| |
Abstract:
Abstract: In processing a visual object we may adopt either a
"local" set (processing components) or "global" set (processing the
object as a whole). Ease of processing may depend on the physical
coherence of the object. In particular, local processing may be
more efficient for fragmented objects whereas global processing may
be more efficient for whole objects. To test these hypotheses we
trained subjects to perform a local (component) task and a global
(whole) task on novel two-dimensional objects. Each object
consisted of two components either joined by a bar (whole object)
or not (fragmented object). In the global task, subjects responded
on the basis of whether particular component pairs were present. In
the local task they responded on the basis of whether particular
individual components were present. In the first experiment
subjects were trained with whole objects, and tested on a mixed set
of whole and fragmented objects. We found that they were more
efficient (fewer errors, shorter reaction times) for whole objects
regardless of task. In subsequent experiments subjects were trained
on fragmented objects only or a mixture of whole and fragmented
objects. We found that the whole-object advantage was dependent on
the training set. We conclude that the impact of physical coherence
on object recognition depends on training condition and little or
not at all on processing task.
|
