| |
Abstract:
Linear and nonlinear categorization rule learning was
investigated in Parkinson's disease (PD) patients. Subjects were
asked to categorize single line stimuli that varied in length and
orientation (separable-dimensions). In the Linear Integration (LI)
condition, correct classification was based on a linear
relationship between the line length and orientation. In the
Nonlinear Integration (NLI) condition, correct classification was
based on a nonlinear relationship between the line length and
orientation. All other procedural aspects were fixed across
conditions (e.g., optimal accuracy, nature of the feedback,
response requirements, etc). Each subject completed 6-100 trial
blocks in each condition. Accuracy-based analyses, as well as
quantitative model-based analyses, were performed. Unlike
accuracy-based analyses, the model-based analyses allow one to
quantify and separate the effects of categorization rule learning
from variability in the trial-by-trial application of the rule. The
results suggested that (a) PD patients were not less accurate than
controls in the LI condition, (b) PD patients were less accurate
than controls in the NLI condition, and (c) the locus of the
accuracy deficit in the NLI condition was due to a deficit in
categorization rule learning, and not to increased variability in
the application of the rule. Overall, these results are consistent
with our previous study that demonstrated a deficit in PD patients
in learning nonlinear rules.
|
