Class Notes (1,100,000)
US (450,000)
UMD (8,000)
INST (100)
Anton (10)
Lecture 6

INST 354 Lecture 6: INST354 Lecture 6: Behavioral Analysis


Department
Information Studies
Course Code
INST 354
Professor
Anton
Lecture
6

This preview shows half of the first page. to view the full 2 pages of the document.
INST354 Lecture 6: Behavioral Analysis
Behavioral analyses of responses to this task focused purely on the reinforcement (being told
“right” or “wrong”) for each choice. Awareness, to the degree that it exists, was assumed not to
affect sorting. Early results appeared to support such analyses. For example, some participants
were able to achieve perfect sorting without being able to verbalize the experimenter’s rule
(although it turned out that they could if pressed, their earlier reluctance apparently resulting
from being unsure of themselves), and in some tasks participants did not achieve the perfect
learning that would be predicted from intellectual insight (but the experimenter’s rules
themselves may have been ambiguous). Moreover, average success in concept identification
across participants appeared to increase gradually, much like the learning of an athletic skill.
However, clever follow-up experiments demonstrated that learning in such tasks was in fact
not gradual but “all or none,” the type of learning predicted on the basis of an active hypothesis-
testing mind that continually searches for the correct rule whenever the experimenter indicates
that an incorrect sorting has been made. First, these investigators analyzed each participant’s
responses separately and determined the pattern of correct and incorrect responses prior to the
last error. If learning was gradual, as predicted by most reinforcement theories, the probability of
a correct sort, within a single participant’s learning trials, should increase gradually from .50 (the
chance probability of being correct). Instead, it was stationary at .50. The gradual increase
found earlier was an artifact of averaging across participants who had identified the correct
concept at different points of time in the experiment. Moreover, patterns of sorting after each
error were indistinguishable irrespective of the point in the experiment at which the error
occurred. By making an error, the participant indicated that he or she didnt get the concept;
hence, performance was at the chance level prior to each error. An error indicated that the
participant had not yet had the insight into the experimenters rule.
Marvin Levine (1975) demonstrated that participants’ conscious beliefs were virtually perfect
predictors of their responses, particular error patterns, and time it took to learn. In an especially
ingenious demonstration, he showed that participants failed to learn very simple concepts (e.g.,
to sort all stimuli to the left), over hundreds of trials, if this concept was unexpected, or absent
from their hypothesis set. Bower and Trabasso (1968) devised a procedure they termed the
alternating reversal shift procedure. Every second time the participant made an error, the rule
was reversed. For example, participants who had initially been told correct” when they placed
You're Reading a Preview

Unlock to view full version