Rational arbitration between statistics and rules in human sequence processing

Abstract

5Detecting and learning temporal regularities is essential to accurately predict the future. 6 Past research indicates that humans are sensitive to two types of sequential regularities: 7 deterministic rules, which afford sure predictions, and statistical biases, which govern the 8 probabilities of individual items and their transitions. How does the human brain arbitrate 9 between those two types? We used finger tracking to continuously monitor the online 10 build-up of evidence, confidence, false alarms and changes-of-mind during sequence 11 learning. All these aspects of behaviour conformed tightly to a hierarchical Bayesian 12 inference model with distinct hypothesis spaces for statistics and rules, yet linked by a 13 single probabilistic currency. Alternative models based either on a single statistical 14 mechanism or on two non-commensurable systems were rejected. Our results indicate 15 that a hierarchical Bayesian inference mechanism, capable of operating over several 16 distinct hypothesis spaces, underlies the human capability to learn both statistics and 17 rules.

Similarity of brain activity patterns during learning and subsequent resting state predicts memory consolidation

“…Thus, greater statistical learning was determined as faster responses to random high compared to random low trials. In sum, serial-order learning quantified the acquisition of the deterministic rule of the sequential pattern, whereas statistical learning captured purely probability-based learning (Nemeth et al, 2013;Maheu et al, 2020). Figure 1.…”

Neurophysiological coding of statistical and deterministic rule information

“…How the human brain encodes regularities of the environment is a current topic in cognitive neuroscientific research. It has been proposed that learning of patterns includes at least two parallel processes (Batterink et al, 2015;Conway, 2020;Maheu et al, 2020;Nemeth et al, 2013a). One of them is called statistical learning and refers to an automatic acquisition of associative relations and frequencies of external stimuli (Conway, 2020).…”

“…Another attentiondependent mechanism plays a role in controlling the learning process, learning of non-adjacent relations, and integrating rule-based regulations. This second process is often labelled as higherorder sequence learning (Howard and Howard, 1997;Nemeth et al, 2013a), deterministic rulelearning (Maheu et al, 2020), or order-based learning (Simor et al, 2019). It has been proposed that humans organise the temporal regularities in their environment in two distinct hypothesis spaces (Conway, 2020;Maheu et al, 2020): one is based on the estimation of probabilities, and works as statistical bias; the other one is based on deterministic relations or rules.…”

“…This second process is often labelled as higherorder sequence learning (Howard and Howard, 1997;Nemeth et al, 2013a), deterministic rulelearning (Maheu et al, 2020), or order-based learning (Simor et al, 2019). It has been proposed that humans organise the temporal regularities in their environment in two distinct hypothesis spaces (Conway, 2020;Maheu et al, 2020): one is based on the estimation of probabilities, and works as statistical bias; the other one is based on deterministic relations or rules. Importantly, in situations where both statistical and rule-like regularities occur in parallel (e.g., in language or music), not only competition can occur but rules can induce statistical biases and vice versa (Maheu et al, 2020).…”

“…It has been proposed that humans organise the temporal regularities in their environment in two distinct hypothesis spaces (Conway, 2020;Maheu et al, 2020): one is based on the estimation of probabilities, and works as statistical bias; the other one is based on deterministic relations or rules. Importantly, in situations where both statistical and rule-like regularities occur in parallel (e.g., in language or music), not only competition can occur but rules can induce statistical biases and vice versa (Maheu et al, 2020). Parallel learning of different types of regularities can be studied with variations of sequence learning tasks, in which rules and inter-stimulus dependencies are repeated in the stimulus stream (Conway, 2020;Howard and Howard, 1997;Nissen and Bullemer, 1987).…”