Abertay Research Collections >
Social & Health Sciences >
Social & Health Sciences Collection >
Please use this identifier to cite or link to this item:
|Title: ||Is visual search a high-level phenomenon? Evidence from structure perception in 3-D scatterplots|
|Authors: ||Shovman, Mark|
Bown, James L.
Scott-Brown, Kenneth C.
|Affiliation: ||University of Abertay Dundee. School of Social & Health Sciences|
|Issue Date: ||2010|
|Rights: ||Published version (c)Pion, available from http://dx.doi.org/10.1068/ava10am|
|Citation: ||Shovman, M., et al. 2010. Is visual search a high-level phenomenon? Evidence from structure perception in 3-D scatterplots. Applied Vision Association Easter Meeting, Liverpool Hope University, Liverpool, UK, 29 March 2010, Abstracts. Perception. 39(8): pp.1149-1150. Available from http://dx.doi.org/10.1068/ava10am|
|Abstract: ||Increasing use of 3-D scatterplots for trend detection in visual analytics, raises the theoretical
question: what constitutes an object in such a task? Previously (Shovman et al, 2008 Perception
37 79 ^ 80) we have shown that detection of a 3-D position of a single outlier point exhibits
characteristics of serial visual search. According to feature integration theory (FIT) (Treisman
and Gelade, 1980 Cognitive Psychology 12 97 ^ 136) that implies that every point is a complex
perceptual object, and therefore detection of trends or patterns in a scatterplot will take longer with increasing number of constituent points. Conversely, according to reverse hierarchy theory
(RHT) (Hochstein and Ahissar, 2002 Neuron 36 791 ^ 804) the object is the highest-level task-
relevant arrangement of points. Therefore, RHT predicts that trend detection will take longer
with more point groups. Participants identified (2 ^ 4AFC task) a group of points whose position-
ing exhibited a 3-D structure when actively rotated. Number of points per group (64, 100, 144
196, 256) and the number of groups (2, 3, or 4) were manipulated independently. The dependent
variable was the scene rotation duration, ie the time when the 3-D structure was potentially
visible. Rotation times increased with number of point groups and decreased with number of
points per group. This is consistent with RHT and contradicts FIT. In conjunction with previous
experiments (Shovman et al, 2009 IEEE Proceedings of International Conference InformationVisual-
isation pp 135 ^ 138), these data support connecting processes of visual search to task-relevant,
high-level semantics of a scene rather than to its low-level visual features.|
|Appears in Collections:||Social & Health Sciences Collection|
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.