Skip to main content
Open Access Publications from the University of California

UC Irvine

UC Irvine Electronic Theses and Dissertations bannerUC Irvine

Deriving Attention Filters from Statistical Summary Representations to Investigate Mechanisms of Feature-Based Attention for Grayscale

Creative Commons 'BY-NC-ND' version 4.0 license

An attention filter is a neural process acting across space to modulate the effectiveness with which a target feature (e.g., blackness) is processed. I investigated the range of attention filters achievable for grayscale in estimating two summary statistics: (1) centroids and (2) mean orientation (MO) judgments. Chapter 1: in estimating centroids of dots varying in grayscale, participants can achieve two types of attention filters: 1) One gives equal weight and 2) one gives weight graded in proportion to contrast intensity. Comparison of results from different display types suggested three grayscale centroid filters accessible: (1) one assigns equal weight to all grayscales; (2) one assigns weight near 0 to negative contrasts, weight proportional to contrast intensity to positive contrasts; (3) and one assigns weight near 0 to positive Weber contrasts, weight proportional to contrast intensity to negative contrasts.

Chapter 2 investigated whether vision can access the same grayscale-selective attention filters for estimating centroids as MO of bars. Each task had three attention conditions: Attend-Light (equal weight to positive contrasts, weight 0 to negative contrasts), Attend-Dark (equal weight to negative contrasts, weight 0 to positive contrasts), and Attend-All (equal weight to all). Centroid filters matched equal targets well; MO filter accuracy was somewhat worse. Efficiencies, or estimated proportion of items used by an ideal detector, were more than 100% higher for centroids than MO. The pattern of Efficiencies across tasks and attention conditions suggested different strategies between tasks: centroid results suggested participants were applying a filter to the input and computing the centroid of the output; MO results suggested participants were selecting a subsample of bars and basing responses on that sample.

Chapter 3: attention filters giving weight inverse to dot contrast intensity can be achieved. This requires preattentive vision to dampen saliency of the extreme contrast and sharply tune to the low contrast. This implicates a preattentive mechanism sharply tuned to dim items, with discrimination dropping more rapidly with increasing contrast intensity. In the Attend -Dark/ -Bright conditions, filters gave high weight to target polarity’s lowest contrast dots, but weight near 0 to the distractor polarity. This implicates separate mechanisms responsive either to only positive or negative contrasts.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View