The retrieval-induced forgetting (RIF) paradigm developed by Anderson, Bjork, and Bjork (1994; see also Anderson, 2005) models the kind of forgetting that occurs unconsciously in response to competition between memories, by practising some memories at the expense of others. Imagine the woman who thinks of her wedding day, and consistently remembers the things that went according to, rather than contrary to, her careful plans. After repeated rehearsals of the things that went right, she is less likely to remember the things that went wrong. Hence, retrieval-induced forgetting avoids cluttering memory with information that is unwanted, redundant, or out-of-date.
In the standard paradigm, participants learn a set of category-exemplar pairs, such as “fruit-apple”, “fruit-banana”, “instrument-flute”, and “instrument- violin”. Participants are then presented with the cue “fruit-a” a number of times, and practise retrieving “apple” repeatedly when presented with this cue. Finally, participants are presented with the categories (fruit, instrument) and asked to recall all the exemplars for each one (see Figure 6.1). Typically, participants are less likely to recall “banana” than they are to remember “flute” or “violin”. This is the RIF effect: retrieval practice reduces recall of unpractised exemplars from the practised category, relative to exemplars from an unpractised category. It has been suggested that when presented with “fruit-a” all the fruit exemplars are activated to some extent, and so successful retrieval practice of “apple” requires the inhibition of the competing, irrelevant fruit exemplar “banana”. This means that “banana” is subsequently more difficult to recall than non-competing irrelevant information (like flute, violin), which was not activated during retrieval practice (see Bjork et al., 1998; see also Levy et al.’s chapter, in Forgetting). It has been argued the RIF impairs both memory accessibility and availability. This is supported by evidence showing that recall of unpractised, related exemplars is still inhibited when tested with a novel, independent cue (Anderson, 2005; Anderson & Spellman, 1995; but see MacLeod, Dodd, Sheard, Wilson, & Bibi, 2003 for a non-inhibitory account).
RIF is considered an automatic, inevitable consequence of practising one piece of information at the expense of another. But researchers have examined whether RIF effects are influenced by motivation. Generally, this has taken the form of comparing RIF for emotional (positive or negative) material with RIF for unemotional material (the standard paradigm uses neutral word pairs).The logic is that people might be motivated to forget certain types of information (e.g., negative information), and so might show greater RIF for these words. Alternatively, people might have difficulty forgetting such information (e.g., in certain clinical populations),
FIGURE 6.1 The retrieval-induced forgetting procedure (Anderson et al., 1994) and so RIF may not occur for emotional material. In other words, are RIF effects selective consistent with the functional view of remembering and forgetting?
Moulds and Kandris (2006) investigated RIF of negative and neutral words in high and low dysphoric participants (dysphoria is a measure of negative mood, and is used as an analogue for depression in non-clinical samples). In general, high dysphoric participants tend to recall more negative than positive memories (Mineka & Nugent, 1995). However, Moulds and Kandris (2006) found that both high and low dysphoric participants showed RIF for neutral but not negative words; that is, in both groups negative words were not forgotten. Similarly, Kuhbandner, Bauml, and Stiedl, (2009) examined RIF for negative pictures and found that the more intensely negative the picture was, the less likely participants were to show RIF for it; this was particularly so for participants in a negative mood. Relatedly, Amir, Coles, Brigidi, and Foa (2001) found that people with generalized social phobia showed RIF for non-social words and positive social words, but not for negative social words; in other words they had difficulty forgetting words that were particularly relevant to their phobia (category-exemplar pairs included, e.g., dating-rejection, dating-clumsy, conversation-babble, conversation-silence). Taken together, these results suggest that motivational factors do influence forgetting in the RIF paradigm: emotionally negative material may be less likely forgotten, and individual memory biases can moderate the effects of retrieval practice.
In updating our review, we note additional findings in the literature about the impact of emotional valence and memory biases on retrieval induced forgetting effects. Dehli and Brennan (2009) reported a RIF effect for neutral, but not positive or negative words. Kobayashi andTanno (2013) reported RIF for neutral words but not negative words, noting that this was due to a reduced baseline in the negative condition. On the other hand, Barber and Mather (2012) reported similar RIF effects for emotional and neutral material, for both older and younger adults. Similarly, Kobayashi andTanno (2015) found RIF effects for both negative and neutral word stimuli, when the retrieval practice involved semantic associates. Overall, with some exceptions, these findings continue to support the general conclusion that emotionality as well as individual characteristics and biases can influence RIF effects, such that RIF effects are not inevitable, but emotional material does not always abolish RIF effects. What then might this predict for RIF of autobiographical memories, which are not only emotional, but are meaningful, complex, and self-relevant?
Macrae and Roseveare (2002) suggested that the personal relevance of the information to be remembered vs. forgotten might influence RIF. In their study, participants learned a list of“gift” words by either imagining themselves purchasing the gift (“self” condition) or imagining another person purchasing the gift (“other” condition). Interestingly, whereas participants in the other condition showed a standard RIF effect, participants in the self condition did not; that is, participants did not forget the gifts they imagined themselves buying, even when these gifts competed for retrieval with practised items. Macrae and Roseveare (2002) argued that self-relevant material might be protected from RIF. Given that autobiographical memories are by definition self-relevant (Conway, 2005), are they susceptible to RIF? Is RIF a good model of autobiographical forgetting?
To test this, Barnier, Hung, and Conway (2004) adapted the RIF paradigm to examine forgetting of positive, neutral, and negative autobiographical memories. In their procedure, participants elicited four memories for each of a number of cues such as “happy”, “tidy”, and “sickness”. Subsequently, participants practised retrieving half their memories in response to half the cues, before being asked to remember all the memories for each cue. Barnier, Hung et al. (2004) found an overall RIF effect. Participants were less likely to recall unpractised memories that competed with practised memories than they were to recall baseline memories. That is, retrieval practice resulted in forgetting of competing, irrelevant autobiographical memories. However, in contrast to RIF research using words and other simple materials, Barnier, Hung et al. (2004) found that emotional valence of the memories did not influence the RIF effect. Rather, independent of retrieval practice, participants were simply less likely to elicit and more likely to forget emotional than unemotional memories.
In a follow-up study, Wessel and Hauer (2006) replicated Barnier, Hung et al.’s (2004) finding of RIF for autobiographical memories. But unlike Barnier, Hung et al., however, they found RIF for negative but not positive memories. In contrast, in a study of individuals experiencing normal and low mood, Harris et al. (2010) found that retrieval-induced forgetting occurred for negative but not positive memories regardless of mood. Stone et al. (2013b) found RIF effects across both positive and negative autobiographical memories, but reported that practising related memories also reduced confidence in the accuracy of positive memories (mirroring the RIF effect), but actually increased confidence in negative memories (opposite to the RIF effect).Therefore the findings about emotion and forgetting of autobiographical memories remain mixed. It may be that manipulating memory valence—positive vs. negative, vs. neutral—does not fully capture memory biases (see Barnier et al., 2007; Harris et al., 2010), and that more subtle manipulations (such as whether memories are personally significant or not and whether memories are self-defining or not) may be required to determine when retrieval practice leads to forgetting of autobiographical memories.