What Determines Visual Statistical Learning Performance? Insights From Information Theory

Abstract

In order to extract the regularities underlying a continuous sensory input, the individual elements constituting the stream have to be encoded and their transitional probabilities (TPs) should be learned. This suggests that variance in statistical learning (SL) performance reflects efficiency in encoding representations as well as efficiency in detecting their statistical properties. These processes have been taken to be independent and temporally modular, where first, elements in the stream are encoded into internal representations, and then the co-occurrences between them are computed and registered. Here, we entertain a novel hypothesis that one unifying construct-the rate of information in the sensory input-explains learning performance. This theoretical approach merges processes related to encoding of events and those related to learning their regularities into a single computational principle. We present data from two large-scale experiments with over 800 participants tested in support for this hypothesis, showing that rate of information in a visual stream clearly predicts SL performance, and that similar rate of information values leads to similar SL performance. We discuss the implications for SL theory and its relation to regularity learning.

Activation of the cognitive control network associated with information uncertainty

“…Therefore, uncertainty processing executed by the CCN may involve a subjective and dynamic mechanism. These results shed light on the theories of higher-level cognitive domains, such as decision-making and statistical learning ( Alexandre et al, 2019 ; Siegelman et al, 2019 ), and advance the methods available to assess the deficit in cognitive control in patients with neuropsychiatric disorders ( Mackie and Fan, 2016 ; Silverstein et al, 2017 ).…”

Attention Shifts to More Complex Structures With Experience

“…The final stream was random; each shape could follow any other (with a sole constraint on repetitions of the same shape), and thus the TP between any two shapes was 0.14, or 1 of 7 (because there were eight shapes total). In the two structured streams with a TP less than 1, when shape A did not precede its paired shape B (20% or 60% of the time for streams with TP = 0.8 and TP = 0.4, respectively), shape B was replaced by the second item from another pair (e.g., shape D, likewise for all other pairs; see also Siegelman et al, 2019). This shape was from each of the other three pairs an even number of times.…”

“…Then, at Trial 2, the transition between the first and second shape was counted and entropy was recalculated. To compute entropy from TP matrices, we used the Markov entropy formula, which estimates the degree of uncertainty (entropy) in a stream given an item’s probability and between-item TPs and is defined mathematically as − ∑ i = 1 n p ( i ) ∑ j = 1 n p ( j | i ) * l o g 0.25em p false( j 0.25em | 0.25em i false) (see also Nastase et al, 2014; Siegelman et al, 2019). To reiterate, this calculation was repeated after each trial across the learning phase (ignoring search trials) for each location and for each participant—thus, because each of our participants saw each stream of shapes in a unique order, this measure was participant, trial, and stream specific.…”

Attention shifts to more complex locations with experience

“…In the two structured streams with a TP < 1, when Shape A did not precede its paired Shape B (20% or 60% of the time for streams with TP=0.8 and TP=0.4, respectively), Shape B was replaced by the second item from another pair (e.g. Shape D, likewise for all other pairs; see also Bogaerts, Siegelman, & Frost, 2016;Siegelman, Bogaerts, & Frost, 2019). This shape was from each of the other three pairs an even number of times.…”

“…Then, at trial 2, the transition between the first and second shape was counted and entropy was re-calculated. To compute entropy from TP matrices, we used the Markov Entropy formula, which estimates the degree of uncertainty (entropy) in a stream given an item's probability and between-item TPs, and is defined mathematically as: − ∑ ( ) ∑ ( | ) * ( | ) / / (Nastase et al, 2014;Siegelman et al, 2019)). To reiterate, this calculation was repeated after each trial across the learning phase (ignoring search trials) for each location and for each subject-thus, because each of our participants saw each stream of shapes in unique order, this measure was subject, trial, and stream specific.…”