That arousal facilitates detection and long term retrieval is not particularly surprising, for most scientific and folk models of memory would suggest that being more alert is a good thing for most cognitive tasks. What is surprising, however, is that increased arousal at storage seems to inhibit immediate retrieval processes. (That it must be a retrieval effect can be understood when the beneficial effects of arousal on long term retrieval are considered.) Finally, most theories of cognitive performance find it difficult to explain the inverted-U relationship between arousal and performance (Hebb, 1955; Humphreys and Revelle, 1984).
Insert Table 1 about here (NOTE--This table is available in the original publication or from W. Revelle. It is not reprinted here).
In further extensions of this work using stimulus materials to induce differential arousal and GSR as a reflection of the arousal, Kaplan, Kaplan & Sampson (1968) replicated the shorter-term recall deficit using two alterations in the design: explicit learning instructions and word-picture pairs as learning materials. Levonian (1966) employed GSR during the screening of a traffic safety film to derive the same type of effect. Jones and colleagues found inferior recall of 9 digit strings for those subjects who evidenced higher baseline EEG activation and higher EEG activation during the first part of the explicit learning trials (Jones, Gale & Smallbone, 1979, Expt. 1).
In an interesting varation of this work, Geen (1973) found that the presence of an observing experimenter led to short-term decrements in recall of nonsense syllable-number pairs learned under explicit instructions. Deffenbacher and colleagues also found that the presence of an observer was sufficiently arousing to produce a highly significant deficit in recall 2 minutes after explicit learning (Deffenbacher, Platt & Williams, 1974). In a later refinement of this work, Geen (1974) found that the presence of an evaluating observer led to a short-term decrement rather than simply an observer per se. Other novel manipulations of arousal that led to successful replication of shorter term deficits include white noise (Berlyne, Borsa, Craw, Gelman & Mandell, 1965, Expt. 3; McLean, 1969, Expt. 1 & Expt. 2); caffeine in humans (Terry & Phifer, 1986) and rats (Terry & Anthony, 1980); exercise (Loftus, 1990, Expt. 1); time of day (Gates, 1916; Folkard & Monk, 1980, Expt. 1; Folkard, Monk, Bradbury & Rosenthall, 1977; Jones et al., 1979, Expt. 3); individual differences (Howarth & Eysenck, 1968; McLaughlin, 1968) and the interaction between time of day and individual differences (Puchalski, 1988).
Attempts to replicate this effect are not without failure, however. Saufley and LaCava's (1977) attempt to replicate using 3-letter trigram-number paired associates and explicit learning instructions failed to find any significant arousal based effects on memory. Schmitt and Forrester (1973) utilized the same experimental materials and paradigm as Butter (1970, Expt. 2). They found a similar decrement in the shorter-term recall of low concreteness-imagery words but garnered no evidence of a matching decrement as a result of GSR to the paired associates. Other failed attempts include Fuller's (1978) use of differences in Introversion/Extraversion and Oakhill (1986) using time of day.
Still other studies that evidenced shorter-term deficits in retention exhibited enhancement of longer-term recall such that the high arousal conditions led to a lesser degree of forgetting than the low arousal conditions. Thus, although Loftus (1990, Expt. 1) found that high self-rated arousal produced by exercise led to inferior shorter term recall, it also led to a greater retention of the initially-recalled word pairs as compared to the low arousal group.
A few researchers have found that high arousal led to benefits both in shorter and longer term retention. Corteen (1969) garnered results that suggested high arousal learning (as measured by GSR) of aurally presented words led to superior shorter- and longer-term recall across testings at immediate, 20 minute and 2 week delays. Similarly, Maltzman, Kantor, and Langdon (1966) found that high arousal learning led to better recall immediately and 30 minutes after learning (This finding represents more of a shorter-term finding given that many studies find the reversal occurs somewhere between 20 and 40 minutes post-learning). Noteworthy here is that these materials were also presented aurally and the materials were once again single words of arousing and nonarousing quality.
Obviously, paradigms involving Sustained Information Transfer (SIT) resource tasks only such as discrimination of degraded digits (Matthews et al., 1990) are less vulnerable to such confounding. In addition, studies that focus on more short-lived state manipulations of arousal and recall intervals that take place safely beyond the range of such a manipulation are also less vulnerable. When the strong influence of individual differences and circadian aspects of arousal on memory are considered, however, it would seem that even longer term retrieval paradigms are placed at risk for this type of confounding.
Some researchers have combined arousal manipulations at learning with arousal manipulations at recall in order to assess potential arousal effects on retrieval probability. Folkard and colleagues (Folkard et al., 1977; Folkard & Monk, 1980) have consistently failed to find any effect of circadian arousal at retrieval upon both longer term recall and recognition tasks. Still others have examined the more direct effect of arousal on retrieval latency and/or probability unconfounded by learning state. M. W. Eysenck (1975a) found that high levels of arousal (a measurement of energetic arousal) as measured before retrieval increased the number of words recalled by extraverts and decreased the number of words recalled by introverts. Moreover, Pascal (1949) found superior recall performance when a relaxation manipulation was given immediately prior to recall. M. W. Eysenck (1975b) also found that moderate levels of arousal (high state arousal extraverts and low state arousal introverts) increased the speed of retrieval when contrasted to low levels (low state arousal extraverts) and high levels (high state arousal introverts) of arousal. Millar et al. (1980) found that retrieval efficiency as measured by retrieval latencies increased during the afternoon when compared to both morning and evening.
A recent study within our own lab (Loftus, 1990, Expt 2) crossed state and trait arousal at learning and immediate recall with that at longer term recall (1 week later). A highly similar pattern was found in both the shorter and longer-term recall performances: high state arousal (measured either by self-report or exercise/relaxation manipulation) tended to increase recall for the high impulsive subjects (low trait arousal) and decrease recall for the low impulsive subjects (high trait arousal). The arousal present at learning was not found to affect recall one week later. The similarity between the two patterns of results is striking and raises the possibility that some sort of retrieval effect was operating at both recall sessions. This pattern of results is reminiscent of an inverted-U relationship (Hebb, 1955) between arousal at retrieval and recall performance.
Another explanation for the effects of arousal on memory is the "tick rate hypothesis" (Humphreys and Revelle, 1984; Revelle, 1989; Revelle & Loftus, 1990) which proposes that arousal increases the rate at which the environment is sampled. In direct analogy to the clock speed of a computer, this hypothesis predicts that increased arousal should lead to a faster rate of response to environmental cues, but to a decrement in availability in immediate memory due to the increased interference associated with a more rapid sampling rate. Finally, by increasing the rate at which the to be learned material is associated with the internal and external context, the model also predicts that high arousal should facilitate long term retrieval.