Distributed attention model of perceptual averaging

Abstract

The visual system efficiently processes complex and redundant information in a scene despite its limited capacity. One strategy for coping with the complexity and redundancy of a scene is to summarize it by using average information. However, despite its importance, the mechanism of averaging is not well understood. Here, a distributed attention model of averaging is proposed. Human percept for an object can be disturbed by various sources of internal noise, which can occur either before (early noise) or after (late noise) forming an ensemble perception. The model assumes these noises and reflects noise cancellation by averaging multiple items. The model predicts increased precision for more items with decelerated increments for large set-sizes resulting from late noise. Importantly, the model incorporates mechanisms of attention, which modulate each item's contribution to the averaging process. The attention in the model also results in saturation of performance increments for small set-sizes because the amount of attention allocated to each item is greater for small set-sizes than for large set-sizes. To evaluate the proposed model, a psychophysical experiment was conducted in which observers' ability to discriminate average sizes of two displays was measured. The observers' averaging performance increased at a decreasing rate with small set-sizes and it approached an asymptote for large set-sizes. The model accurately predicted the observed pattern of data. It provides a theoretical framework for interpreting behavioral data and leads to an understanding of the characteristics of ensemble perception.

Roles of saliency and set size in ensemble averaging

“…Another prediction is that observers sample the items equally regardless of their saliency (middle prediction panel of Figure 1). This is consistent with either the totally distributed, parallel mode of sampling ENSEMBLES: ATTENTION AND CAPACITY 8 (Baek & Chong, 2020), or with limited-capacity sampling, but in both cases no amplification is supposed. This mode of sampling predicts the unbiased error distribution with the large SD.…”

“…Therefore, we conclude that saliency is an important but not an absolute determinant of what is efficiently integrated for an ensemble percept. In addition, we demonstrated that averaging precision tended to grow with set size (probably as a result of noise cancellation, Baek & Chong, 2020), which suggests that the efficiently integrated set size can be quite large. Overall, our data are consistent with a high-capacity averaging process modulated by attentional amplification of a subsample of items.…”

“…However, there is still no strong consensus regarding the true computational capacity of a mechanism beyond this seemingly efficient ensemble representation. According to one theory, ensemble summary statistics are accomplished via parallel exhaustive processing that have no restrictions on the number of items taken into account (Ariely, 2001;Attarha, Moore, & Vecera, 2014;Baek & Chong, 2020;Chong et al, 2008;Chong & Evans, 2011; ENSEMBLES: ATTENTION AND CAPACITY 4 2005; Utochkin & Tiurina, 2014). Presumably, a low-resolution but high-capacity mode of processing involving globally distributed attention can be a proper mechanism for that (Alvarez, 2011;Chong et al, 2008;Chong & Treisman, 2005;Treisman, 2006).…”

“…An alternative account that captures the observed pattern including both growing precision and amplification is high-capacity sampling with attentional modulation. In this account, the increase of the efficient set size causes the decreased averaging error via noise cancellation (Baek & Chong, 2020). Amplification can act upon a limited subsample within this large efficient set size giving more weight to subsample members.…”

“…This flexible mode can work to explain performance in tasks where attention is summoned by particular items via bottom-up saliency cues (as in our study or in Kanaya et al, 2018) or via task demands (as in De Fockert & Marchant, 2008). Whether it works the same way in tasks where no obvious attentional priority is set is a further research question (for example, see Baek & Chong, 2020 suggesting a distributed attention model of ensemble processing).…”