VSS 2010 - Vision Sciences Society 2010 Conference Abstracts

Aks, Haladjian, Kosmides, Annamraju, Kourtev, & Pylyshyn (2010). Blink-induced masking and its effect on Multiple Object Tracking: It's easier to track those that stop during interrupted viewing. Vision Sciences Society 2010, Naples, FL.

When tracking multiple objects, does the visual system encode the location and trajectory of tracked objects? Is encoding only triggered from the abrupt changes that typically occur in the real world such as when objects disappear behind other objects? We extend our 2009 work examining the role of location-coding in Multiple Object Tracking (MOT) using a novel blink-contingent method, enabling us to control simultaneously: item disappearance and abrupt transitions.[1] Here, we introduce backward-masking to the eye-blink paradigm to further control onset transitions. Observers were instructed to blink their eyes when a brief tone was presented midway into each 5s trial of tracking (4 of 8 circles). Eye-blinks induced two events: item disappearance (for 150, 450, or 900 ms), and onset of a mask, which occluded the entire display of items (either for the full disappearance time, or 75ms plus a blackout for the remaining interrupt). During their disappearance, objects either continued moving along their trajectories, or halted until their reappearance. Therefore, "move" objects reappeared further along their trajectory while "halt" objects did not. Results replicate Keane & Pylyshyn, (2006); and Aks et al., (2009) with better tracking when items halt (but here only reliably in the 450 and 900 ms trials). These trends indicate that trajectory information is not encoded during tracking, and the visual system may refer back to past position samples as a ‘best guess’ for where tracked items are likely to reappear. Importantly, the "halt" advantage occurred in both blocked and randomized forms of object motion, suggestive of an automated and data-driven tracking mechanism; one not inclined to predict objects’ trajectories even when presented in a repeated, and thus, predictable context.

[1. Although an eye-blink is a sudden event optically, we are typically unaware of it and it is likely not encoded as a transient event.]

Haladjian & Pylyshyn (2010). Enumeration by location: Exploring the role of spatial information in numerosity judgments. Vision Sciences Society 2010, Naples, FL.

Enumerating a set of visual objects requires the individuation of these items, which inherently relies on location information. To examine the role of location memory in small-number judgments (subitizing), we devised a task that presented observers a brief display of small discs and then required them to mark the location of each disc on a blank screen. In doing so, observers provided an indirect measure of their representation of the numerosity of the display. Observers were tested on three stimulus durations (50, 200, 350 ms) and eight numerosities (2-9 discs); the black discs were approximately 1 degree in visual angle and placed randomly on a gray screen. Following a full-screen mask, observers marked the disc locations on a blank screen by using a mouse pointer to place markers ("X") for each disc. This provided a measure of recall for object locations and display numerosity. ANOVAs on enumeration performance revealed significant main effects for numerosity and display duration (with interactions). High enumeration accuracy was observed for displays containing up to six discs (>90% of trials with perfect recall); error rates increased rapidly for larger numerosities. When observers made counting errors, they were generally underestimates. In the location analysis, error was measured as the distance between a stimulus disc and a paired response disc (discs were paired using nearest-neighbor methods). Location errors were significantly worse in the 50-ms presentation duration and for larger numerosities. We speculate that the process of adding markers for each object provided a way to keep track of which objects had already been counted and thus improved enumeration accuracy. The methodology for this new subitizing task and the implications of the current findings will be discussed.