Handbook of Psychology, Volume 4: Experimental Psychology

Mood Dependence 75

Affect Infusion at Retrieval

Affect

Infusion

at

Encoding

High

High

Low

Low





Figure 3.3 Circumstances under which evidence

of mood-dependent memory is likely (+) or unlikely

(–) to emerge.

reminders or cues (see Bower, 1981; Eich, 1980). Thus, free
recall seems to be a much more sensitive measure of MDM
than is recognition memory, which is why the former was the
test of choice in all three of the autobiographical memory
studies reported by Eich et al. (1994).
According to the network model, “recognition memory
for whether the word lakeappeared in the list [of probes]
simply requires retrieval of the lake-to-list association; that
association is not heavily overloaded at the list node, so its re-
trieval is not aided by reinstatement of the [event generation]
mood” (Bower & Forgas, 2000, p. 98). In contrast, the AIM
holds that recognition memory entails direct-access thinking,
Forgas’s (1995) term for cognitive processing that is simpler,
more automatic, and less affectively infused than that re-
quired for free recall.
In terms of their overall explanatory power, however, the
AIM may have an edge over the revised network model on
two accounts. First, although many studies have sought,
without success, to demonstrate mood-dependent recognition
(see Bower & Cohen, 1982; Eich & Metcalfe, 1989; for ex-
ceptions, see Beck & McBee, 1995; Leight & Ellis, 1981),
most have used simple, concrete, and easily codable stimuli
(such as common words or pictures of ordinary objects) as
the target items. However, this elusive effect was revealed in
a recent study (Eich et al., 1997; described in more detail
later) in which bipolar patients were tested for their ability to
recognize abstract, inchoate, Rorschach-like inkblots, exactly
the kind of complex and unusual stimuli that the AIM sug-
gests should be highly infused with affect.
Second, although the network model deals directly with
differences among various explicit measures of mood depen-
dence (e.g., free recall vs. recognition memory), it is less
clear what the model predicts vis-à-vis implicit measures.
The AIM, however, implies that implicit tests may indeed be
sensitive to MDM, provided that the tests call upon substan-
tive, open-ended thinking or conceptually driven processes
(see Roediger, 1990). To date, few studies of implicit mood
dependence have been reported, but their results are largely
in line with this reasoning (see Kihlstrom, Eich, Sandbrand,
& Tobias, 2000; Ryan & Eich, 2000).
Before turning to other factors that figure prominently in
mood dependence, one more point should be made. Through-
out both this section and the previous one, we have suggested
several ways in which the affect infusion model may be
brought to bear on the basic problem of why MDM occurs
sometimes but not others. Figure 3.3 tries to tie these various
suggestions together into a single, overarching idea—specif-
ically, that the higher the level of affect infusion achieved
both at encoding and at retrieval, the better the odds of
demonstrating mood dependence.

Although certainly simplistic, this idea accords well with

what is now known about mood dependence, and, more

important, it has testable implications. As a concrete exam-

ple, suppose that happy and sad subjects read about and

form impressions of fictional characters, some of whom ap-

pear quite ordinary and some of whom seem rather odd.

As discussed earlier, the AIM predicts that atypical, un-

usual, or complex targets should selectively recruit longer

and more substantive processing strategies and correspond-

ingly greater affect infusion effects. Accordingly, odd char-

acters should be evaluated more positively by happy than

by sad subjects, whereas ordinary characters should be per-

ceived similarly, a deduction that has been verified in sev-

eral studies (Forgas, 1992, 1993). Now suppose that the

subjects are later asked to freely recall as much as they can

about the target individuals, and that testing takes place

either in the same mood that had experienced earlier or in

the alternative affect. The prediction is that, relative to their

mismatched mood peers, subjects tested under matched

mood conditions will recall more details about the odd peo-

ple, but an equivalent amount about the ordinary individu-

als. More generally, it is conceivable that mood dependence,

like mood congruence, is enhanced by the encoding and

retrieval of atypical, unusual, or complex targets, for the

reasons given by the AIM. Similarly, it may be that judg-

ments about the self, in contrast to others, are more con-

ducive to demonstrating MDM, as people tend to process

self-relevant information in a more extensive and elaborate

manner (see Forgas, 1995; Sedikides, 1995). Possibilities

such as these are inviting issues for future research on mood

dependence.

Strength, Stability, and Sincerity

of Experimentally Induced Moods

To this point, our discussion of MDM has revolved around

the idea that certain combinations of encoding tasks and re-

trieval tests may work better than others in terms of evincing

robust and reliable mood-dependent effects. It stands to rea-

son, however, that even if one were to able to identify and