ArticlePDF Available

Abstract and Figures

Conditioned suppression is a useful technique for assessing whether subjects have learned a CS-US association, but it is difficult to use in humans because of the need for an aversive US. The purpose of this research was to develop a non-aversive procedure that would produce suppression. Subjects learned to press the space bar of a computer as part of a video game, but they had to stop pressing whenever a visual US appeared, or they would lose points. In Experiment 1, we used an A+/B- discrimination design: The US always followed Stimulus A and never followed Stimulus B. Although no information about the existence of CSs was given to the subjects, suppression ratio results showed a discrimination learning curve-that is, subjects learned to suppress responding in anticipation of the US when Stimulus A was present but not during the presentations of Stimulus B. Experiment 2 explored the potential of this preparation by using two different instruction sets and assessing post-experimental judgements of CS A and CS B in addition to suppression ratios. The results of these experiments suggest that conditioned suppression can be reliably and conveniently used in the human laboratory, providing a bridge between experiments on animal conditioning and experiments on human judgements of causality.
Content may be subject to copyright.
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 1996, 49B (3), 27 283
A Behavioural Preparation for the Study of
Huma n Pavlovian Con ditioning
Franc isco Arcediano, Nuria Ortega, and Helena Matute
Deusto University, Bilbao, Spain
Conditioned suppression is a useful technique for assessing whether subjects have learned a
CS± US association, but it is dif® cult to use in humans because of the need for an av ersive
US. The purpose of this research was to develop a non-aversive procedure that would
produce suppression. Subjects learned to press the space bar of a computer as part of a
video gam e, but they had to stop pressing whenever a visual US appeared, or they would lose
points. In Experiment 1, we used an A+/B
2
discrimination design: The US always followed
Stimulus A and never followed Stimulus B. Although no information about the existence of
CSs was given to the subjects, suppression ratio results showed a discrimination learning
curveÐ that is, subjects learned to suppress responding in anticipation of the US when
Stimulus A was present but not during the presentations of Stimulus B. Experiment 2
explored the potential of this preparation by using two different instruction sets and assess-
ing post-experimental judgements of CS A and CS B in addition to suppression ratios. The
resu lts of these experiments suggest that conditioned suppression can be reliably and con-
veniently used in the human laboratory, providing a bridge between experiments on animal
condition ing and experiments on human judgements of causality.
A general assuption in Pavlovian research is that common associative processes can
explain both animal and human learning. However, the majority of the data ava ilable
are based on non-human animals. There are no obvious theoretical reasons for this
strong preference for the use of animal subjects. Instead, the relatively small number
of human e xperiments seems to be due to the lack of convenient behavioural preparations
for use with humans.
The traditional techniques used in human Pavlovian research, such as electrodermal
conditioning (e.g. Hinchy, Lovibond, & Ter-Horst, 1995), have not become p opular
because they are expensive, complicated, and sometimes ethically problematic. Some
researchers have tried to solve the methodological problem in human research by focus-
Requests for reprints should be sent to Helena Matute, Departamento de Psicologõ
Â
a, Universidad de Deusto,
Apartado 1, 48080 Bilbao, Spain. Email: univem05@sarenet.es
Support for this research was provided by Grant PB91± 0288 from Direccio n General de Investigacio n
Cien ® ca y Te cnica (Spain). We are grateful to Angel Baquedano and Ralph Miller for helpful comments
and suggestions on a previous draft. We also wo uld like to thank Abraham Arias, Martha Escobar, and Oskar
PinenÄ o for running the subjects of Experiment 1, and Mirko Gerolin for assistance in writing the computer
program.
q
1996 The Experimental Psychology Society
ing on the idea that studies of human j udgements of causality can be viewed as analogues
of animal conditioning (e.g. Allan, 1993; Shanks & Dickinson, 1987; Wasserman, 1990a).
The basic idea underlying this approach is that what an animal learns during the pairings
of the conditioned stimulus (CS) and the unconditioned stimulus (US) in Pavlovian
conditioning is very similar to what a h uman learns during the pairings of a cause and
an effect during causal learning. In a typical study in the causal learning tradition, the
subject learns a relationship between the CS or potential cause (e.g. a ® ctitious allergen)
and the US or potential effect (e.g. a ® ctitious allergy), and this learning is usually
assessed through the subjects verbal judgement of the degree to which the target CS
is the cause of the US or effect (e.g. Wasserman, 1990b). The use of causal judgements as
a tool to explore associative learning in humans is becoming very popular because
assessing them is much simpler than assessing autonomic conditioning.
Although we agree with the overall assumption that causal judgement studies with
humans may be viewed as analogues of Pavlovian conditioning, we think that the analogy
would be more straightforward if behavioural rather t han judgemental responses were
used to assess conditioning in humans. Verbal judgements are known to be in¯ uenced by
many variables in addition to the critical ones under study. For example, very slight
differences in the wording of the test questions or in the names used for the ® ctitious
causes and effects in verbal causal preparations substantially in¯ uence the results (Matute,
Arcediano, & Miller, 1996). In our view, behavioural responses are more direct and
perhaps more ecologically valid estimates of whether subjects have learned the CS± US
relationship. Therefore, our goal was to develop a simple preparation in which beha-
vioural responses could be used to assess whether human subjects have learned an
association between a potential cause (CS) and a potential effect (US).
One of the most common t echniques used to assess Pavlovian conditioning in animals
is conditioned suppression (Estes & Skinner, 1 941). In one variant of this technique,
called on-line conditioned suppression, subjects are reinforced with food or water f or
barpressing while, at random intervals, they are exposed to a CS (e.g. light) that signals a
US (usually a footshock). The footshock US induces unconditioned suppression of bar-
pressing (i.e. rats s top barpressing d uring and brie¯ y after the footshock). When the
subjects learn that the CS will be followed by the footshock US, conditionedsuppression
is evidenced in that subjects also stop barpressing when the innocuous CS is presented.
Presumably, at some level they anticipate the US and, as a consequence, the suppression
response that initially accompanied the US is now also given in the presence of the CS.
This technique is common in animal research because it offers a reliable and convenient
means for the study of P avlovian conditioning.
Our goal was to adapt this preparation to the human laboratory to make it as similar as
possible to that used with animal subjects. However, many pilot experiments conducted in
our laboratory suggested that USs that were s uf® ciently ``mild to be ethically acceptable
in human research (e.g. loud noise of up to 112 dB A scale; electric shock wit h subjects
setting the intensity level) were not effective in causing unconditioned suppression
beyond the ® rst few trials. Other attempts to use conditioned suppression techniques
with human subjects have been made in the past (e.g. Di Giusto, Di Giusto, & King, 1974)
but then abandoned, presumably because of the ethical problems involved in exposing
HU M AN PA V LOV IA N C ON DITIO NIN G 271
human subjects to a US that would consistently cause unconditioned suppression over a
series of trials.
An interesting behavioural alternative that used an ethically acceptable US was devel-
oped long ago by Ivanov-Smolensky (1927). He instructed children to squeeze a ball
every time an innocuous US (a piece of chocolate) was visible. The ball was connected to
an apparatus that could me asure reaction time as well as the magnitude (i.e. force) of the
response. The interesting aspect o f the procedure was that after a series of trials, this
response transferred to a CS (a b ell) that was paired with the US. Ivanov-Smolensky
reported simple excitatory conditioning, as well as generalization, discrimination, extinc-
tion, and several other phenomena using this instructed-US preparation with children.
More recently, Perruchet (1985) has also developed a procedure that uses an instructed
US (either visual stimulation or an airpuff) with adult humans and keypress reaction time
as the dependent variable. Although we were not successful at obtaining sensitive r esults
with direct adaptations of t he instructed procedure to simple computer-based tasks, a
combination of the instructed procedure and the conditioned suppression preparation
used in animal research yielded what we think is a very sensitive and convenient prepara-
tion. Thus, the p reparation that we describe below, like the procedures described by
Ivanov-Smolensky (1927) and Perruchet (1985), uses an instructed US that is not
biologically signi® cant but is ethically acceptable in human research. Our US was a
visual stimulus that suppressed responding through prior explicit instructions telling
the subjects to s top responding when they saw the US. Our critical question was whether
subjects would learn to anticipate the US and would show this learning through condi-
tioned suppression to a CS that predicted the US.
For this purpose, we used a simple A+/B
2
discrimination procedure in which
Stimulus A was always followed by the US, and Stimulus B was never followed by the
US. In Experiment 1, subjects were not given any information concerning the existence
of these stimuli. We expected that subjects would learn to anticipate the US in the
presence of A but not in the presence of B, and our primary interest was whether this
learning would be manifested through conditioned suppression. In Experiment 2 we
explored whether adding instructions concerning the signi® cance of t he conditioned
stimuli would affect learning, and we also explored the correspondence between our
behavioural data and the more traditional causal judgement measurements.
EXPERIM ENT 1
M ethod
Subjects and Apparatus
Sixteen undergraduate non-psychology students volunteered for the study. Subjects were run
individually, using a personal computer, in a room measuring about 2
3
2 m. The experimenter
stayed in the room with the subject during the pre-training phase and until the subject understood
the instructions for the Pavlovian phase. After that, the experimenter stayed in an adjacent room.
272 AR CE DIAN O , OR TE G A, M A TU TE
Depe ndent Variable
The dependent variable was a suppression ratio, which is the conventional dependent variable in
animal conditioned suppression studies and is computed as X/(X+Y ), where X is the number of
barpresses during the CS, and Y is the number of barpresses in a period of time that is identical to the
durat ion of the CS and that immediately precedes the presentation of the CS. A suppression ratio of
0 indicates that conditioned suppression to the CS was complete. By contrast, if the CS does not
affect behaviour, th e number of responses during the CS and that during the period of time
imm ediately preceding the CS are identical; therefore, the resultant suppression ratio is 0.5 and
indicates that c onditioning has not taken place.
Procedure
Pre-traini ng. Subjects ® rst completed a pre-training phase. The purpose of this phase was to
teach each subject to barpress consistently using the space bar of the computer keyboard. No CSs o r
USs were presented at this stage. A translat ion of the instructions used for this phase reads as follows:
Your task is to prevent Ma rtians fromlanding. Every half-second one new Martian will try to land.
In order to destroy them, you must use your laser-gun ( the spa ce ba r) before they can see yo tha t
is, just before you can see them. But do not shoot too early, beca use you onl y ha ve one shot per
Martian. At the end of this phase, we will tell you the percenta ge of Ma rtians that you ha ve killed.
THE PLANET DEPENDS ON Y OU! ! DO NOT ALLOW THEM TO LAND! !
Marti ans were represented by ASCII 002 (
r
) in graph ics mode and were approximately 7 mm in
diameter. They appeared at intervals of 0.3 sec on a black background. A space of approximately 14
mm separated each Martian from the previous one. If subject p ressed the space bar before a Martian
was d isplayed, an explosion (represented by ASCII 015 [
U
] w ith the same format), rather than a
Marti an, was displayed. The objective was to have a screen ® lled with explosions, not with Marti ans.
Only one barpress per Martian was allowed, and it could be made at any time during the 0.3-sec
interval. The instruction that responding too early was not desirable was used to emphasize that only
one response per Martian was allowed, and to obtain, therefore, a regu lar rate of responding through
the task (as well as to prevent t he subjects from holding the space bar pressed down). If more than one
press was recorded during the 0.3-sec interval (e.g. if subjects held the space bar pressed dow n), a
Marti an rather than an e xplosion was printed. All this was shown by the experimenter, who
perform ed the ® rst few r esponses and then passed the keyboard to the subject. The screen was
® lled when 80 Martians or explosions were visible (10 in each of 8 lines), but no breaks between
screens were used. Instead, when the screen was ® lled, it moved up progressively to make room for
new Martians. One hundred Martians were depicted a s attempting to land during this p hase. The
score (percentage of explosions) was shown at the end.
Pav lovia n Conditioning. Following pre-training, the training phase was begun. This phase
consiste d of the critical Pavlovian conditioning, and it was superimposed on the operant task.
That is, the C Ss and USs were presented during the Martian task. A translation of the instructions
used for this phase reads as follows:
Now the Ma rtians have developed a powerful anti-laser shield. You must continue using your laser
to prevent their landing. BUT BE CAREFUL, because if you shoot your laser-gun when the shield
is connected, your shot will re¯ ect back to you, thousands of Ma rtians will land safely immedia tely,
HU M AN PA V LOV IA N C ON DITIO NIN G 273
and you will not be able to stop that inva sion. You will know tha t the shield is connected when you
see a WHITE INTERMITTENT FLASHING on the screen.
Remember, just one shot while the shield is connected ( WHI TE FLASHI NG SCREEN) and
the Martians will successful ly invade! !
The US was a white ¯ ashing screen (12 ¯ ashes/sec) presented for 0.4 sec if the subject did not
respond . But if the subject responded during the US, the ¯ ashes remained on for an additional 5-sec
period, during which Martians invaded the screen (at a rate of 20, rather than 10, Martians per line,
and a between-Martian space of 7 mm). The experimenter used a demonstration variant of the
program to show the subject the difference between a short (without punishment) and a long
(punishing) US (with no CSs being presented during this demonstration). Following this, the
experimenter left the room and the subject started the Pavlovian phase.
Now Martians appeared at 0.2-sec intervals. CSs (A and B) were blue and yellow backgrounds,
counterbalanced. CS A was always followed by the US; CS B was never followed by the US. There
were 20 trials for each CS. Their distribution was pseudorandom and identical for all subjects,
starting with a B trial, and with no more than 3 trials of the same type in succession. CSs duration
was 1 sec, but for every fourth A trial and every fourth B trial the CS duration was 3 sec. These 3-sec
trials were intended as test trials for which suppression ratios could be calculated. Thus, suppres sion
ratio was recorded for Trials 1, 4, 8, 12, 1 6, and 20 for each of the two CSs. The background colour
remained black during the inter-trial intervals (ITI). The duration of the ITI was pseudorandom,
with a range between 5 and 10 sec, and a mean of 7.5 sec.
Results and D iscussion
A learning curve for the discrimination t ask was observed through increased conditioned
suppression to A but not to B during the Pavlovian training phase. Figure 1 depicts the
learning curveÐ that is, mean conditioned suppression to CSs A and B on Trials 1, 4, 8 ,
12, 16, and 20, which were the longer trials that had been inserted for assessment
purposes during the discrimination training phase. Suppression to CS B was stronger
than suppression to CS A only on Trial 1, which probably r ected the surprise produced
by the very ® rst presentation of a CS (i.e. the sequence of trials always started with a B-
trial). A 2 (stimulus)
3
6 (test trial) ANOVA yielded a main effect for stimulus, F(1, 15) =
11.93, p < .01, and trial, F(5, 7 5) = 4 .63, p < .001. The Stimulus
3
Trial interaction was
also signi® cant, F(5, 75) = 19.14, p < .001, re¯ ecting gradual learning of the A+/B
2
discrimination. Planned comparisons showed that suppression to A was signi® cantly
stronger than suppression t o B on Trial 8, F(1, 15) = 8.88, p < .01, Trial 12, F(1, 15) =
16.84, p < .001; Trial 16, F(1, 15) = 21.32, p < .001; and Trial 20, F(1, 15) = 30.66,
p < .001. That i s, after 8 training trials with each CS, the suppression ratio showed that
subjects had learned to predict the occurrence of the US in the presence of A, as opposed
to B.
274 AR CE DIAN O , OR TE G A, M A TU TE
EXPERIM ENT 2
Experiment 1 showed that students who were trained individually learned to discriminate
between the stimulus that predicted the US and the one that did not, as evidenced
through a conditioned suppression learning curve. Experiment 2 was designed to replic-
ate and extend those results in several ways. First, in order to increase the potential of this
preparation, we thought it important to replicate the results in a group setting rather than
in the individual setting used in Experiment 1. Second, in order to explore the corres-
pondence between our behavioural data and more traditional judgemental measures of
associative learning in humans, in Experiment 2, in addition to suppression ratios, we also
assessed the subjects judgements of CS A a nd CS B once the Pavlovian phase was
completed. Finally, we explored t he effect of different instruction sets. That is, although
Experiment 1 showed t hat learning took place even though subjects had not been warned
of the existence of CSs and their potential relation to the US, it is possible that more
explicit instructions would make the task work better. Thus, in Experiment 2, half of the
subjects (uninstructed group) received instructions that were identical to the instructions
used in Experiment 1 (i.e. uninstructed with respect to CSs), and the other half of the
subjects received explicit instructions concerning the CSs (instructed group).
HU M AN PA V LOV IA N C ON DITIO NIN G 275
FIG. 1.
Discrimination learning curve for Stimuli A+ and B
2
during the Pavlovian phase of Experiment 1,
which was r un in dividually and in which no instructions on the existence of CSs were given. The dependent
variable was mean suppression ratio; thus, a lower value represents st ronger conditioning.
M ethod
Subjects and Apparatus
Forty non-psychology undergraduate students volunteered for the study. None of them had
participated in E xperiment 1. Subjects were randomly assigned in equal numbers to the instructed
and the uninstructe d group. The study was run in a group setting in a large laboratory containing 70
personal computers. Subjects were separated from each other by about 1 m, and each subject was
exposed to a different experimental condition (and counterbalancing of stimuli) from the two
subjects sitting next to him or her. Data from one subject from group instructed were eliminated
from the study due to equipment failure.
Procedure
Pre-traini ng. The pre-training phase was identical to that in Experiment 1 , except for the
changes required t o adapt it to the group setting used in this exper iment. After subjects had read
the instructions, one of the experimenters explained them while performing the pre-training task on
a computer connected to a large screen that could be seen by all subjects. Once the experimenter had
completed this phase, any questions were answered aloud, so that all subjects would have the same
inform ation. Finally, each subject completed pre-training, which ® nished with a screen showing the
percentage of Martians that subject had destroyed. Questions were allowed at this point, but none
were asked.
Pav lovia n Conditioning. Subjects were then asked to go on to the next screen, read it carefully,
and wait until all subjects had read it. This screen contained the instructions for the Pavlovian phase,
and subjects were told to pay no attention to their neighbours instructions, because each of them was
being instructed to do a different thing. Subjects in the uninstructed group received the same
instructions as those used in the Pavlovian phase in Experiment 1 (i.e. no instructions about the
CSs). Subjects in the instructed group received the same instructions, but with the following
paragrap h inserted between the ® rst and second paragraphs:
Some INDICATORS will help you predict when the shield is about to be connected, but there will
also be some false cues. If you learn to distinguish between the correct and the false indica tors, you
will always be able to a void the shield. Otherwise, each t ime the Ma rtians connect the shield, you will
be shooting, a nd they will invade you.
Once subjects had read the instructions, the experimenter performed a variant of the program in
which only two trials, containing one US each but no CSs, were shown. As in pre-training, the
experimenter’s screen could be seen by all subjects. The purpose of this demonstration was to show
subjects the difference between a large (punishing) an d a short (non-punishing) US (white ¯ ash).
Because subjects had been warned that there were many different instructions, the experimenter
explained that only the aspects common to all instructions were being explained during this demon-
strat ion and asked subjects to make sure they understood any additional details provided by their own
instructional screen. No questions were asked at this stage. After all subjects said they had under-
stood their own instructions, the experimenter told them the password that was needed to begin the
experiment. This password was needed s o that no one could start the experiment before the
demonstrat ion was ® nished. The critical Pavlovian training and testing was identical to that in
Experiment 1.
276 AR CE DIAN O , OR TE G A, M A TU TE
J udgemental Tests. The judgemental tests occurred with no interruption after subjects had
completed the latest Pavlovian trial. The following instru ction was then shown on a black background:
We will soon show you your score, but ® rst we would like to know if you have been able to distinguish
what the different colour s indica ted. That is, which ones warned you that the shield was about to be
connected and which ones were false cues. For this purpose we need to ask you a couple of questions.
Press <ENTER> to continue.
When subjects pressed the ``enter key, the background colour changed to either the A or the B
colour (counterbalanced for order), and the following question was printed in the middle of the
screen:
To what degree do you think that this colour wa s a n indica tor that the shield was about to be
connected? ( Please give a number between 0 [ none] and 9 [ very high] )
After subjects had responded to this question, the screen changed to the other colour (A or B),
and the same question was asked again. Finally, the next screen showed the percentage of Martians
the subject had destroyed, and subjects were carefully debriefed, with special emphasis on the
difference between the instructed and the uninstructed conditions, which might have resulted in
different degrees of task dif® culty for different subjects.
Results and D iscussion
The results of this experiment replicated those of Experiment 1 in a group rather than an
individual setting. Additionally, they showed that the instructed condition was even more
effective than the uninstructed one in producing a faster and more pronounced discrim-
ination, although good p erformance was also observed under the uninstructed condition
that replicated Experiment 1. Finally, the post-experimental judgemental questions
re¯ ected the same pattern of results as the behavioural data. That is, group instructed
was superior to group uninstructed, but both groups showed very good d iscrimination
through their judgements of CSs A and B.
Behavioura l Data. Figure 2 depicts the conditioned suppression discrimination
learning curves for group uninstructed (left panel) and group instructed (right panel).
Inspection of Figure 2 reveals that both groups showed good discrimination learning but
also that the discrimination was faster and more pronounced in group instructed
compared to group uninstructed. Consistent with t his, a 2 (group)
3
2 (stimulus)
3
6
(test trial) ANOVA yielded a main effect for stimulus, F(1, 37) = 35.03, p < .001, and trial,
F(5, 185) = 5.37, p < .001, and no main effect for group, p > .5. The Group
3
Trial,
Group
3
Stimulus, Stimulus
3
Trial, and Group
3
Stimulus
3
Trial interaction
F(5, 185) = 2.44, p < .05; F(1, 37) = 0.8, p > .1; F(5, 185) = 21.74, p < .001; F(5, 185)
= 2.18, p = .05, respectivelyÐ indicated that both groups learned progressively to
discriminate b etween A and B, and that the discrimination learning was faster and
more pronounced in the instructed group compared to the uninstructed group. Planned
comparisons showed that suppression to A differed sign cantly from suppression to B
as early as in Tr ial 4 for group instructed, F(1, 37) = 14.42, p < .001, but not for group
HU M AN PA V LOV IA N C ON DITIO NIN G 277
uninstructed, p > .1, in which the discrimination was signi® cant after Trial 8, F(1, 37)
= 11.74, p < .01. Thus, group uninstructed replicated the discrimination learning
observed in Experiment 1, and group instructed enhanced those results.
It should be noted, however, that although the statistical results of group uninstructed
replicated those of Experiment 1, the shape of the learning curve suggests some differ-
ences, particularly during the early trials. These differences should probably be attributed
to the individual versus group s etting, w hich was the only difference between Experiment
1 and the uninstructed group in this experiment (with the individual setting used in
Experiment 1 involving a greater experimente subject interaction during the
pre-training and instructional stages, as well as fewer distractions and a greater personal
involvement on the part of the subject during the actual experiment).
J udgemental Da ta. Figure 3 depicts subjects post-experimental judgements o f CSs A
and B. As can be seen in Figure 3, the judgemental data re¯ ected the same pattern of
results as the behavioural data. That is, both the uninstructed and instructed groups
showed very good discrimination through their post-experimental judgements of A and
B, although this discrimination was better in the instructed group. This was con® rmed by
a 2 (group)
3
2 (stimulus) ANOVA conducted on s ubjects’ judgements. This yielded a
main effect for stimulus, F(1, 37 ) = 113.05, p < .001, no main effect for group, F(1, 37) =
0.2, p > .5, and a sign cant Group
3
Stimulus interaction, F(1, 37) = 4.26, p < .05. This
suggests that although the difference between the post-experimental judgement of A and
that of B was more pronounced in group instructed than in group uninstructed, both
groups discriminated very well between A and B in their verbal ratingsÐ F(1, 37) =
278 AR CE DIAN O , OR TE G A, M A TU TE
FIG. 2.
Discrimination learning curves for Stimuli A+ and B
2
during the Pavlovian phase of Experiment 2,
which was run in a group setting. Group uninstructed (left panel) received the instructions used in Experiment
1. Group instructed (right panel) was given instructions on the existence of CSs. The dependent variable was
mean suppression ratio; thus, a lower value represents stronger conditioning.
37.67, p < .001 for the uninstructed group; F(1, 37) = 78.59, p < .001 for the instructed
group.
G EN ERA L DIS CUSSION
The results of Experiment 1 indicated t hat even though subjects had not been told that
CSs would be presented, they learned progressively to discriminate between the two CSs,
anticipating the US in the presence of CS A and n ot in the presence of CS B. Most
importantly, this learning was evidenced behaviourally through conditioned suppression.
Experiment 2 further explored the potential of this preparation by replicating the results
of Experiment 1 in a group setting and by showing that even better discrimination can be
obtained if explicit instructions on the existence of CSs are used. Additionally, Experi-
ment 2 showed a very high correspondence b etween our behavioural data and more
traditional causal judgement assessment procedures. This provided a reliable and very
convenient methodology for the study of human conditioning, which can also be used to
assess human causal judgements.
On the other hand, our observation of similar behavioural and judgemental patterns of
results could be interpreted as suggesting that the development of a behavioural prepara-
tion was an unnecessary complication over already es tablished causal judgement assessment
tools. However, there are at least two arguments against such conclusion. First, under the
very simple A+/B
2
discrimination procedure that we used, the similar patterns of results
yielded by the behavioural and the judgemental data is not surprising. Under more complex
designs, as mentioned in t he introduction, causal judgement studies are in¯ uenced by many
variables in addition to the critical ones under study. For example, the speci® c wording of
HU M AN PA V LOV IA N C ON DITIO NIN G 279
FIG. 3.
Post-experimental judgements of CSs A and B under the uninstructed and instructed conditions used
in Experiment 2 . Judgements were given in a 0± 9 scale and represent the degree to which subject expected the
US to occur after the CS.
the assessment question frequently in¯ uences the results in those studies (see, e.g., M atute
et al., 1996). A dditionally, and at a more theoretical level, many researchers have argued that
independent and dissociable learning systems may be responsible for verbal and perfor-
mance data. If this were true, data from experiments using verbal reports as the dependent
variable would provide evidence for a learning system that is not necessarily the same as that
involved in conditioning phenomena. Although this is very controversial (see Shanks & St.
John, 1994, for a review), that being the case, the development of a behavioural preparation
to be used with humans b ecomes potentially signi® cant.
Related to the debate on dissociable learning systems is the issue of learning without
awareness (see Shanks & St. John, 1994, for a recent review). For example, some
researchers have argued that although verbal reports cannot occur without the subjects
awareness o f the contingencies, behavioural responses can occur before t he subject is
aware of the contingencies or even in the absence of awareness (e.g. Lewicki & Hill,
1989; Reber, 1989). Our use o f instructed versus uninstructed conditions could perhaps
be interpreted in relation to this problem. The purpose of the present research, however,
was to develop a behavioural preparation, and, in our view, it should not be interpreted in
relation to the issue of learning without awareness. Although we did observe a very similar
pattern of results for the judgemental and the behavioural data, judgements were assessed
only after the Pavlovian phase had been completed, and thus our data do not s peak to the
question of whether conditioned responses (CRs) or awareness occurred ® rst. What our
results show is, on the one hand, that explicit instructions favour the development of CRs
as well as t he subjects awareness and, on the other hand, that instructions are not
necessary for CRs and awareness to occur: Subjects receiving no instructions on the
existence of CSs also learned to respond differentially to the two CSs and were aware
of the contingencies (at least at the end of training).
Our use of a US that is not biologically sign cant may be viewed as problematic by
researchers who focus on the Pavlovian d nition of conditioning, according to which the
US must be biologically signi® cant. Indeed, although some current accounts of Pavlovian
phenomena interpret conditioning as a mechanism for identifying c ausal relationship
between any events, regardless of their nature (e.g. Dickinson, 1980), according to
more traditional interpretations we m ight expect conditioning to occur only in situations
where the predicted event is biologically signi® cant (e.g. Pavlov, 1927). However, it should
be noted that sever al traditional Pavlovian procedures involve the use of US surrogates
that are not biologically signi® cant (e.g. sensory pre-conditioning and second-order
conditioning). Similarly, many accepted applications of Pavlovian research (e.g. system-
atic desensitization) involve the use of instructed USs (in this case, instructions to relax).
Thus, our procedure is c onsistent with a broader de® nition of conditioning, according to
which organisms learn a relationship between events, regardless of whether they are
neutral, as in sensory precondition ing, or biologically signi® cant, as in more traditional
forms of conditioning, and this learning is evidenced through a response that is given to a
CS that initially did not elicit the response.
It is true, however, that the relationships between these two types of learning (with and
without biologically signi® cant USs) have not yet been adequately established. Is it just a
matter o f different d nitions of conditioning, or do these two situations involve different
processes? In our view, most instances of conditioning involve two different components: (a)
280 AR CE DIAN O , OR TE G A, M A TU TE
the learning of a relationship between two events (see, e.g., Dickinson, 1980); and ( b) t he
second event almost always being biologically sign cant (e.g. Pavlov, 1927). The develop-
ment of a CR to the CS, which is the way in which conditioning is usually assessed, has been
regarded as a consequence either of the ® rst (e.g. Dickinson, 1980) or the second (e.g., Pavlov,
1927) component, bu t investigation of the separate role of each of them, or even of whether
those two co mponents can be explained through the same l earning process, did not start until
very recently. Recent studies suggest that t he biological signi® cance of the US could be a
source of some of the differences observed between animal conditioning (which generally
uses biologically signi® cant USs) and human causal learning research (which uses US
surrogates that are biologically neutral). For example, Miller and Matute (in press) reported
that several phenomena that had been observed in human causal learning b ut not in animal
conditioning (e.g. backward blocking) could be o btained in animals if they were treated
analogously to human subjects in terms of being exposed to biologically neutral rather
than biologically signi® cant USs (e.g. through the use of sensory p reconditioning). Thus,
the question of whether the same process is responsible for the acquisition of associations
between neutral events and for the acquisition of biological signi® cance by the CS is an open
but empirical one, which, in our view, should be addressed in future research.
With regard to the present research, o ne implication of t his is that phenomena that are
relevant with respect to current causal-learning interpretations of conditioning and that
were dif® cult to observe in conditioning research because of the use of biologically
signi® cant USs, should be easier to demonstrate using the present biologically neutral
behavioural preparation. The other implication is that the role of the biological signific-
ance component of conditioning cannot be investigated using the present preparation. On
the other hand, if the need to separate the role of both components of conditioning is
accepted, t he present preparation provides a unique opportunity to study the develop-
ment of associations between neutral events and the conditioning of behavioural
responses that are not mediated by the biological signi® cance of the US. Thus, the
suppression observed in the present research should be regarded as an index of whether
subjects have learned an a ssociation between two events, but not as an index of fear or of
the t ransfer of biological signi® cance to the CS.
Another aspect of our procedure also deserves some comment. As previously men-
tioned, ethical constraints i n the use of biologically s ign cant USs with humans prompted
us to use a US that was given motivational value through instructions and punishment. In
the present experiments, subjects were instructed to stop respondingÐ and were punished
if they respondedÐ in the presence of a visual US. Thus, it could be argued that the
present research dealt with punishment rather than with Pavlovian conditioning. Note,
however, that subjects were not punished for responding during the CS but during the US.
Pavlovian learning was evidenced in that subjects progressively suppressed responding,
not only in the presence of the US, but also in the presence of the CS that predicted it.
Admittedly, this can be interpreted as avoidance learning. However, it is the Pavlovian
component (prediction of the US) that we think was in need of a m ethodological devel-
opment. Although some studies in the human causal learning tradition had previously used
behavioural responses rather tha or in addition toÐ verbal judgements, to our knowl-
edge the majority of those studies had been concerned with instrumental rather than with
Pavlovian conditioning (e.g. Matute, 1995; Shanks & Dickinson, 1991; Wasserman, 1990a;
HU M AN PA V LOV IA N C ON DITIO NIN G 281
but see Perruchet, 1985, for a Pavlovian preparation). In our view, the present preparation
includes the necessary components for the study of Pavlovian conditioning in humans by
allowing simple manipulations of the CS± US contingencies a nd by making use o f a very
sensitive dependent variableÐ suppression ratioÐ in order to study how behavioural
responses are transferred to a CS as a function of such manipulations between CSs and
USs. This allows almost any Pavlovian design to be run in human subjects.
As mentioned in the Introduction, more traditional preparations for the study of
human Pavlovian conditioning have involved either causal judgements (e.g. Shanks &
Dickinson, 1987) or psychophysiological measures (e.g. Hinchy et al., 1995) as d epend-
ent variables. The latter offer a non-verbal assessment tool that is probably a more direct
estimate of conditioning than is causal judgement, but the former is widely used because
it is more convenient and simple. The conditioned suppression preparation that we
described is similar to techniques used in animal research and combines the higher
reliability of behavioural assessment with the simplicity in application that we think is
also important for the development of research.
REFERENCES
Allan, L.G. (1993). Human contingency judgments: Rule based or associative? Psychological Bulletin,
114 , 435± 448.
Dickinson, A. (1980). Contempora ry animal learning theory. Cambridge: Cambridge University Press.
Di G iusto, J.A., Di Giusto, E.L., & King, M.G. (1974). Heart rate and muscle tension correlates of
conditioned suppression in humans. J ournal of Experimenta l Psychology, 103, 515± 521.
Estes, W.K., & Skinner, B.F. (1941). Some quantitative properties of anxiety. J ournal of Experimental
Psychology, 29, 390± 400.
Hinchy, J., Lovibond, P.F., & Ter-Horst, K.M. (1995). Blocking in human electrodermal conditioning.
Qua rterly J ournal of Experimenta l Psychology, 48B, 2± 12.
Ivanov-Smolensky, G.A. ( 1927). On the methods o f examining the conditioned food re¯ exes in children
and in mental disorders. Brain, 5 0, 138± 141.
Lewicki, P., & Hill, T. (1989). On the status of nonconscious p rocesses in human cognition: Comment on
Reber. J ournal of Experimental Psychology: General, 118, 239± 241.
Matute, H. (1995). Human reactions to uncontrollable outcomes: Further evidence for superstitions
rather than helplessness. Qua rterly J ournal of Experimenta l Psychology, 48B, 142± 157.
Matute, H., Arcediano, F., & Miller, R.R. (1996). Test question modulates cue competition between
causes and between effects. J ournal of Experimental Psychology: Learning, Memory, and Cognition, 22,
182± 196.
Miller, R.R., & Matute, H. (in press). Biological signi® cance in forward and backward blocking:
Resolution of a discrepancy between animal conditioning and human causal judgment. J ournal of
Experimenta l Psychology: Genera l.
Pavlov, I. (1927) . Conditioned re¯ exes. London: Clarendon Press.
Perruchet, P. (1985). Expectancy for airpuff and conditioned eyeblinks in humans. Acta Psychologica, 58,
3 44.
Reber, A.S. (1989). Implicit learning and tacit knowledge. J ournal of Experimental Psychology: Genera l,
118 , 219± 235.
Shanks, D.R., & Dickinson, A. (1987). Associative accounts of causality judgment. In G.H. Bower (Ed.),
The psychology of learning a nd motiva tion (Vol. 21, pp. 22 261). San Diego, CA: Academic Press.
Shanks, D.R., & Dickinson, A. (1991). Instrumental judgment and performance under variations in
action-outcome conti ngency and contiguity. Memory & Cognition, 19, 353± 360.
Shanks, D.R., & St. John, M.F. (1994). Characteristics of dissociable human learning systems. Behavioral
and Brain Sciences, 17, 367± 447.
282 AR CE DIAN O , OR TE G A, M A TU TE
Wasserman, E.A. (1990a). Detecting response-outcome relations: Tow ard an understanding of the causal
texture of the environment. In G.H. Bower (Ed.), The psychology of learning and motiva tion (Vo l. 26,
pp. 27± 82). San Diego, CA: Academic Press.
Wasserman, E.A. (1990b). Attribution of causality to common an d distinctive elements of compound
stimuli. Psychological Science, 1, 29 302.
Manuscript received 8 August 1995
Accepted revision received 18 Ma rch 1996
Une proce
Â
dure com portementale pour le
Â
t ude du
conditionnem ent pa vlovien chez l’hom m e
La suppression conditionne e est une technique utile pour examiner si des sujets ont acquis
une association SC-SI (CS-US), mais elle est di cile a
Á
utiliser chez l’homme parce quelle
ne cessite l’emploi d un SI aversif. Le but de cette recherche e tait de mettre au point une
proce dure non-aversive induisant la suppression conditionne e. ApreÁ s avoir appris aÁ appuyer
sur la barre d espacement d un ordinateur dans le cadre dun jeu vide
Â
o, les sujets devaient
interr ompre lappui chaque fois quun SI vi suel apparaissait, au risque de perdre des points.
Une condition de descrimination A+/B
2
a e te utilise e lors de la premieÁ re