These are the main areas in which we are now conducting research
Emergent Features and Configural Superiority: Complex patterns are built up from many smaller elements, as when a photograph arises from pixels or dots, or when a face arises from the juxtaposition of eyes, nose, and mouth. Often the parts look different, sometimes unrecognizable, in isolation or in other contexts. We study such Gestalt or context effects with visual configurations. Our criterion for spotting emergent features is the Configural Superiority Effect (CSE), where two elements are more readily distinguished within a context than without one. For example, people are better telling apart two diagonal line segments when they are in the company of L intersections arranged to create arrows and triangles. To date we have uncovered over a dozen CSEs, each demonstrating the presence of one or more emergent features or Gestalts.
Building Gestalts from the Ground Up: We would very much like to know how complex configurations like faces and real-world scenes are perceived, but in truth we don’t fully understand how people perceive a simple shape like a square! Even simpler 3-line configurations have proven quite complex, so Mary Portillo launched an effort to build simple Gestalts from the “ground up”, beginning with patterns as rudimentary as two dots. Each dot has its x and y position in space, but a pair of dots has two additional features: the interdot distance and angle. Her work has demonstrated that these two EFs both show configural superiority: It is easier to tell apart two dot patterns that differ in either of these EFs than to distinguish two dots differing only in position, even if the difference involved is exactly the same in the two cases! She has taken this further with 3-dot configurations, uncovering evidence for the Emergent Features of linearity and symmetry. Finally she has shown that those EFs can also produce flat search slopes – popout – and search asymmetries, whereby it is easier to detect a pattern X in a field of Ys than the reverse. Collectively these results suggest that EFs – Gestalts – are basic features in human vision.
Explorations of Two-line Space: If two lines are drawn randomly on a page, will they form a true pattern or Gestalt, or will they simply be two independent lines that happen to be near one another? Anna Stupina has begun research to answer that question, starting with a uniform sampling of the space of all possible 2-line configurations. Starting by determining how well people can spot a horizontal line in a field of vertical lines, say, she then tests that same discrimination ability when a second line is added to each original line to create 2-line configurations. For example, adding a horizontal line to a discrimination between horizontal vs. vertical can convert it into a + sign vs. = sign discrimination. By moving the position of these added context lines, a stimulus space is swept out, and variations in performance over the surface of this space can tell us what the effective emergent features are for humans.
((, )(, )), )(, (), )), ((, )(, ((, (), ))….
on the basis of the direction in which the lefthand parenthesis curves, the task is much harder when the righthand parenthesis varies randomly compared with when it stays the same (GI). But the task is no harder when the righthand element curves oppositely to the lefthand one (no SI). Although it is hard to attend selectively to an individual element when it is part of a larger group, it is easy to spread attention across the whole group. For example, with the sequence of stimuli above, people are faster telling whether the two members of each parenthesis pair are the same or different (parallel or nonparallel, symmetric or asymmetric) than they are at telling which way either curve is facing
Negative Search Slopes: Generally, it takes you longer to find something the more things you must search through. So it takes longer to find a particular pair of socks in a crowded sock drawer than in a nearly empty one. An exception to this rule occurs when the item you’re looking for differs from all the others in one simple way, e.g. a pair of red socks in a drawer full of black socks. Now your search time doesn’t grow with the number of socks to be searched (so long as they all remain in view of course!) In our research, we find even more extreme examples, cases when search becomes faster – much faster – the more items you search through. This finding, sometimes called “supercapacity,” suggests that the background items can group together into a single background field, almost like a textured surface, so that the target then pops out like a spot on a clean, blank wall.
False Popout: A single black sheep will usually pop out from a field of white sheep with no search required. It’s automatic! With pattern elements that configure differently than sheep do in a flock, however, we have found a strange phenomenon we call “false popout.” It occurs when one of the homogeneous background items – one of the white sheep, if you will – pops out. It is easy in the following sequence to determine which is the odd item: X0X. Clearly it is the second item, the 0. But the task is harder with this sequence: )(). Again the second item is the odd one out, but perceptually it appears to group with the third, leaving the first one perceptually isolated. Thus, that first item falsely pops out. It also occurs with four of these six sets of patterns:
Redundancy Losses: Generally people do better at performing perceptual tasks when they are given redundant information. It’s easy to tell a large red square from a small green circle because they differ in color and shape as well as in size. In general, the more different two items are in terms of their physical features, the more dissimilar they will appear to the human eye and the easier they will be to tell apart. We call this result a redundancy gain. With some configurations possessing emergent features, however, the opposite occurs: now stimuli that differ in two respects look more similar than stimuli differing only in one. In some ways this parallels the emergence of color from physical wavelengths, in that the two edges of a rainbow – representing the most extreme wavelengths corresponding to red and indigo – look somewhat similar to one another, as Newton himself noted centuries ago.
Pointing: What is it that makes a visual pattern like an arrow appear to point? We see directional indicators everywhere we look, from signs telling us a street is one-way, though the up and down buttons on an elevator, to the play button on an iPod. Three icons are used most often as pointers: an arrow, a V, and a triangle. What is it about these three icons that make them appear to point clearly and unambiguously? The T symbol does not point well, nor do virtually any of the other characters on a computer keyboard. We view pointing as another example of an emergent feature, and we are trying to isolate what combination of simpler features, such as specific intersections, asymmetry, elongation, are responsible for pointing. We are also looking at whether pointing is something that can or must be learned, rather than being “wired in” from birth. This project has implications for human factors systems, including the design of GPS navigational systems.