PURPOSE: Part-based theories of object recognition have been around for
nearly 20 years (Biederman 1987 Psych.Rev. 94 115-147) and there is recent
evidence for part-based attention (Vecera/et.al. 2001 Percep.&Psych. 63,
308-321), but viewing strategies in object identification tasks have not
been studied. What strategies do subjects employ when identifying objects?
How are these strategies affected by the appearance of parts? We
investigate these questions with a word identification task.
METHODS: Subjects' horizontal and vertical eye positions were measured
using the SRI Dual Purkinje Image Eye tracker. Words were presented with
letters subtending 1 deg of visual angle to create a closer analogy to
object identification tasks. Subjects returned to a central fixation marker
after identifying the word that appeared in a random location. Four word
conditions were used: unmasked (control), front, middle and end masked.
RESULTS: In the control condition, subjects were able to identify the word
in ~1 second. In all conditions, subjects first fixated at 35% of the word
length, consistent with the optimal word reading location (O'Regan 1992 In
Rayner: Eye movements and visual cognition p333-354). When the central
region was masked, the fixation pattern was identical to that of the
control condition, but dwell times increased by 50%. When the beginning and
end regions were masked, additional fixations were made at 20% and 80% of
the word length, respectively. CONCLUSIONS: First fixations fall on the
most semantically meaningful part of the object, in our case, the root of
the word. When the appearance of parts outside the fovea is inconsistent
with the object hypothesis, subjects modify their strategy to include a
fixation in the ambiguous region.