The roles of order, distance, and interstitial items in temporal visual statistical learning

Abstract

Humans are adept at learning regularities in a visual environment, even without explicit cues to structure and in the absence of instruction-this has been termed "visual statistical learning" (VSL). The nature of the representations resulting from VSL are still poorly understood. In five experiments, we examined the specificity of temporal VSL representations. In Experiments 1A, 1B, and 2, we compared recognition rates of triplets and all embedded pairs to chance. Robust learning of all structures was evident, and even pairs of non-adjacent items in a sequentially presented triplet (AC extracted from a triplet composed of ABC) were recognized at above-chance levels. In Experiment 3, we asked whether people could recognize rearranged pairs to examine the flexibility of learned representations. Recognition of all possible orders of target triplets and pairs was significantly higher than chance, and there were no differences between canonical orderings and their corresponding randomized orderings, suggesting that learners were not dependent upon originally experienced stimulus orderings to recognize co-occurrence. Experiment 4 demonstrates the essential role of an interstitial item in VSL representations. By comparing the learning of quadruplet sets (e.g., ABCD) and triplet sets (e.g., ABC), we found learning of AC and BD in ABCD (quadruplet) sets were better than the learning of AC in ABC (triplet) sets. This pattern of results might result from the critical role of interstitial items in statistical learning. In short, our work supports the idea of generalized representation in VSL and provides evidence about how this representation is structured.

General precedes specific in memory representations for structured experience.

“…Additionally, while our experimental design was tailored to assess forward transitions (the probability that A is followed by B), we also consider bidirectional associations (given exposure to MEMORY FOR GENERAL AND SPECIFIC STATISTICS 5 "AB," whether "BA" is endorsed as old at test) in follow-up analyses in the online supplemental materials (online supplemental materials, "Bidirectional Item-Item Links Do Not Predict Responses Better Than Forward Transitions"). Note that the bidirectional associations we interrogate here (as in Park et al, 2018) are in contrast to the more frequently investigated-and similarly named, but distinct-concept of backward transitions (the probability that B was preceded by A; see e.g., Endress et al, 2020;Tummeltshammer et al, 2017), which our study is not poised to measure.…”

“…We show old-new memory decisions reflect these links, building on the past statistical learning literature which has largely measured relative oldness using forced choice tests (cf. Bays et al, 2016; Isbilen et al, 2017; Park et al, 2018; Turk-Browne et al, 2005). This departure from standard testing means we are now able to show that people are sensitive to even small variations in the extent to which test sequences match their prior experience.…”

“…Because we tested each participant on all three exposure triplets (instead of just one), this resulted in a total of 20 test sequences per participant. Of these 20, three were original exposure triplets (meaning they were a sequence of three shapes which had always appeared together in the same order during exposure: ABC [Position 1–2–3], DEF [1–2–3], and GHI [1–2–3]); five were the remaining sequences which included three shapes from the same exposure triplet, but in a shuffled order (e.g., BAC [2–1-3] or IGH [3–1–2]); six were sequences with two shapes from one exposure triplet (e.g., AEC [1–2–3] or IDE [3–1–2], a subset of these were the part-word foils mentioned above); and six were sequences with one shape from each exposure triplet, akin to the “position-matched foils” commonly used in visual statistical learning paradigms (e.g., Park et al, 2018; Schlichting et al, 2017; Turk-Browne et al, 2005; e.g., AEI [1–2–3]; these could also be position-matched to a shuffled sequence: IGB [3–1–2]).…”

Memories of structured input become increasingly distorted across development - Working Paper

“…The fact that here, and in past work, position was not used as a signal for oldness (even after extended exposure, Forest et al, 2021), may seem surprising given other empirical work suggesting the importance of position coding in sequence memory (Clewett et al, 2020;Kalm & Norris, 2017;Kikumoto & Mayr, 2018;Pathman & Ghetti, 2015) and of boundary events in event segmentation (DuBrow & Davachi, 2013;Kosie & Baldwin, 2019;Kurby & Zacks, 2008). Indeed, we included position lures partly inspired by these literatures, and because position has frequently been held constant relative to exposure in past visual statistical learning research in adults (Batterink & Paller, 2019;Luo & Zhao, 2018;Park et al, 2018;Turk-Browne et al, 2005) and children (Arnon, 2020;Jung et al, 2020;Schlichting et al, 2016). Thus, including this lure type provided a methodological link to past studies.…”

Spatial variability induces generalization in contextual cueing.

“…In terms of the flexibility of learned representation, contextual cueing seems to have somehow different characteristics from visual statistical learning (VSL), which is another form of incidental learning. The previous studies of VSL have demonstrated that representations generated by VSL are flexibly expressed (Fiser & Aslin, 2002; Park, Rogers, & Vickery, 2018; Turk-Browne & Scholl, 2009). VSL for a structured sequence (ABC) generalized to the reversed (CBA) and shuffled sequences (e.g., CAB or BAC).…”