Behavioural Brain Research 107 (2000) 1–8
Research report
Associative learning in humans — conditioning of sensory-evoked
brain activity
W. Skrandies *, A. Jedynak
School of Medicine, Justus-Liebig Uni!ersity, Aulweg 129, 35392 Giessen, Germany
Received 9 March 1999; received in revised form 14 June 1999; accepted 14 June 1999
Abstract
A classical conditioning paradigm was employed in two experiments performed on 35 human volunteers. In nine subjects, the
presentation of Landolt rings (conditioned stimuli, CS+ ) was paired with an electric stimulus (unconditioned stimuli, UCS)
applied to the left median nerve. Neutral visual control stimuli were full circles (CS− ) that were not paired with the UCS. The
skin conductance response (SCR) was determined in a time interval of 5 s after onset of the visual stimuli, and it was measured
in the acquisition and test phase. Associative learning was reflected by a SCR occurring selectively with CS+ . The same
experiment was repeated with another group of 26 adults while electroencephalogram (EEG) was recorded from 30 electrodes. For
each subject, mean evoked potentials were computed. In 13 of the subjects, a conditioning paradigm was followed while the other
subjects served as the control group (non-contingent stimulation). There were somatosensory and visual brain activity evoked by
the stimuli. Conditioned components were identified by computing cross-correlation between evoked somatosensory components
and the averaged EEG. In the visual evoked brain activity, three components with mean latencies of 105.4, 183.2, and 360.3 ms
were analyzed. Somatosensory stimuli were followed by major components that occurred at mean latencies of 48.8, 132.5, 219.7,
294.8, and 374.2 ms latency after the shock. All components were analyzed in terms of latency, field strength, and topographic
characteristics, and were compared between groups and experimental conditions. Both visual and somatosensory brain activity
was significantly affected by classical conditioning. Our data illustrate how associative learning affects the topography of brain
electrical activity elicited by presentation of conditioned visual stimuli. © 2000 Elsevier Science B.V. All rights reserved.
Keywords: Brain electrical topography; Classical conditioning; Human learning; Somatosensory evoked potential (SEP); Visual evoked potential
(VEP)
www.elsevier.com/locate/bbr
1. Introduction
Non-invasive recordings of brain electrical activity
with a time resolution in the order of milliseconds in
response to external stimuli yield neuronal indicators of
human information processing [21]. Scalp potential
fields can be used to characterize human brain activity,
and components of evoked potentials have been shown
to be affected by physical stimulus characteristics [20],
and also by attention and task relevance of the stimuli
[19,26]. There are systematic changes in the scalp to-
pography of event-related brain activity during cogni-
tive processing, and with language stimuli, subtle
differences in semantic meaning are reflected in the
topographic pattern of brain electrical activity [22]. In
event-related potential studies, memory processes, as
well as learning to discriminate meaningful stimuli and
recognition of material, are followed by electrophysio-
logical changes. Landis et al. [15] have demonstrated
how learning to perceive a hidden figure in a noisy
background is accompanied by significant alterations in
visually evoked potentials. In a similar line, human
perceptual learning is associated with significant
changes in brain electrical activity: when normal adults
learn to improve their hyperacuity thresholds, changes
* Corresponding author. Tel.: +49-641-99-47-270; fax: +49-641-
99-47-279.
E-mail address: wolfgang.skrandies@physiologie.med.uni-giessen
.de (W. Skrandies)
0166-4328/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S0166 -4328 (99 )00096 -0

W. Skrandies, A. Jedynak / Beha!ioural Brain Research 107 (2000) 1–82
in evoked activity are observed [23,25]. Subjects who
succeed in learning to see 3-D structures in dynamic
random dot stereograms display significant differences
in evoked brain electrical activity before and after
learning [24]. In the present paper, we use the method
of topographic electroencephalogram (EEG) analysis in
order to detect neurophysiological descriptors of effects
of associative learning in healthy adult humans.
Classical conditioning of brain electrical activity has
a long tradition, especially in the Russian literature, as
is reviewed by Bechtereva [1]. Associative pairing of
electrical stimulation of visual areas with stimulation of
the motor cortex of cats yields conditioned responses
reflected in behavioral and electrophysiological mea-
sures [6]. In invasive studies on monkeys, it has also
been shown that classical conditioning systematically
changes visual and auditory elicited brain activity
[5,10]. In a similar line, the electrical responses of the
olfactory bulb of rabbits have extensively been studied
in classical conditioning experiments illustrating the
occurrence of specific spatio-temporal patterns associ-
ated with the recognition of odor [9].
From early work on the EEG of human subjects, we
know that the frequency content of spontaneous EEG
is influenced by conditioning and learning mechanisms.
Jasper and Shagass [11] and Knott and Henry [13] were
the first to show that alpha activity can be modified.
Later, in clinical applications of biofeedback in operant
conditioning experiments, it was demonstrated that
epileptic or paralyzed patients can learn to change the
rhythms of their EEG in order to control seizure activ-
ity [27,32] or to interact with their environment [14,28].
In a similar line, early studies also illustrated that brain
activity evoked by sensory stimuli can be systematically
influenced in classical conditioning paradigms [8,10].
Operant conditioning of visual evoked activity (e.g.,
[18]) or cortical slow waves [2] has also been reported.
Classical conditioning of the human visual system
was described by Wagner [30], who demonstrated that
visual stimuli can be used in a conditioning paradigm
for clinical purposes. The recording of electrodermal
activity showed that the contingent pairing of Landolt
rings with electrical shocks is followed by a conditioned
skin conductance response (SCR). These experiments
had been performed on monkeys as well as on human
patients, and Wagner [30] tried to use SCR signals as
indicators for visual acuity of patients.
In the present paper, we employ a similar experimen-
tal paradigm in order to test how electrical brain re-
sponses are influenced by associative learning. In one
experiment, it will be shown that pairing visual stimuli
with electrical peripheral nerve stimulation establishes a
conditioned electrodermal reaction measured as SCR.
In a series of topographical multichannel EEG experi-
ments performed on a different group of 26 subjects, we
will then illustrate how electrical brain activation is
modulated by associative learning.
With classical conditioning and associative learning,
a new functional connection between brain regions
responsive to the conditioned (CS+ ) and uncondi-
tioned stimuli (UCS) is established. We investigate
whether such a relationship can be demonstrated in
humans by scalp recordings of brain electric activity. As
mentioned above, several studies on operant condition-
ing had demonstrated that feedback on brain signals
can be applied in order to modify electrical activity
measured as spontaneous EEG or in evoked potentials.
So far, however, there has been only little systematic
research on classical conditioning of brain activity elic-
ited by sensory stimuli in humans [4]. The experimental
design of the present study involves the presentation of
visual and somatosensory stimuli, which yield visual-
evoked or somatosensory-evoked potentials, respec-
tively (VEP and SEP). Landolt rings were paired with
electrical stimuli in an acquisition phase, and brain
activity could be compared to that obtained in a non-
contingent test phase. Comparison of electrophysiologi-
cal effects seen in an experimental group and in a
control group in which no contingent stimulation oc-
curred allowed further to characterize functional corre-
lates of associative learning in the human brain. In
these topographical experiments, we will illustrate how
successful conditioning changes the neural responses
caused by a previously neutral visual stimulus as there
occur different patterns of activation in brain regions
conventionally subserving somatosensory stimuli.
2. Materials and methods
A differential conditioning design with a random
control group was used. Stimuli were presented in two
different phases to each group of subjects, in an acqui-
sition phase and a test phase. Electrical stimuli applied
to the median nerve occurred during the acquisition
phase but not during the test phase.
2.1. Stimuli
For visual stimulation, Landolt rings were designed
according to DIN 58220 with gaps pointing to one of
four different orientations. These rings or full rings
were presented in the center of a computer monitor for
350 ms and yielded VEPs. At a viewing distance of
150 cm, all stimuli subtended 1°; mean luminance was
9 cd/m2, and a contrast of 95% was used.
Stimulation of the median nerve of the left arm was
used as UCS that elicited SEPs. Bipolar stimulation
electrodes were placed with a distance of about 1 cm on
the inner side of the lower forearm over the median
nerve. Stimulus duration was 2 ms; the intensity was
determined individually for each subject, just below the
motor threshold of the thumb.