Three theories of forgetting

• consolidation failure
• interference
• discrimination failure

Consolidation

Consolidation refers to an encoding process that continues for some period of time after the initial interaction with the stimulus. Under some conditions, then, forgetting might be attributed to the interruption of this consolidation process. By this view:

• memory performance should be better if the consolidation process is not interrupted
• subjects perform better at a recall task after periods of sleep than after periods of wakefulness (Jenkins & Dallenbach, 1928; see Figure 7.1 in text)
• conditioning in rats can be disrupted by administration of electroconvulsive shock (ECS); the closer in time the shock is delivered after the conditioning, the less conditioning is observed, suggesting that the ECS may disrupt consolidation (see Figure 7.2 in text)
• if consolidation is interrupted, the implication is that poor memory performance occurs because of a disruption of encoding
• however, increases in the interval between ECS and the "memory test" results in a return of conditioned responses. Therefore, the conditioned response was not obliterated by the ECS afterall, arguing against a consolidation account of the disruptive effects of ECS

Interference

Interference theories of forgetting are often contrasted with the notion that forgetting occurs because a memory trace decays passively over time. The interference theorist would argue that time itself could not possibly cause forgetting. Instead, something else that occurs with the passage of time must cause forgetting. Therefore, the critical task is to understand why memory performance often deteriorates with increases in the time between encoding and retrieval, rather than using time as an explanation for this effect.

Mechanisms of interference

• response competition: Studying paired associates such as TABLE-DOG and TABLE-AIRPLANE can produce interference when retrieval is cued with TABLE because DOG and AIRPLANE compete with one another in the retrieval process. Interestingly, one might predict that any errors in recall of DOG would correspond to the other associate, AIRPLANE. However, this does not appear to be the case.
• altered context: As noted in the section on context-dependent memory in the preceding chapter, memory performance can be made worse by changing the context between the encoding and retrieval events. In that discussion, context was referred to as the environment in which an event occurred (e.g., under water versus on land); call this "noninteractive context" or "context Alpha." However, context can be used to refer to the interaction of one stimulus with another. This second use of the term context, call it "interactive context" or "context Beta", is made clear by considering the distinction often made between a nominal and a functional stimulus. When presented with the word DOG to study, the nominal stimulus is the letter string itself, which remains stable across presentation to different individuals, in different contexts. The functional stimulus, however, includes processing of the word DOG that may be specific to an individual or a context. For example, when presented with the nominal stimulus DOG, I may think of a dog that lives in the house beside mine, or I may think about a dog that I owned when I was a kid. The functional stimulus, then, includes idiosyncratic aspects of processing that occur upon presentation of a nominal stimulus, and can include both noninteractive and interactive aspects of the context in which an item was presented. Memory performance is hurt by altered context, in the broad sense implied by notions of a functional stimulus
• set: The notion of inappropriate retrieval strategies, or sets, was also used by interference theorists to explain decrements in memory performance. Consider going to the shopping mall to buy milk and eggs at the grocery store, and aspirin at the drug store. You encode this shopping list by thinking "two things at one place and one thing at the other place." After stopping at the drug store to buy aspirin, you also pick up a newspaper, and then head off to the grocery store. Now, after picking up the milk at the grocery store you may find yourself paying for the milk and going home. However, the failure to bring home the eggs does not mean that you could not remember to buy eggs, but rather that your retrieval set, or strategy, prevented you from triggering the cues necessary to remember to buy eggs as well as milk; two things had been bought at one store, and milk qualified as the single thing at another store.

Decay versus Interference

Recall that Peterson and Peterson (1959) reported a classic experiment that examined the cause of short-term forgetting. As in our classroom demonstration, the subject was asked to remember three consonsants while counting backwards by threes for varying amounts of time. The result reported by Peterson and Peterson (1959) was that the retention of the consonants depended on the retention interval, with good performance over short intervals and poor performance after intervals that were just 18 seconds in duration (see Figure 4.3 in text).

Peterson and Peterson (1959) attributed this forgetting effect to decay rather than interference. However, Keppel and Underwoord (1962) demonstrated that there is no forgetting of the consonants even after an 18 second retention interval on the first trial in such short-term forgetting experiments (see Figure 7.4 in text). Across trials 1 to 3, proactive interference appears to build, as does the forgetting forget across increasing retention intervals. Therefore, the conclusion that this short-term forgetting effect is due to decay rather than to interference appears to be incorrect.

Release from proactive interference

The release from proactive interference effect refers to the recovery in memory performance in a short-term forgetting experiment when a switch occurs in the category of items studied. For example, if three color names are studied on three consecutive trials there will be a buildup of proactive interference across those trials. If names of pieces of furniture are used for the the fourth trial, then performance will return near to the level of the first trial; a release from proactive interference triggered by use of a different category. Importantly, it has been shown that this effect can occur even when subjects don't detect the category change until they are told of it at the time of retrieval (see Figure 7.6 in text). This property of the release from PI effect suggests that it is best explained by reference to processes that occur at retrieval.

Although the bulk of data argue against a decay interpretation of forgetting, and in favor of an interference theory of forgetting, the principles sometimes used by interference theorists to account for forgetting across time (from notes for Chapter 4) seem somewhat strained:

• new associations are assumed to obliterate old ones
• with time, old associations spontaneously recover
• this recovery over time underlies the forgetting that occurs with increases in retention interval

The notion that memory performance depends on a discrimination process at the time of retrieval is consistent with data that support interference theory over decay theory, and suggests a way in which forgetting can occur with increasing retention intervals without postulating the recovery of previously obliterated associations.

Relative distinctiveness: Forgetting as discrimination

Turvey, Brick, and Osborn (1970) conducted a short-term forgetting experiment that differed from others in how the retentional interval was manipulated. Rather than use different retention intervals on trials within a session, chosen at random, they used a single retention interval for all four trials within a first session of their experiment. For one group this retention interval was 5 seconds, for another it was 10 seconds, for another 15 seconds, for another 20 seconds, and for a final group it was 25 seconds. They found that forgetting was pretty much the same across groups even though the retention interval differed. This result contradicts the decay explanation of forgetting, but the important issue concerns why there was no retention interval effect for these trials. This issue is answered by considering the results of a fifth trial, in which the retention interval was set at 15 seconds. In this last trial, performance was best for the groups who had previously been tested with a long retentional interval (20 or 25 seconds), intermediate for the group tested previously with a 15 second retention interval, and worst for the groups tested previously with a short retention interval (5 or 10 seconds). These results can be explained by considering the difficulty of discriminating the memory trace of the presented items from recently presented similar traces (see Table 7.4, and note the errors in this table). According to this view, memory performance depends on the ratio of two intervals: the interval between presentation of the studied items for the fourth trial (P4) and the time of retrieval of the studied items for the fifth trial (R5), and the interval between presentation of the studied items for the fifth trial (P5) and the time of retrieval of those same items (R5; the current retention interval). When the first interval is much longer than the second interval it suggests that the items that are currently the target of retrieval are easy to discriminate from the studied items from the prior trial.

One final result that argues in favor of the discrimination account of forgetting is the finding that memory performance can get better with increases in retention interval. This result clearly cannot be accounted for by a decay explanation of forgetting. Neath and Knoedler (1994) demonstrated that the primacy effect in a recognition task of nonverbal material can actually get larger with increases in retention interval. This result was explained by reference to increases in the temporal distinctiveness of the primacy items with increases in retention interval. A simple method of calculating temporal distinctiveness is provided in the text (see page 151).