Research Article: The processing of inter-item relations as a moderating factor of retrieval-induced forgetting

Date Published: August 21, 2012

Publisher: University of Finance and Management in Warsaw

Author(s): Tobias Tempel, Werner Wippich.


We investigated influences of item generation and emotional valence on
retrieval-induced forgetting. Drawing on postulates of the three-factor theory
of generation effects, generation tasks differentially affecting the processing
of inter-item relations were applied. Whereas retrieval-induced forgetting of
freely generated items was moderated by the emotional valence as well as
retrieval-induced forgetting of read items, even though in the reverse direction
(Experiment 1), fragment completion eliminated the moderation of
retrieval-induced forgetting by emotional valence (Experiment 2). The results
corroborate the assumption that the processing of inter-item relations is
crucial for the immunization against retrieval-induced forgetting. Moreover,
differential processing of inter-item relations may clarify the mixed results on
moderating factors of retrieval-induced forgetting that have been reported.

Partial Text

Memory for generated items is mostly better than memory for read items (Bertsch, Pesta, Wiscott, & McDaniel, 2007).
Generation effects have been very thoroughly investigated. Up to the present time,
it remains an open question, however, as to how retrieval affects previously
generated items as opposed to read items.

In Experiment 1, we used the retrieval-practice paradigm, but modified the encoding
phase by letting participants freely generate exemplars in response to category
names as the sole cues in one experimental condition. Four negative and four neutral
categories were used. In the retrieval-practice phase, half of the exemplars from
two negative and two neutral categories had to be retrieved. Subsequently, recall
for all items was assessed. To promote the processing of inter-item relations in
order to intensify a potential moderation of RIF by emotional valence, the items
were presented in blocked order. In the reading condition, all exemplars of one
category were presented at the same time, and in the free-generation condition, all
exemplars of one category had to be generated consecutively. Assuming that
influences of emotional valence were mediated by integration or distinctiveness, in
the reading condition, we therefore expected less RIF of exemplars of negative
categories than of neutral categories or no RIF of exemplars of negative categories
at all (cf. Dehli & Brennen, 2009). In the
free-generation condition, we expected a general immunization against RIF
independent of emotional valence due to the additional promotion of processing of
inter-item relations by the act of generating several exemplars in response to the
same category name. However, the special material of individually generated
exemplars could also suggest the alternative expectation that RIF would emerge for
negative but not for neutral categories. RIF of generated exemplars might be
influenced by the valence of the category in a parallel way as RIF of
autobiographical memories. Perhaps generating autobiographical memories as items and
generating the more abstract material of exemplars of semantic categories have in
common that the generated material is highly self-relevant. Negative (but not
neutral or positive) material, then, might be prone to forgetting.

In Experiment 2, we investigated influences of fragment completion on RIF using
identical items as in the reading condition of Experiment 1. Since free generation
can be considered as the clearest operationalization of a generation task
establishing congruent item-cue and inter-item relations, and our main interest was
to examine the influences of item generation on RIF, we decided to use a generation
task that would establish the incongruency of item-cue and inter-item relations.
Using fragment completion also allowed us to control the time spent on generation.
In contrast to the non-time-limited free generation in Experiment 1, potential
differences in encoding time between the categories thus could not exert any

A total of 40 psychology students at the University of Trier, 31 women and 9 men,
participated in the experiment. They received course credits for their
participation. Two variables, emotional valence (negative, neutral) and item type
(Rp+, Nrp, Rp-), were manipulated within subjects. The stimuli were the same 10
category names and 60 exemplars used in the reading condition of Experiment 1.

RIF of generated items differed from RIF of read items. The aim of our investigation
was to test predictions derived from the three-factor theory of generation effects
(McDaniel, Waddill, & Einstein, 1988;
Steffens & Erdfelder, 1998). On the
one hand, based on the assumption that the processing of inter-item relations
accounted for immunization against RIF, we expected to find a general immunization
for items generated in a task promoting the processing of inter-item relations or
the emergence of RIF only for negative but not for neutral items. The latter result
would parallel studies that demonstrated RIF for negative but not for positive
autobiographical memories (Harris et al.,
2010; Wessel & Hauer, 2006).
The self-relevance of freely generated material in general might determine that
negative material is forgotten easily, but neutral and positive material is not. On
the other hand, no immunization for items generated in a task preventing the
processing of inter-item relations was expected.