Le Pelley 2004 PDF

This article was downloaded by: [130.132.173.
182]
On: 06 November 2014, At: 11:18
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK
The Quarterly Journal of Experimental Psychology

Section B: Comparative and Physiological Psychology
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/pqjb20
The role of associative history in models of

associative learning: A selective review and a hybrid
model
M. E. Le Pelley
Cardiff University , Cardiff, UK

Published online: 13 May 2010.
To cite this article: M. E. Le Pelley (2004) The role of associative history in models of associative learning: A selective
review and a hybrid model , The Quarterly Journal of Experimental Psychology Section B: Comparative and Physiological
Psychology, 57:3, 193-243
To link to this article: http://dx.doi.org/10.1080/02724990344000141
PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Q232738QJEP(B)03b14 / Jun 1, 04 (Tue)/ [51 pages 2 Tables 8 Figures 6 Footnotes 0 Appendices] .

Centre single caption cf. [no comma] Shortcut keys UK Spelling
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2004, 57B (3), 193243
The role of associative history in models of

associative learning: A selective review and a
hybrid model
M. E. Le Pelley
Downloaded by [] at 11:18 06 November 2014
Cardiff University, Cardiff, UK

Associative learning theories strive to capture the processes underlying and driving the change in
strength of the associations between representations of stimuli that develop as a result of
experience of the predictive relationships between those stimuli. Historically, formal models of
associative learning have focused on two potential factors underlying associative change, namely
processing of the conditioned stimulus (in terms of changes in associability) and processing of the
unconditioned stimulus (in terms of changes in error). This review constitutes an analysis of the
proper role of these two factors, specifically with regard to the way in which they are influenced by
associative history (the prior training undergone by cues). A novel hybrid model of associative
learning is proposed and is shown to provide a more satisfactory account of the effects of
associative history on subsequent learning than any previous single-process theory.
It has long been a goal of experimental psychologists to discover how animals and humans are
able to learn about relationships between stimuli and events in the world around them. For it is
this ability to learn about predictive relationships that enables organisms to adapt and survive
in a changing environment. Over a century has passed since Thorndike (1898) proposed the
first theory of associative learning, and debate over the proper way to characterize the learning
ability of animals shows little sign of abating. In recent years this debate has tended to be
directed towards two outstanding issues relevant to any model of associative learning: (1) the
way in which stimuli are represented by the learning system (either as sets of independent
elements or as more holistic configurations); and (2) the mechanics of the learning process
itselfthat is to say, the factors that determine the amount of associative change (i.e.,
learning) that a given cue will undergo on a learning episode. The former topic has been the
subject of several recent reviews (Pearce, 1994; Pearce & Bouton, 2001; Wagner, 2003;
Wagner & Brandon, 2001). This paper, in contrast, concentrates almost exclusively on the
second issuethat is, on the factors that influence the extent to which a representation of a
Correspondence should be addressed to Mike Le Pelley, School of Psychology, Cardiff University, Wales, CF10
3YG, UK. Email: lepelleyme@cardiff.ac.uk
I would like to thank John M. Pearce, I. P. L. McLaren, and S. M. Oakeshott for their helpful advice during the
preparation of this article. This work was supported by a grant from the ESRC.
2004 The Experimental Psychology Society
http://www.tandf.co.uk/journals/pp/02724995.html
DOI:10.1080/02724990344000141
194
LE PELLEY
given cue (or configuration of cues) will engage the learning process. More specifically, this
review seeks to address the question of how best to characterize the effect of associative history
(the prior training that a stimulus has received) on learning by examining its influence in
various representative formal models. We begin by looking at the role of associative history in
models of learning based on processing of the unconditioned stimulus (US), before providing
empirical evidence to suggest that such theories do not go far enough in their characterization
of how the previous training undergone by a conditioned stimulus (CS) affects subsequent
learning about that stimulus. Instead, it seems that the processing received by a stimulus can
vary as a function of the associative history of that stimulus: This leads on to discussion of two
of the most influential CS-processing models of learning, the Mackintosh (1975) theory and
Pearce and Halls (1980) model.
The view of associative history taken by CS-processing models differs profoundly from
that taken by US-processing models. Moreover, the approaches taken by different CSprocessing models differ from one another: The Mackintosh and PearceHall models, for
instance, take (in some sense) opposing views of the way in which processing of a stimulus
changes as a result of experience of predictive relationships involving that stimulus. Given the
existence of evidence supporting each of these conflicting views of the role of associative
history, it rapidly becomes clear that none of these approaches alone is sufficient to account for
the range of effects observed empirically. As a possible solution to this problem, I present a
hybrid model of associative learning that attempts to reconcile the various empirical
demonstrations of the effects of associative history by borrowing from all of the previously
mentioned approaches. This model is shown to have more explanatory power than any of its
predecessors, providing a more satisfactory account of an array of experimental findings than
any existing single process model.
Before going any further, it would seem sensible to set out the limitations of this article. A
complete review of the effects of previous experience of stimuli on the subsequent learning
that they undergo, and the models that have been used to describe these effects, would take far
more space than is available here. As a consequence the current analysis must be somewhat
selective in its focus. For example, there will be only brief discussion of the effects of
nonreinforced preexposure to stimuli on the subsequent learning that they undergo, either in
terms of subsequent conditioning (which makes contact with the phenomenon of latent inhibition) or in terms of discrimination learning (perceptual learning). For an in-depth discussion of empirical studies of preexposure effects in relation to various models of associative
learning, the interested reader is referred to Hall (1991). On the theoretical side, the discussion
here is confined to acquisition-based models of associative learning, and more specifically
models that view effects of associative history as acquisition basedthat is, based on a property that changes incrementally as a result of experience with stimuli and their predictive
relationships. This precludes discussion of models of associative interference and the characterization of associative history effects that they offer (e.g., Bouton, 1993, 1994), or models of
response competition (e.g., Miller & Matzel, 1988). Despite this necessarily narrow focus, this
review aims to introduce a number of general principles of associative history and its effects on
learning, and to relate these principles to various formal models as a first step towards understanding how learning in the past affects learning in the future.
ASSOCIATIVE HISTORY
195
US-PROCESSING MODELS
Separable error term models
Among the earliest formal models of associative learning are those based on the standard linear
operator (Bush & Mosteller, 1951; Estes, 1950; Kendler, 1971), in which the increment in
associative strength undergone by a cue is a linearly decreasing function of the strength of that
cue. For example, Bush and Mosteller proposed that the associative change undergone by a
CS, A, on a given learning episode was defined by:
VA = A( VA)
where VA is the associative strength of Cue A; A represents the associability of A and is a

function of the cues intensity or salience; is a learning-rate parameter reflecting the intensity of the US occurring on that trial; and is the asymptote of conditioning supportable by
that US.1
According to this early view, then, modulation of associative change is determined solely by
changes in processing of the US. The error term, ( VA) represents the extent to which the
US occurring on a given trial is predicted by Stimulus A. If the US is not well predicted by a
CS then it is able to support more learning with respect to that CS than if it is already well
predicted. The contribution of the CS to learning is fixed, taking a value determined by its
intensity. The Bush and Mosteller model thus assumes that the associative change occurring
on a given trial depends only on the current associative strength of a stimulus and the asymptote of conditioning supportable by the US presented on that trial, not on how the current
associative strength was reached (the associative path)this is known as the assumption of
path independence.
Moreover, Bush and Mostellers (1951) theory assumes that stimuli are treated wholly
independently of one another in the determination of associative change. If two cues, A and B,
are presented on the same learning episode, the error term for A will be ( VA) while that for
B will be ( VB): This model employs a separate error term for each presented stimulus. As
such, the associative change undergone by each cue will be independent of the current associative strength of the other.
This assumption of cue independence has been challenged by a number of studies demonstrating that cues presented in compound can, and will, interact and compete for associative
strength. This is most powerfully demonstrated in the phenomenon of blocking (Kamin,
1969), which refers to the finding that the gain in excitatory strength of a cue, B, following
reinforcement of an AB compound is much reduced if Cue A has previously been trained as
being a good predictor of that US. Learning does not simply progress with each cue independently. Instead the two cues seem to compete for a limited amount of associative strength.
1
In fact Bush and Mostellers model described the change in probability with which a conditioned response would
be emitted in the presence of a given CS, rather than the change in associative strength of the CS. As long as it is
assumed that probability of conditioned response is monotonically related to associative strength, however, Bush and
Mostellers model can reasonably be rephrased as Equation 1 (this is also the formulation offered by Kendler, 1971).
196
LE PELLEY
The RescorlaWagner (1972) summed error term model

This idea of competition for associative strength is encapsulated by the model of learning
proposed by Rescorla and Wagner (1972) which states that:
VA = A( V)
where V is the summed associative strength of all currently presented cues. Hence the
RescorlaWagner model states that the error governing associative change for any cue on a
trial is based on the combined associative strength of all cues present on that trial, allowing it to
account for a number of observations of interaction between cues in the acquisition of associative strength. For example, the use of this summed error term is essential to the models
explanation of blocking. According to Equation 2, when the AB compound is followed by reinforcement, the associative change undergone by B will be determined by the discrepancy
between and the combined associative strengths of A and B. As a result of pretraining with A
(such that VA ), though, this discrepancy will be near zero, and hence B will gain little associative strength.
A second strength of the formulation offered by Rescorla and Wagner (1972) is its ability
to account for the phenomenon of conditioned inhibition (Hearst, 1972; Pearce, Nicholas, &
Dickinson, 1982; Rescorla, 1969a, 1969b). In the first stage of the experiment by Pearce et al.
(1982) rats experienced trials on which a clicker signalled electric shock in a conditioned
suppression procedure. In the second stage, trials on which the clicker was followed by
shock were intermixed with trials on which a compound of the clicker and a light was not
followed by shock. Thus the light signalled the absence of a shock that was otherwise
predicted by the presence of the clicker. Following this training, presentation of the clicker
suppressed lever pressing for food, indicating that the rats had learnt the clicker shock
association. The compound of the clicker and the light, on the other hand, produced considerably less suppression of lever pressing, indicating that the rats had learnt that the light
signalled the absence of the shock, and that this learning opposed the excitatory learning about
the clicker. In support of the idea that the light had acquired inhibitory potential as a result of
this training, Pearce et al. found that when the light was subsequently paired with shock,
conditioning proceeded more slowly than for a cue that had not previously received this inhibitory training (the retardation test for conditioned inhibition). Furthermore, they demonstrated that a light given conditioned inhibition training would also counteract the
conditioned suppression produced by a CS that was trained separatelythat is, it was able to
transfer its inhibitory potential to a novel excitor (the summation test for conditioned
inhibition).
Conditioned inhibition provides another demonstration of the interaction of cues in the
learning process. In standard inhibitory conditioning procedures, a cue, A, will only become a
conditioned inhibitor as a result of nonreinforcement in the presence of an excitor, B. If cues
are treated entirely independently in the determination of associative change, there can be no
way for the excitatory potential of B to influence the inhibitory learning undergone by A. The
RescorlaWagner model, on the other hand, by specifying competition between cues for a
limited resource (amount of available learning) by means of a summed error term, is able to
account for the development of conditioned inhibition. The model assumes that will be
zero on nonreinforced AB trials, while VB > 0 as a result of separate excitatory training of B.
ASSOCIATIVE HISTORY
197
Therefore A will acquire negative associative strength on these trials. As a result, excitatory
conditioning with A will occur slowly if it is subsequently paired with the US, and the
presence of A will be able to suppress responding to any other excitatory CSA will pass the
retardation and summation tests as described above.
The RescorlaWagner theory shares with the earlier Bush and Mosteller model the idea
that learning is wholly governed by changes in the effectiveness of the US, with a surprising
outcome supporting more learning than an outcome that is predicted, while the contribution
of the CS to learning is fixedas such it too is a USprocessing model of learning. And like
Bush and Mosteller (1951), Rescorla and Wagner (1972) also make the assumption of path
independence, wherein the associative change undergone by a cue is independent of the associative history of that cue.
In fact, the RescorlaWagner model takes an even more extreme view of the assumption
of path independence than its forerunner: The use of a summed error term extends the idea
of path independence to compounds of cues. In other words, the associative change undergone by the elements of a compound depends only on the current associative strength of that
compound and the current outcome, not on the individual strengths of the elements of
that compound, and not on how those associative strengths were reached. The validity
of this assumption was tested recently by Rescorla (2000), who investigated the distribution
of associative change between the elements of a compound made up of an excitor and an
inhibitor occurring as a result of reinforcement of that compound. His Experiment 1a
employed a magazine approach paradigm, in which rats learnt that certain stimuli (tones,
clickers, lights, etc.) predicted the delivery of food. In the first stage of this experiment, rats
experienced trial types A+, C+, X+, BX, DX. Thus Cues A and C were initially trained
as equivalent excitors (i.e., they predicted the delivery of food), whereas B and D were
trained as equivalent inhibitors (i.e., they predicted the absence of the US that would otherwise have been expected on the basis of the presence of X). In Stage 2 a compound of an
excitor (A) and an inhibitor (B) was reinforced (AB+). The question of interest was whether
this reinforcement would lead to equal-sized increments in the associative strengths of Cues
A and B, or whether one cue would undergo a greater increment than the other. Rescorla
addressed this question by looking at responding to compounds AD and BC following Stage
2 training. If these compounds were compared in the absence of AB+ training, they should
yield equal responding: Each is comprised of one excitor (A or C) and one inhibitor (B or D).
If AB+ trials led to equal-sized increments in the strengths of A and B, then responding to
the AD and BC compounds would remain equal (as each starts Stage 2 at the same level and
undergoes the same change). If, however, AB+ trials produced a greater associative increment in the excitatory A than the inhibitory B, then the test should reveal greater responding
to AD than to BC. Conversely, a greater increment in VB than VA would be evidenced by
greater responding to BC than to AD.
The use of the summed associative strength of all presented cues in the error term of the
RescorlaWagner model means that all cues presented on a given trial must have an identical
error term. Applying Equation 2 to Cues A and B on the AB+ trials in Stage 2 of Rescorlas
(2000) experiment gives:
VA = A( (VA + VB))
and
198
LE PELLEY
VB = B( (VA + VB))
It is easy to see that if the salience () of A and B is equal (ensured empirically by appropriate
counterbalancing of stimuli), Equations 3 and 4 will be identical. In other words, the Rescorla
Wagner model is constrained to predict that A and B will undergo identical increments in
associative strength as a result of AB+ trials, despite the fact that A (an excitor) and B (an
inhibitor) begin these trials with very different associative strengths (VA > 0, VB < 0). As stated
above, the use of a summed error term means that the individual associative strengths of the
elements are unimportant in the determination of associative changeall that matters is the
overall strength of the cue compound presented on each Stage 2 trial.
Contrary to this fundamental prediction of the RescorlaWagner model, Rescorla (2000)
found significantly greater responding to BC than to AD on test, indicating that the inhibitory
B had undergone a greater increment in associative strength than the excitatory A over the
AB+ trials. This finding provides strong evidence against the idea that the assumption of path
independence applies to compounds of cuesthe distribution of associative change among
the elements of a reinforced compound is not independent of the associative status of those
separable elements. The results of Rescorlas study instead fit better with the predictions of
Bush and Mostellers (1951) earlier model, which employs separable error terms for the
different elements present on a given trial and hence predicts that the stimulus whose associative strength is more discrepant from that supportable by the outcome of the trial will undergo
the greater associative change. Excitatory training of A in the first stage will ensure that it
becomes a good predictor of the US. As such it will have a small error term when it is paired
with the US on Stage 2 AB+ trials and hence will undergo little associative change over these
trialsits associative strength will already be near asymptote. B, on the other hand, is trained
to predict the absence of the US during Stage 1. Hence the occurrence of the US following
presentation of B on AB+ trials will be very surprising (i.e., B will have a large error term), and
as a result B will undergo a large increment in associative strength.
A model combining separable and summed error terms

So the results of Rescorlas (2000) experiment provide support for a model incorporating separable error terms in its mechanism governing associative change (see also Rescorla, 2001). But
we saw earlier that treating cues independently of one another in the determination of associative change leads to problems in accounting for various phenomena of cue competition such as
blocking, and it was for this reason that the RescorlaWagner model incorporated a summed
error term. It is possible to resolve this apparent conflict within a US-processing model of
learning by effectively combining the Bush and Mosteller and RescorlaWagner theories, to
yield a model of learning in which the effectiveness of a summed error term in governing the
associative change undergone by a cue is modulated by the separable error for that cue alone. It
is relatively straightforward to formulate such a model (cf. Brandon, Vogel, & Wagner, 2003).
One of the simplest approaches is outlined here.
The associative change undergone by cue A on a trial on which ( V ) > 0 (i.e., a trial that
will support excitatory learning) is given by:
VA = A ( V) | VA|
ASSOCIATIVE HISTORY
199
and the associative change on a trial on which ( V) < 0 (i.e., a trial that will support inhibitory learning) is given by:
VA = A ( V) |V + + VA|
where V is the summed associative strength of all excitatory cues (V > 0) present.
The separable error term modulators in these equations ensure that the associative
change undergone by a cue is influenced by the discrepancy between its own associative
strength and the current outcome. The | VA| term modulating excitatory learning in
Equation 5 represents the discrepancy between the magnitude of the US occurring on this
trial and the current associative strength of Cue A. The better Cue A predicts the current US,
the smaller this modulator and hence the smaller the increment in associative strength undergone by A. On inhibitory trials, inhibitory learning is driven by the presence of cues predicting
the occurrence of the US while no US actually occurs. That is, the driving force behind inhibitory learning is the prediction of the US made by all presented excitatory cues, V +. The
|V + + VA| term modulating inhibitory learning in Equation 6 represents the discrepancy
between this inhibitory potential and the current associative strength of Cue A. In this
model inhibitory cues have a negative associative strength. As such, the better Cue A correctly
predicts the absence of the US in the presence of excitatory cues, the smaller this modulator
and hence the smaller the decrement in As associative strength.
These separable error term modulators ensure that for a compound cue the element that is
the poorer predictor of the outcome will undergo the greater associative change. As a result
this model is well equipped to account for the results of Rescorlas (2000) investigation. This
was confirmed by computational simulation of this study. Parameters used for this simulation
were: = .5, = .3, (US present) = .8, (US absent) = 0. It should be noted, however, that
the predictions made by the model with regard to this study are parameter independent.
Following the parameters of Rescorlas experiment, Stage 1 training proceeded for 100 blocks,
followed by eight AB+ trials in Stage 2. The results of the simulation are shown in Figure 1
(data for Cues C and D of Rescorlas experiment are omitted from this figure for clarity). As
Figure 1. Simulation of Rescorlas (2000) Experiment la with a US-processing model that combines separable and
summed error terms. Data for Cues C and D are omitted for clarity. Left panel: Mean associative strengths for Stage 1.
Right panel: Mean associative strengths for Stage 2. During Stage 2, B undergoes a greater increment in associative
strength than does A.
200
LE PELLEY
expected, over the course of Stage 1 the associative strength of A increases, and that of B
becomes increasingly negative to counter the excitation caused by the presence of X on BX
trials. And, as expected, on Stage 2 AB+ trials there is a greater increase in the associative
strength of B than in that of A, a result of the fact that | VB| > | VA| on these trials.
Following AB+ training, the associative strength for compound AD (given by VA + VD) stands
at .18, while that for BC is .64. This is, of course, the pattern indicated by the empirical results
of Rescorlas experiment.
Further support for this model integrating a summed error term with separable error term
modulators comes from the results of Rescorlas (2000) Experiment 2. This study employed
exactly the same Stage 1 training regime as that for the experiment outlined above (A+, C+,
X+, BX, DX), but the subsequent AB compound trials were nonreinforced. A and B might
be expected to undergo a decrement in associative strength as a result of nonreinforcement;
the question of interest is which undergoes the greater decrement. On test, conditioned
responding to compound BC was greater than that to AD, indicating that the excitatory A
underwent a greater decrement in strength than the inhibitory B. Once again this rules out the
RescorlaWagner models assumption that path independence applies to compounds, and
again it fits better with a model employing separable error terms. During the first stage, A is
consistently paired with the US in isolation and hence will rapidly gain excitatory strength.
The acquisition of inhibitory strength by B, on the other hand, is driven by its pairings with
the excitatory X. Given that X is paired with the reinforcer on only one third of its presentations, it will be much slower than A to develop excitatory strength. As a consequence, inhibitory learning about B (driven by the excitatory strength of X) will be relatively slow, and
certainly slower than development of excitatory strength by A (as shown in Panel A of Figure
1). On AB trials, the US is not presented, and hence = 0. Given that VB is relatively small
and negative, and VA is larger and positive, the separable error term modulator for B on AB
trials, |0 VB|, will be smaller than that for A, |0 VA|, with the result that A will undergo
the greater decrement in associative strength on these trials. This in turn will lead to a greater
associative strength for BC than for AD, as indicated by the empirical data. Computational
simulation of this experiment confirms this prediction: Using exactly the same parameters as
employed above yields an associative strength for compound AD of .05, and a strength for BC
of .10.
So this combined US-processing model is able to account for the results of Rescorlas
(2000) study as a result of its use of separable error term modulators. The presence of a
summed error term also allows the model to explain cue competition effects such as blocking
in much the same way as does the RescorlaWagner modelby attaching an importance to the
associative strength of the compound in the determination of associative change. Computational simulation using the parameters above confirms this. Thus 20 A+ trials, followed by 8
AB+ trials intermixed with 8 CD+ trials, result in VB = .12 and VD = .32. Thus prior conditioning of one element of a stimulus compound blocks learning about the other element when
that compound is reinforced, as compared to a control compound made up of novel elements.
By combining separable and summed error terms in a single algorithm, we are able to
resolve the potential conflict between (1) Rescorlas (2000) finding that the distribution of
associative change between the elements of a reinforced compound depends on the training
history of the separable elements making up that compound, and (2) demonstrations of cue
competition effects such as blocking (indicating that it is insufficient to view separable
ASSOCIATIVE HISTORY
201
elements independently in the determination of associative change). The resulting model, like
its antecedents (Bush and Mostellers, 1951, standard linear operator and Rescorla and
Wagners, 1972, summed error term theory), specifies associative change to be governed
solely by changes in processing of the US.
CS-PROCESSING MODELS
The Mackintosh (1975) model

There is, however, another possibility, and this is to ascribe cue competition effects to competition between CSs for processing power, rather than to variations in processing of the US.
The amount of processing power secured by a given CS is reflected in its associability, . In
the Bush and Mosteller and RescorlaWagner models outlined above, the associability of a cue
is simply a fixed parameter depending on its intensity or salience. In a CS-processing model of
learning, on the other hand, associability is a variable, able to change as a result of experience
with a cue and with that cues predictive abilities.
One of the most influential models incorporating this notion of variable associability is the
theory of selective attention proposed by Mackintosh (1975). This states that:
VA = SA( VA)
where S is a learning rate parameter. This equation is clearly very similar to that for the Bush
and Mosteller model outlined above. The difference is that Mackintosh allows the associability of a cue, A, to change as a result of experience of a cues predictiveness, with animals
proposed to devote more processing power to stimuli that are uniquely successful in their
predictions. Specifically, Cue A maintains a high to the extent that it is a better predictor of
the outcome of the current trial than are all other cues present. Conversely, will decrease if
the outcome is predicted by other events at least as well as by A. The extent to which the
outcome is predicted by A is represented by the absolute value of the error term ( VA).
Hence on each learning episode n, following the adjustment of weights according to Equation
7, the associability for each presented cue is updated according to the following rules:
nA > 0 if | n VAn 1|<| n VXn 1|
nA > 0 if | n VAn 1|| n VXn 1|
where nA is the change in associability of Cue A on trial n, n is the magnitude of the

outcome occurring on trial n,VAn 1 is the associative strength of Cue A on trial n 1 (i.e., the
associative strength of A before it was updated on trial n), andVXn 1 is the associative strength of
all stimuli other than A present on trial n before associative strengths were updated on that
trial. In other words, the change in associability is determined by how good a cue was at
predicting the outcome of the current trial at the outset of that trial, before the associative
strengths were adjusted. Mackintosh suggested that the size of the change in A on each trial
should be proportional to the magnitude of these two inequalities, but gave no specific
algorithm for computing this change.
According to Mackintosh (1975), then, there are two factors governing the change in associative strength of a cue. First there is the error term, given by the discrepancy between and
the individual strength of that cue and hence driven by changes in processing of the US. The
202
LE PELLEY
influence of this error term, as we saw with the Bush and Mosteller model, is modulated by the
current associative strength of the cue, but not by how that associative strength was reached
the error term is path independent. Second there is , reflecting changes in CS processing and
determined by the past relative predictive power of the cue. As such the influence of is
entirely determined by the cues associative history. Allowing associability to vary in this
manner ensures that associative change is not solely determined by a cues associative
strengthit is also affected by how that associative strength was achieved. The Mackintosh
model does not make the assumption of path independence.
Furthermore, it is this term that allows for the interaction of cues in the determination of
associative change: Mackintoshs (1975) model specifies competition between cues in terms of
associability, rather than error. Consider again the blocking paradigm. Pretraining of A will
establish it as a good predictor of the outcome. The novel cue, B, presented on subsequent
AB+ trials will therefore be a poorer predictor of the outcome than will A, and so its associability will fall over the course of these trials (according to Equation 8), such that changes in VB
on every trial but the first (when its associability has not yet had a chance to decline) will be
smaller than those in a control group that has not had pretraining with A to establish it as a
good predictor of the outcome.
By specifying competition between cues in terms of associability, rather than error, Mackintosh (1975) removed the immediate need for a summed error term in his calculation of V
(required implicitly by any US-processing model that purports to explain cue competition
effects). Instead the model uses a separable error term of the form proposed by Bush and
Mosteller (1951), and as a result it is in principle able to account for the results of Rescorlas
(2000) study of the distribution of associative change amongst the elements of a reinforced
compound. In the first stage of this study A is consistently followed by an event of significance
(reinforcement), and so is B (the absence of a predicted reinforcer). A and B are both good
predictors of their respective outcomes (which for B is actually nonreinforcement), and as a
result both will maintain a high over the trials of Stage 1. On the initial AB+ trial in Stage 2,
then, both A and B will be well processed (as the calculation to update associability is
conducted after each trial). However, B will have a much larger separable error term on this
trial than will A (see the discussion of Bush and Mostellers model earlier). Indeed, if Stage 1
training is sufficient to bring As associative strength near to asymptote (), its error term will
be near zero, and as a result any increase in VA on this trial will be only very slight. Of course,
following the updating of weights on this first trial, the modulation of associability outlined
above will occur. The associability of A (a good predictor of occurrence of the US) will remain
high, while that for B (a poor predictor of the occurrence of the US) will be reduced. So subsequent changes in VB will become increasingly smaller as B falls. However, given that VA was
already near asymptote at the start of Stage 2, it will undergo little further increase over Stage 2
trials. In other words, the effect of Stage 1 training outweighs any influence of attentional
modulation in Stage 2, such that the effective changes in the model are dominated by the Bush
and Mosteller type error term.
The extended Mackintosh model

So the Mackintosh model is, in principle, able to account for the results of Rescorlas (2000)
experiment. In practice, however, a major obstacle remains. How can Mackintoshs model
ASSOCIATIVE HISTORY
203
account for the development of conditioned inhibition by B in the first place? Its use of a separable error term for each cue means that there is no way for the excitatory strength of X to drive
the development of inhibitory strength by B as a result of nonreinforcement of the BX
compound. It is hard to see how any theory of associative learning could provide a satisfactory
account of conditioned inhibition without the use, either implicitly or explicitly, of a summed
error term. Schmajuk and Moore (1985; see also Moore & Stickney, 1985) attempted to
modify the Mackintosh model to allow it to account for conditioned inhibition and various
related phenomena by adding a summed error term to the expression governing associative
change and by allowing for the development of antiassociations, representing the prediction
of nonreinforcement by a CS. This idea stems from Konorskis (1967) suggestion that inhibitory learning reflects development of an association between a representation of the CS and a
no-US centre (US), and that there exists an inhibitory relationship between US and US
centres such that if both are activated simultaneously, activity in the US centre will inhibit
activity in the US centre. It is these CSUS associations that Schmajuk and Moore refer to as
antiassociations. However, their formulation only allows for the development of
antiassociations on trials on which the US is not presented. As such, it is unable to account for
the well-established phenomenon of overexpectation (Khallad & Moore, 1996; Kremer, 1978;
Lattal & Nakajima, 1998; Rescorla, 1970; Wagner, 1971): When two stimuli that have been
separately paired with a US are presented in compound along with the same US, there is a
reduction in the associative strength of each element. This reduction occurs despite the fact
that these compound trials are reinforced, and hence it is outside the scope of Schmajuk and
Moores model. Moreover, Kremer (1978) demonstrated that a novel cue presented on the
reinforced compound trials of an overexpectation study would take on inhibitory properties,
again indicating that reinforced trials can lead to the development of antiassociations.
Furthermore, the model is unable to account for demonstrations of inhibitory conditioning
resulting from a reduction in the magnitude of reinforcement. That is, if A is paired with a
strong outcome, and AB is paired with a weak outcome, B is typically seen to acquire inhibitory properties despite the fact that it is consistently paired with an outcome, albeit a weak one
(Cotton, Goodall, & Mackintosh, 1982; Mackintosh & Cotton, 1985; Wagner, Mazur,
Donegan, & Pfautz, 1980). It is not easy to modify Schmajuk and Moores model to explain
these findings without leading to problems with its account of other phenomena, such as
blocking. As a result a new formulation is proposed here that is loosely based on Schmajuk and
Moores model but borrows from the approach to inhibition offered by Pearce and Hall (1980)
in their own model of associative learning (discussed later in this paper).
In this extended Mackintosh model, the effective strength of the reinforcer, R, on a trial
is given by:
R = (V V )
where V is the summed associative strength of all presented stimuli for the US representation. On the compound trials of a standard inhibitory conditioning procedure, where the US is
not presented, this value will of course reduce to V V .
If R is positive (i.e., the predicted intensity of reinforcement is less than the actual magnitude of reinforcement such that this is a trial that will support excitatory learning), the
strength of the CSUS association is increased according to the equation:
204
LE PELLEY
VA = AE (1 VA +V A) |R|
10
where E is a learning-rate parameter for excitatory learning.

If R is negative (i.e., the predicted intensity of reinforcement is greater than the actual
magnitude of reinforcement such that this is a trial that will support inhibitory learning), the
strength of the CSUS antiassociation is increased by:
V A = AI (1 V A + VA) |R|
11
where I, is a learning-rate parameter for inhibitory learning. In order to account for demonstrations of conditioned responding resulting from lean schedules of partial reinforcement, it
must be assumed that E > I such that increases in associative strength on reinforced trials
outweigh decreases on nonreinforced trials.
The net associative strength of a cue, VNET, which determines the level of conditioned
responding to that cue, is then given by:
VANET = VA VA
12
Note that this new formulation (like that of Schmajuk & Moore, 1985) moves away from the
schema offered by Mackintosh (1975) in that it employs a form of summed error term in the
value of R: Equation 9 involves a comparison of the extent to which the US is predicted by all
currently present cues, and the strength of the US actually presented. This summed error
term determines the type of learning that occurs on a given trial [by dictating whether learning
is excitatory (Equation 10) or inhibitory (Equation 11), and it also modulates the amount of
learning undergone (the |R| factor in Equations 10 and 11)]. In common with Mackintoshs
original formulation, though, this extended model also employs a separable error term in the
equations governing associative change (1 VA +V A in Equation 10 and 1 V A + VA in Equation 11). This ensures that the net associative strength of each individual element of a
compound, and not just the summed strength of that compound, has an effect on the associative change undergone by that element.2
All that remains is to specify the algorithm governing the change in associability of a cue on
a given trial. Equation 8 sets out the general framework suggested by Mackintosh (1975), in
which Cue As associability increases if it is a better predictor of the outcome of the current
trial than are all other cues present and decreases if the outcome of the current trial is predicted
by other stimuli at least as well as by A. There are of course a huge number of different algorithms for manipulating associability that are consistent with these principles, and there is
little direct evidence to support one approach over another. One of the simplest practicable
algorithms that remains close to the principles suggested by Mackintosh, and the one
employed in the present formulation, is outlined below.
Note that the specific separable error terms employed here require that the maximum possible value of for any
US is 1. As such, the simulations presented with this model employ (US present) = .8, reflecting the fact that, while
the USs typically used in these experiments (shock, food etc.) are certainly potent, more potent USs are surely possible. For the sake of consistency, this parameterization is used in all other simulations employed here where standard
USs are employed.
ASSOCIATIVE HISTORY
205
After updating the associative strengths on a given trial according to Equations 10 or 11

above, the associability of each cue is adjusted. When R > 0, the change in associability of Cue
A on trial n is defined by:
nA = E (| n VAn 1 + VAn 1|| n VXn 1 + VXn 1|)
13
and when R < 0 the change in associability is given by:
nA = I (||R n |VAn 1 + VAn 1|||R n |VXn 1 + VXn 1|)
14
The n 1 suffixes are intended to indicate that the values of the associative strengths before
they are updated on trial n are used to determine the change in associability of Cue A on trial n.
VX represents the combined associative strengths of all stimuli other than A presented on trial
n (andV X represents the combined inhibitory strength of all stimuli other than A). E and I
are learning-rate parameters for changes in on excitatory and inhibitory trials respectively.
Consider the A+, AB trials of a standard conditioned inhibition contingency. A will rise on
A+ trials, where A is the best predictor of reinforcement, and will fall on AB trials, where it is
a poorer predictor of the absence of reinforcement than is B. Hence in order to allow A to
maintain a high associability across these trials, we must assume that E > I, such that the
increases in A on A+ trials outweigh the decreases on AB trials.
Finally, we limit to be in the range from .05 to 1. In other words, following the calculation
in Equation 13 or 14:
If A 1 then
If A .05 then
A = 1
A = .05
15
A lower limit of .05, rather than zero, is used in order to prevent stimuli from ever becoming
completely frozen out of the learning process.
This modified Mackintosh model behaves very much like the original with regard to
contingencies that would not allow for the development of antiassociations (e.g., simple acquisition, overshadowing, blocking). This was confirmed by computational simulation. Figure 2
shows the results of a simulation of a blocking experiment (20 A+ trials, followed by 8 AB+
trials intermixed with 8 CD+ trials) using the model outlined above. Parameters used for this
simulation, and all other simulations with this model presented in this paper, were: E = .3, I
= .1, E = .3, I = .1, starting value of = .8, (US present) = .8, (US absent) = 0. Note,
however, that within the constraint that E > I and E > I, the predictions made by the model
are parameter independent. Panel A of Figure 2 shows the changes in associative strength for
Cues A, B and C/D (these latter cues are equivalent and hence have been averaged). Note that
this graph shows the net associative strength of these cuesthat is, VNET for Cue A is given by
VA V A. As outlined above, prior conditioning of A establishes it as a better predictor of the
US than is B on compound trials, causing a rapid decline in Bs associability across Stage 2
AB+ trials and hence reducing its ability to engage the learning process. This can be compared
with Cues C and D, which are novel on Stage 2 CD+ trials. Neither cue is a poorer predictor of
the US than the other, and so both maintain a high associability. The result is a greater
excitatory associative strength for C/D than for B.
Crucially, this extended Mackintosh model is now able to provide a satisfactory account of
the results of Rescorlas (2000) study of associative change in excitors and inhibitors conditioned in compound, as it is equipped to explain the development of conditioned inhibition.
206
LE PELLEY
Figure 2. Simulation of a blocking experiment using the extended Mackintosh model. Upper panel: Net associative strength (VNET = V V ) for Cues A, B, and C/D (Cues C and D undergo identical training and hence are equivalent to one anotherconsequently the results for these two cues have been averaged). During Stage 2, B gains less
excitatory strength than the control cues, C/D, thus demonstrating blocking. Lower panel: Associability changes for
A, B, and C/D. B falls during Stage 2 as it is a poorer predictor of the US than is A.
Figure 3 shows results of a simulation of this experiment: The data for Cues C and D of
Rescorlas experiment are omitted from this figure for clarity. During Stage 1 (left panels), X+
trials result in this cue taking on net excitatory strength, and this drives the development of net
inhibitory strength by Cue B. As A is the best predictor of the US on A+ trials, it maintains a
high associability. Likewise, on X+ trials the associability of X will increase. However, on BX
and DX trials, X is a poorer predictor of the outcome (nonreinforcement) than is B or D, and
hence on these trials its associability will fall (while that for B and D will rise, such that B and
D remain high throughout Stage 1). Given the parameterization outlined above (with E >
I), once X starts to become reasonably excitatory and B/D reasonably inhibitory, the
increases in X tend to outweigh the decreases, and X comes to maintain a high associability.
On the initial AB+ trials of Stage 2 (right panels), the greater separable error term for B (1 VB
+V B in Equation 10) results in a greater increment in the net associative strength of this cue
than for A. Given that B is a poorer predictor of reinforcement than is A on AB+ trials,
however, its associability undergoes a rapid decline, greatly reducing subsequent increments
in its excitatory strength. The overall result, though, is a greater associative change for B than
for A: Net associative strength for the AD compound is .26, while that for BC is .58.
ASSOCIATIVE HISTORY
207
Figure 3. Simulation of Rescorlas (2000) Experiment la with the extended Mackintosh model. Left panels show
data for Stage 1; right panels for Stage 2. Upper panels show net associative strength; lower panels show associability.
During Stage 2, B undergoes a greater increment in net associative strength than does A.
This extended Mackintosh model is able to account for the results of Rescorlas (2000)
Experiment 2, in which the AB compound was nonreinforced in Stage 2, in similar fashion.
On these Stage 2 AB trials, A, fuelled by its greater separable error term, undergoes
greater associative change than does B (while A declines as it is a poorer predictor of
nonreinforcement than is B). Simulation confirms this pattern of resultsfollowing Stage 2,
compound AD yields net associative strength of .06, while BC yields .14.
Unlike Schmajuk and Moores (1985) model, this implementation is also able to account
for experiments demonstrating reductions in net associative strength of cues on reinforced
trials. For instance, it provides a good account of overexpectation. Panel A of Figure 4 shows
simulation results for an overexpectation preparation, in which 100 blocks of intermixed A+
and B+ trials are followed by 8 AB+ trials.3 On compound trials, the combined associative
strength of A and B is greater than that supported by the outcome of the trial (i.e., V V )
< 0), such that these trials will support inhibitory learning (development of CSUS
antiassociations). Hence compound trials lead to a reduction in net associative strength of A
As a result of space constraints, the data for associability changes are not shown for any of the experiments whose
simulation results are shown in Figure 5. In an overexpectation paradigm, the associability of A and B remains high
throughout as both are good predictors of reinforcement in the first stage, and neither is a better predictor than the
other in the second stage. In Kremers (1978) overexpectation preparation, the associability of C falls on Stage 2
ABC+ trials, as it is a poorer predictor of reinforcement than A or B. In a reduced reinforcement inhibition design,
both A and B will ultimately maintain a high associability, although A will take longer to achieve this as it is only
inconsistently followed by a strong reinforcer. Finally, in a superconditioning design (A+, AB, then AB+) the
associability changes are much as for Cues A and B in Rescorlas (2000) experiment, as shown in Figure 3.
208
LE PELLEY
Figure 4. Further simulations using the extended Mackintosh model. Panel A: Overexpectation (A+ B+ then
AB+). Panel B: Kremers (1978) study of overexpectation (A+ B+ then ABC). Panel C: Conditioned inhibition
resulting from a reduction in reinforcer magnitude on compound trials (A strong outcome, AB intermediate
outcome). Panel D: Supernormal conditioning, following Pearce and Redhead (1995).
and B, despite the fact that these trials are reinforced. Panel B shows data from a simulation of
Kremers (1978) overexpectation experiment, in which 100 intermixed A+ and B+ trials are
followed by 8 ABC+ trials. The negative summed error term on these latter trials causes C to
take on net inhibitory properties, again despite the fact that this cue is consistently paired with
the US, in agreement with the results of Kremers empirical study. In addition, the model is
able to account for demonstrations of inhibitory conditioning resulting from a reduction in
reinforcer magnitude. Panel C shows data from a simulation of a typical study, employing 100
trials on which A is paired with a strong outcome ( = .8) intermixed with 100 trials on which
AB is paired with an intermediate outcome ( = .4). B can be seen to take on net inhibitory
strength as a result.
The use of a summed error term also allows this extended Mackintosh model to account
for the related phenomena of superconditioning and supernormal conditioning. Super-
ASSOCIATIVE HISTORY
209
conditioning refers to the observation that the effectiveness of excitatory conditioning is

enhanced when it takes place in the presence of a conditioned inhibitor. Thus greater conditioned responding is observed following AB+ training if B has previously been trained as an
inhibitor of that US, compared to the situation in which B is neutral (Navarro, Hallam,
Matzel, & Miller, 1989; Pearce & Redhead, 1995; Rescorla, 1971; Wagner, 1971; Williams &
McDevitt, 2002). Supernormal conditioning refers to the observation that, in certain circumstances, reinforcement of a CS in the presence of a conditioned inhibitor can result in that CS
acquiring a greater-than-asymptotic level of associative strength (Pearce & Redhead, 1995).
If A has asymptotic excitatory strength (i.e., VANET = ), and B has net inhibitory strength
(VBNET < 0), the overall strength of the AB compound will be less than . Thus the inhibitory
strength of B ensures that the summed error term (R) on AB+ trials is greater than zero,
allowing excitatory learning on these trials such that the associative strength of A will grow to a
level that exceeds the asymptote set by the US. Once again the models ability to account for
superconditioning and supernormal conditioning was assessed by means of computational
simulation. In the first stage, simulated subjects received trial types A+, AB, C+, DE. This
should endow B with net inhibitory strength, while leaving E neutral. Stage 2 comprised AB+
and CE+ trials. The presence of the inhibitor, B, should enhance As associative strength on
AB+ trials as compared to C, trained in the presence of a neutral cue. Moreover, reinforcement in the presence of B should allow A to develop an associative strength greater than .
Following the parameters of Pearce and Redheads study, Stage 1 training proceeded for 120
blocks, while Stage 2 comprised 18 blocks. Panel D of Figure 4 shows the averaged results for 8
simulated subjects (VDNET remained at 0 throughout). Superconditioning is demonstrated in
that, following Stage 2, A has achieved a higher excitatory strength than C. Supernormal
conditioning is demonstrated in that the net associative strength of A (.90) at this point is
greater than (.8).
Finally, the combination of summed and separable error terms in the extended Mackintosh model allows it to account for the results of a further study of the distribution of
associative change among the elements of a compound by Rescorla (2002). In the first stage
of this study, A was paired with a strong reinforcer, while AB was nonreinforced, such that
B became a conditioned inhibitor. Following this training, the AB compound was paired
with a reinforcer of intermediate strength. Rescorla noted that this treatment caused both
A and B to undergo increases in excitatory associative strength, but that the increase for B
was greater than that for A. Figure 5 shows simulation results for this study using the
extended Mackintosh model. This simulation employed 64 blocks in each stage, as for
Rescorlas empirical study; parameters were as described above, with = .8 for the strong
reinforcer and = .4 for the intermediate reinforcer. During Stage l, A takes on excitatory
strength and B inhibitory strength. As a result, the overall associative strength of the AB
compound following Stage 1 training tends to an asymptote of 0 with increased training, as
B counteracts the excitatory influence of A. Notably the associative strength of the AB
compound will rapidly fall below the asymptote of strength supported by the intermediate
reinforcer (.4). Hence on Stage 2 trials the summed error term (.4 VANET VBNET ) is
positive, promoting excitatory learning about A and B. However, the separable error term
is greater for B (the poorer predictor of reinforcement), and hence this cue undergoes
greater associative change on the initial Stage 2 trials. Given that B is a poorer predictor
of reinforcement than A, its associability declines rapidly, leading to a reduction in
210
LE PELLEY
Figure 5. Simulation of Rescorlas (2002) study with the extended Mackintosh model. Upper panel shows net associative strength, lower panel shows associability. During Stage 2, both A and B undergo an increase in net associative
strength, but the change for B is greater than that for A.
subsequent increments in VBNET . Nevertheless, the impact of the early trials is such that,
overall, a greater associative increment is seen for B than for A. So by combining separable
and summed error terms, this model is again able to provide a better characterization of the
effects of associative history on the associative change undergone by the elements of a
compound than a model employing only a summed error term (which would incorrectly
predict equal change for A and B) or only a separable error term (which could not explain
the development of conditioned inhibition in the first place).
So far, then, we have two options for reconciling the independence of the elements of a
compound in the determination of associative change, as demonstrated by Rescorla (2000),
with the interaction of cues demonstrated by blocking, conditioned inhibition, and related
phenomena of cue competition. One is to combine a separable error term with a summed error
term in a US-processing model of learning. The other is to allow for variations in the
processing of the CS as an explanation of cue competition effects, as suggested by Mackintosh
(1975). But we saw above that if the approach offered by Mackintosh is to provide a satisfactory account of conditioned inhibition, a summed error term must be built into the model, if
not as explicitly as for the RescorlaWagner formulation. Given that the combination of a
summed and separable error term (as employed by the extended Mackintosh model) is able to
resolve the conflict presented by the results discussed so far, is there any reason to specify a
further parameter, , allowing for competition between CSs for processing power?
EVIDENCE FOR ASSOCIABILITY PROCESSES

Studies of discrimination learning
One common approach to investigating the effects of previous learning on the associability of
stimuli has been to examine the extent to which the learning of one discrimination transfers to
ASSOCIATIVE HISTORY
211
a subsequent novel discrimination in which the same, or similar, stimuli remain relevant. The
problem with this approach lies in ensuring that any positive transfer (better performance)
observed cannot be accounted for solely in terms of direct transfer of learning about the stimuli
involved in the original discrimination, rather than an increase in the associability of those
stimuli (i.e., their readiness to engage the learning process in the novel discrimination).
Among the earliest studies to address this problem successfully were those looking at the
rate of acquisition of reversals of previously learnt discriminations. Consider the situation in
which rats are trained on a blackwhite discrimination in which black is the reinforced stimulus (S+), and white is nonreinforced (S). This should result in black acquiring excitatory
associative strength and white acquiring inhibitory strength. The discrimination is then
reversed, such that white becomes the S+ and black the S. Any transfer of associative
strength from the original discrimination to this reversed version, it is argued, should lead to
negative transfer (poorer performance), as the stimuli predict inappropriate responses immediately following reversal. Reid (1953) conducted just such an experiment. Rats were trained
on a blackwhite discrimination until they reached a criterion of 9 correct responses in 10
trials. One group was then immediately transferred to the reversed version of this contingency, while the other received a further 150 overtraining trials on the original discrimination before reversal. Surprisingly, Reid found that rats overtrained on the original
discrimination were faster to learn its reversal than those rats trained only to criterion. This
facilitated learning clearly could not be a result of direct transfer of associative strength from
the original discrimination: On this basis, extended training on the original discrimination
would be expected to result in increased negative transfer, rather than the positive transfer that
was observed. A detailed empirical analysis of this overtraining reversal effect (ORE) by
Mackintosh (1969) established it as a reliable phenomenon of discrimination learning, whilst
also indicating that the circumstances under which it would be observed were rather specific.
Mackintosh (1969; see also Sutherland & Mackintosh, 1971) noted that the ORE could be
accounted for by a theory allowing for learned changes in the associability of stimuli. During
the first stage, black and white are established as good predictors of reinforcement and
nonreinforcement, respectively, and hence their associability, or the attention paid to them
(Mackintosh uses the terms associability and attention interchangeably), might be
expected to increase. Any increase in associability of these stimuli would be expected to facilitate reversal, as the stimuli will develop associations to the new response assignments more
rapidly. As long as the positive effects of associability changes outweigh the negative effects of
direct transfer of associative strengths from the original discrimination to the reversal, net
positive transfer to the reversed discrimination will be observed. Overtraining might be one
factor influencing this balancewe need only assume that overtraining has a greater effect on
strengthening associability than it does on strengthening specific stimulusresponse associations. Hence the extra negative transfer produced by the slightly stronger stimulusresponse
associations in the overtrained group will be outweighed by the much higher associability of
S+ and S in this group. Moreover, Sutherland and Mackintosh suggested that might be
sticky, such that high values of can be reduced only slowly. This suggestion further aids
explanation of the ORE. Immediately following reversal, the former S+ and S are now poor
predictors of their respective outcomes and hence might be expected to lose associability until
the new stimulusresponse mappings have been learnt sufficiently. However, if the higher
values of reached by the overtrained group are sticky, the reduction in associability on
212
LE PELLEY
reversal will be reduced for this group, thus facilitating more rapid learning of the reversal.
Suret and McLaren (2003) present another model based in part on Mackintosh (1975), which
incorporates this notion of sticky alpha in order to account for their results concerning
discrimination reversal learning following overtraining in humans.
Experiments on spatial learning can also be seen as studies of discrimination learning:
Animals must learn that a certain location (defined by a particular configuration of landmarks)
is rewarded, while other locations are not. And, just as for more standard discrimination
learning paradigms, learned associability seems to play an important role in spatial learning.
Prados, Redhead, and Pearce (1999) demonstrated that rats were faster to learn the location of
a hidden platform in a Morris swimming pool when this location was defined by a configuration of landmarks if these landmarks had previously been consistent predictors of a different
platform location than if they were novel. Likewise, Redhead, Prados, and Pearce (2001)
demonstrated that rats were slower to learn the relationship between landmarks and platform
location if the landmarks had previously been inconsistent predictors of platform location than
if they were novel. Taken together, these results provide strong evidence for a role of learned
predictiveness in determining the rate of spatial learning. Moreover, this learned predictiveness effect seems to operate in accordance with the Mackintosh (1975) model. Pretraining of
the landmarks as consistent predictors of platform location would allow them to maintain a
high associability, promoting learning about the relationship between the landmarks and the
new platform location on test. Pretraining of the landmarks as inconsistent predictors of platform location would cause their associability to decline, slowing learning about the relationship between these cues and the new platform location on test.
Further powerful evidence for a role of associability processes in discrimination learning
comes from studies of discriminational shifts, in which the response requirements are maintained between training and transfer discriminations, but the stimuli are changed in such a
way that any influence of direct transfer of associative strengths is eliminated. Consider the
case of stimuli that can vary on two independent dimensions, say colour and shape. In the
initial, training, discrimination, the stimuli can be either blue or yellow and either a circle or a
triangle. In the second, transfer, discrimination, the stimuli can be either red or green, and
either a square or a diamond. In an extradimensional (ED) shift, training and transfer discriminations are conducted along different dimensions. In the training discrimination the circular
stimulus might serve as S+ and the triangular stimulus as S, with the colour of the stimulus
being irrelevant. In the transfer discrimination it is now the other dimension, colour, that is
relevant, with red as S+ and green as S, while shape is irrelevant. In an intradimensional (ID)
shift, the same dimension is relevant for both training and transfer discriminations. In the
training discrimination the blue stimulus might serve as S+ and the yellow stimulus as S,
with the shape of the stimulus being irrelevant. The transfer discrimination is exactly the same
as that for the ED shift group (red as S+, green as S). So for the ID shift group colour is relevant in both discriminations. The fact that different colours are used in training and transfer
discriminations (along with suitable counterbalancing of the stimuli serving as S+ and S)
ensures that direct transfer of associative strength cannot influence the rate of learning of the
transfer discrimination. Hence any advantage in learning the second discrimination shown by
the ID group, as compared to the ED group, can only be accounted for in terms of changes in
the associability of, or attention paid to, relevant or irrelevant stimuli. And such an advantage
for intradimensional shifts over extradimensional shifts has been observed in a large number
ASSOCIATIVE HISTORY
213
of studies, using monkeys (Roberts, Robbins, & Everitt, 1988; Shepp & Schrier, 1969), rats
(Oswald, Yee, Rawlins, Bannerman, Good, & Honey, 2001; Schwartz, Schwartz, & Teas,
1971), pigeons (George & Pearce, 1999; Mackintosh & Little, 1969), and humans (Whitney &
White, 1993).
The ID/ED shift effect raises the question of the level at which associability applies to
stimuli. In their analyser model, Sutherland and Mackintosh (1971) proposed that associability applies to whole dimensions, such that animals may learn to attend to one dimension
(e.g., colour) rather than another (e.g., shape) as a result of experience. This approach lends
itself easily to explanation of the ID/ED shift effect. During Stage 1 animals in the ID shift
group will learn to attend more to colour than to shape. Given that colour is also relevant in the
transfer discrimination, this learned attention will facilitate learning of the transfer discrimination compared to an ED shift group for which initial training will result in shape (irrelevant
on the transfer discrimination) receiving more attention than colour. We saw above that the
Mackintosh (1975) model, in contrast, states that associability applies to individual cues or
features, rather than whole dimensions. In order to account for the superiority of ID shifts
(where the stimuli are changed between training and transfer discriminations) Mackintosh
suggested that associability could generalize from one stimulus to another as a function of their
similarity. It seems reasonable to assume that features from the same dimension will tend to be
more similar to one another than are features from different dimensionsred is more similar
to yellow than it is to a triangle. As a result there will be a greater generalization of associability
from the features relevant to the training discrimination to those relevant to the transfer
discrimination in the ID group (as the relevant cues from the two discriminations have
relatively high similarity) than in the ED group (as the relevant cues from training have lower
similarity to the relevant cues on transfer).
Recent evidence from a study of spatial learning by Trobalon, Miguelez, McLaren, and
Mackintosh (2003) indicates that associability applies to individual stimuli rather than to
dimensions. In each of their experiments they trained two groups of rats in a radial arm maze.
One group was trained on a visuo-tactile discrimination between two distinctive floor coverings. Reinforcement was always associated with a particular floor covering, while the direction
that the arm pointed to was irrelevant. The other group was trained on a spatial discrimination, in which the reinforced (S+) and nonreinforced (S) arms were defined by the directions
they pointed to, regardless of the floor coverings. Both groups of rats were then transferred to a
novel spatial discrimination, in which S+ and S pointed in two new directions never
experienced during training. Hence for rats originally trained on a visuo-tactile discrimination
this constitutes an ED shift, while for rats originally trained on a spatial discrimination it is an
ID shift.
In Experiment 1 rats in the ID group were trained to discriminate an arm pointing north
from arms pointing east or west. These rats were faster to learn a subsequent spatial discrimination between arms pointing south-east and south-west than rats from an ED group who had
received a visuo-tactile discrimination in the first stage. The superior performance of the ID
group could result from the rats learning to attend to the dimension of spatial landmarks
during training (the approach taken by Sutherland & Mackintosh, 1971). Alternatively, the
rats may learn to attend to specific landmarks that define the positions of S+ and S (those
defining north, east, and west) and to ignore other cues that were irrelevant to the solution of
the problem. These irrelevant cues would include any landmarks that do not serve to
214
LE PELLEY
differentiate between S+ and S (e.g., those lying midway between the two arms, defining
north-east and north-west). Greater attention to landmarks defining east and west might then
facilitate subsequent discrimination of arms pointing south-east and south-west.
In order to discriminate between these alternative hypotheses, Trobalon et al. (2003)
conducted a further experiment in which rats in the ID group were initially trained to discriminate an arm pointing south from arms pointing east or west. These rats were substantially
slower to learn a subsequent discrimination between south-east and south-west arms than
were ED rats initially trained on a visuo-tactile discrimination. It is hard to see how Sutherland and Mackintoshs (1971) view of dimensional associability can account for this result.
Just as in the previous experiment, initial training for the ID group should result in greater
attention paid to the spatial dimension, which should facilitate subsequent learning of the
transfer discriminationSutherland and Mackintoshs view predicts that ID training should
always result in superior transfer performance to ED training. Mackintoshs view of stimulus
specific associability, on the other hand, provides a ready explanation. During initial training,
ID rats will learn to attend to landmarks defining south, east, and west, and to ignore those
defining south-east, and south-west (as these landmarks do not serve to differentiate south,
east, and west arms and are therefore irrelevant to this discrimination). These ignored landmarks, however, are exactly those that are relevant to solution of the transfer discrimination
between south-east and south-west arms. Moreover the landmarks defining south, which the
rats will have learnt to attend to strongly in Phase 1, are common to both south-east and southwest arms in Phase 2 and so might be expected to interfere with learning of this transfer
discrimination. As a result Mackintosh (1975) is able to account for the finding that, under
some conditions, an ID shift will be learnt more slowly than an ED shift. These results imply
that, in the spatial domain at least, associability applies to individual cues, rather than to
dimensions.
Studies of conditioning
We saw above that phenomena of discrimination learning such as the overtraining reversal
effect and intradimensional shift advantage provide strong evidence for a role of associability
processes in learningthe processing received by a cue can be influenced by the associative
history of that cue. Moreover the results support the view of associability suggested by Mackintosh (1975), wherein good predictors of an outcome maintain high associability, while the
associability of poorer predictors declines. A number of studies of Pavlovian conditioning
have also indicated that it is insufficient to view learning as driven solely by changes in
processing of the USprocessing of CSs plays its part too.
One of the most well-known phenomena indicating a role of associability in conditioning is
that of learned irrelevance. The RescorlaWagner model predicts that if a CS and a US are
positively correlated, excitatory conditioning to the CS will occur; if they are negatively correlated, inhibitory conditioning will be observed. If the two are uncorrelated, such that the US is
as likely to occur in the presence of the CS as in its absence, then the CS will end up with zero
associative strength and hence be indistinguishable from a novel stimulus. Contrary to this
prediction, Mackintosh (1973; see also Baker & Mackintosh, 1977) demonstrated that rats
given uncorrelated exposure to a tone and water were slower to learn about a contingent tone
water relationship in a subsequent conditioning phase than rats given no preexposure to either
ASSOCIATIVE HISTORY
215
tone or water. Something must have been learnt during uncorrelated exposure to the CS and
US. Could it be that uncorrelated CS/US exposure leads to development of an inhibitory CS
US association, which would then slow acquisition of excitatory conditioning in Phase 2? This
seems unlikely, as uncorrelated exposure also leads to a retardation in subsequent inhibitory
conditioning of the CS (Baker & Mackintosh, 1979; Bennett, Wills, Oakeshott, & Mackintosh,
2000; but see Bonardi & Ong, 2003). It would seem that past experience of the tones irrelevance with respect to the delivery of water results in the representation of the tone becoming
less ready to enter into association (either excitatory or inhibitory) with the representation of
water. This is in line with the predictions of Mackintoshs (1975) model. During uncorrelated
exposure the tone is a poorer predictor of the delivery of water than is the training context
itself; as such its associability will fall, with the result that it will be slower to enter into subsequent association with water than if it were novel. On this view, then, while uncorrelated exposure does not result in development of a CSUS association, it is effective in causing changes
in the processing power devoted to that CS that will affect the subsequent development of
such an association.
There is, however, an alternative explanation of the learned irrelevance effect. It has long
been known that prior exposure to a stimulus leads to a retardation in the rate of subsequent
conditioning to that stimulus. This CS preexposure effect is known as latent inhibition, and it
has been reliably demonstrated in a large number of animal species (see Lubow, 1989). In addition, preexposure to the US alone can also lead to retardation of subsequent conditioning
(Randich & LoLordo, 1979). Could learned irrelevance be a simple consequence of the CS
preexposure and/or US preexposure effects, negating the need for explanation in terms of
associability changes resulting from experience of a noncontingent relationship between CS
and US? In their studies of learned irrelevance Mackintosh (1973) and Baker and Mackintosh
(1977) also included groups exposed to either CS alone or US alone, and showed that animals
given uncorrelated CS/US exposure were much slower to condition than either. A number of
subsequent studies have provided further evidence that learned irrelevance represents more
than the sum of CS and US preexposure effects, supporting the idea of a reduction in CS associability as a result of uncorrelated CS/US exposure (Baker & Mackintosh, 1979; Bennett,
Maldonado, & Mackintosh, 1995; Bennett et al., 2000; Matzel, Schachtman, & Miller, 1988;
but see Bonardi & Hall, 1996; Bonardi & Ong, 2003).
We saw earlier that the RescorlaWagner and Mackintosh models take very different views
of the processes underlying blocking. RescorlaWagner explains blocking in US-processing
termsprior learning that A is a good predictor of the US reduces the surprise caused by
presentation of the US on AB trials, and so (given that it is surprise that drives learning) little is
learnt about B. The Mackintosh model instead explains blocking in terms of associability
changes. Prior conditioning of A means that A is a better predictor of the US on AB trials than
is B. As a result the associability of B will decline, leading to a retardation in learning about B.
So is there any evidence for a role of associability processes in blocking?
A number of experiments employing unblocking procedures indicate that the answer is
yes. If A is followed by a single delivery of food (US1), and AB is then followed by two successive food deliveries (US1 . . . US2), more excitatory conditioning to B is observed than for a
standard blocking contingency, in which A and AB are both followed by a single US (or one in
which A and AB are both followed by two food deliveries); (Dickinson & Mackintosh, 1979;
Holland, 1984, 1988; a similar experiment employing shock instead of food USs is presented
216
LE PELLEY
by Dickinson, Hall, & Mackintosh, 1976). This unblocking as a result of an unexpected

upshift in food deliveries on compound trials is consistent with the view of blocking taken by
the Mackintosh model. The added US (US2) on the first compound trial is surprising and is
predicted by B as well as it is by A. Consequently, US2 will serve to maintain Bs associability,
promoting its ability to enter into excitatory associations with US1.4 Unblocking by addition
of a surprising reinforcer is, however, also in line with the predictions of the RescorlaWagner
model. As US2 is surprising it will support additional excitatory conditioning, some of which
will accrue to B. According to this approach, then, unblocking is a result of excitatory conditioning of B with respect to the surprising US2. Evidence against this latter view is provided
by Hollands (1988) study in which US2 was qualitatively different to US1 (sucrose solution
versus solid food pellets) and generated a conditioned response with a different response
topography, such that on test it was possible to determine the nature of learning about the
added CS on compound trials. This experiment indicated that the major function of the added
US2 was to enhance the development of associations between the added CS and the original
reinforcer, US1. This finding supports the idea that unblocking by addition reflects the operation of associability processes, wherein the added US2 serves to maintain the associability of
the added CS and hence promotes the formation of associations between that CS and US1.
Further evidence in support of a CS-processing approach to unblocking by addition is
provided by Mackintosh (1978), who demonstrated that inserting trials on which the
compound AB is followed by US1 alone before AB US1 . . . US2 trials resulted in very little
conditioning accruing to B on these latter trials. In other words, the inclusion of trials on which
the AB compound was followed by the same reinforcement as was A alone attenuated the
unblocking effect that would normally be expected as a result of the addition of US2 on the
latter compound trials. Quite why this should be so from the US-processing viewpoint typified by the RescorlaWagner model is unclear. The occurrence of US2 on AB US1 . . .
US2 trials is just as surprising following A US1, AB US1 training as it is following A
US1 trials alone, and hence it should support just as much learning about B. The CSprocessing account of unblocking fares better. Over the course of AB US1 trials the associability of B will decline (as it is a poorer predictor of US1 than is A). It will therefore begin AB
US1 . . . US2 trials with a low associability and as a result will be unable to enter into significant associations with either US1 or US2 over the course of these trials. Mackintosh and
Turner (1971) obtained similar results when conditioning to B (i.e., unblocking) was achieved
by increasing the strength of a single outcome on compound trials, rather than by adding a
second outcome. If weak shock is denoted us, and strong shock US, then significant
conditioning to B is observed if A us trials are followed by AB US trials; if however, four
AB us trials are inserted before the compound trials with the stronger shock, little
responding to B is observed. The associability of B must have declined during the course of
these compound trials with the weaker shock, presumably as that shock was better predicted
by the presence of A.
4
In fact, unblocking is still predicted even if generalization of Stage 1 AUS1 learning means that A begins Stage 2
as a slightly better predictor of US2 than is B. While this will result in a decline in Bs associability, this decline will be
slower than for a control group in which AB is followed by US1 only, this being well predicted by the presence of A.
ASSOCIATIVE HISTORY
217
But perhaps the most powerful evidence for a role of associability processes in blocking
comes from demonstrations of unblocking following the omission of an expected reinforcer
(Dickinson et al., 1976; Dickinson & Mackintosh, 1979; Holland, 1984, 1988). In these studies
trials on which A is followed by US1 . . . US2 precede trials on which the compound AB is
followed by US1 alone. Just as in the case of unexpected addition of US2 on compound trials,
omission of the expected US2 on compound AB trials leads to unblocking (greater excitatory
conditioning to B than that observed in an A US1 . . . US2, AB US1 . . . US2 control
group). This is clearly incompatible with the US-processing view of learning taken by the
RescorlaWagner model: Omission of an expected outcome on compound trials should
produce inhibitory, not excitatory, conditioning. Instead it would seem that the surprising
omission of US2 somehow facilitates the association of B with US1. This is again predicted by
the view of associability taken by the Mackintosh model. On the initial AB trial, US2 is absent
but expected on the basis of A. B is a better predictor of the absence of this outcome than is A,
and hence (according to Equations 8 or 14) its associability will be maintained at a higher level
across compound trials than if US2 had been presented on those trials.
Demonstrations of unblocking by reinforcer omission are counterintuitive, in that most
theories predict that a downshift in the overall magnitude of reinforcement occurring on
compound trials will endow the added CS, B, with inhibitory properties, whereas in the
studies outlined above the opposite result was seen. Even the extended Mackintosh model
described earlier predicts that any positive effect of reinforcer omission on Bs associability
(enhancing the formation of BUS1 associations) will be countered to some extent by
formation of inhibitory BUS2 associations. Consistent with this idea, Holland (1988)
demonstrated that downshifts in reinforcer number (from US1 . . . US2 to US1 alone) did
indeed result in both enhancement of formation of BUS1 associations and development of
inhibitory BUS2 associations. Perhaps, then, in the studies mentioned so far it is the
former effect that dominates, leading to net excitation. This is not always the case. In the
studies described above, moving from two temporally distinct USs to one on compound
trials resulted in net excitation. In contrast, we saw earlier that decreasing the intensity of a
single shock on compound trials established inhibition to the added cue (Cotton et al., 1982;
Wagner et al., 1980). These results imply that temporal separation of the two USs is important if the added cue is to become an excitor. In line with this idea, Holland (1988) found that
downshift procedures generated net inhibition with short US1US2 intervals, but net excitation with longer US1US2 intervals. This observation is consistent with the idea that the
indirect effects of reinforcer omission on the added cue (enhancement of its ability to
enter novel associations) and the direct effects (establishment of conditioned inhibition)
are described by different temporal gradients. That is, the influence of reinforcer omission
on associability seems to be effective over a wider temporal window than does the influence
of omission on associative strength. These differing temporal gradients for associability and
associative strength are clearly not captured by any of the trial-based models presented here,
in which temporal factors are disregarded in exchange for the simplicity of viewing each
learning episode as a separate event. Nevertheless, these findings have important implications for the characterization of associability and associative strength changes in any realtime model of learning.
218
LE PELLEY
The PearceHall (1980) model

So far we have focused on phenomena of conditioning (learned irrelevance, unblocking by
addition or omission of reinforcers) providing evidence for a role of associability processes in
the determination of associative change, specifically phenomena in line with the approach to
associability taken by Mackintosh (1975). However, there exist other demonstrations of CSprocessing modulation that present more of a challenge to this model. One such challenge is
seen in the phenomenon of HallPearce negative transfer. Hall and Pearce (1979) demonstrated that rats were slower to develop conditioned suppression to a tone paired with a strong
shock if they had previously been exposed to consistent pairings of that tone with a weak shock
than if they had received no prior exposure to the tone. Once again, this finding is problematic
for theories specifying learning to be governed solely by changes in processing of the US. If
anything, pairing of the tone with a weak shock would be expected to result in more rapid
acquisition of conditioned responding as a result of pairing the same tone with a strong shock,
as the tone will begin these latter trials with a certain amount of excitatory strength, thus giving
it a head-start. Instead this effect seems to require that consistent pairings of the tone with a
weak outcome result in changes in the processing of that tone, altering its ability to enter into
association with a stronger version of the same outcome. The view of learned associability
taken by the Mackintosh model seems to fare no better than the US-processing view, however.
During the first phase, consistent pairings of the tone and weak shock should allow the tone to
maintain a high associability (as it is the best available predictor of this weak shock). As a result,
processing of the tone on tonestrong-shock trials should be at least as strong as for the control
group given no preexposure to the tone (for whom the tone has not previously been established
as a good predictor), with the attendant prediction that conditioned responding to the tone
should develop at least as rapidly, if not more so, in the experimental group.
Instead demonstration of HallPearce negative transfer seems to imply a view of associability processes that is diametrically opposite to that taken by Mackintosh (1975). That is, it
implies that processing of the tone is reduced as a result of consistent pairings of that tone with
an outcome (weak shock). This makes contact with the phenomenon of latent inhibition,
wherein repeated nonreinforced preexposure to a stimulus (establishing that stimulus as a
good predictor of no outcome) produces a retardation in the rate of subsequent conditioning
of that stimulus (the CS preexposure effect discussed earlier; see Lubow, 1989). Effects such
as HallPearce negative transfer and latent inhibition led Pearce and Hall (1980) to develop
their own CS-processing model of learning. They argued that, while reliable predictors of
outcomes should be able to control an animals behaviour, there is little point in allocating a
large proportion of the processing capacity for learning to events that are involved in stable
relationships. Instead, it would seem to make more sense to devote processing power to stimuli
whose predictive status is currently unclear in an attempt to learn more rapidly about the true
significance of those stimuli. In other words, unlike the Mackintosh model, Pearce and Hall
argue that stimuli that are unreliable predictors of outcomes will maintain a higher associability than stimuli that are reliable predictors.
It should be noted, however, that the PearceHall formulation is not the only CSprocessing model that is able to account for latent inhibition. Wagners (1981) standard operating procedures (SOP) model proposes that processing of a stimulus depends upon whether
or not that stimulus is surprising (see also McLaren, Kaye, & Mackintosh, 1989). Repeated
ASSOCIATIVE HISTORY
219
exposure to a CS in a given context will encourage the growth of associations between the
context and that CS. Thus the context comes to predict the presence of the CS, rendering it
less surprising and hence less able to engage in subsequent conditioning, yielding the latent
inhibition result. The model is able to account for HallPearce negative transfer in similar
fashion (Swartzentruber & Bouton, 1986). A prediction of SOPs account is that latent inhibition should be context specific: If latent inhibition relies on the development of associations
between exposure context and CS, conducting conditioning trials in a different context should
restore the surprisingness of the CS and hence abolish the latent inhibition effect. This
prediction, which lies beyond the scope of the PearceHall model, has received empirical
support (e.g., Channell & Hall, 1983; Hall & Minor, 1984: Lovibond, Preston, & Mackintosh,
1984; Rosas & Bouton, 1997). Wagners model, on the other hand, is unable to account for the
effects of predictive accuracy on associabilitythat is, it fails to provide a satisfactory account
of the way in which experience of relationships between stimuli and outcomes affects the
subsequent processing of those stimuli (Hall, 1991). The PearceHall model is more
successful in this respect. Given that it is these effects of associative history on learning that
provide the basis for the present article, in the following discussion I will focus on the approach
to CS-processing offered by the PearceHall model in preference to Wagners theory. For a
more detailed discussion of the similarities and differences of the PearceHall and Wagner
models, see Hall (1991).
Specifically, the PearceHall model5 states that, following the adjustment of associative
strengths on trial n (see below), the associability of each presented stimulus is updated
according to the equation:
n 1
| + (1 ) n 1
nA = | n VNET
16
is a parameter (which can vary between 0 and 1) that determines the extent to which is
determined by the events of the immediately preceding trial. If 1 then is determined
almost solely by the events of the immediately preceding trial, with earlier trials having little
effect. Conversely, if 0 then is determined largely by earlier trials, with the immediately
n 1
is defined by:
preceding trial having little effect. VNET
n 1
VNET
= V n 1 V
n 1
17
whereV , as earlier, refers to the strength of a CSno-US association (i.e., an inhibitory associn 1
represents the extent to which the US occuration or antiassociation). In other words, VNET
ring on the current trial was predicted by all stimuli presented on that trial.
5
In the original PearceHall (1980) model, the associability of a stimulus was determined entirely by events occurring on the immediately preceding trial. Pearce, Kaye, and Hall (1981) noted that this analysis had a number of unfortunate consequences and therefore amended the model such that associability was determined by the average of the
values for a number of preceding trials, thereby damping changes in associability. Specifically their inclusion of a
parameter implemented an exponentially weighted moving average across trials for . So the current value of is
strongly influenced by the events of the immediately preceding trial, less so by the trial before that, less still by the trial
before that and so on. Pearce, Kaye, and Hall also proposed inclusion of a reinforcer-related learning-rate parameter,
, which could take different values depending on the nature of the reinforcer used (as employed by the extended
Mackintosh model in Equations 10 and 11). It is this amended version of Pearce and Halls original model that is
presented here.
220
LE PELLEY
Now we need the equations governing the change in associative (and antiassociative)
strength on trial n (note that these changes occur immediately preceding the change in associability according to Equation 16 above). Just as for the extended Mackintosh model presented
earlier, the associative change undergone on a given trial depends on the nature of learning that
that trial will support. So as before we calculate:
R n = n (V n 1 V
n 1
18
If R is positive (i.e., this is a trial that will support excitatory learning), the strength of the CS
US association is increased according to the equation:
VAn = E nA1 n
19
If R is negative (i.e., this is a trial that will support inhibitory learning), the strength of the
CSUS association is increased according to the equation:
VAn = I nA1 |R n |
20
where E and I, are learning-rate parameters for excitatory and inhibitory learning,
respectively.
We can see from Equation 19 above that the PearceHall model goes one step further than
Mackintosh (1975) in its view of the role of associability processes in learning. For excitatory
learning at least, this model places the entire burden of modulation of associative change on
processing of the CSthere is no error term in the calculation of associative change. That is to
say, processing of the US does not change (as a result, say, of changes in its surprisingness)
over the course of learning.6
In order to see how the PearceHall model can function in the absence of an error term in
the equation for associative change, it is perhaps easiest to apply it to a simple example.
Consider pairing a novel Cue, A, with an outcome. On the initial conditioning trials A is a
poor predictor of the US. As a result the value of | VNET| in Equation 16 will be high,
such that A maintains a high associability, which drives increases in VA according to
Equation 19. However, as conditioning proceeds the associability of A will decline (as VA
increases, | VNET| decreases), with changes in associative strength becoming increasingly
small. Ultimately, when VA = the associability of A will be negligible, and hence learning will
effectively come to a stop (if VA should overshoot and increase slightly above then the
inhibitory process described by Equation 20 will kick in to reduce it again). At this point, then,
A is able to control the animals behaviour (in that it will give rise to strong conditioned
responding), but little processing power is devoted to learning about it. An account of Hall
Pearce negative transfer follows naturally from this idea. Consistent pairing of the tone and
weak shock will establish the tone as a good predictor of this weak shock and will hence lead to a
decline in its associability, such that little is learnt on subsequent pairings of the tone and
strong shock. Consistent with this idea, Hall and Pearce (1982) found that inserting two trials
on which the tone was followed by no shock immediately prior to tonestrong-shock pairings
6
The same principle could be seen as applying to inhibitory learning. However the R term in Equation 20 is a
summed error term. As such, changes in processing of the US will affect the change in strength of CSno-US
associations.
ASSOCIATIVE HISTORY
221
reduced the negative transfer observed (i.e., led to more rapid conditioning on these tone
strong-shock trials). The PearceHall model views these inserted trials as restoring the associability of the tone by virtue of the fact that the outcome on these trials is surprising (no shock
occurs where a weak shock is predicted).
We saw earlier that the Mackintosh model specifies changes in a cues associability to be
governed by a separable error term ( VA in Equation 8, with analogues in Equations 13 and
14). Hence the associability of each element of a stimulus compound is determined by how
well that element alone predicts the current outcome. In contrast, the PearceHall model
specifies associability changes to be governed by a summed error term (| VNET| in Equation 16). This has important ramifications for compound conditioning, as it means that the
associability of each element of a compound is determined by how well that compound predicts
the current outcome.
For example, the use of a summed error term to determine associability allows the Pearce
Hall model to account for blocking. During the first phase, A+ trials will establish A as a good
predictor of the outcome (VA ), and A will decline correspondingly. In the second phase,
the AB compound is followed by the US. The presence of A will ensure that the US is already
well predicted on these trials, however; the outcome following AB trials is not surprising.
Therefore little processing power will be devoted to the elements of this compoundthe low
value of | (VA + VB)| will cause B to decline rapidly. As a result little will be learnt about B,
compared to a control group not given pretraining with A such that the occurrence of the US
on AB compound trials is more surprising. Figure 6 shows the results of a simulation of a
blocking experiment with the PearceHall model. The experimental design employed was
exactly the same as that modelled earlier by the extended Mackintosh model: 20 A+ trials,
followed by 8 AB+ trials intermixed with 8 CD+ trials. Parameters used for this simulation
were: E = . 1, I = .01, = .4, starting value of = .8, (US present) = .8, (US absent) = 0,
although the prediction of blocking is parameter independent. Blocking is observed in that the
associative strength attained by B following Stage 2 is lower than that for C or D.
The PearceHall model is also able to account for the various manipulations described
earlier that produce an unblocking effect. Both addition of an unexpected reinforcer and
omission of an expected reinforcer on Phase 2 compound trials mean that the outcome occurring on these trials is in some way surprising (i.e., | VNET| will be greater than if the
outcome had remained the same in the two cases), and hence Bs associability will be maintained as compared to a control group experiencing the same outcome in both phases.
Further evidence supporting the PearceHall view of associability is provided by the
observation that, under certain conditions, learning about a stimulus is more rapid when that
stimulus is an inaccurate predictor of the events that follow it than when it is an accurate
predictor (Kaye & Pearce, 1984; Swan & Pearce, 1988; Wilson, Boumphrey, & Pearce, 1992).
For example, Wilson et al. trained rats on a partially reinforced serial conditioning procedure.
A light was followed by a tone; on half of the trials this tone terminated with delivery of a food
US, and on the other half no food US was provided. This procedure resulted in only minimal
conditioned responding (magazine orienting) to the light. These rats were then divided into
two groups. For group consistent training continued as before. For group shift, the tone was
omitted on all nonreinforced trials. Hence while the relation between the light CS and the food
US was maintained for group shift, its relation to the tone was changed from 100% to 50%.
According to the PearceHall model, this will result in a higher associability for the light in
222
LE PELLEY
Figure 6. Simulation of a blocking experiment using the PearceHall model. Upper panel: Net associative strength
(V V ) for Cues A, B, and C/D. During Stage 2, B gains less excitatory strength than the control cues, C/D, thus
demonstrating blocking. Lower panel: Associability changes for A, B, and C/D. B falls during Stage 2 as the occurrence of the US on AB+ trials is unsurprising, being already predicted by A.
group shift, where the event following the light (the tone) is unpredictable, as compared to
group consistent, where the tone consistently followed the light. In line with this idea Wilson
et al. found that, when the light was subsequently paired directly with food, rats in group shift
were faster to develop conditioned responding to the light than those in group consistent. This
seems to run against the principles of the Mackintosh model, which would, if anything,
predict that degrading the lighttone contingency would lead to a decrement in the associability of the light.
Human causal learning

Studies of animal conditioning are in some sense analogous to studies of causal learning in
humans. Both involve arranging for a contingent relationship between stimuli (CSs and USs
in animal conditioning; cues and outcomes in causal learning), and the measurement of
learning about that relationship (conditioned responding or a judgement of the causal
strength of a cue). In support of the idea of a parallel between these two preparations, similar
factors are seen to influence both in similar waysanimal conditioning and human causal
ASSOCIATIVE HISTORY
223
judgement show similar sensitivity to temporal contiguity and contingency, for instance.
These parallels led Dickinson, Shanks, and Evenden (1984) to suggest that similar processes
might underlie animal conditioning and human causal judgement, with the attendant implication that models of animal conditioning might be used to describe the acquisition of human
causal judgements (see De Houwer & Beckers, 2002, for a review of recent studies of parallels
between conditioning and causal learning).
Given this possibility of a common process underlying animal conditioning and human
causal learning, and the weight of evidence indicating a role for associability processes in
animal learning, we might expect human causal learning to show a similar sensitivity to associative history. In line with this idea, several recent studies have demonstrated that the
processing of cues in a causal learning task can vary depending on the training history of those
cues.
We saw earlier that the US-processing account typified by the RescorlaWagner model
and the CS-processing account offered by the Mackintosh model differ in their explanation of
the blocking effect. The former explains blocking as an effect of reduced processing of the US
(the outcome is unsurprising on AB compound trials as it is already predicted by A), the latter
as an effect of reduced processing of the added CS (Bs associability falls as it is a poorer
predictor of the US than is A). Kruschke and Blair (2000) investigated the source of the
blocking effect in human causal learning and found evidence favouring this latter approach.
They employed a multiple-outcome medical diagnosis paradigm, in which participants had to
decide which of several possible diseases (outcomes) a patient was suffering on the basis of the
symptoms exhibited by that patient (cues). In the blocking contingency, A was established as a
good predictor of Outcome 1 before pairings of an AB compound with Outcome 1. In the
control contingency, a compound CD was paired with Outcome 2, neither element having
received prior training. A blocking effect was observed in that, following this training, B was
perceived to be a weaker predictor of Disease 1 than was D of Disease 2. This blocking effect is,
of course, predicted by either class of theory mentioned above. However, Kruschke and Blair
demonstrated that participants were slower to learn a subsequent relationship between B and a
novel outcome (Disease 3) than they were to learn an identical relationship between a novel cue
(E) and a novel outcome (Disease 4), indicating that the blocking treatment had led to a decline
in Bs associability as compared to this novel control cue. This finding would seem to lie
beyond any theory that attempts to explain the blocking effect solely in terms of changes in
outcome processingit requires that the processing of the blocked stimulus must change as a
result of the training it undergoes.
It should be acknowledged, however, that it may be possible to explain the results of
Kruschke and Blairs (2000) experiment in terms of differential proactive interference acting
on experimental (B) and novel control (E) cues. A further study providing evidence for a role
of associability processes in human causal learning by Le Pelley and McLaren (2003; see also
Lochmann & Wills, 2003) is harder to interpret in terms of proactive interference. The basic
design of their experiment is shown in Table 1. This experiment employed an allergy prediction paradigm, in which participants play a food allergist looking at the causes of allergic
reactions in fictitious patients. In this design, Cues AY represent different foods, and the
numbers 14 represent different types of allergic reaction that patients could suffer as a result
of eating these foods. On each trial of Stage 1, participants were told the contents of a meal
eaten by Mr. X, and they were asked to predict the type of allergic reaction that he would
224
LE PELLEY
TABLE 1
Basic design of study by Le Pelley and
McLaren (2003)
Stage 1
Stage 2
Test
AV 1
BV 2
AW 1
BW 2
CX 2
DX 1
CY 2
DY 1
AX 3
BY 4
CV 3
DW 4
AC
BD
VX
WY
Note: AY = foods; 14 = allergic reactions. Filler

trials omitted for clarity.
suffer as a result, given a choice of Allergy 1 or Allergy 2. In this stage Cues A and D consistently indicated the occurrence of Allergy 1, Cues B and C consistently indicated the occurrence of Allergy 2, and Cues VY provided no basis for discrimination between the two
outcomesthey were paired with Allergies 1 and 2 an equal number of times. According to
the Mackintosh model, then, Cues AD should maintain a high associability over Stage 1
trials, as they are better predictors of the outcome on each trial than are the cues with which
they are paired. Conversely, the associability of Cues VY will decrease over the course of
Stage 1, as they are poorer predictors of the outcome on each trial than are the cues with which
they are paired. In Stage 2, participants were given information regarding foods and allergies
for a new patient, Mr. Y. On each of the Stage 2 trial types shown in Table 1, a good
predictor from Stage 1 (A, B, C, or D) is paired with a poor predictor (V, W, X, or Y), and
this compound is paired with a novel outcome: Compounds AX and CV are paired with
Allergy 3, while BY and DW are paired with Allergy 4. A subsequent causal judgement test
revealed that participants had learnt more about the relationships between the good predictors
(good with respect to Stage 1 outcomes) and the novel outcomes than between the poor
predictors and the novel outcomes during Stage 2. That is, following Stage 2, compound AC
(composed of two good predictors from Stage 1 that were paired with Allergy 3 in Stage 2) was
rated as being strongly causative of Allergy 3, and BD (good predictors paired with Allergy 4 in
Stage 2) was rated as being strongly causative of Allergy 4, while VX (composed of poor
predictors paired with Allergy 3 in Stage 2) and WY (poor predictors paired with Allergy 4 in
Stage 2) were rated as being only weak causes of their respective Stage 2 outcomes. In line with
the Mackintosh view, it would seem that the cues that were experienced as being good predictors during Stage 1 (AD) commanded more processing in Stage 2 than did the cues that were
experienced as being poor predictors during Stage 1 (VY).
The outcome specificity of learned associability

Kruschke and Blair (2000), Le Pelley and McLaren (2003), and Lochmann and Wills (2003)
employ a common fundamental approach to investigating associability processes in humans:
All train stimuli as being predictive or nonpredictive of certain outcomes and then examine
ASSOCIATIVE HISTORY
225
changes in the associability of those stimuli by studying how rapidly these cues will enter into
associations with novel outcomes. The fact that these studies do show associability effects
indicates that a cues associability is not entirely outcome specific. That is, changes in a cues
associability brought about by experience of its predictive relationship with one outcome will,
under some circumstances at least, affect subsequent learning about a relationship between
that cue and a different outcome. This concept of the outcome specificity of associability
makes contact with earlier studies of animal conditioning. For instance, Mackintosh (1973)
found that giving rats uncorrelated exposure to a tone CS and water US led to retardation (i.e.,
a learned irrelevance effect) in subsequent conditioning of a tonewater relationship, but had
no effect on conditioning of a toneshock relationship. Likewise, uncorrelated exposure to
tone and shock led to a retardation in tone-shock conditioning but had no effect on learning
about a tonewater relationship. In this case, then, changes in the associability of the tone as a
result of experience of its relationship with one US were able to modulate its ability to enter
into associations with that same US, but did not affect learning about the relationship between
that cue and a different US. This experiment indicates that associability is an outcomespecific property. It is tempting to compare the human and animal studies in an attempt to
elucidate the true specificity of associability. In Le Pelley and McLarens human experiment
the outcomes used in the two stages of the experiment, while qualitatively different, had many
similarities (both were types of allergic reaction; more generally, both were aversive), whilst in
the Mackintosh study of animal conditioning, the USs used were very different (one
appetitive, the other aversive). It is possible that, while associability is not completely outcome
specific, the associability developed by a cue with respect to a particular outcome will generalize only to similar outcomes, perhaps only to those of the same affective class. Evidence
consistent with this idea comes from Hollands (1988) study of unblocking caused by the
surprising addition of a post-trial reinforcer. As mentioned earlier, the added US2 on
compound trials was qualitatively different to the original US1 (sucrose solution vs. solid food
pellets) but, in contrast to Mackintoshs experiment, was from the same affective class (both
appetitive). Holland found that addition of this different US2 enhanced the development of
associations between the added CS and US1 at least as well as if US1 and US2 were identical.
In line with the idea that associability will transfer between outcomes from the same affective class (as indicated by the human studies of Kruschke & Blair and Le Pelley & McLaren),
experience of the predictive relationship between the added CS and appetitive US2 was able to
modulate learning about that CS with respect to appetitive US1, despite the fact that these
outcomes were qualitatively different.
RECONCILING EFFECTS OF ASSOCIATIVE HISTORY:

A HYBRID MODEL
In summary, then, there is a wealth of evidence in support of the idea that the amount of
processing received by a CS on a given learning episode depends to at least some extent on
the associative history of that CS. Phenomena of discrimination learning such as the overtraining reversal effect and intradimensional shift advantage, and phenomena of conditioning such as learned irrelevance, unblocking by surprising omission of a reinforcer, and
HallPearce negative transfer, rule out the view taken by the RescorlaWagner modelthat
226
LE PELLEY
associative change on a given trial is determined solely by changes in the processing received
by the US. Moreover, similar effects of training history are seen to affect the learning undergone by cues in studies of human causal learning, in line with Dickinson et al.s (1984)
argument that common associative processes may underlie animal conditioning and human
causal learning.
The question now becomes one of how best to interpret this mass of evidence supporting
a role of associability processes in learning. The problem faced in attempting to construct a
unified theory of associability is that the experiments outlined above often conflict in the
view of associability that they support. For example, the overtraining reversal effect,
intradimensional shift advantage, and learned irrelevance all support an approach wherein
good predictors of an outcome maintain a high associability, while the associability of poorer
predictors falls; this is the approach developed in the Mackintosh model. The results of Le
Pelley and McLarens (2003) study of human learning also provide unique support to this
view. However, HallPearce negative transfer and the fact that, under some circumstances,
learning is seen to proceed faster with stimuli that are inaccurate rather than accurate
predictors of the events that follow them (Kaye & Pearce, 1984; Swan & Pearce, 1988;
Wilson et al., 1992) support an opposing view, wherein poor predictors maintain a higher
associability than good predictors; this is the approach developed in the PearceHall model.
The PearceHall model also provides a more satisfactory account of latent inhibition (the
detrimental effect of nonreinforced preexposure to a stimulus on the rate of subsequent
conditioning of that stimulus) than does the Mackintosh model (see Hall, 1991, for an
exhaustive discussion of PearceHall and Mackintosh models in relation to latent inhibition
and other phenomena of exposure learning). In addition, there exist other bipartisan
phenomena that are in line with both views of associability processesfor example, both can
account for unblocking by addition or omission of post-trial events, although they do so in
quite different ways.
Given this conflict in the empirical data, with reliable evidence supporting two opposing
views of associability, it seems unlikely that either the Mackintosh or the PearceHall model
alone will be able to provide a full account of the way in which the processing afforded to a CS
changes as a result of experience. Maybe a more fruitful approach would be to combine the
ideas encapsulated in those two models in an attempt to capture the strengths of each.
Perhaps the simplest way to reconcile the two theories of associability mentioned here is to
see them as describing two different properties of a cue, rather than being rival descriptions of
the same property. The Mackintosh model can be viewed as measuring the weight that should
be afforded to a particular stimulus in the learning process as compared to other potential
stimuli. In some sense, then, the Mackintosh model describes an attentional associability,
determining which stimuli should have access to the learning process and which should not.
The PearceHall model, on the other hand, can be seen as measuring the rate with which each
stimulus will be learnt about on the basis of the exposure history of that stimulus, regardless of
its attentional weight. As such, the PearceHall formulation indexes a property that might be
called salience associability. So we can see the Mackintosh alpha as allowing an animal to
pick out which stimuli it should learn about and the PearceHall alpha as determining how
much should be learnt about those stimuli. Rather than being at loggerheads, on the contrary it
seems that these theories could be made to work in concert and, in doing so, provide a more
satisfactory account of the effects of associative history on the associative change undergone by
ASSOCIATIVE HISTORY
227
a given stimulus on a given trial. Below I present one way in which this combination could be
achieved, but clearly many others are possible.
Given the proposed difference in the quantities described by the Mackintosh and Pearce
Hall models, it makes little sense to label them both with the same symbol. Hence in the
following discussion I refer to the attentional associability of the Mackintosh model as ,
and the salience associability of the PearceHall model as . The simplest way to incorporate these two properties into a model of associative learning is simply to insert them as multiplicative factors in the equation for associative change, yielding a hybrid model of
associability. The idea of a hybrid model of associability has previously been suggested by
Pearce, George, and Redhead (1998) and Rodriguez, Lombas, and Alonso (2002). The model
offered in the current paper represents an advanced implementation of this earlier suggestion.
The approach taken here is to add the PearceHall into the extended Mackintosh model
described earlierthe opposite approach (adding the Mackintosh into the PearceHall
model) would presumably also be possible. Insertion of a multiplicative factor into Equations 10 and 11 yields the following equations.
If R is positive (i.e., this is a trial that will support excitatory learning), the strength of the
CSUS association is increased according to the equation
VA = AAE (1 VA +V A) |R|
21
If R is negative (i.e., this is a trial that will support inhibitory learning), the strength of the CS
US antiassociation is increased by:
V A = AAI (1 V A + VA) |R|
22
where E and I are learning-rate parameters for excitatory and inhibitory learning,
respectively.
The equations for changing are exactly the same as those of the extended Mackintosh
model (Equations 13 and 14), and the equation for changing is exactly the same as that of the
PearceHall model (Equation 16). Crucially, in this hybrid model attentional associability
operates over a wider range than does salience associability. Attentional associability, , is
constrained to lie between 1 and .05, while salience associability, , is constrained to lie
between 1 and .5. That is:
If A 1 then
If A .05 then
A = 1
A = .05
If A 1 then A = 1
If A 1 then A = .5
23
24
These constraints afford attentional associability a potentially greater importance than

salience associability. Reductions in attentional associability can effectively halt learning about
a stimulus regardless of its salience associability (if = .05 in Equations 10 and 11 there can be
relatively little change in associative strength, even for a stimulus with high ). Reductions in
salience associability, on the other hand, while attenuating the rate of learning about a stimulus, will not prevent that learning to nearly the same extent (a stimulus with = .5 can still
undergo relatively large changes in associative strength, especially if it has high ). Remember
that, according to the Mackintosh model, stimuli that are good predictors of outcomes will
228
LE PELLEY
maintain high attentional associability, while the attentional associability for poorer predictors
declines. The PearceHall model effectively states the oppositestimuli will maintain a high
salience associability to the extent that the events that follow them are surprising. Given the
constraints on and as explained above, this means that a stimulus that is a poor predictor of
an outcome (low Mackintosh , high PearceHall ) will be learnt about only very slowly,
while a stimulus that is a good predictor of an outcome (high Mackintosh , low PearceHall
) will be learnt about more rapidly. This assumption seems intuitively plausiblea stimulus
that is unattended should receive very little processing regardless of the accuracy of the
predictions that it makes, while a stimulus that is attended to but whose predictions are
unsurprising should be given greater weight in the learning mechanism. Nevertheless, in the
absence of experimental evidence bearing on the issue of the minimum values of attentional
and salience associability, the assumption that min < min should perhaps be treated with some
caution.
The final assumption made in the model presented here is that the starting values of and
for a novel stimulus are near their maximum: A value of .9 is used for each in the simulations
discussed below. As a result there is the potential for a stimulus to undergo a large decline in its
ability to activate the learning process relative to this starting point, but the potential for only a
very small increase in this ability. That is to say, this parameterization assumes that associability effects seen empirically will generally reflect reductions in the ability of stimuli to
engage the learning process relative to their ability when novel. With regard to the salience
associability envisaged by the PearceHall model, this is not a particularly contentious issue
in general that model views the ability of a stimulus to engage the learning process as starting
from a high value and falling as a result of any kind of exposure to that stimulus (though some
kinds of exposuree.g., partial reinforcement trainingwill tend to maintain a high salience
associability for longer). Indeed, Pearce and Hall (1980) do not mention the possibility of an
increase in the associability of a novel stimulus as a result of any kind of exposure to that
stimulus (although following a decline a stimuluss associability may rise again as a result of
manipulations making the events following it more surprisinge.g., addition of an unexpected post-trial reinforcer).
From the point of view of the Mackintosh model, however, the suggestion that empirical
associability effects generally reflect decreases, and not increases, in the attentional associability afforded to stimuli is perhaps more debatable. Certain phenomena seem to demand that
the attentional associability of stimuli can fall as a result of experience of the predictive history
of those stimuli. For instance, it is hard to account for learned irrelevance without allowing for
a decrease in the associability of the CS as a result of uncorrelated CS/US preexposure. Recall
also Mackintosh and Turners (1971; see also Mackintosh, 1978) study of unblocking by
increasing reinforcer magnitude. A us (weak shock) training followed by AB US (strong
shock) trials results in greater conditioned responding to B than that seen in a standard
blocking group (A us then AB us, or A US then AB US). Mackintosh and Turner
found that insertion of four compound trials, during which the AB compound was paired with
weak shock, before compound training with the strong shock attenuated this unblocking
effect, revealing little evidence of conditioning to B. The associability of B must have declined
during the course of the four compound trials on which AB was paired with weak shock,
presumably because it was a poorer predictor of this weak shock than was A. Seraganian (1979)
ASSOCIATIVE HISTORY
229
provides further evidence demonstrating decreases in attentional associability in a study of

discrimination learning.
So there exists persuasive evidence supporting the idea that the attentional associability of
stimuli can fall from some starting value as a result of experience of those stimuli. But can we
take this idea one step further and state that all associability effects reflect decreases in associability of nonpredictive stimuli, or must we allow for increases in the associability of predictive
stimuli over and above this starting value? Consider the superior discrimination learning typically observed following an intradimensional (ID) shift in the rewarded and nonrewarded
stimuli, as compared to following an extradimensional (ED) shift (George & Pearce, 1999;
Mackintosh & Little, 1969; Oswald et al., 2001; Schwartz et al., 1971; Shepp & Schrier, 1969).
Let us imagine a study in which the stimuli can vary on two dimensions, colour and shape. For
the ID shift group, colour is the relevant dimension for both training and transfer discriminations, while for the ED shift group shape is relevant during training, and colour is relevant
during transfer. Typically, more rapid learning of the transfer discrimination is observed for
animals undergoing an ID shift than for those undergoing an ED shift. One might view this ID
shift advantage as resulting from an increase in the attentional associability of the predictive
features (colours) of the original training stimuli: This increased associability of the trained
colours then generalizes to the colours of the transfer stimuli, facilitating acquisition of this
latter discrimination as these predictive features are more ready to engage the learning
process. On this view, the ID shift advantage reflects an increase in the associability of predictive features during training of the original discrimination. An alternative view, however, is
that the ID shift advantage reflects a decrease in the associability of nonpredictive features.
During training for the ED subjects, the colour of the stimuli is uncorrelated with reinforcement. If the associability of the colours used (or the general dimension of colour, following
Sutherland & Mackintosh, 1971) declines as a result, this will have a detrimental effect on
subsequent acquisition of a discrimination in which colour is relevant. Likewise during
training for the ID subjects, the shape of the stimuli is irrelevant: A decline in the associability
of this feature will facilitate acquisition of a subsequent discrimination in which shape is again
irrelevant. Thus any difference in rate of learning of the transfer discrimination could be
ascribed to decreases in the associability of irrelevant stimuli. In fact, there is empirical
evidence to support this latter approach to the ID shift advantage: Turrisi, Shepp, and Eimas
(1969) found no difference in acquisition of a discrimination following an ID or ED shift if the
irrelevant dimension had not been present in the first problem.
Likewise, it is possible to account for all of the empirical phenomena described here in
support of the view of attentional associability envisaged by the Mackintosh model as
reflecting decreases in the associability of nonpredictive stimuli, as opposed to increases in the
associability of predictive stimuli. In fact, while evidence supporting decreases in attentional
associability for nonpredictive cues is well established, as yet there exists no evidence
providing clear and unequivocal support for an increase in the associability of a predictive
stimulus. It is, of course, quite possible that such evidence will be revealed in future. Hence the
model presented here does not completely rule out the idea that the associability of predictive
cues might increase above some starting valuethe associability of a predictive cue is
permitted to increase from .9 to a maximum of 1. Nevertheless, the potential for decreases in
associability (and consequently the potential for these decreases to effect empirical
230
LE PELLEY
observations) is much greaterthe associability of a nonpredictive cue can decrease from .9 to

a minimum of .05.
Simulations reveal that the hybrid model outlined above can account for all of the standard phenomena of associative learning. Thus acquisition of simple conditioning follows a
negatively accelerated function, as does subsequent extinction of that conditioning. This
model combines separable error terms (1 VA + V A in Equation 10 and 1 V A + VA in
Equation 11) with a summed error term, |R| in a similar fashion to the extended Mackintosh
model presented earlier. As such it is able to account for the development of standard conditioned inhibition, conditioned inhibition from a reduced reinforcer, overexpectation, superconditioning, and supernormal conditioning in a very similar way to this earlier model. In
addition the model explains the results of Rescorlas (2000, 2002) studies of the distribution of
associative change between the elements of a compound in the same way as this earlier model,
again by virtue of combining separable and summed error terms. These predictions have all
been confirmed by computational simulations employing exactly the same model parameters
as outlined belowspace constraints preclude the inclusion of the results of those simulations
here.
These are phenomena that any model attempting to provide a satisfactory account of associative change must address. Nevertheless, any model incorporating both separable and
summed error terms could account for these phenomena in terms of changes in processing of
the US. The true power of the hybrid model lies instead in its ability to account for a wide
variety of associability effects: that is, phenomena reflecting changes in processing of CSs as a
function of the associative history of those CSs. Here I describe the application of the model to
a subset of the most diagnostic associability effects as described earlier; application of the
model to the remaining preparations, while in some cases perhaps requiring extensions (e.g.,
inclusion of a temporal component in a real-time implementation), would be a matter of
relative simplicity.
Learned irrelevance refers to the finding that uncorrelated exposure to a CS and US
results in slower subsequent conditioning to that CS when it is consistently paired with the
US than if it were novel. The hybrid model provides a clear account of the learned irrelevance effect. During uncorrelated CS/US exposure, the CS is a poorer predictor of the US
than is the experimental context, and hence the attentional associability (CS, determined by
the Mackintosh equations) of the CS will fall. Given that even the CS + context
compound is a relatively poor predictor of the US, the salience associability of the CS (CS,
determined by the PearceHall equation) will tend to be maintained at a relatively high
value. Nevertheless, the low of the CS following uncorrelated CS/US exposure will
ensure that, when that CS is subsequently paired with the US, learning about it will be slow
(although CS will increase steadily over the course of conditioning training as the CS is
established as a good predictor of reinforcement). This can be compared with a control
group for whom the CS is novel at the outset of conditioning, such that it will enter conditioning trials with high attentional and salience associabilities (CS = CS = .9), promoting
rapid learning about that stimulus.
The earlier discussion of learned irrelevance raised the possibility that it reflected nothing
more than the sum of CS and US preexposure effects. The balance of experimental evidence
presented there suggested that learned irrelevance is more than the sum of its parts. It is
important to verify, then, that the hybrid model sees learned irrelevance as more than a
ASSOCIATIVE HISTORY
231
preexposure effect. None of the components of this model (summed error term, separable
error term, Mackintosh, and PearceHall) allow for changes in the processing of a US as a
result of simple exposure to that US, and hence the hybrid model has no scope for explaining a
US preexposure effect. It does, however, predict that CS preexposure will have an effect on
subsequent conditioning. While preexposure to a CS in the absence of reinforcement will not
affect CS, it will lead to a decline in CS, as the events following the CS (nothing) are not
surprising. Hence nonreinforced exposure will have a detrimental effect on the ability of the
CS to enter into associations with the US on subsequent conditioning trials, as compared to a
novel CS that has not undergone this decrease in salience associability. In other words, inclusion of a variable salience associability (following the PearceHall approach) allows this hybrid
model to account for the well-established empirical phenomenon of latent inhibition (see
Lubow, 1989). Note, however, that the retardation in learning resulting from nonreinforced
exposure to a stimulus will not be as great as that resulting from uncorrelated CS/US exposure. In the case of nonreinforced exposure the attentional associability of the CS remains high
(CS = .9), while its salience associability declines to the lower limit of .5. During uncorrelated
CS/US exposure, however, the attentional associability of the CS will decline towards its
lower limit of .05, while salience associability will also decline, if only slightly (as the CS +
context compound is a relatively poor predictor of the outcome). Hence in the former case, the
factor limiting the rate of conditioning will be salience associability; in the latter case, it will be
attentional associability. And as attentional associability is more potent than salience associability in terms of its ability to slow learning (as discussed earlier), uncorrelated CS/US exposure will result in slower conditioning than nonreinforced CS exposure, with both giving
slower conditioning than a novel CS.
These predictions were confirmed by simulation of a simplified learned irrelevance experiment. For the irrelevance condition, preexposure was to randomly intermixed X+, X, AX+,
and AX- trials: In this design, the occurrence of the US is not correlated with the presence or
absence of A. This preexposure continued for 20 blocks, with each of the four trial types occurring once per block. The CS-exposure condition employed the same trial types, but with no
USs occurring. In the novel control condition there was no preexposure. All groups then
received eight A+ conditioning trials. Figure 7 shows the results of this simulationthe data
for each group represent the average values for eight simulated subjects. Parameters used for
this simulation, and for all other simulations run with this model, were: E = .5, I = .1, E = .8,
I, = .1, = .1, starting value of = .9, starting value of = .9, (US present) = .8, (US
absent) = 0. It should be noted, however, that the models predictions are robust against even
large changes in these parameters. Panel A shows the changes in attentional and salience associability respectively undergone by A during the preexposure phase for the irrelevance and
CS-exposure groups. Panel B shows the changes in attentional and salience associability
during the subsequent conditioning phase for all three groups, and Panel C shows the changes
in associative strength undergone by A during this phase.
The results of this simulation confirm the predictions set out above. Nonreinforced
preexposure results in a steady decrease in while leaving unaffected, while uncorrelated
CS/US exposure drives to a very low value, with maintained at a higher level for longer.
As a result, nonreinforced preexposure to a CS results in slower conditioning of that CS than if
it were novel, but uncorrelated CS/US exposure leads to slower conditioning still (a result of
the fact that, at the outset of conditioning, is lower for the irrelevance group than is for the
232
LE PELLEY
Figure 7. Simulation of a simplified learned irrelevance study with the hybrid model. Panel A: Associability
changes for irrelevance and CS-exposure groups during Stage 1. Panel B: Associability changes for irrelevance, CSexposure, and novel control groups during Stage 2. In Panels A and B, open symbols represent salience associability
(), and closed symbols represent attentional associability (). Panel C: Net associative strength changes during Stage
2. Development of A US association is slowest in group irrelevance, followed by group CS-exposure, with fastest
learning in group novel control. Note that A does not enter Stage 2 with zero associative strength for group irrelevance, despite having a noncontingent relationship with the US during Stage 1: This maintenance of weak excitatory
strength by a noncontingent cue is a general problem for any model containing a separable error term.
CS-exposure group). The value of A rises for all conditions over the course of conditioning,
as A is the best predictor of the outcome on these trials; A rises slightly, as the occurrence of
the US on these trials is initially surprising, and then falls again as A comes to predict the US.
The hybrid model also provides a clear account of blocking. On AB+ trials of an A+, AB+
blocking procedure, three separate mechanisms combine in the hybrid model to yield blocking
of learning about the added cue, B. The |R| term in Equation 10 is a summed error term and
dictates that the associative change undergone by a cue is, in part, influenced by processing of
the US. In a blocking contingency, the level of reinforcement occurring on compound trials is
already predicted by A; hence |R| will be low, and so little learning about B will occur.
Second, as A is a better predictor of the level of reinforcement than is B on compound trials,
ASSOCIATIVE HISTORY
233
the attentional associability of B, B, will decrease rapidly, drastically reducing the extent to
which it engages the learning process. Third, the fact that the outcome is well predicted on
compound trials by the presence of A means that the salience associability of B will decline
over the course of compound training, as this associability is determined by a summed error
term (Equation 16), again slowing learning about B. In a sense, then, the blocking predicted by
this model is an effect of RescorlaWagner, Mackintosh, and PearceHall models combined.
In addition, the hybrid model is able to account for unblocking by an increase in reinforcer
magnitude, again as an effect of RescorlaWagner, Mackintosh, and PearceHall components
combined. That is, the increased magnitude of the reinforcer on compound trials (1) produces
an increased magnitude of the summed error term, |R| (RescorlaWagner), (2) leads to an
increase in the salience associability of A and B, as the level of reinforcement on compound
trials is only poorly predicted by the AB compound (PearceHall), and (3) ensures that the
difference in predictiveness of A and B for the intensified US on compound trials will be
smaller than that for a control condition employing the same reinforcer magnitude in both
stages, such that B in the unblocking condition will fall more slowly than that in the control
condition (Mackintosh). Moreover, the hybrid model can account for Mackintosh and
Turners (1971) demonstration that insertion of trials on which the AB compound is followed
by the weaker reinforcer before AB strong US trials results in very little conditioning
accruing to B on these latter trials. On these added trials, A is a much better predictor of the
level of reinforcement than is B, leading to a sharp decline in B, thus preventing B from
engaging the learning process to any great degree on subsequent trials with the stronger US.
Computational simulations have confirmed the ability of the hybrid model to account for
blocking, unblocking by increase in reinforcer magnitude and the reblocking effect demonstrated by Mackintosh and Turner.
So far we have looked at the ability of the hybrid model to account for phenomena that
could be just as easily explained by the extended Mackintosh model, presented earlier. The
advantage of the hybrid model is that it permits explanation of results that lie beyond the scope
of this earlier model. For instance, the inclusion of a variable salience associability allows the
hybrid model to account for HallPearce negative transfer (Hall & Pearce, 1979). Figure 8
shows simulation results for a typical HallPearce experiment. Group pretrained receives 20
A us trials followed by 8 A US trials; group control receives 8 A US trials in the
Figure 8. Simulation of HallPearce negative transfer with the hybrid model. This figure shows the net associative
strength of A over the course of Stage 2 A US training. Learning of the A US association is more rapid if A is
novel (group control) than if it has previously been consistently paired with a weak US (group pretraining).
234
LE PELLEY
absence of any pretraining. For the former group, during the course of pretraining A will
decline as the occurrence of the outcome comes to be predicted by the presence of A. A is the
best available predictor of the outcome, and hence its attentional associability will rise: Given
that the starting value of A is near its maximum, however, the effect of this increase will be
only slight. On subsequent A US trials, group pretraining will have a considerably lower A
than group control, and only a slightly higher A. Consequently, the model predicts that
conditioned responding to A should be slower to develop following A us pretraining than if
A were novel: This is, of course, the result seen in Hall and Pearces (1979) empirical study.
One area in which the hybrid model might, at first sight, appear to run into trouble is with
respect to the various studies demonstrating that, under certain conditions, learning about a
stimulus is more rapid when that stimulus is an inaccurate predictor of the events that follow it
than when it is an accurate predictor (Kaye & Pearce, 1984; Swan & Pearce, 1988; Wilson et al.,
1992). This would seem to run against the idea that the poor predictors maintain a low , effectively turning off learning about them. However, let us consider again the study by Wilson et
al., which is fairly typical of this type of preparation. To recap, this study employed a partially
reinforced serial conditioning procedure in which a light was followed by a tone, and on half of
the trials this tone terminated with delivery of the food US. Following this training, subjects
were split into two groups: group consistent, for whom training continued as before, and
group shift, for whom the tone was omitted on all nonreinforced trials. In the former group the
light remained a perfect predictor of the tone, whereas in the latter group the contingency
between light and tone moved from 100% in Stage 1 to 50% in Stage 2. On a subsequent test
stage in which the light was paired directly with food, Wilson et al. found more rapid conditioning in group shift than in group consistent, indicating that degrading the contingency
between light and tone had enhanced the ability of the light to engage the learning process
relative to the situation in which that contingency was maintained.
Given the current status of the hybrid model (and indeed the original PearceHall and
Mackintosh models) it is not possible to simulate this experiment: These theories are designed
to model the development of associations between CSs and USs, not between CSs and other
CSs (and it must be the difference in relationship between light and tone that generates the
pattern of results seen in Wilson et al.s, 1992, experiment). Nevertheless, it is possible to make
certain observations regarding this experiment from the point of view of the hybrid model.
During the first stage, experience of the consistent relationship between light and tone will
lead to a decline in the salience associability of the light, as it is an accurate predictor of the
events that follow it, but will allow the light to maintain a high attentional associability, as it is
the best available predictor of the tone. In the second stage, degrading the contingency
between light and tone will act to restore the salience associability of the light to some extent, as
it is no longer such an accurate predictor of the events that follow. It is, however, still the best
available predictor of the tone on trials on which light and tone occur (note that the tone never
occurs in the absence of the light), and hence it will still maintain a high throughout this
latter stage. Consequently at the outset of the test phase, for group consistent the light will
have a high and a low , while for group shift the light will have a similar but a higher (as
salience associability has been restored in this group). The result will be more rapid learning of
the lightfood relationship in group shift, as observed experimentally. The issue of the
outcome specificity of associability is also germane to this analysis. Recall Mackintoshs (1973)
learned irrelevance study, indicating that attentional associability is, to some extent, outcome
ASSOCIATIVE HISTORY
235
specific. Suppose, then, that the attentional associability developed by the light with respect to
the tone does not transfer to subsequent learning about the light with respect to food, as tone
and food are such dissimilar outcomes. This will attenuate any effect of the difference in light
tone relationship for the two groups on the of the light that is effective on lightfood trials. In
other words, even if shifting from a 100% to a 50% contingency between light and tone does
have an impact on the of the light with respect to tone, this will not be manifest in the rate of
learning of a lightfood association. And as long as we assume that the salience associability of a
stimulus is not outcome specific, but is instead a general property of that stimulus, we are still
able to account for the more rapid learning of lightfood for group shift.
All of these results taken in combination reveal the power of the hybrid model: By integrating two different approaches to changes in CS processing as a result of experience, the
model is able to reconcile a number of seemingly opposing demonstrations of the effects of
associative history on subsequent learning. Moreover, combining these associability-based
processes with mechanisms allowing for modulation of learning in terms of changes in
processing of the US extends the model still further. In general, then, the model can be broken
down as follows:
1. Attentional associability, (cf. Mackintosh, 1975). CS-processing, allows explanation
of learned irrelevance, Mackintosh and Turner (1971) reblocking, overtraining
reversal effect, ID shift advantage, etc.
2. Salience associability, (cf. Pearce & Hall, 1980). CS-processing, allows explanation of
latent inhibition, HallPearce negative transfer, and better learning, under certain conditions, when a stimulus is a poor predictor of following events (e.g., Wilson et al.,
1992).
3. Separable error term (1 VA +V A in Equation 10, 1 V A + VA in Equation 11); (cf. Bush
& Mosteller, 1951). US-processing, allows explanation of greater associative change for
poorer predictor of outcome as shown by Rescorla (2000, 2001, 2002).
4. Summed error term, |R| (cf. Rescorla & Wagner, 1972). US-processing, allows explanation of conditioned inhibition, overexpectation, superconditioning, and supernormal
conditioning.
Future developments of the hybrid model

Alternative or additional components
It should he noted that the precise formulation of the hybrid model offered here is relatively
unimportant. That is to say, there are a number of other ways in which these ideas of changes
in CS- and US-processing could be integrated. Moreover, the approach taken by the model is
not tied to the mechanisms for associability change outlined in this review. I have used the
Mackintosh and PearceHall models as components of the hybrid model to illustrate how
reconciliation could be achieved simply for convenience: A number of alternative component
models are available, and each may bring its own advantages. For example, it would be
perfectly possible to employ Wagners (1981) SOP model or McLaren et al.s (1989) elemental
model instead of the PearceHall theory to describe changes in salience associability. Either of
these components would allow the resulting hybrid model to account for the US-preexposure
effect (habituation), the context specificity of latent inhibition, and demonstrations of
236
LE PELLEY
perceptual learning (see Hall, 1991): All of these phenomena lie beyond the present implementation integrating Mackintosh and PearceHall models. That said, the use of a Pearce
Hall component allows the hybrid model to account for effects of predictive accuracy on
salience associability that lie beyond Wagners model. Similarly, the attention-shifting mechanisms employed by Kruschkes (2001) ADIT or mixture of experts models could be
adopted in place of the Mackintosh theory for modelling changes in attentional associability.
These models differ from Mackintoshs model in that attentional associability is updated
before associative strengths are adjusted. In addition, in Kruschkes models attention influences performance as well as learning. That is, while in Mackintoshs model responding to a
cue is simply given by the cues associative strength, in Kruschkes model responding to a cue
is modulated by the attention paid to that cue.
The purpose of the current paper is not to support a particular implementation of the
hybrid model over any other. Instead, my aim in presenting a hybrid model is to demonstrate
that a number of seemingly opposing effects of associative history on later learning need not be
seen as irreconcilable. While attempting to account for all of these effects with a single associability mechanism may prove impossible, if we abandon this extreme parsimony and accept
that these preparations may engage a number of different mechanisms, each of which plays its
own part in the learning process, a successful reconciliation may well be possible to achieve.
That said, the discussion above indicates that a clear direction for future development of the
hybrid model is in the identification and refinement of the most suitable components to use.
Should associative strengths and associabilities interact to determine performance? Should
the model employ a PearceHall module, a Wagner (1981) module, or separate modules
implementing the mechanisms of associability change suggested by both theories? This latter
suggestion raises the question of when the increased explanatory power offered by the hybrid
approach justifies the added complexity and reduction in parsimony that goes with it. Given
the conflicting empirical evidence presented in the current paper, the theoretical distinction
between attentional and salience associabilities employed in the current hybrid model seems
well justified on these grounds. Whether the available evidence demands further distinctions
within the domain of CS-processing mechanisms remains to be seen.
Novel predictions
This agnosticism with regard to the best possible implementation of the hybrid model
means that novel predictions derived from the current implementation should be regarded
with some caution. That is, a failure of the specific Mackintosh / PearceHall hybrid model
offered here need not undermine the idea of a hybrid model of associability processes in
general. For instance, the current implementation makes a novel prediction with regard to the
phenomenon of relative validity and related contingencies (Wagner, Logan, Haberlandt, &
Price, 1968; Wasserman, 1974). Consider the three training conditions below (all trials in each
condition are intermixed, +/ indicates partial reinforcementi.e., reinforcement on 50% of
presentations):
1. AX + BX
2. AX+/ BX+/
3. AX+ BX+
ASSOCIATIVE HISTORY
237
The question of interest is what effect this training has on the processing of cue X. In Conditions 2 and 3, X is the best available predictor of reinforcement (reinforcement sometimes
occurs in the absence of A or B, but never in the absence of X; as such the contingency between
X and US is higher than that between A and US or B and US). In Condition 1, A and B are
better predictors of the following events (reinforcement and nonreinforcement, respectively)
than is X. As a result, in Conditions 2 and 3 X will maintain a high attentional associability,
while in Condition 1 this will fall. What about salience associability? In Condition 3 all of A, B,
and X are consistently reinforced, and as such the salience associability of all three will fall
rapidly. In Condition 1, A and B are consistent predictors of following events, but the fact that
X is inconsistently reinforced will mean that the compounds are slower to become accurate
predictors than in Condition 3, and as such salience associability will fall more slowly in
Condition 3 than in Condition 1. Finally, partial reinforcement in Condition 2 will ensure that
the events following AX and BX compounds remain relatively poorly predicted throughout
training. Hence X will maintain a higher salience associability in this condition than in either
of the others.
According to the Mackintosh model, processing of X is solely a function of its attentional
associability. As such this model predicts greatest processing of X following training in Conditions 2 and 3, and less processing in Condition 1. The PearceHall model, relying solely on
salience associability, predicts greatest processing of X following training in Condition 2, less
in Condition 1, and less still in Condition 3. The hybrid model, on the other hand, employs
both attentional and salience associability to determine processing of X. As a result it predicts
that while a cue that is a better predictor of following events than are other presented cues will
maintain strong processing, stronger processing will be maintained if these following events
are surprising. This seemingly paradoxical situation is manifest in Condition 2: X is a better
predictor of following events than is either A or B, but these events are themselves surprising.
As such the hybrid model predicts greatest processing of X following Condition 2, less
following Condition 3 and less still following Condition 1 (the ordering of these latter conditions being determined by the greater importance afforded to attentional associability than
salience associabilitysee above). Hence the three theories considered (Mackintosh, Pearce
Hall, and hybrid model) make different predictions about the ordering of processing of X
following training in these three conditions. If some way of testing this processing of X
following training, independent of Xs current associative strength, could be found (perhaps
by training with a novel reinforcer or a reinforcer of increased magnitude), this novel prediction of the hybrid model would be open to test.
A further novel prediction of the hybrid model can be made with regard to the training
conditions shown in Table 2: The two conditions in this table bear similarities to the
TABLE 2
Basic design of a study to test a novel prediction of the hybrid
model
Condition
Stage 1
Stage 2
Stage 3
1
2
A us
A us
A us
AB us
AB US
AB US
Note: us = weak outcome; US = strong outcome.
238
LE PELLEY
experimental contingencies employed by Hall and Pearce (1979) and Mackintosh and Turner
(1971), respectively. The focus of this design is the relative magnitude of associative change
undergone by A and B during Stage 3. In Condition 1, A is trained as a consistent predictor of a
weak outcome in Stages 1 and 2. According to the Mackintosh model, this will allow A to
maintain a high attentional associability. In Stage 3 a compound of A and a novel cue, B, is
paired with a stronger outcome. Given As high attentional associability at the end of Stage 1,
the Mackintosh model is constrained to predict that the associative change undergone by A
will be at least as great as that undergone by B (which, as a novel cue, will have the starting
value of at the outset of Stage 3) on these AB US trials. In Condition 2, during Stage 2 B is
a poorer predictor of the weak outcome than is A. This should lead to a decline in B, while A
is maintained at a higher level. As a result, A should undergo greater associative change than B
on Stage 3 trials.
The PearceHall model makes different predictions. In Condition 1, consistent pairings of
A with the weak outcome will lead to a decline in As salience associability. Thus on the first
Stage 3 AB US trial (before the associabilities of A and B are updated) B will undergo
greater change than will A. Recall that a summed error term is used to determine salience associability in the PearceHall model (Equation 16). If = 1 (i.e., associability is dictated only by
events on the immediately preceding trial), then following this first trial A and B will have
equal associabilities, such that all subsequent changes in the associative strengths of A and B
will be equal. If < 1 then Bs advantage in terms of associability will persist for longer, gradually decreasing as the moving window average of associability moves along. So unlike the
Mackintosh model, PearceHall predicts if anything a greater associative change in B than in
A over Stage 3 as a whole (with smaller values of giving rise to greater differences between B
and A). In Condition 2, A will begin Stage 2 AB us trials with a lower salience associability
than B. Again bearing in mind the summed error term determining associability, if = 1 then
the salience associability of A and B will be equal following Stage 2, and hence both cues will
undergo identical associative change during Stage 3. If < 1, then Bs higher associability at
the outset of Stage 2 will persist for longer, such that B could begin Stage 3 with a higher
salience associability than A and hence will undergo greater associative change than A on
AB US trials.
To summarize, in Conditions 1 and 2 of Table 2 the Mackintosh model predicts. if
anything, VA > VB, while in both conditions the PearceHall model predicts, if anything,
VB > VA. The hybrid model, on the other hand, predicts different results for Conditions 1
and 2. Recall that in the hybrid model as presented here, the attentional associability of good
predictors does not rise greatly from its starting value. Consequently, in Condition 1, A and B
will begin Stage 3 with similar attentional associabilities, but the salience associability of A will
be lower than that of B (see discussion of PearceHall model with regard to this condition
above). Hence the hybrid model predicts VB > VA during Stage 3 of Condition 1. Just as for
the Mackintosh model as discussed above, Stage 2 AB us trials of Condition 2 will lead to a
rapid decline in B. While the salience associability of B may end Stage 2 slightly higher than
that of A (as for the PearceHall model), this will be easily outweighed by the difference in
attentional associabilities, with A > B. Accordingly the hybrid model predicts VA > VB
during Stage 3 of condition 2. So by combining Mackintosh and PearceHall components in a
single model, the hybrid model is able to predict a pattern of results that neither theory alone
could generate. Once again, if the associative change undergone by A and B during Stage 3
ASSOCIATIVE HISTORY
239
could be measured (perhaps using the technique developed by Rescorla, 2000), this novel
prediction would be open to test.
Associability and representation

Both Mackintosh and PearceHall models of CS processing employ elemental schemes of
representation. That is, a compound AB is viewed as being composed of separable A and B
elements, each of which has its own associability (and associative strength). This elemental
view of representation is also incorporated into the hybrid model as presented here. In recent
work, however, George and Pearce (1999; see also Oswald et al., 2001) have demonstrated that
an extreme elemental view of associability (in which each cue develops its own, independent,
associability) may not provide the most satisfactory approach to modelling empirical effects of
associative history on subsequent learning. Their experiment employed a biconditional
discrimination of the form AB+ CD+ AC BD, in which the outcome of a trial is not determined by the presence or absence of individual stimuli. According to the elemental view of
associability outlined above, this discrimination should result in a low attentional associability
to each of the four stimuli. Nevertheless, all four stimuli are relevant to the solution of the
discrimination in that they belong to configurations that are predictive of trial outcome; while
neither A nor B alone is predictive of outcome, for example, the AB configuration certainly
is. Contrary to the predictions of the elemental view George and Pearce (1999) found evidence
that training on a biconditional discrimination allowed stimuli AD to maintain a high
attentional associability. This finding instead supports the idea that configurations of stimuli
can develop their own associability, and this associability will then generalize to the constituent elements of that configuration. Thus it would seem that, at some level, associability is a
property of configurations, rather than (or, perhaps more likely, in addition to) individual
cues. Clearly these results are beyond the scope of the purely elemental hybrid model as it
currently stands. As such, future development of the hybrid model will need to address the
issue of the proper characterization of stimulus representation with regard to associability (see
Buhusi & Schmajuk, 1996; Kruschke, 1992; Pearce et al., 1998, for potential solutions to the
problem of configural associability that might guide future development of the hybrid
model in this regard).
CONCLUSION
Associative learning theory has come a long way since the early linear operator models proposed
by Bush and Mosteller (1951), Estes (1950), and their contemporaries. In developing new
models, the (very sensible) tendency has been to focus on one aspect of learning and build a
model centred on that aspect. For instance, the RescorlaWagner model focuses exclusively on
changes in US-processing, while the PearceHall model focuses exclusive on changes in CSprocessing. These models, while able to provide an account of phenomena within their relatively
narrow scope of expertise, fail to provide a full and satisfactory account of the varying effects of
associative history on the associative change undergone by cues on a given learning episode.
With the wealth of experimental evidence that has now built up on this subject, it may be time to
take the next step in the development of associative learning theories. That is, perhaps it is time
to begin integrating the domain-specific models of learning developed over the last 50 years in an
240
LE PELLEY
attempt to generate holistic models that are consequently able to capture more of the data. The
hybrid model presented here represents one effort to do just that.
REFERENCES
Baker, A. G., & Mackintosh, N. J. (1977). Excitatory and inhibitory conditioning following uncorrelated
presentations of CS and UCS. Animal Learning & Behavior, 5, 315319.
Baker, A. G., & Mackintosh, N. J. (1979). Preexposure to the CS alone, or CS and US uncorrelated: Latent inhibition,
blocking by context or learned irrelevance? Learning and Motivation, 10, 278294.
Bennett, C. H., Maldonado, A., & Mackintosh, N. J. (1995). Learned irrelevance is not the sum of exposure to CS and
US. Quarterly Journal of Experimental Psychology, 48B, 117128.
Bennett, C. H., Wills, S. J., Oakeshott, S. M., & Mackintosh, N. J. (2000). Is the context specificity of latent inhibition
a sufficient explanation of learned irrelevance? Quarterly Journal of Experimental Psychology, 53B, 239253.
Bonardi, C., & Hall, G. (1996). Learned irrelevance: No more than the sum of CS and US preexposure effects?
Journal of Experimental Psychology: Animal Behavior Processes, 22, 183191.
Bonardi, C., & Ong, S. Y. (2003). Learned irrelevance: A contemporary overview. Quarterly Journal of Experimental
Psychology, 56B, 8089.
Bouton, M. E. (1993). Context, time, and memory retrieval in the interference paradigms of Pavlovian learning.
Psychological Bulletin, 114, 8099.
Bouton, M. E. (1994). Conditioning, remembering, and forgetting. Journal of Experimental Psychology: Animal
Behavior Processes, 20, 219231.
Brandon, S. E., Vogel, E. H., & Wagner, A. R. (2003). Stimulus representation in SOP: I. Theoretical rationalization
and some implications. Behavioural Processes, 62, 525.
Buhusi, C. V., & Schmajuk, N. A. (1996). Attention, configuration and hippocampal function. Hippocampus, 6, 621
642.
Bush, R. R., & Mosteller, F. (1951). A mathematical model for simple learning. Psychological Review, 58, 313323.
Channell, S., & Hall, G. (1983). Contextual effects in latent inhibition with an appetitive conditioning procedure.
Animal Learning & Behavior, 11, 6774.
Cotton, M. M., Goodall, G., & Mackintosh, N. J. (1982). Inhibitory conditioning resulting from a reduction in the
magnitude of reinforcement. Quarterly Journal of Experimental Psychology, 34B, 163180.
De Houwer, J., & Beckers, T. (2002). A review of recent developments in research and theories on human
contingency learning. Quarterly Journal of Experimental Psychology, 55B, 289310.
Dickinson, A., Hall, G., & Mackintosh, N. J. (1976). Surprise and the attenuation of blocking. Journal of Experimental
Psychology: Animal Behavior Processes, 2, 313322.
Dickinson, A., & Mackintosh, N. J. (1979). Reinforcer specificity in the enhancement of conditioning by posttrial
surprise. Journal of Experimental Psychology: Animal Behavior Processes, 5, 162177.
Dickinson, A., Shanks, D. R., & Evenden, J. L. (1984). Judgement of actoutcome contingency: The role of selective
attribution. Quarterly Journal of Experimental Psychology, 36A, 2950.
Estes, W. K. (1950). Toward a statistical theory of learning. Psychological Review, 57, 94107.
George, D. N., & Pearce, J. M. (1999). Acquired distinctiveness is controlled by stimulus relevance not correlation
with reward. Journal of Experimental Psychology: Animal Behavior Processes, 25, 363373.
Hall, G. (1991). Perceptual and associative learning. Oxford: Oxford University Press.
Hall, G., & Minor, H. (1984). A search for contextstimulus associations in latent inhibition. Quarterly Journal of
Experimental Psychology, 36B, 145169.
Hall, G., & Pearce, J. M. (1979). Latent inhibition of a CS during CSUS pairings. Journal of Experimental
Psychology: Animal Behavior Processes, 3, 3142.
Hall, G., & Pearce, J. M. (1982). Restoring the associability of a pre-exposed CS by a surprising event. Quarterly
Journal of Experimental Psychology, 34B, 127140.
Hearst, E. (1972). Some persistent problems in the analysis of conditioned inhibition. In M. S. Halliday (Ed.),
Inhibition and learning (pp. 539). London: Academic Press.
Holland, P. C. (1984). Unblocking in Pavlovian appetitive conditioning. Journal of Experimental Psychology: Animal
ASSOCIATIVE HISTORY
241
Holland, P. C. (1988). Excitation and inhibition in unblocking. Journal of Experimental Psychology: Animal Behavior
Processes, 14, 261279.
Kamin, L. J. (1969). Predictability, surprise, attention and conditioning. In B. A. Campbell & R. M. Church (Eds.),
Punishment and aversive behavior. New York: Appleton-Century-Crofts.
Kaye, H., & Pearce, J. M. (1984). The strength of the orienting response during Pavlovian conditioning. Journal of
Experimental Psychology: Animal Behavior Processes, 10, 90109.
Kendler, T. S. (1971). Continuity theory and cue dominance. In J. T. Spence (Ed.), Essays in neobehaviorism: A
memorial volume to Kenneth W. Spence. New York: Appleton-Century-Crofts.
Khallad, Y., & Moore, J. (1996). Blocking, unblocking, and overexpectation in autoshaping with pigeons. Journal of
the Experimental Analysis of Behavior, 65, 575591.
Konorski, J. (1967). Integrative activity of the brain. Chicago: University of Chicago Press.
Kremer, E. F. (1978). The RescorlaWagner model: Losses of associative strength in compound conditioned stimuli.
Kruschke, J. K. (1992). ALCOVE: An exemplar-based connectionist model of category learning. Psychological
Review, 99, 2244.
Kruschke, J. K. (2001). Towards a unified model of attention in associative learning. Journal of Mathematical
Psychology, 45, 812863.
Kruschke, J. K., & Blair, N. J. (2000). Blocking and backward blocking involve learned inattention. Psychonomic
Bulletin & Review, 7, 636645.
Lattal, K. M., & Nakajima, S. (1998). Overexpectation in appetitive Pavlovian and instrumental conditioning. Animal
Learning & Behavior, 26, 351360.
Le Pelley, M. E., & McLaren, I. P. L. (2003). Learned associability and associative change in human causal learning.
Quarterly Journal of Experimental Psychology, 56B, 6879.
Lochmann, T., & Wills, A. J. (2003). Predictive history in an allergy prediction task. Proceedings of EuroCogSci 03:
The European Conference of the Cognitive Science Society (pp. 217222). Mahwah, NJ: Lawrence Erlbaum
Associates Inc.
Lovibond, P. F., Preston, G. C., & Mackintosh, N. J. (1984). Context specificity of conditioning and latent inhibition.
Lubow, R. E. (1989). Latent inhibition and conditioned attention theory. Cambridge, UK: Cambridge University Press.
Mackintosh, N. J. (1969). Further analysis of the overtraining reversal effect. Journal of Comparative and Physiological
Psychology, 67, No. 2, Part 2.
Mackintosh, N. J. (1973). Stimulus selection: Learning to ignore stimuli that predict no change in reinforcement. In
R. A. Hinde & J. S. Hinde (Eds.), Constraints on learning (pp. 7596). London: Academic Press.
Mackintosh, N. J. (1975). A theory of attention: Variations in the associability of stimuli with reinforcement.
Psychological Review, 82, 276298.
Mackintosh, N. J. (1978). Cognitive or associative theories of conditioning: Implications of an analysis of blocking. In
H. Fowler, W. K. Honig, & S. H. Pulse (Eds.), Cognitive processes in animal behavior (pp. 155175). Hillsdale, NJ:
Lawrence Erlbaum Associates, Inc.
Mackintosh, N. J., & Cotton, M. M. (1985). Conditioned inhibition from reinforcement reduction. In N. E. Spear
(Ed.), Information processing in animals: Conditioned inhibition. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Mackintosh, N. J., & Little, L. (1969). Intradimensional and extradimensional shift learning by pigeons. Psychonomic
Science, 14, 56.
Mackintosh, N. J., & Turner, C. (1971). Blocking as a function of novelty of CS and predictability of UCS. Quarterly
Journal of Experimental Psychology, 23, 359366.
Matzel, L. D., Schachtman, T. R., & Miller, R. R. (1988). Learned irrelevance exceeds the sum of CS-preexposure
and USpreexposure deficits. Journal of Experimental Psychology: Animal Behavior Processes, 14, 311319.
McLaren, I. P. L., Kaye. H., & Mackintosh, N. J. (1989). An associative theory of the representation of stimuli:
Applications to perceptual learning and latent inhibition. In R. G. M. Morris (Ed.), Parallel distributed processing:
Implications for psychology and neurobiology (pp. 102130). Oxford, UK: Oxford University Press.
Miller, R. R., & Matzel, L. D. (1988). The comparator hypothesis: A response rule for the expression of associations.
The Psychology of Learning and Motivation, 22, 5192.
Moore, J. W., & Stickney, K. J. (1985). Antiassociations: Conditioned inhibition in attentionalassociative networks.
In R. R. Miller & N. E. Spear (Eds.), Information processing in animals: Conditioned inhibition. Hillsdale, NJ:
Lawrence Erlbaum Associates, Inc.
242
LE PELLEY
Navarro, J. I., Hallam, S. C., Matzel, L. D., & Miller, R. R. (1989). Superconditioning and overshadowing. Learning
and Motivation, 20, 130152.
Oswald, C. J. P., Yee, B. K., Rawlins, J. N. P., Bannerman, D. B., Good, M., & Honey, R. C. (2001). Involvement of
the entorhinal cortex in a process of attentional modulation: Evidence from a novel variant of an IDS/EDS
procedure. Behavioral Neuroscience, 115, 841849.
Pearce, J. M. (1994). Similarity and discrimination: A selective review and a connectionist model. Psychological
Review, 101, 587607.
Pearce, J. M., & Bouton, M. E. (2001). Theories of associative learning in animals. Annual Review of Psychology, 52,
111139.
Pearce, J. M., George, D. N., & Redhead, E. S. (1998). The role of attention in the solution of conditional
discriminations. In N. A. Schmajuk & P. C. Holland (Eds.), Occasion setting: Associative learning and cognition in
animals. Washington, DC: American Psychological Association.
Pearce, J. M., & Hall, G. (1980). A model for Pavlovian conditioning: Variations in the effectiveness of conditioned
but not of unconditioned stimuli. Psychological Review, 87, 532552.
Pearce, J. M., Kaye, H., & Hall, G. (1981). Predictive accuracy and stimulus associability. In M. L. Commons, R. J.
Herrnstein, & A. R. Wagner (Eds.), Quantitative analyses of behavior: Acquisition (Vol. 3). Cambridge, MA:
Ballinger.
Pearce, J. M., Nicholas, D. J., & Dickinson, A. (1982). Loss of associability by a conditioned inhibitor. Quarterly
Pearce, J. M., & Redhead, E. S. (1995). Supernormal conditioning. Journal of Experimental Psychology: Animal
Prados, J., Redhead, E. S., & Pearce, J. M. (1999). Active preexposure enhances attention to the landmarks
surrounding a Morris swimming pool. Journal of Experimental Psychology: Animal Behavior Processes, 25, 451
460.
Randich, A., & LoLordo, V. M. (1979). Associative and non-associative theories of the UCS preexposure
phenomenon: Implications for Pavlovian conditioning. Psychological Bulletin, 5, 2528.
Redhead, E. S., Prados, J., & Pearce, J. M. (2001). The effects of pre-exposure on escape from a Morris pool. Quarterly
Reid, L. S. (1953). The development of noncontinuity behavior through continuity learning. Journal of Experimental
Psychology, 46, 107112.
Rescorla, R. A. (1969a). Conditioned inhibition of fear resulting from negative CSUS contingencies. Journal of
Comparative and Physiological Psychology, 67, 504509.
Rescorla, R. A. (1969b). Pavlovian conditioned inhibition. Psychological Bulletin, 72, 7794.
Rescorla, R. A. (1970). Reduction in the effectiveness of reinforcement after prior excitatory conditioning. Learning
and Motivation, 1, 372381.
Rescorla, R. A. (1971). Variations in the effectiveness of reinforcement and nonreinforcement following prior
inhibitory conditioning. Learning and Motivation, 2, 113123.
Rescorla, R. A. (2000). Associative changes in excitors and inhibitors differ when they are conditioned in compound.
Rescorla, R. A. (2001). Unequal associative changes when excitors and neutral stimuli are conditioned in compound.
Quarterly Journal of Experimental Psychology, 54B, 5368.
Rescorla, R. A. (2002). Effect of following an excitatory-inhibitory compound with an intermediate reinforcer.
Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: Variations in the effectiveness of
reinforcement and non-reinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: Current
research and theory (pp. 6499). New York: Appleton-Century-Crofts.
Roberts, A. C., Robbins, T. W., & Everitt, B. J. (1988). The effects of intradimensional and extradimensional shifts on
visual discrimination learning in humans and non-human primates. Quarterly Journal of Experimental Psychology,
40B, 321341.
Rodriguez, G., Lombas, S., & Alonso, G. (2002, March). Previous blocking trials impede conditioning of the added
element when US intensity is increased. Poster presented at the Sixth Associative Learning Symposium, Gregynog,
Wales.
ASSOCIATIVE HISTORY
243
Rosas, J. M., & Bouton, M. E. (1997). Additivity of the effects of retention interval and context change on latent
inhibition: Toward resolution of the context forgetting paradox. Journal of Experimental Psychology: Animal
Schmajuk, N. A., & Moore, J. W. (1985). Real-time attentional models for classical conditioning and the
hippocampus. Physiological Psychology, 13, 278290.
Schwartz, R. M., Schwartz, M., & Teas, R. C. (1971). Optional intradimensional and extradimensional shifts in the
rat. Journal of Comparative and Physiological Psychology, 77, 470475.
Seraganian, P. (1979). Extradimensional transfer in the easy-to-hard effect. Learning and Motivation, 10, 3957.
Shepp, B. E., & Schrier, A. M. (1969). Consecutive intradimensional and extradimensional shifts in monkeys. Journal
of Comparative and Physiological Psychology, 67, 199203.
Suret, M. B. S., & McLaren, I. P. L. (2003). Representation and discrimination on an artificial dimension. Quarterly
Sutherland, N. S., & Mackintosh, N. J. (1971). Mechanisms of animal discrimination learning. New York: Academic
Press.
Swan, J. A., & Pearce, J. M. (1988). The orienting response as an index of stimulus associability in rats. Journal of
Experimental Psychology: Animal Behavior Processes, 4, 292301.
Swartzentruber, D., & Bouton, M. E. (1986). Contextual control of negative transfer produced by prior CSUS
pairings. Learning and Motivation, 17, 366385.
Thorndike, E. L. (1898). Animal intelligence: An experimental study of the associative processes in animals.
Psychological Review Monograph Supplements (No. 8) [entire issue].
Trobalon, J. B., Miguelez, D., McLaren, I. P. L., & Mackintosh, N. J. (2003). Intradimensional and extradimensional
shifts in spatial learning. Journal of Experimental Psychology: Animal Behavior Processes, 29, 143152.
Turrisi, F. D., Shepp, B. E., & Eimas, P. D. (1969). Intra- and extra-dimensional shifts with constant- and variableirrelevant dimensions in the rat. Psychonomic Science, 14, 1920.
Wagner, A. R. (1971). Elementary associations. In J. T. Spence (Ed.), Essays in neobehaviorism: A memorial volume to
Kenneth W. Spence. New York: Appleton-Century-Crofts.
Wagner, A. R. (1981). SOP: A model of automatic memory processing in animal behaviour. In N. E. Spear & R. R.
Miller (Eds.), Information processing in animals: Memory mechanisms (pp. 547). Hillsdale, NJ: Lawrence Erlbaum
Associates, Inc.
Wagner, A. R. (2003). Context-sensitive elemental theory. Quarterly Journal of Experimental Psychology, 56B, 729.
Wagner, A. R., & Brandon, S. E. (2001). A componential theory of Pavlovian conditioning. In S. B. Klein (Ed.),
Handbook of contemporary learning theories. Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Wagner, A. R., Logan, F. A., Haberlandt, K., & Price, T. (1968). Stimulus selection in animal discrimination
learning. Journal of Experimental Psychology, 76, 171180.
Wagner, A. R., Mazur, J. E., Donegan, N. H., & Pfautz, P. L. (1980). Evaluation of blocking and conditioned
inhibition to a CS signalling a decrease in US intensity. Journal of Experimental Psychology: Animal Behavior
Processes, 6, 376385.
Wasserman, E. A. (1974). Stimulusreinforcer predictiveness and selective discrimination learning in pigeons.
Journal of Experimental Psychology, 103, 284297.
Whitney, L., & White, K. G. (1993). Dimensional shift and the transfer of attention. Quarterly Journal of Experimental
Psychology, 46B, 225252.
Williams, B. A., & McDevitt, M. A. (2002). Inhibition and superconditioning. Psychological Science, 13, 454459.
Wilson, P. N., Boumphrey, P., & Pearce, J. M. (1992). Restoration of the orienting response to a light by a change in its
predictive accuracy. Quarterly Journal of Experimental Psychology, 44B, 1736.
Original manuscript received 30 June 2003
Accepted revision received 1 September 2003

Le Pelley 2004 PDF

Caricato da

Informazioni sul documento

Descrizione originale:

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Le Pelley 2004 PDF

Caricato da

Copyright:

Formati disponibili

This article was downloaded by: [130.132.173.

The Quarterly Journal of Experimental Psychology

The role of associative history in models of

Cardiff University , Cardiff, UK

PLEASE SCROLL DOWN FOR ARTICLE

Q232738QJEP(B)03b14 / Jun 1, 04 (Tue)/ [51 pages 2 Tables 8 Figures 6 Footnotes 0 Appendices] .

The role of associative history in models of

Downloaded by [] at 11:18 06 November 2014

Cardiff University, Cardiff, UK

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

where VA is the associative strength of Cue A; A represents the associability of A and is a

The RescorlaWagner (1972) summed error term model

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

A model combining separable and summed error terms

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

The Mackintosh (1975) model

where nA is the change in associability of Cue A on trial n, n is the magnitude of the

Downloaded by [] at 11:18 06 November 2014

The extended Mackintosh model

Downloaded by [] at 11:18 06 November 2014

where E is a learning-rate parameter for excitatory learning.

Downloaded by [] at 11:18 06 November 2014

After updating the associative strengths on a given trial according to Equations 10 or 11

and when R < 0 the change in associability is given by:

Downloaded by [] at 11:18 06 November 2014

nA = I (||R n |VAn 1 + VAn 1|||R n |VXn 1 + VXn 1|)

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

conditioning refers to the observation that the effectiveness of excitatory conditioning is

Downloaded by [] at 11:18 06 November 2014

EVIDENCE FOR ASSOCIABILITY PROCESSES

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

by Dickinson, Hall, & Mackintosh, 1976). This unblocking as a result of an unexpected

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

The PearceHall (1980) model

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Human causal learning

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

Note: AY = foods; 14 = allergic reactions. Filler

The outcome specificity of learned associability

Downloaded by [] at 11:18 06 November 2014

RECONCILING EFFECTS OF ASSOCIATIVE HISTORY:

Downloaded by [] at 11:18 06 November 2014

Downloaded by [] at 11:18 06 November 2014

These constraints afford attentional associability a potentially greater importance than