Hancock, T. W., Hicks, J.
L., Marsh, R. L., and Ritchel, L.
(2003). Measuring the activation
level of critical lures in the Deese-Roediger-McDermott paradigm. American Journal of Psychology, 116,
1-14.
Background: In the DRM paradigm, participants are shown a list of
thematically related words. Often,
these lists will elicit a false memory for a critical lure that was never
actually presented. For example,
individuals who study bed, rest, awake, slumber, and nap may later
report having seen sleep (Deese, 1959; Roediger & McDermott, 1995). These results are consistent even in
experiments where the participants are forewarned about the purpose of the
study (Gallo et al., 1997; McDermott & Roediger, 1998). Even source-monitoring instructions were
only capable of limiting false recall (Hicks & Marsh, 1999 &
2001).
Several theories have been attempted to
explain the occurrence of false memories.
For example, the implicit associative response (IAR)
theory suggests that the critical lure comes to mind consciously during the
study of lists (Gallo et al., 2001; Roediger et al., 1998). Then, do to a failure in source-monitoring,
participants are unable able to distinguish between those words which were
presented, and those which were internally activated. The fuzzy-trace theory suggests that participants
are relying on gist, as opposed to verbatim traces, or that they are simply
having a “phantom” recollection for the critical lure (Reyna & Lloyd,
1997).
Despite
all the research that has examined the conditions that reduce false memories,
there has been little to examine the underlying mechanisms responsible for
their creation. Hancock et al. feel
that focusing on manipulations that affect the magnitude of the false memory
will give greater insight as to their underlying causes. The method they chose to follow was that of
Robinson and Roediger (1997), who found that the critical lure was more likely
to be evoked when studying lists with a greater number of related items.
The
goal of this study was to compare the activation level of the critical lure
with that of truly studied items. To
exam this, they used a lexical decision task which relies on recognition times
for words vs. nonwords. More highly
activated words are responded to more quickly than the less activated
items. Thus, shorter latencies of
identification indicate a high activation.
So, if critical items are identified as words faster than non-presented
control items, it may mean that the word was consciously activated during study
and an IAR is occurring.
Experiment
One: In this experiment, 24 lists containing the
top 14 associates of the critical item were used. Simple addition problems were administered before the lexical
decision task to serve as a distraction period. During the lexical decision task, participants were exposed to
presented words, non-presented words, nonwords, and the critical lure of each
list. The average response latency was
used to determine the activation level for each word.
Results:
· Nonwords received the longest latencies.
· The
critical lures were responded to much more quickly than their control matched
counterparts, which suggests they were highly activated through study of the
DRM lists.
· The
critical lures were even responded to faster than the actually studied items
suggesting the activation was very strong for the critical lures, maybe even
occurring multiple times. This becomes
referred to as the superadditive priming effect.
Experiment Two: In this experiment, all procedures were the same except that
participants were tested over items in the 2nd and 13th
position because of their higher and lower backward associations to the
critical lure. If reaction times are
different between the two words, it can be safe to assume that activation is
proportional to how related the words are to the list theme.
Results:
· The reaction time to the 2nd and 13th items
did not differ, meaning that reaction time does not depend on how related an
item is to the critical lure, just that it was studied.
· They also replicated the superadditive priming effect found in
Experiment 1.
Experiment Three: Going back to Robinson and Roediger’s findings, Hancock et al.
manipulated the number of words presented to measure its effect on activation
of the critical lures. So, they
presented only the top three items from each word list and tested over items in
positions 2 and 3.
Results:
· The critical lures received faster responses than their
control-matched words, suggesting they did accrue activation from the
presentation of the three associates.
· This experiment, however, lacked the superadditive priming effect
supporting Robinson and Roediger’s theory about total association strength.
Experiment Four: The same procedure as Experiment 3 was replicated, but the bottom
three associates of each list were used.
If the lower associates have a lower probability of evoking an IAR, then
latency to the critical lures should be slower than the presented items.
Results:
· Once again, latencies to the critical lure were shorter than
those to the control-matched words. So,
even the three lowest associates can create activation for the critical lure.
· Also, the critical lure is as activated as the truly studied
items, but not higher as in Experiments 1 and 2.
General Discussion: So, when a full list was studied, they found a superadditive
priming effect on the critical lures.
Moreover, this was true of items from any point in the study lists as
all appeared to be equally activated.
When fewer items were presented, the superadditive priming was lost, but
critical lures were still found to be more highly activated that than the
non-presented control items. These
results indicate that activation levels were related to the number of items
presented. So, Hancock et al. believe
their results suggest that the degree of activation in the critical lure
determines it’s availability, which in turn may be the leading cause in the
lure becoming a candidate for becoming a false memory.
Hancock et al. found their
data to support the IAR account that critical lures come to mind during
study. However, they believe that
source-monitoring could be one way of sorting them out. Therefore, the availability of the lures may
just be the first stage in creating a false memory, but they must also make it
past any monitoring that may occur during retrieval (McDermott and Watson,
2001; Schacter ,1999). This is referred
to as a two-process theory of recognition.
A single-process model, on the other hand, would suggest that only
activation level, or familiarity, would be enough to elicit a false memory
(Underwood, 1965). These theories fall short in several explanations, however
(Seamon et al., 2002). So, Hancock et
al. feel that the best theory appears to be that items are initially activated
during encoding, and then are left to pass some sort of monitoring test before
becoming a false memory.