Roediger, H.L., McDermott, K.B., Pisoni, D.B. & Gallo, D.A. (2004). Illusory
recollection of voices. Memory,
12, 586-602.
False memories in the DRM
paradigm are often experienced in a subjectively compelling manner. One piece
of evidence consistent with this comes from the voice attribution paradigm
originally used by Payne et al. (1996). Participants listen to DRM lists
presented by two or more speakers (e.g. male and female speaker). At test, if they
indicate that they remember a word, they are asked to indicate if they can
remember which speaker said it. Participants are instructed not to guess
and to only indicate a speaker if they are absolutely certain they can remember
the speaker. The interesting finding is that participants indicate that
they remember a speaker for a large proportion of false memory items,
suggesting a suprisingly rich memorial experience.In the present research the authors tested the
effects of prior recall attempts on false voice attributions as well as the
effect of matching study and testing modalities.
Experiment 1
In Experiment 1
participants studied a series of DRM lists, half presented by a male speaker
and half presented by a female speaker. Half of the lists of each type
were followed by a recall test and the other half were followed by a math
test. Participants then took a recognition memory test in which they also
made voice attributions. Their response options were "Old/Male",
"Old/Female", "Old/Don't Know", or "New".
There was no significant
effect of prior recall on false recognition. Participants made a voice
attribution for about 40% of items from studied lists and did so at about the
same rate for targets and critical lures. There was no significant effect of
prior recall attempts on attributions.
Experiment 2
Experiment 2 used a mixed
list procedure. Half of the items on each DRM list were presented by a male
speaker and half the items were presented by a female speaker.
Prior recall attempts
increased overall recognition of list items (the effect was small for critical
lures; 3%). Voice attributions were more common for targets than for critical
lures. Prior recall attempts did not significantly influence voice attributions
(p=.06). Surprisingly, across experiments accurate voice attributions were as
common for the mixed lists as for the pure lists. The authors argued that this
means that the attributions are being based on item specific information rather
than relying on overall information about the list (e.g. "Everything on
that list was read by the guy/woman").
Experiment 3
Experiment 3 was like
Experiment 1 except that test items were presented auditorially
so that the study and test modalities matched. This should increase memory for
targets and might arguably also influence false memories for the critical
lures. Like Experiment 1 each DRM list was presented either in a male voice or female voice. For the recognition phase of the
experiment participants had these response options: OLD/SAME VOICE, OLD/DIFFERENT
VOICE, OLD, NEW.
There was an overall
effect of prior recall on recognition of list related items. Attributions were
not effected by prior recall. More attributions were made with the test
item was presented by the same voice as it had been at study.
Additionally, participants were more likely to make attributions for both
targets and critical lures that were consistent with the list the item was
associated with.
Experiment 4
Experiment 4 was like
Experiment 3 except that it used the mixed list procedure. Overall
recognition of list associated items was associated with prior recall
attempts. There was no significant effect of prior recall on
attributions. Voice attributions tended to match the voice used at test
regardless of what voice had been presented at study. Unlike the pattern in
Experiments 1 and 2, attributions were more common for pure lists than for
mixed lists.
Discussion
Participants
quite willing to attribute a voice to a false memory item in the DRM paradigm. Large numbers of attributions are
made both in recall and recognition experiments, with or without prior recall,
with pure lists and mixed lists, and even when the modality at test matches the
modality at study. The authors discuss their results in terms of discrepancy
attribution accounts, fuzzy trace theory, and activation monitoring theory. The
authors support activation monitoring accounts. Let's see what you think during
our discussion!