Ru Qi Yu

Statistical regularities reduce perceived numerosity

Numerical information can be perceived at multiple levels (e.g., one bird, or a flock of birds). The level of input has typically been defined by explicit grouping cues, such as contours or connecting lines. Here we examine how regularities of object co-occurrences shape numerosity perception in the absence of explicit grouping cues. Participants estimated the number of colored circles in an array. We found that estimates were lower in arrays containing colors that consistently appeared next to each other across the experiment, even though participants were not explicitly aware of the color pairs (Experiments 1a and 1b). To provide support for grouping, we introduced color duplicates and found that estimates were lower in arrays with two identical colors (Experiment 2). The underestimation could not be explained by increased attention to individual objects (Experiment 3). These results suggest that statistical regularities reduce perceived numerosity consistent with a grouping mechanism.

Object representations are biased toward each other through statistical learning

ABSTRACT The visual system is remarkably efficient at extracting regularities from the environment through statistical learning. While such extraction has extensive consequences on cognition, it is unclear how statistical learning shapes the representations of the individual objects that comprise the regularities. Here we examine how statistical learning alters object representations. In three experiments, participants were exposed to either random arrays containing objects in a random order, or structured arrays containing object pairs where two objects appeared next to each other in fixed spatial or temporal configurations. After exposure, one object in each pair was briefly presented and participants judged the location or the orientation of the object without seeing the other object in the pair. We found that when an object reliably appeared next to another object in space, it was judged as being closer to the other object in space even though the other object was never presented (Experiments 1 and 2). Likewise, when an object reliably preceded another object in time, its orientation was biased toward the orientation of the other object even though the other object was never presented (Experiment 3). These results demonstrated that statistical learning fundamentally shapes how individual objects are represented in visual memory, by biasing the representation of one object toward its co-occurring partner. Importantly, participants in all experiments were not explicitly aware of the regularities. Thus, the bias in object representations was implicit. The current study reveals a novel impact of statistical learning on object representation: spatially co-occurring objects are represented as being closer in space, and temporally co-occurring objects are represented as having more similar features.

The consistency of the subjective concept of randomness

A pervasive bias in the subjective concept of randomness is that people often expect random sequences to exhibit more alternations than produced by genuine random processes. What is less known is the stability of this bias. Here, we examine two important aspects of the over-alternation bias: first, whether this bias is present in stimuli that vary across feature dimensions, sensory modalities, presentation modes and probing methods, and, second, how consistent the bias is across these stimulus variations. In Experiment 1, participants adjusted sequences until they looked maximally random. The sequences were presented as temporal streams of colors, shapes, auditory tones or tiled as spatial matrices. In Experiment 2, participants produced random matrices by adjusting the color of each cell. We replicated the findings using a within-subjects design in Experiment 3. We found that participants judged and produced over-alternating stimuli as the most random. Importantly, this bias was consistent across presentation modes (temporal vs spatial), feature dimensions (color vs shape), sensory modalities (visual vs auditory), speed (fast vs slow), stimulus size (small vs large matrices) and probing methods (adjusting the generating process vs individual bits). Overall, the results suggest that the subjective concept of randomness is highly stable across stimulus variations.

Statistical learning generates implicit conjunctive predictions

The cognitive system readily detects statistical relationships where the presence of an object predicts a specific outcome. What is less known is how the mind generates predictions when multiple objects predicting different outcomes are present simultaneously. Here we examine the rules with which predictions are made in the presence of two objects that are associated with two distinct outcomes. In three experiments, participants first implicitly learned that an object predicted a specific target location in a visual search task. When two objects predicting two different target locations were present simultaneously, participants were reliably faster to find the target when it appeared in the conjunctive location than in disjunctive locations. This was true even if participants were not consciously aware of the association between the objects and target locations. The results suggest that in the presence of multiple predictors, statistical learning generates implicit expectations about the outcomes in a conjunctive fashion.

Alternation blindness in the representation of binary sequences.

Binary information is prevalent in the environment and contains 2 distinct outcomes. Binary sequences consist of a mixture of alternation and repetition. Understanding how people perceive such sequences would contribute to a general theory of information processing. In this study, we examined how people process alternation and repetition in binary sequences. Across 4 paradigms involving estimation, working memory, change detection, and visual search, we found that the number of alternations is underestimated compared with repetitions (Experiment 1). Moreover, recall for binary sequences deteriorates as the sequence alternates more (Experiment 2). Changes in bits are also harder to detect as the sequence alternates more (Experiment 3). Finally, visual targets superimposed on bits of a binary sequence take longer to process as alternation increases (Experiment 4). Overall, our results indicate that compared with repetition, alternation in a binary sequence is less salient in the sense of requiring more attention for successful encoding. The current study thus reveals the cognitive constraints in the representation of alternation and provides a new explanation for the overalternation bias in randomness perception.

Numerosity perception is distinct from mean or sum perception

The visual system is efficient at extracting a range of ensemble statistics. Most research has independently focused on the estimation of the number, the average, or the sum. Since these processes have been studied separately, their relationship is not well understood. Here we explore the interaction among numerosity, mean, and sum perception in one paradigm. In each trial, observers viewed an array of circles varying in size, and estimated the number, the mean size, or the total size of circles in each array in separate blocks (order counterbalanced across observers). Thus, for every array we obtained numerosity, mean, and sum estimates from the same observer. We noticed that there was consistent underestimation in the number, the mean, and the sum judgments. For every array, we also derived the arithmetic number (the estimated sum/the estimated mean), the arithmetic mean (the estimated sum/the estimated number), and the arithmetic sum (the estimated mean*the estimated number) for each observer. We found that the arithmetic mean was significantly closer to the estimated mean than to the objective mean, and the arithmetic sum was significantly closer to the estimated sum than to the objective sum. However, the estimated number was closer to the objective number than to the arithmetic number. This dissociation suggests that observers may have implicitly followed the arithmetic model for mean and sum estimation, but not for number estimation. Moreover, the errors in the sum estimates were highly correlated with the errors in the mean estimates, but the errors in the number estimates were not correlated with either the errors in the sum or the errors in the mean estimates. This provides further evidence that numerosity was calculated independently from the mean or the sum. Taken together, the results suggest that numerosity perception operates in a distinct manner from mean or sum perception. Meeting abstract presented at VSS 2015.

The perception of multi-dimensional regularities

Regularities are prevalent in many aspects of the environment. How does the visual system extract structured information from multiple sources? One possibility is that the visual system selectively focuses on one source. Alternatively, it may incorporate all sources to form a weighted representation of the regularities. To address this question, we generated matrices containing cells that varied independently on the color dimension (red/blue) or the shape dimension (circle/square). Each matrix could be divided into two equal halves either horizontally or vertically. One half was fully random, whereas the other half was structured (i.e., organized in chunks). Observers discriminated the boundary between the two halves in three conditions (Experiment 1). In the color condition, the cells were structured only on the color dimension; in the shape condition, the cells were structured only on the shape dimension; and in the color+shape condition, the cells were structured both on the color and the shape dimensions. Importantly, each dimension contained an equal amount of regularities. We found that the boundary discrimination accuracy was higher in the color+shape condition than that in the shape condition, but not different from the color condition. This suggests that color was prioritized when regularities were present in both dimensions. To examine whether this prioritization was specific to color, we introduced a new surface dimension (solid/hollow) in the matrices (Experiment 2). Now the boundary discrimination accuracy was the highest in the surface condition, compared to the other dimensions. Critically, the accuracy was equally high when the cells were structured on all three dimensions. This suggests that the surface dimension was prioritized over the others. These findings demonstrate that the visual system relies on one feature dimension to extract regularities, even though every dimension is equally informative. Moreover, such extraction did not benefit from the presence of multiple sources of regularities. Meeting abstract presented at VSS 2015.