Information Technology Reference
In-Depth Information
Improving Control at Retrieval Via Experience
Just as learners can be induced via instructions to shift their response bias in a
recognition task, they have also been shown—in some circumstances—to adaptively
adjust their bias across experience with a task. For example, Benjamin (2001; see
also Benjamin & Bawa, 2004) found that presenting a word list three times, rather
than only once, led young adult learners to adopt a more conservative response bias
and thus to less frequently falsely endorse unstudied items that were highly related
to studied items. Han and Dobbins (2009) found that learners shifted their bias in
response to experience with misleading feedback. Learners who were told that they
were correct when they replied “new” to a studied item adopted a more conservative
bias (increasing misses), while learners who were told that they were correct when
they replied “old” to an unstudied item adopted a more liberal bias (increasing false
alarms). However, whether a learner engages in a response bias shift and whether
that shift increases accuracy depends on a host of as-yet unidentified factors, and
there are numerous cases in which such strategic shifts are not obtained (Healy &
Kubovy, 1977; Stretch & Wixted, 1998).
In free recall tests, learners tend to output the most recently studied items first
(Deese & Kaufman, 1957). Furthermore, learners increasingly adopt this retrieval
strategy across experience with multiple study-test cycles (Huang, 1986; Huang,
Tomasini, & Nikl, 1977). This effect can be seen as learners learning to take advan-
tage of the fact that not only are the most recently studied items better recalled than
older items on an immediate test (Murdock, 1962), but this recency effect quickly
evaporates (Jahnke, 1968). This would be consistent with the findings of Castel
(2008): learners' JOLs reflected an improved appreciation for serial position effects
(the benefits of primacy and recency) when learners were given experience across
multiple study-test cycles and when serial position was made salient by either col-
lecting JOLs prior to presenting each item or by explicitly presenting each item's
serial position during study.
When a subset of studied material is again presented at a free recall test, osten-
sibly to help the learner remember the rest of the material, these cues can actually
impair that performance. This is known as the part-list cuing or part-set cuing effect
(e.g., Nickerson, 1984). Liu, Finley, and Benjamin (2009) investigated whether
learners would come to appreciate the potentially deleterious effects of part-list
cues across five study-test cycles in which learners were allowed to choose how
many cues they would receive on the test. In each cycle, learners first studied a list
of 30 words presented one at a time. At the end of this presentation, learners chose
how many of the words (from 0 to 15) they wanted to be given as cues on the test to
help them remember the rest of the words. Finally, learners were given a free recall
test that represented the number of cues they had requested and instructed learn-
ers to recall the non-cue words. Learners indeed chose fewer cues across cycles,
demonstrating a strategic improvement in their choices of testing condition. This
is consistent with work by Rhodes and Castel (2008) which showed that learners'
