Hannigan, S.L. & Reinitz,
M.T. (2003). Migration of objects and inferences across episodes. Memory
& Cognition, 31(3), 434 – 444.
Past research has demonstrated that information gained
during one episodic event can be falsely remembered as having occurred in
another event. For example, someone
seen on the day a crime is witnessed might later be falsely recalled to have
been at the crime scene (Ross, Ceci, Dunning, & Toglia, 1994).
The present study deals with this type of false memory
and looks at two questions:
1.
1.
Will these episodic conjunction errors occur more frequently when the
two events are
Anchored in the same schema (i.e.
two separate events of going to a restaurant) as opposed to when they are
different schemas (i.e. going to a restaurant and waking up in the morning)?
2.
2.
Do these types of cross-episodic migrations even happen with inferred
events (i.e.
those that are
not displayed in the actual event sequence, but could be inferred either
causally or
effectually).
For the second question, the researchers want to test if
these inferred events act in a manner similar to other items in memory in so
much that they too can be transferred across episodes to form false memories.
Subjects were exposed to 6 episodes during study with
each episode being 7 slides long depicting an event in normal progression. The first and the last events were fillers
to control for primacy and recency effects and were not involved in the test
portion of the experiment. The
remaining 4 episodes consisted of two episodes of eating at a restaurant
and two episodes of getting up in the morning.
Within the same schemas, the environments and actors were very different
so to provide a good deal of contrast so the episodes wouldn’t run
together. Each slide was presented for
4 seconds, with a 3.5 second break between each one, and a 12 second break
between episodes. This was followed by
a 20-minute distracter task. The test
consisted of 12 scenes. Four of these
scenes were “old”, and the rest were new broken down into 2 conjunctions, 2 controls, and 4 previously unseen filler
items. The conjunction scenes were
created by taking a critical target object from one episode and putting it into
the other episode, within-schema for one of the conjunctions and
between-schemas for the second. The
control foils contained a critical object, but not one shown during the study
phase. New items were just completely
new slides but contained the same actors in a previously seen context.
The results came out as expected and a significant effect
of the test condition was found.
Critical objects migrated freely between schemas that were thematically
similar. A significant effect was not
found for the migration of critical objects in the between episode
condition. Confidence ratings for the
within schema condition were significantly higher than those in the control
condition.
For this experiment, there were 8 episodes in all
comprised of an A and B version of 4 themed episodes. Each episode had 14 slides out of which 4 were made up of 2
casual pairs (table 1), 4 were schema
based slides (2 high- and 2 low-schema relevant), and 6 non-experimental. At study, subjects viewed 8 episodes, 9
slides each. Each action sequence in
each episode was designed to have 2 effect scenes which could lead to backward
inference/same episode and backward inference/conjunction errors. After study and one of 3 delays, subjects
received a 24-item recognition test using the same confidence scale as the
first experiment.
This test includes old items, half high-schema, half
low-schema, and then new items in 8 different condition which are:
backward-inference/same episode, backward-inference/conjunction,
backward-inference/control, new high-schema relevant, new low-schema relevant,
high-schema relevant/conjunction, and low-schema relevant/conjunction.
Confidence ratings are shown in
figure 2.
An ANOVA was done to compare
the backward-inference/same episode and backward-inference/conjunction
conditions with the control. The results of the ANOVA show that in both of
these experimental conditions, participants responded “old” significantly more
often than to the control condition.
Significantly higher confidence ratings were given to higher-schema
relevant conjunction items over the new high-schema relevant items.
-Objects are
semantically constrained so that they migrate freely only across episodic
events that are of the same schema.
-Effect scenes showed during an episode lower the confidence of
the subject when rejecting new cause scenes from a schematically related
episode.
-Backward inferences drawn from one episode can produce false
memories by migrating to a different episode with the same schema.
-Backward inference errors occur more significantly within an
episode as opposed to across episodes.
-High schema relevant conjunction errors occurred more often
than low schema relevant conjunctions.
-Confidence ratings for same-episode and between-episode
inference conditions as well as the ratings for high and low schema relevant
source misattribution increased significantly from the 15 min delay to the 1
day delay.