Watson, J. M,
Balota, D. A., and Roediger III, H. L.
(2003). Creating false memories
with hybrid lists of semantic and phonological associates: Over-additive false memories produced by
converging associative networks. Journal
of Memory and Language, 49, 95-118.
Background:
The DRM paradigm (Deese, 1959; Roediger &
McDermott, 1995) has proven that semantically related word lists are capable of
producing both errors of false recall and false recognition. This is accomplished by providing subjects
with word lists that are all centered around one, nonpresented theme. Not only are Ss very likely to falsely
recall or recognize this critical item, but they are willing to do so with a
great deal of confidence.
Other researchers have taken this paradigm and
altered it so that the lists are centered along a phonological theme rather
than semantic. For example, where the
semantic associate of “dog” may be “cat”, the phonological associate would be
“log”. Note that the phonological
associates may also be orthographically similar. Similar to the semantic false memories in the DRM paradigm,
phonological word lists are also prone to cause false recall and
recognition. Nevertheless, the two list
types have very different effects on the phenomenological experience of the
subject. A higher proportion of
“remember” judgments usually characterizes semantic false recognition, whereas
phonological recognition is more likely to be judged as a “know”
experience.
In the present paper, Watson et al. evaluate the
influence of combining both phonological and semantic associates to induce
false memories. The idea is when
subjects study semantic word lists they may use sensory information to prevent
false memories, even when an abundance of conceptual information exists. By mixing in phonological associates,
subjects may become less capable of distinguishing between what they thought
and what they actually saw. For that
reason, Watson et al. are predicting that Ss will be more prone to false recall
or recognition when studying hybrid lists than either pure lists of semantic
associates or pure lists of phonological associates.
Experiment 1:
Method- The goal of experiment 1 was to
determine the effect of adding phonological associates to existing DRM
lists. Each list contained 15 words
with a minimum of 10 semantic associates and a maximum of 13. The lists containing phonological associates
included either 1, 2, or 3 associates with the rest of the spots filled by
unrelated words. Recall tests were
given after each list.
Results- There was a dramatic increase in
false recall due to adding 1, 2, or 3 phonological associates to the lists (see
fig. 1). Not only was the increase of
false recall greater by adding 3 phonological associates, it nearly doubled its
baseline rate. False recall due to
adding 3 semantic associates increased slightly.
Experiment 2:
Method- The main goal was to specify
whether hybrid lists of phonological and semantic associates would have an
under-additive, additive, or over-additive effect. To address this, subjects were given six 72-item lists. Two of their lists consisted of 18 semantic
associates and the rest were unrelated words.
Two of the lists contained 18 phonological associates, the rest
unrelated items. The final two lists
consisted of both 18 semantic associates and 18 phonological associates. Once again, they followed each list with a
test of recall.
Results- As in experiment 1, false recall
increased dramatically in the hybrid lists containing both semantic and
phonological associates. Furthermore,
the false recall rate of the hybrid lists was greater than the predicted sum of
both types of associates proving that an over-additive effect existed. (And, if you are interested, they believe
this may be the highest level of false recall in a DRM list since the beginning
of man.)
Experiment 3:
Method- Because of possible confounds,
Watson et al. felt it necessary to replicate the results of experiment 2 with
lists containing an equal number of associates (in this case 16). So, Ss were presented with twenty-four
16-item lists. Each list consisted
entirely of semantic associates, phonological associates, or a hybrid of the
two item types. After each list, they
were given one of three conditions: a recall test, a fun-filled algebra test,
or nothing. Finally, they were given a
surprise recognition test consisting of 108 items. For this, subjects were asked to give remember/know judgments
according to whether or not they could recall specific aspects of the word’s
presentation, or if it just seemed familiar.
Results- The results of experiment 3
supported those of the previous experiment only now they could be certain that
the over-additive effect was due to a combination of the two types of
associates and not just the fact that there were more presented. Furthermore, consistent with previous
research (Schacter et al., 1997), false recognition of semantic associates was
commonly referred to as a “remember” judgment, whereas false recognition of
phonological associates was more likely to be labeled a “know” judgment. Lastly, the false recognition of hybrid
lists was most likely to be called a “remember” judgment regardless of whether
recognition was preceded by a recall or algebra task.
General Discussion:
·
The main thing to understand here is that by combining both semantic and
phonological associates, you are more likely to get a false memory than by
doing either one alone.
· Experiment 1 shows us that adding 3 phonological
associates to a gives a stronger effect than adding 3 more semantic associates
to a DRM list.
· Experiment 2 produced a higher level of false recall
with hybrid lists than could have been expected by adding the effect of pure
semantic lists with pure phonological lists.
·
Experiment 3 extended the over-additive effect to a situation where the total
number of associates was equal across all lists.
There are several possible frameworks that could
explain the over-additive effect found by Watson et al.
· First, the spreading activation model
could explain this by assuming that there are distinct lexical representations
with varying levels of sensitivity. So
by increasing the number of associates in that network, activation may become
more likely.
· Second, an interactive activation model
assumes there are two stages to lexical access. The first stage is a top-down activation in which the word is
viewed conceptually and mapped onto the semantic network. The second stage is a bottom-up approach in
which the word is viewed phonologically and mapped into a phoneme network. Because both of these processes are
occurring, a convergent item has a greater chance of being mistakenly
activated, at least more so than if only semantic or phonological associates
were studied.
· A final suggestion is that false memories are occurring
as the result of source monitoring failure. Pure lists of semantic associates may be difficult
to distinguish between targets and the critical lure in the subject’s semantic
network. However, they may be able to
distinguish between the two on a orthographic or phonological level. When hybrid lists are studied it may be
harder to decide if the critical item was presented or whether items that
looked, sounded, or meant the same thing were presented. So, hybrid lists provide the subject with
both perceptual and conceptual support for the critical item.