368 Conditioning and Learning
contrary to some theoretical formulations (e.g., Rescorla &
Wagner, 1972).
There are several different procedures that appear to pro-
duce conditioned inhibition (LoLordo & Fairless, 1985).
Among them are (a) explicitly unpaired presentations of the
cue (inhibitor) and outcome (described in objective contin-
gency on pp. 361–363); (b) Pavlov’s (1927) procedure in
which a training excitor (Y) is paired with an outcome, inter-
spersed with trials in which the training excitor and intended
inhibitor (X) are presented in nonreinforced compound; and
(c) so-called backward pairings of a cue with an outcome (out-
come→X; Heth, 1976). What appears similar across these
various procedures is that the inhibitor is present at a time that
another cue (discrete or contextual) signals that the outcome is
apt to occur, but in fact it does not occur. Conditioned inhibi-
tion is stimulus-specific in that it generates relatively narrow
generalization gradients, similar to conditioned excitation
(Spence, 1936). Additionally, it is outcome-specific in that an
inhibitor will transfer its response-attenuating influence on
behavior between different cues for the same outcome, but not
between cues for different outcomes (Rescorla & Holland,
1977). Hence, conditioned inhibition, like conditioned excita-
tion, is a form of stimulus-specific learning about a relation-
ship between a cue and an outcome. But because it is neces-
sarily mediated (the cue and outcome are never paired),
conditioned inhibition is more similar to second-order condi-
tioning than it is to simple (first-order) conditioning. More-
over, just as responding to a second-order conditioned stimu-
lus not only appearsas ifthe subject expects the outcome at a
time and place specified conjointly by the spatiotemporal re-
lationships between X and Y and between Y and the outcome
(e.g., Matzel, Held et al., 1988), so too does a conditioned in-
hibitor seemingly signal not only the omission of the outcome
but also the time and place of that omission as well (e.g., Den-
niston, Blaisdell, & Miller, 1998).
One might ask about the behavioral consequences for
conditioned inhibition of posttraining extinction of the medi-
ating cue. Similar to corresponding tests with second-order
conditioning, the results have been mixed. For example,
Rescorla and Holland (1977) found no alteration of behavior
indicative of inhibition, whereas others (e.g., Best, Dunn,
Batson, Meachum, & Nash, 1985; Hallam, Grahame, Harris, &
Miller, 1992) observed a decrease in inhibition. Yin, Grahame,
and Miller (1993) suggested that the critical difference be-
tween these studies is that massive posttraining extinction of
the mediating stimulus is necessary to obtain changes in be-
havioral control by an inhibitor.
Despite these operational and behavioral similarities
of conditioned inhibition and second-order conditioning,
there is one most fundamental difference. Responding to a
second-order cue is appropriate for the occurrence of the out-
come, whereas responding to an inhibitor is appropriate for
the omission of the outcome. In sharp contrast to second-
order conditioning (and sensory preconditioning), which are
examples of positive mediation(seemingly passing informa-
tion, so to speak, concerning an outcome from one cue to a
second cue), conditioned inhibition is an example of negative
mediation(seemingly inverting the expectation of the out-
come conveyed by the first-order cue as the information is
passed to the second-order cue). Why positive mediation
should occur in some situations and negative mediation in
other apparently similar situations is not yet fully understood.
Rashotte, Marshall, and O’Connell (1981) and Yin, Barnet,
and Miller (1994) have suggested that the critical variable
may be the number of nonreinforced X-Y trials. A second dif-
ference between inhibition and second-order excitation that
is likely related to the aforementioned one is that nonrein-
forced exposure to an excitor produces extinction, whereas
nonreinforced exposure to an inhibitor not only does not re-
duce its inhibitory potential, but also sometimes increases it
(DeVito & Fowler, 1987).Retrospective RevaluationMediated changes in stimulus control of behavior can often
be achieved by treatment (reinforcement or extinction) of a
target cue’s companion stimulus either before, during, or
after the pairings of the target and companion stimuli (rein-
forced or nonreinforced). Recent interest has focused on
treatment of the companion stimulus alone afterthe comple-
tion of the compound trials, because in this case the observed
effects on responding to the target are particularly problem-
atic to most conventional associative theories of acquired
behavior. A change in stimulus control following the termi-
nation of training with the target cue is called retrospective
revaluation.Importantly, both positive and negative media-
tion effects have been observed with the retrospective reval-
uation procedure. Sensory preconditioning is a long-known
but frequently ignored example of retrospective revaluation
in its simplest form. It is an example of positive retrospec-
tive revaluation because the posttarget-training treatment
with the companion stimulus produces a change in respond-
ing to the target that mimics the change in control by the
companion stimulus. Other examples of positive retrospec-
tive revaluation include the decrease in responding some-
times seen to a cue trained in compound when its companion
cue is extinguished (i.e., mediated extinction; Holland &
Forbes, 1982). In contrast, there are also many reports of