastery of non-native contrasts.

scholarly journals Twelve-Month-Old Infants Benefit From Prior Experience in Statistical Learning

2008 ◽  
Vol 19 (12) ◽  
pp. 1247-1252 ◽  
Author(s):  
Jill Lany ◽  
Rebecca L. Gómez
Keyword(s):  
Statistical Learning ◽  
Verbal Memory ◽  
Prior Experience ◽  
Memory Capacity ◽  
Artificial Language ◽  
Learning Mechanisms ◽  
Language Structure ◽  
Statistical Regularities ◽  
Nonadjacent Dependencies ◽  
Enhanced Learning

A decade of research suggests that infants readily detect patterns in their environment, but it is unclear how such learning changes with experience. We tested how prior experience influences sensitivity to statistical regularities in an artificial language. Although 12-month-old infants learn adjacent relationships between word categories, they do not track nonadjacent relationships until 15 months. We asked whether 12-month-old infants could generalize experience with adjacent dependencies to nonadjacent ones. Infants were familiarized to an artificial language either containing or lacking adjacent dependencies between word categories and were subsequently habituated to novel nonadjacent dependencies. Prior experience with adjacent dependencies resulted in enhanced learning of the nonadjacent dependencies. Female infants showed better discrimination than males did, which is consistent with earlier reported sex differences in verbal memory capacity. The findings suggest that prior experience can bootstrap infants' learning of difficult language structure and that learning mechanisms are powerfully affected by experience.

scholarly journals Examining Three-way Binding as a Constraint on Statistical Learning

2018 ◽  
Author(s):  
Hyungwook Yim ◽  
Simon Dennis ◽  
Vladimir Sloutsky
Keyword(s):  
Statistical Learning ◽  
Explicit Memory ◽  
Paired Associate ◽  
Empirical Studies ◽  
Learning Task ◽  
Paired Associate Learning ◽  
The Other ◽  
Learning Paradigm ◽  
Associate Learning ◽  
Memory Structures

Models of statistical learning do not place constraints on the complexity of the memory structure that is formed during statistical learning, while empirical studies using the statistical learning task have only examined the formation of simple memory structures (e.g., two-way binding). On the other hand, the memory literature, using explicit memory tasks, has shown that people are able to form memory structures of different complexities and that more complex memory structures (e.g., three-way binding) are usually more difficult to form. We examined whether complex memory structures such as three-way bindings can be implicitly formed through statistical learning by utilizing manipulations that have been used in the paired-associate learning paradigm (e.g., AB/ABr condition). Through three experiments, we show that while simple two-way binding structures can be formed implicitly, three-way bindings can only be formed with explicit instructions. The results indicate that explicit attention may be a necessary component in forming three-way memory structures and suggest that existing models should place constraints on the representational structures that can be formed.

scholarly journals Order in the Statistical Learning of Phonotactics

2020 ◽  
Author(s):  
Peter Richtsmeier
Keyword(s):  
Statistical Learning ◽  
Temporal Structure ◽  
Order Effects ◽  
Learning Paradigm ◽  
Adult Participants ◽  
Learning Research ◽  
Strong Order ◽  
With Memory

A premise of statistical learning research is that learners attend to and learn the frequencies of repeating or co-occurring elements in the input. When the input is a series of words, participants readily learn the frequencies of phoneme sequences, that is, to learn phonotactic frequencies. Inherent to the concepts of both frequency and phonotactics is order, or the temporal structure of the input. Order is similarly inherent to statistical learning, yet the effect of order on statistical learning is not well understood. In the present study, adult participants learned the relative frequencies of eight item-medial consonant sequences, for example, the /mk/ in /nʌmkət/. Across five ordering conditions, both familiarization and test stimuli were independently ordered and randomized, thus allowing for a relatively broad search for order effects in an established statistical learning paradigm. Participants learned the target frequencies equivalently across the five ordering conditions, indicating no modulating effect of order. Nevertheless, participants also approached the task by applying idiosyncratic, structured orders to their responses. The result is an unexpected but robust effect of order. Both the results and the design of the study also allow for increased integration of statistical learning with memory and other aspects of cognition.

scholarly journals Reinforced statistical learning of auditory categories

2020 ◽  
Author(s):  
Bob McMurray ◽  
Samantha Chiu
Keyword(s):  
Language Acquisition ◽  
Unsupervised Learning ◽  
Supervised Learning ◽  
Statistical Learning ◽  
Relevant Dimension ◽  
Learning Paradigm ◽  
Statistical Structure ◽  
Critical Step ◽  
Specific Category ◽  
L2 Learning

A critical step in language acquisition is learning phoneme categories. While L1 learning has been thought to use unsupervised learning (using the distributional statistics of cues), recent research raises the possibility of supervised learning (using teaching signals). Similarly, L2 learning is studied with supervised learning, but unsupervised may also contribute. We developed the reinforced statistical learning paradigm to examine their interaction. Participants first underwent unsupervised learning, hearing a series of non-linguistic sounds whose statistical structure reflected two categories. In subsequent supervised learning, categories either matched or mismatched. Supervised learning was faster when phases matched, though benefits were limited to specific category configurations. Unsupervised learning did not affect the steepness of categorization along the relevant dimension, but it helped subjects learn to ignore irrelevant dimensions. Unsupervised learning may set the stage for supervised learning, but its role may be to determine which dimensions are important, and not to directly acquire categories.

Sign in / Sign up

Export Citation Format

Share Document