Exp Brain Res (1996) 112:298-312
9 Springer-Verlag 1996
H e r t a F l o r . Niels B i r b a u m e r 9 Larry E. Roberts B e r n d Feige 9 Werner L u t z e n b e r g e r Christiane H e r m a n n 9 B r u n o K o p p
Slow potentials, event-related potentials, "gamma-band" activity, and motor responses during aversive conditioning in humans Received: 12 June 1995 / Accepted: 16 May 1996
Abstract We examined slow potentials, transient event-
related potentials, and oscillatory-like responses in the electroencephalogram during aversive conditioning in humans, in order to determine what is happening in the neocortex when behavioral adaptations are learned. Pictures of an angry and a happy human face served as reinforced (CS+) and unreinforced (CS-) conditioned stimuli, respectively, in one group, and either the reversed condition or two discriminably different neutral faces in two other groups (total n=48 subjects). The unconditioned stimulus (US) was intracutaneous shock delivered to the left hand 5 s after CS+ onset. The electroencephalographic (EEG) activity was recorded from Fz, Cz, Pz, C3, and C4, electromyographic (EMG) activity from bilateral forearm and corrugator muscles, and skin conductance from the right hand. During acquisition a negative slow potential developed after CS+ (not CS-), which was more pronounced when a neutral face served as CS+. Early (iCNV, initial contingent negative variation) H. Flor (~)1. B. Kopp Department of Psychology, Humboldt-University,Berlin, Germany N. Birbaumer 9W. Lutzenberger Institute of Medical Psychology and Behavioral Neurobiology, University of Ttibingen, Germany N. Birbaumer Dipartimento di Psicologia Generale, Universita Degli Studi, Padova, Italy L.E. Roberts Department of Psychology, McMaster University, Hamilton, Ontario, Canada B. Feige Institute of Experimental Audiology, University of Mtinster, Germany C. Hermann Department of Psychology, State University of New York, Albany, N.Y., USA Present address:
1Institut ffir Psychologie, Humboldt-Universit~itzu Berlin, Hausvogteiplatz 5-7, D- 10117 Berlin, Germany; Fax: +49-30-20377-308; e-mail:
[email protected]
and late (tCNV, terminal contingent negative variation) components of the slow-potential response were positively related to the magnitude of conditioned EMG responses. Differentiation of tCNV was larger when neutral faces signaled the US; iCNV persisted during extinction when a happy face served as CS+. Late-occurring event-related potentials (ERPs) elicited by the US diminished over conditioning, whereas short-latency US components and ERPs elicited by CS events did not. Fourier analysis revealed oscillatory ("gamma-band") activity between 30 and 40 Hz, which persisted up to 3 s after US delivery and diminished as conditioning progressed. Our findings indicate that learning is expressed in neocortical structures at the earliest stages of conditioning. The functional roles of the three types of EEG response in learning are discussed. Key w o r d s Pavlovian conditioning 9 Slow potentials Event-related potentials - Gamma-band activity 9 Skin conductance 9Motor conditioning - Cholinergic modulation. Human
Introduction Much is known from animal studies about the role of paleocortical and allocortical structures in the formation of discrete conditioned responses (CRs; Davis et al. 1991; Thompson and Krupa 1994) and about synaptic processes that appear to be involved in associative conditioning (Kandel 1991; Lynch and Granger 1992). Less is known, however, about the role of the neocortex in conditioning. Although several types of associative learning are demonstrable without a neocortex (Grau et al. 1990; Weinberger et al. 1990; Thompson and Krupa 1994), neocortical lesions have been reported to interfere with discriminative conditioning (Jarrell et al. 1987) and to prolong extinction when the signal value of a stimulus is changed (LeDoux 1995). Damage to the hippocampus, which is believed to deliver a training input to the neocortex, has also been reported to interfere with learning
299 when configural cues are used as conditioned stimuli (CS; Sutherland and Rudy 1989). These findings suggest that neocortical structures may be concerned more with perceptual and/or inhibitory aspects of learning than with stimulus-response adaptations, which are organized in phylogenetically older regions of the brain. Neocortical structures are informed about stimuli that are processed at lower levels of the neuraxis during conditioning (Thompson and Krupa 1994), which implies that significant functions are served by the neocortex in conditioning. In principle, the role of neocortical processes can be investigated by recording electroencephalographic (EEG) potentials from the scalp or surface of the brain during conditioning in humans and animals. Calculations based on biophysical models suggest that the EEG activity recorded from the scalp reflects the polarization of extracellular spaces in the neocortical laminae, particularly the plexiform layer and superficial pyramidal cells, which receive input from deeper cortical layers and from thalamic and basal forebrain structures (Braun et al. 1990). Measurement of scalp potentials and deep sources by indwelling electrodes also supports a neocortical source for EEG responses. Although brain potentials originating in subcortical or brain stem nuclei are volume-conducted to the scalp and are detectable in the electroencephalogram by averaging techniques, their amplitude is attentuated relative to the contribution of sources in the neocortical layers by a factor of 10 or more (Altafullah et al. 1986). Therefore, EEG potentials in the microvolt range are likely to arise from the activity of pyramidal cells in the neocortex and may provide a selective picture of neocortical dynamics during associative learning. Those dynamics appear to be generated by networks of neurons whose global activity may not be as easily described by recording from indwelling multipleor single-unit electrodes in animal or human surgical preparations. Recent animal experiments have confirmed that at least three categories of EEG response originate from the neocortex during associative learning. EEG responses of the first category, surface-negative slow potentials, have been differentially conditioned in several animal preparations and are attended by increases in the firing rate of a majority of cortical neurons (82%) that discriminate between CS+ and CS- trial types (Pirch et al. 1986). Slowpotential responses are also well developed in transcortical measurements, further confirming a neocortical basis for these responses (Hablitz 1973). Two components appear to be distinguishable in the conditioned slow-potential response: (1) an early component that increases gradually during training and is dominant bilaterally at frontal and central sites, and (2) a later component that becomes focussed frontally with overtraining (Skinner 197l) or contralaterally over motor cortex if a discrete motor response is induced by conditioning (Nakamura et al. 1993). Although in principle slow-potential responses could reflect a passive summation of excitatory postsynaptic potentials (EPSPs) elicited by task stimuli, most ac-
counts propose an active role for the brain sources of slow potentials in information processing (Birbaumer et al. 1990). The conditioned slow-potential responses of the rat neocortex are abolished ipsilaterally by cholinergic blockade or electrolytic lesion of the basal forebrain (nucleus basalis magnocellularis), indicating that these potentials depend on subcortical cholinergic mechanisms and are more than a nonspecific arousal effect (Pirch et al. 1986). Responses of the second EEG category, transient event-related potentials (ERPs), which are elicited by signaling and rewarding stimuli and are time-locked to these stimuli, also appear to originate from neocortical sources (Altafullah et al. 1986) that receive input from mechanisms residing more deeply in the brain. Although the specific properties of ERPs (amplitude, latency, morphology) depend on the particular stimulus contingencies used, there is reason to suggest that certain regularities may transcend different conditioning arrangements. For example, Nakamura et al. (1993) and Begleiter and Platz (1969) found that long-latency ERPs elicited by the CS (N200, P350) were affected more by conditioning than were short-latency ERPs (Ns0), presumably because longer latencies allowed greater opportunity for expression of neuromodulators and/or for analysis of the significance of task stimuli. Nakamura et al. (1993) also reported that long-latency ERPs elicited by the unconditioned stimulus (US) were more likely to change over conditioning than were long-latency ERPs elicited by CS events. The latter result could reflect differences in the predicitability of US and CS events from their preceding contexts, and/or changes in the behavioral responses elicited by these stimuli over the course of conditioning. However, contrary to this principle, Weinberger et al. (1990) recently documented short-latency (
CZ
0 . . . . .
.%...."
"'.....,...'"
GS10H
I
I
I
I
r
I
I
A
A
A
A
A
A
E
10
H
-20
I
I
I
I
I
I
I
I
A
A
A
A
A
A
E
E
E
E
E
~_0 PZ
PZ
-10
10
> ,.'..
0
..,'..
H
A
A
A
A
A
E
A
E
E
10 H
A
A
A
A
A
A
E
_>0
-20
C3
C3 10
-10
.. .............
>
...."" ..........................
./
=5-
0
10H
I
I
I
I
I
I
I
I
A
A
A
A
A
A
E
E
H
E
A
A
I
I
I
I
I
I
A
A
A
A
E
E
E
2O
-20
C4
C4 -10
..
".
.
::
> A ..."
" , . . ~ "9
j
10H
~. 9
-,...,...,.." I
I
I
A
A
A
I
"....~/
v
~
..... ".%..-"
.. ............... I
0
"'...
]
A A A Trial blocks
I
I
E
E
E
Fig. 5 Changes in the initial and terminal contingent negative variation (CNV) over habituation (H), acquisition (A), and extinction (E) for the three conditioning groups combined. Responses are measured from a pretrial baseline
10
H
I
I
I
I
I
I
I
I
A
A
A
A
A
A
E
E
E
Trial blocks
Oscillatory responses elicited by the CS and US Oscillatory responses were evaluated by Fourier analysis. It will be recalled that prior to this analysis we removed time-locked activity (see Fig. 4) from the data of individual subjects by subtracting their mean EEG activity from the EEG traces of successive trials, leaving a re-
307 CS-
CS+
Difference [CS+ - CS-]
R1
g o
20 15 10 5 0 -5 -10
~
20 15 10 5 0 -5 -10
R3
03 O
20 15 10 5 0 -5 -10 0
',.2.., 03 E r-o
1 ~
3
Block A1
0 78
~U 50
~
0
r
~x 20 E 15 10 5 >, 0 -5 -10
..Q O
0
n BlockA6
20 15 10 5 0 -5 -10
Is
-10 ~ .~,4
5 ~-"V----.,~,40 6
lime Is]
Fig. 6 Baseline-corrected spectral responses to CS+ and CS- in the first (top row) and last (bottom row) block of acquisition trials. The difference between contours for CS+ and CS- trials is given at the right. Zero on the abscissa denotes trial onset (slide projector operation). The two lines shown in the lower plane of each panel denote, consecutively, the appearance of differential CS+/CS- information at 1.6 s into the trial and US onset. Regions of spectral power exceeding P
