Set-size effects for spatial frequency change and discrimination in multiple targets

Wright, M. J., Alston, L., & Popple, A. V. (2002). Set-size effects for spatial frequency change and discrimination in multiple targets. Spatial Vision, 15(2), 157-170.

Abstract:
In visual search tasks with a near-threshold target among distracters, log detection thresholds rise in proportion to the log of the number of stimuli. Previous research has shown a very steep slope for this set-size effect where the target is a change in spatial frequency (SF) across an ISI, suggesting a low-level explanation for 'change blindness.' In this study, we analyse stimulus and task variables in order to determine the contributions of stimulus detection and attention processes using the authors as the sample (N=3). Stimuli consisted of 2 150 ms frames each containing 1 to 4 Gabor targets, with an ISI of 250 ms. In a 2AFC detection task with uniform distracters, slopes of 0.23-0.52 were found, in line with visual search results. 2AFC SF discrimination tasks gave slopes of 0.68, 0.69 with homogeneous distracters and 0.76-0.96 with inhomogeneous distracters, consistent with averaging of stimuli within a frame. Conclusions suggest, under conditions where a stimulus array can be analyzed globally, change detection performance is limited by signal detection mechanisms, rather than limited capacity attention or memory mechanisms. However, where this is prevented, limitations due to attention or memory produce an even steeper set-size effect.

Note:
This paper investigated how people discriminate the spatial frequency changes in multiple gabor patches. They conducted a signal detection approach to explain how decision may play a role in change detection. Change detection is limited by the signal detection mechanism when there is one object or when the distractors are homogeneous. Attention and memory may limit the performance when more than one object are presented and distractors are heterogeneous.

The authors used the idea from Palmer (1994) (see the information integration theory). They suggested that the change detection becomes difficult when multiple stimuli are presented because every stimulus is potentially mistaken as the target (source of noisy information).
Scott-Brown and Ordbach (1998) used uniform distractors which made the task context as simple detection task. (If the distractors are homogeneous, CD is like a visual search with flat slope; f the distractors are heterogeneous, CD is set-size dependent with a steep slope)

*Discrimination is more difficult than detection (Slope (discrimination) > Slope (detection) as a function of set size).
*The ISI between the pre- and post-change displays does not affect the CD

In the present study, they manipulated the type of task (discrimination), distractor homogeneity (uniform, varied in SF, orientation, or phase, or all) (In Exp1, with uniform distractors; In Exp 2, with non-changing distractors; In Exp 3, with changing distractors), presence of pre-cue (allowing to compare the relevant set size with display set size), and the presence of post cue (compare a global processing, or elaborative encoding à post-cue makes participants to build the exact stimulus-location binding). Participants were asked to discriminate the SF of the gabors (2AFC detection, 2AFC discrimination).

This paper did quite a lot of Exps, but the presentation of the data is not very clear and it’s difficult for me to catch the main idea proposed by the authors. I just understand those mentioned in the abstract. Distractor homogeneity may affect the detection slope. Still, I think these manipulations can be applied to my experimental context.

* They used the increment of threshold as a function of the set size. I think I can test this processing architecture by using the test of overall processing time.