366 Conditioning and Learning
(i.e., X→AB), responding on tests of the X→A relationship
often yield less responding than that of a control group for
which B was omitted, provided A and B are sufficiently dif-
ferent. Such a result might be expected based on either
distraction during training or response competition at test,
both of which are well-established phenomena. However,
some studies have been designed to minimize these two po-
tential sources of outcome competition. For example, Burger,
Mallemat, and Miller (2000) used a sensory preconditioning
procedure (see this chapter’s section entitled “Second-Order
Conditioning and Sensory Preconditioning”) in which the
competing outcomes were not biologically significant; and
only just before testing did they pair A with a biologically
significant stimulus so that the subjects’ learning could be as-
sessed. As neither A nor B was biologically significant during
training, (a) distraction by B from A was less apt to occur
(although it cannot be completely discounted), and (b) B con-
trolled no behavior that could have produced response
competition at test. Despite minimization of distraction and
response competition, Burger et al. still observed competition
between outcomes (i.e., the presence of B during training at-
tenuated responding based on X and A having been paired).
To our knowledge, no one to date has reported facilitation
from the presence of B during training. But analogy with the
multiple-cue case suggests that facilitation might occur if the
two outcomes had strong within-compound links (i.e., A and
B were similar or strongly associated to each other).
Multiple Outcomes Trained Apart With a Single Cue:
Counterconditioning. Just as multiple cues trained apart
with a common outcome can result in an interaction, so too
can an interaction be observed when multiple outcomes are
trained apart with a common cue. Alternatively stated, re-
sponding based on X→A training can be disrupted by X→B
training. The best known example of this is countercondi-
tioning(e.g., responding to a cue based on cue→food train-
ing is disrupted by cue→footshock training). The interfering
training (X→B) can occur before, among, or after the target
training trials (X→A). Although response competition is a
likely contributing factor, there is good evidence that such
interference effects are due to more than simple response
competition (e.g., Dearing & Dickinson, 1979). Just as inter-
ference produced by Y→A in the X→A, Y→A situation can
be due in part to degrading the X-A contingency, so atten-
uated responding produced by X→B in the X→A, X→B
situation can arise in part from the degrading of the X-A
contingency that is inherent in the presentations of X during
X→B trials. However, research has found that the response
attenuation produced by the X→B trials is sometimes greater
than that produced by X-alone presentations; hence, this sort
of interference cannot be treated as simply an instance of de-
graded contingency (Escobar, Arcediano, & Miller, 2001).Resolving AmbiguityThe magnitude of the interference effects described in the
two previous sections is readily controlled by conditions
at the time of testing. If the target and interfering treatments
have been given in different contexts (i.e., competing ele-
ments trained apart), presentation at test of contextual cues
associated with the interfering treatment enhances interfer-
ence, whereas presentation of contextual cues associated
with target training reduces interference. These contextual
cues can be either diffuse background cues or discrete stimuli
that were presented with the target (Escobar et al., 2001).
Additionally, more recent training experience typically dom-
inates behavior (i.e., a recency effect), all other factors being
equal. Such recency effects fade with increasing retention
intervals, with the consequence that retroactive interfer-
ence fades and, correspondingly, proactive interference in-
creases when the posttraining retention interval is increased
(Postman, Stark, & Fraser, 1968).
Notably, the contextual and temporal modulation of inter-
ference effects is highly similar to the modulation observed
with degraded contingency effects (see this chapter’s section
entitled “Factors Influencing Aquired Stimulus Control of
Behavior”). This similarity is grounds for revisiting the issue
of whether interference effects are really different from de-
graded contingency effects. We previously cited grounds for
rejecting the view that interference effects were no more than
degraded contingency effects (see this chapter’s section on
that topic). However, if the training context is regarded as
an element that can become associated with a cue on a cue-
alone trial or with an outcome on an outcome-alone trial,
contingency degrading trials could be viewed as target cue-
context or context-outcome trials that interfere with behavior
promoted by target cue-outcome trials much as Y-outcome or
target-B trials do within the interference paradigm. In princi-
ple, this allows degraded contingency effects to be viewed as
a subset of interference effects. However, due to the vague-
ness of contextas a stimulus, this approach has not received
widespread acceptance.MediationMediated changes in control of behavior by a stimulus refers
to situations in which responding to a target cue is at least
partially a function of the training history of a second cue
that has at one time or another been paired with the target.
Depending on the specific situation, mediational interaction