Attentional Capture - the Relationship Between Feature Salience and Change Detection

Attentional Capture – the relationship between feature salience and change detection

Previous work has demonstrated that change detection is a central determinant of directional attention (Jonides & Yantis 1988), but has failed to clarify the importance of feature salience on the visual search process. In the present study 392 first year undergraduate psychology students were marked on their success at identifying specified alpha numerals on 54 sudden-onset displays, randomly drawn from two set pools (set size 3; set size 7). Subjects were found to score significantly higher for target sudden-onsets (mean error rate 37%) than for distractor sudden-onsets (48.5%). The difference in salience between set size 3 (higher salience) and set size 7 (lower salience) appeared to have a limited impact - with a mean error rate differential of only 2.33%, suggesting that change detection contributes significantly more to directional attention than an object’s salience or conspicuousness. Furthermore the study does state that for these results to be more reliably interpreted in further testing, stimuli variance must increase, and a more sound theoretical understanding of salience should be developed.

Attentional Capture – the relationship between feature salience and change detection

The determinants of visual search have yet to reach consensus, and pivotal to this discussion has been the relationship between feature salience and change detection. Jonides & Yantis (1988) stipulated from their results that change detection is key in determining directional attention. Contrarily Mitroff & Simons (2002) concluded that elemental to visual search and attention capture was the clarity of features within a scene, rendering change localisation as essentially arbitrary.

However, experimentation within this field (Motter & Holsapple 2001; Serences & Yantis 2006; Jonides & Yantis 1988; Mitroff & Simons 2002; Motter & Belky 1998) has been challenging due to numerous unseen obstacles. (Motter & Holsapple 2001). Yantis & Hillstrom (1994) offered evidence for attention capture between new and old objects, but interpretation was convoluted by the somewhat recurrent prevalence of capture errors for equiluminant objects on motion and depth displays. Likewise Mitroff & Simons (2002) and Motter & Belky (1998) both too hastily stated that preattentive mechanisms are arbitrary, consequently ignoring the impact of implicit attentive systems in their results.

Motter & Holsapple (2001), as a consequence of their own annulled hypothesis, later took part in research examining attentional processing during active visual search and concluded that this activity was always undermined by chance contributions. However the chance contributions of the experiment, that lead to this conclusion, were significantly increased by design faults.

Those visual search studies (such as Motter & Holsapple 2001) that used retinal eccentricity as the attention discriminator were impaired by logistical difficulties; for example they were generally unable to use large sample populations. The Motter & Holsapple (2001) and Motter & Belky (1998) studies used only two rhesus monkeys, which may have attributed to the data's statistical insignificance. Alternatively, those experiments using variants of the “flicker” task benefited from much larger sample populations (Mitroff & Simons 2002) and decreased probability of chance contributions.

This particular study was designed upon the “flicker” task model due to this consideration, wherein a sudden-onset (suddenly appearing stimuli) appeared with the target display and subjects were instructed to find specific alpha numerals. This satisfied the aim to observe whether change detection is key in determining directional attention, as consistently reiterated by Yantis (Jonides & Yantis 1988; Yantis & Hillstrom 1994; Serences & Yantis 2006). In order to investigate the varying impacts of salience and change on attention, two set pools were needed, the smaller set condition to represent high target salience, and the larger set condition to represent a display of lower target salience.

Therefore it was hypothesized that when target numerals appeared as sudden-onset, error rates would be significantly lower than conditions with no-sudden onset (control condition) or distractor sudden-onsets (which were expected to mislead subjects), and; that error rates between set sizes (salience) would be proportionally less then those between target, baseline and distractor conditions (change detection).