Anne Treisman’s most important contribution to science is her feature integration theory of visual attention, which is arguably one of the most important works in the history of cognitive psychology and visual science. In brief, this theory proposed a two-level model. Preattentive features are extracted in parallel in the early visual system so that the search for a feature-defined target (e.g., a red target among many other green items) is easy and effortless, and the difficulty of this task remains constant regardless of the number of non-targets. However, as the name of the theory indicates, the integration of multiple features requires attention, so that the search for an integration-defined target (e.g., a red circle among green circles and red crosses) is effortful, and the difficulty of this task grows with the number of non-targets. In four decades, the feature integration theory has inspired thousands of follow-up studies. Some of these follow-up studies have attempted to revise or clarify various aspects of the feature integration theory. Nevertheless, almost all of them have inherited Treisman’s fundamental insight on the important role of “features” as well as the difficulty associated with the “integration of features”.
What is a preattentive feature? To put it simply, a feature is a type of visual information that can be used to guide visual attention. For a recent example, in a one-to-one comparison task, the preattentive shape features were processed no better than Ts in different orientations (thresholds 48 vs. 47 ms). However, in a visual search task that required processing eight items, the former was processed dramatically better than the latter (thresholds 81 vs. 335 ms). Features play a unique role in attentional processing and are not reducible to merely “discriminability”.
In 1985, Treisman and Souther reported an interesting phenomenon termed “search asymmetry” which is diagnostic for new preattentive features. In a classic example, it is much easier to search for a Q among a set of Os than to search for an O among a set of Qs. This intriguing asymmetry in search performance cannot be explained by the discriminability between target and non-targets. The discriminabilities are equal in these two cases (O-Q vs. Q-O). Instead, this asymmetry is explained in terms of the presence/absence of a feature: It is easy to search for the presence of a feature (i.e., line intersection in this case) but difficult to search for the absence of it.
After the birth of feature integration theory, many researchers have joined Treisman in exploring the list of preattentive features. According to a recent review, researchers have identified a total of two dozen undoubted and probable preattentive features (i.e., described as guiding attributes) including color, orientation, shape, etc. A new member has been added to this list in a recent study by Hayden Schill and her collaborators in the special issue of Attention, Perception, and Psychophysics in Honor of the Contributions of Anne Treisman. Using the classic paradigm of search asymmetry, they argued that the axis of rotation is a basic feature in visual search.
In a set of five experiments (see the example displays per experiment below), Hayden Schill and her collaborators found a search asymmetry between two axes of rotation: search for a rolling target among spinning distractors is more efficient than search for a spinning target among rolling distractors. Specifically, rolling objects rotated about a horizontal axis (e.g., a piece of meat on a skewer, rotating over the fire), and spinning objects rotated about a vertical axis (e.g., child’s toy tops). In Experiment 1-2, they established this search asymmetry in complex objects (Experiment 1) and in spheres which are simple in structure (Experiment 2).
One may attribute the observed asymmetry to physical plausibility. Specifically, in Experiments 1-2, the stimuli were presented in the background of a horizontal plane, and that may have led the observers to a specific expectation on the axis of rotation. Perhaps, rolling stimuli have been expected to move on that plane in a way that spinning stimuli would not. This issue was addressed in Experiments 3-4. In Experiment 3, the background horizontal plane was removed. In Experiment 4, each sphere was pierced by a horizontal or vertical rod to create an expected axis. In both experiments, the search asymmetry (i.e., rolling advantage) persisted, ruling out the physical plausibility as a potential confounding factor.
One may also attribute the observed asymmetry to 2-D texture motion rather than to axis of rotation. Perhaps, one can find the rolling stimuli more easily not because of the axis of rotation per se, but because of the direction of texture motion. In other words, this is essentially an advantage for vertical texture motion rather than for rolling. To address this potential factor, Experiment 5 used checkerboard stimuli that had strong 2-D texture motions but no rotations. Unlike in the previous experiments, Experiment 5 did not show a search asymmetry. Observers were not more efficient searching for the vertically moving texture (roll) among horizontally moving textures (spin). The observed search asymmetry in Experiment 1-4 is due to the axis of rotation rather than to 2-D texture motions.
Together, Schill and her collaborators identified a new search asymmetry between the axes of rotation. How to interpret this result? In Treisman and Souther’s original terms, this would imply that the rolling is a feature and the spinning is the absence of that feature. What does this mean in terms of the underlying mechanism? The authors noted that perhaps the processing (rejection) of the spinning stimuli demands less attention than that of the rolling stimuli, making search through spinning distractors for a rolling target easier than the opposite search.
Psychonomics Article featured in this post:
Schill, H. M., Cain, M. S., Josephs, E. L., & Wolfe, J. M. (2020). Axis of rotation as a basic feature in visual search. Attention, Perception, & Psychophysics, 82(1), 31-43. https://doi.org/10.3758/s13414-019-01834-0
