Cortical neural evoked correlates of somatosensory stimulus detection in the rhesus monkey

78

Electroencephalography and Clinical Neurophysiology, 1982, 53:78--93 Elsevier/North-Holland Scientific Publishers, Ltd.

CORTICAL NEURAL EVOKED CORRELATES OF SOMATOSENSORY STIMULUS DETECTION IN THE RHESUS MONKEY 1 ALBERT T. KULICS

Neurobiology Program, Northeastern Ohio Universities College of Medicine, Rootstown, Ohio 44272 (U.S.A.) (Accepted for publication: September 11,1981)

A powerful strategy for elucidating the neural correlates of sensory function is the combination of both psychophysical and neurophysiologic methods within a single experimental framework (Sutton 1969; Libet 1973; Uttal 1973; Vaughan and Ritter 1973; Mountcastle 1976). By utilizing such an approach which permitted simultaneous observation of neural evoked responses and psychophysical reports in monkeys, we found that the amplitude of a late evoked response component, at approximately 50--100 msec poststimulus, correlated with somatosensory psychophysical performance measures (Kulics 1977; Kulics et al. 1977). The psychophysical model used to interpret the results suggested that somatosensory evoked potential (SEP) amplitude could reflect neural activity important for somatic sensation in monkey. Because of demonstrated similarities in performance of monkey and man in this psychophysical paradigm (Kulics and Lineberry 1977), it appeared possible that the experimental approach developed for monkeys would be directly applicable to the study of the neural correlates of somatosensory experience in man.

lobe, correlates with reports of conscious somatosensory experiences in normal and brain-damaged humans (Giblin 1964; Libet et al. 1967; Williamson et al. 1970; Tsumoto et al. 1973). Relatively little is known of the origin or distribution of these late components of the somatosensory evoked potential in monkey (however, see Arezzo et al. 1981). Since these may have significance for understanding somatic sensation the following experiment was performed. The cortical neural evoked responses to electrical cutaneous stimuli at intensities near the limits of detection and at successively more detectable levels were studied in the performing monkey. This was done to establish the generality of previous findings over a continuum of psychophysical performance, in the context of both faint and strong sensory signals. In addition, neural evoked activity was recorded from a number of sites simultaneously to observe how activity is distributed over postcentral gyms and to establish whether correlations between evoked response amplitude and psychophysical performance are apparent at a number of locations.

This view is supported by previous demonstrations that the appearance of late-occurring evoked components, originating in parietal

Methods

Subiects 1 This research was supported by Grant BNS 79-24952 from the National Science Foundation.

Three adult male rhesus monkeys (Macaca mulatta) designated F, L and B served as

0013-4649/82/0000--0000/$02.75 © 1982 Elsevier/North-Holland Scientific Publishers, Ltd.

SEP CORRELATES OF STIMULUS DETECTION IN MONKEY experimental subjects. They were trained and tested in a primate chair designed for temporary restraint (Carlson 1972) and were returned after each daily session to their home cages.

Apparatus Experimentation was conducted within a ventilated wooden test chamber ( 1 . 8 m X 1.8 m X 1.5 m) containing a loudspeaker for white noise presentation and a television camera for closed-circuit monitoring of the subjects' behavior. A constant-current stimulator presented the discriminative stimuli which were single, anodal, constant-current, 0.5 msec duration pulses of variable intensity. These were delivered via a stainless-steel electrode to the 6 m m diam. center hole of a Scholl's no. 327 corn pad; the hole having been filled previously with electrode paste after the pad had been attached to the thenar eminence of the restrained left hand. The intensities employed in the present study were well below those observed to elicit pain in man (Gibson 1968) or escape behavior in m o n k e y (Lineberry and Kulics 1978). A large (6 cm X 4 cm) metal plate served as the indifferent electrode which was attached to the shaved right leg. AC shock was delivered to the right leg by two screw heads embedded in plastic and separated by 2.5 cm. A circular plastic plate (5 cm diam.) connected to a microswitch served as a discriminative response manipulandum and was placed within easy reach of the subject's right hand. Solid state logic circuitry in conjunction with a paper-tape reader controlled stimulus presentation and reinforcement contingencies and recorded behavioral responses and response latency. Bipolar, concentric electrodes were used for neurophysiologic recordings during behavioral test sessions. These consisted of a 0.25 mm diam. Teflon-coated stainless-steel inner wire with a 1 m m bared tip, within 22gauge stainless-steel hypodermic tubing; the tip and barrel separation being 4 mm. Neuro-

79

physiologic activity was amplified and tape recorded (8-track FM tape recorder). Analysis of recorded neural activity was performed by a digital signal analyzer.

Procedure Behavioral

training. The monkeys were brought to the test chamber daily, placed in the chair, and trained to perform conditioned avoidance responses to the electrical cutaneous stimuli. A trial consisted of the presentation of a white warning light within the darkened test chamber 2 sec prior to stimulus delivery. A bar press within 1.5 sec following stimulus presentation (hit) prevented the occurrence of a noxious AC shock to the right leg. No response to a stimulus presentation was termed a miss. Responses on trials when no stimulus was presented (false alarms) also resulted in an AC shock. When consistent avoidance response levels were established, a blue light was used to signal errors at the end of a trial. Shock was programmed to occur only after 2 or 3 errors in succession. This manipulation (i.e., substituting a conditioned cue for shock reinforcement) stabilizes behavior, and as a result, performance measures become consistent over time. After reliable performance was established, stimulus intensity was gradually reduced by 0.2 mA decrements over a further course of training until response probability dropped to approximately 0.10. This stimulus intensity represented the lowest used in subsequent training and testing. Following this, 3 stimulus intensities ($3, $2, $1) separated by 0.2 mA increments ($1 being the lowest intensity) and a null stimulus (So) were chosen for further training. Probability of presentation equalled 0.25 for each stimulus condition. Surgical preparation. After training, 12 cortical recording electrodes were implanted under aseptic conditions in each m o n k e y ; the six located in the region of right postcentral gyms were pertinent to this report. Each subject was anesthetized (20--30 mg/kg phenobarbital i.v.) and placed in a stereotaxic apparatus. At a predetermined location (AP +12;

80

L 20) a hole was drilled in the skull to define the center of the array of postcentral gyrus electrodes (site 1) for each subject. Five other holes were then drilled in a cluster surrounding this central point. With the aid of a dissecting microscope and electrode carrier, the electrode tips were driven through holes made in the dura into underlying cortex until the electrode barrel just contacted the dura. The electrodes were cemented into place with dental acrylic. The lead wires from the tips and barrels were fed to a 25-pin connector which was also affixed to the skull with dental acrylic. Training and testing resumed after a minimum of 2 weeks recovery. Neural response recording. Differential recordings (bandpass-- 1--1000 Hz), between the tip and barrel, were made from the 6 postcentral gyrus electrodes simultaneously while the subjects were performing the detection task. Neurophysiologic data were collected during two sessions of 400 trials per day; 100 trials each of $3, $2, $1, and So in a constrained random order. $1, $2, and $3 equaled 0.8, 1.0 and 1.2 mA, respectively for subjects F and B, and 0.6, 0.8 and 1.0 mA for L. During another two sessions, stimulus intensity was increased by 0.2 mA for each stimulus condition. Hits, false alarms and misses were recorded along with response latency for each trial. Histology. Upon completion of data collection the m o n k e y s were deeply anesthetized and marking lesions were made by passing a 1 mA current for 1 sec between the tip and barrel of each implanted electrode. The subjects were perfused through the heart with physiologic saline followed b y a fixative consisting of formalin (10%) and giutaraldehyde (1.25%) in a phosphate buffer. Brains were removed, photographed and blocked for sectioning. Serial frozen sections of 33 pm were made through the entire region of electrode implants and stained with cresyl violet (L and B) or thionin iF). Data interpretation. The theoretical model used to interpret neurophysiologic and behavioral results in the present study is illustrated

A.T. KULICS d"

/

.....(), 0

/

I

i

il /i

'~i ,,

!

', / ')

Xso Xs,

Xsks Cx

Magnitude of Sensory E v e n t ( X )

Fig. 1. The theoretical model employed to interpret neurophysiologic and behavioral results. The 4 hypothetical distributions represent probability density functions of sensory event (X) magnitude elicited by $3, $2, S1 stimuli and a null stimulus, So. Cx represents the subject's response criterion. The d' measures, based on hit and false alarm proportions, are thought to reflect the difference between the means

of the signal (X--s3, X--s., Xsl) and noise (Xs0) distributions in S.D. units."

in Fig. 1. It is an extension of that reported in previous publications (Kulics 1977; Kulics and Lineberry 1977; Kulics et al. 1977; Lineberry and Kulics 1978) which is derived from signal detection theory (Green and Swets 1966; Swets 1973). The presentation of stimuli ($3, $2, $1) elicit sensations (Xs3, Xs2, Xs,) or sensory events which are assumed to be statistically distributed along a sensory dimension, e.g., a dimension of sensory magnitude when the stimuli differ along the intensive dimension, as in the present study. Degree of overlap or 'confusion' determines h o w reliably, on the average, stimuli can be detected relative to background noise (Xs0) or differentiated from one another. The subject faced with the detection task is thought to a d o p t a response criterion or 'cut-off' point (Cx) along the intensive dimension such that whenever an elicited subjective sensation exceeds the criterion sensory magnitude, a response is emitted. Hit and false alarm proportions are thought to reflect the proportions of X that exceeded the Cx sensory magnitude. Esti-

SEP CORRELATES OF STIMULUS DETECTION IN MONKEY

mates of sensory sensitivity or detectability, d', which are obtained from the response proportions reflect the relative difference, in standard deviation units, between the means of the underlyin_g hy__pothetical distributions of sensations (Xs3, Xs 2, Xsx) and the occasions when no stimulus was presented

of these neural events is u n k n o w n at this time. However, a working hypothesis employed in this and past investigations is that the amplitude of specific features of the somatosensory evoked response could be a reflection of the magnitude of neural events necessary for somatic sensation. Similarly, psychophysical responses under controlled conditions may reflect subjective sensory event magnitude. Thus the demonstration of orderly functional relationships between the amplitude of specific SEP peaks and behavioral response latency could serve as the basis for the tentative identification of neural events which underlie somatic sensation. In the present study, an inverse monotonic relationship between neural and behavioral measures would be of greatest interest.

(Xs0). Response latency is incorporated within this explanatory framework by assuming that the faster a hit or false alarm is emitted, the greater the degree by which X exceeded the Cx magnitude. Thus the subjective magnitude of a sensation is assumed to control both the choice made and the latency of this choice. Therefore, hit and false alarm response latency distributions for each stimulus condition should reflect the proportions of sensations and the relative degree by which these sensations exceeded the criterion sensory magnitude. The assumption, which links the physiologic and psychologic response domains, is that neural events give rise to these sensations within the central nervous system. The nature

Results

Psychophysical performance Human subjects report that these electrical cutaneous stimuli produce discrete painless L

F 1.0

d"

1 . 0- -~ l

S3 So~2-5 S~- So=0"3

81

t S3

d" S~-S°>4"2

~

P

B _~ ~ ~

~

1.o

SI~So=3.1

d" S3 s°=3.O Sl- So= 1.1

~ ~..

¢~

S~

Sz S (~ 0.5

0.5

0.5

1 i

JJ: o SEC.

Fig. 2. Bar press latency distributions plotted as cumulative probabilities for two combined sessions at the same intensity levels (S~, $2, Sz equaling 1.2, 1.0 and 0.8 mA, respectively) for each monkey subject. At the top left portion o f each panel are the d' values derived from the cumulative probabilities of hits to $3, $2, and $1 and false alarms to So at 1.5 sec (i.e., the end of trial interval).

82

A.T. KULICS

touch sensations over a broad intensity range (Gibson 1968 and personal observations). Fig. 2 shows the cumulative probability of hits and false alarms of increasing latency for each m o n k e y to the same stimulus intensities ($3, $2, $1 = 1.2, 1.0, 0.8 mA, respectively). Fig. 2 demonstrates that response probability increases as a function of stimulus intensity but not at the same rate for each subject. Variability in response bias among subjects tends to mask similarities in sensory sensitivity

(Green and Swets 1966; Swets 1973). A biasfree estimate of the detectability of the 0.8 mA ($1) stimulus was obtained for each subject by subtracting the z-score for false alarm probability from the z-score for hit rate probability at the 0.8 mA intensity (see Kulics 1977; Kulics et al. 1977; Lineberry and Kulics 1978; for discussion of this procedure). The resultant difference, in standard deviation units, is the measure of sensory sensitivity or detectability, d'. For F, L and B, d'

\

i I / /

F

\\ ! I I

\ \

L

B

Fig. 3. Photographs of postcentral gyms region which contain the penetration sites (nos. 1--6) of the recording electrodes. Landmarks indicated are central (C) and intraparietal (IP) sulci.

SEP CORRELATES OF STIMULUS DETECTION IN MONKEY

Histology

equaled 0.29, 3.11 and 1.14, respectively, suggesting that substantial differences in sensitivity are apparent at this stimulus intensity even after the effects of response bias have been accounted for. Detectability of other intensities was determined by the same method. These d' indices for each subject also appear in Fig. 2. The results of Fig. 2 in the context of other behavioral results for these subjects suggested that the lower limit for stimulus detection is near 0.6 mA for L and near 0.8 mA for F and B.

F

83

Fig. 3 illustrates the electrode entry sites on the cortical surface of the 3 subjects. For subject F, all 6 penetrations are in postcentral gyms. For L, sites 1, 3, 4 and 5 are within postcentral gyms. For B, penetrations 1, 3 and 4 entered postcentral gyrus while 5 and 6 entered near intraparietal and central sulci, respectively. Microscopic inspection of stained serial sections containing electrode tracks revealed that electrodes at sites 1 through 5 for F; 1, 3, 4, 5 and 6 for L; 1 and

B

L q

1 +

i

3

4

i

i i

i i

4

5 ,i i

5

6

i

6

2

I

0

.1

L,l,ll[,IJI .2 .3

I I L , l , I , .1; l l L I

0

.2

.3

[,

0

L[LILILILI .2

.3

SEC.

Fig. 4. Average (n = 100) SEPs obtained during performance to the same stimulus intensity (1.2 mA). Relative surface negativity is upward. SEP amplitude is scaled in S.D. units of ECoG background amplitude (see text) at that site; the bars around each wave form representing +1 S.D. The vertical dashed isolatency lines are taken from N1 and P2 peaks at site 1 for each subject.

84 4 for B were transcortical since the tips rested in underlying white matter or in deepest cortical layers. Electrode tips at sites 6 for F, 2 for L, and 2, 3 and 6 for B were located in more superficial rather than deep cortical layers. The track of penetration 5 for B could not be accurately traced because of cortical damage during implant removal.

Neural recordings The average SEP elicited by 1.2 mA stimuli at each of the 6 electrode sites in or near postcentral gyms is illustrated in Fig. 4 for each subject during discrimination performance. SEP amplitude is scaled in units of the estimated standard deviation (S.D.) of the electrocorticogram (ECoG) amplitude obtained at each respective site in the absence of any stimulus. This background ECoG amplitude ranged from +100 to -+200 pV at the various sites. The S.D. for each site was determined by calculating the square root of the average variance of every fourth millisecond of a 256 msec post-stimulus interval (N = 64) for the So trials of a session. The SEPs obtained during sensory performance fell into two groups (Fig. 4). The first group, comprised of wave forms obtained at sites 1 through 5 for F; 1 and 3 through 6 for L; and 1, 4 and 5 for B, represents the most c o m m o n type of response encountered. This group of recordings was obtained from electrodes judged to be transcortical (except for number 5 of B). The prominent features are a relative cortical-surface negativity (N1) in the range of 50--65 msec followed by a relative positivity (P2) at 105--130 msec post. stimulus. Early positivities at approximately 12 msec (P1) and 20 msec are apparent at sites 1 and 5 for F and 1, 4 and 6 for L. Other small early positivities such as those seen at sites 3 and 5 o f L and 4 of B do n o t exceed 1 S.D. of the background ECoG. In general, the latency to N1 peak was shorter at those sites which also displayed the most prominent P1 response (i.e., sites 1 and 5 for F, 1 for L). With the exception of 5 for F, N1 peak latency was consistently

A.T. KULICS longer at sites surrounding site 1 for each monkey. These latency differences averaged 6 msec and were reliably observed in oscilloscope traces on a trial-by-trial basis. The second grouping o f responses are those seen at site 6 for F, 2 for L, and 2, 3 and 6 for B. These display an early relative cortical surface positivity between 30 and 60 msec followed by negative (100--120 msec) and positive (220--240 msec) fluctuations. At sites 2, 3 and 6 for B, SEP amplitude barely exceeds + 1 S.D. of the background activity at

F d/ S 3 - 2.2

-1--

S 2 - 1.4

0-

S ~ - 0,2

+1--

S 3 - 3.4 S 2 - 3.1 S~ - 0 . 9

Sz-3.2 $2-2.6 S t - 1.2

J 0

,

[

~ I

,

r

J

f .2

~ j

,

i .3

SEC.

Fig. 5. Average (n = 100) SEPs to $3 (solid line); S 2 (dashed line); and Sj (dotted line) stimuli at site 1 in sessions where $1 elicited a relatively low response probability ($3, $2, and SI equaled 1.2, 1.0 and 0.8 m A for F and B; and 1.0, 0.8 and 0.6 m A for L). The ordinate is scaled as in Fig. 4. The d' values are derived from hit and false alarm probabilities during the same sessions.

SEP CORRELATES OF STIMULUS DETECTION IN MONKEY

85

tudes both decrease; the change in detectability being roughly proportional to changes observed in SEP amplitudes for the same stimuli. A comparison between SEP amplitude at site 1 and psychophysical performance was made on a trial-by-trial basis. The individual response trials were digitized and peak negative and positive amplitudes were measured within -+5 msec of peak latencies observed for the average SEP of the entire group. N1 and P2 peak amplitudes were scaled relative to the S.D. of ECoG background activity and plotted as a function of behavioral response latency (Fig. 6). Misses were assigned a latency value of >1.5 sec, the end of trial interval. P1 peaks were not measured since they could not be reliably observed on these individual trials. The horizontal parallel lines through the scatter plots represent median amplitude of N1 and P2 for trials on which misses were observed. Although there is a great deal of amplitude variability on single trials, it is evident that on trials where hits were

each site. Since such small amplitude responses cannot be observed reliably, activity at these sites was not considered for further analysis. The negativities observed at site 6 for F ( N l 1 6 ) and 2 for L ( N l 1 4 ) temporarily coincide with P2 observed simultaneously at other sites, while the early positive (p60) peak latency for F is similar to N1 peak latencies at other sites. Histology showed that the electrode tips for these t w o sites contacted surface rather than deep cortical layers. This finding, together with other evidence presented later, suggests that the t w o responses which appear anomalous may in fact be polarity-reversed versions of SEPs comprising the first group. Neural-behavioral correlates The average SEP to $3, $2, and $1 at site 1 for each subject is presented for one session in Fig. 5. The d' values calculated during that test period are also displayed so that a comparison between psychophysical performance and SEP amplitude can be made. The measure d' decreases as N1 and P2 peak ampli-

F

B

10 •° 8-

• ~. •

•°°

:

=•

• ,•

6

! o• o•°

•••

•

.r.

| :':.

.*

• °.'..

4-

°. ••

° ••

•

• •

.! •

s•

•

•

4,

2-

u~

o

•

-~

• • o.

E

2

.

.:•

.

..

•

".% ~ • "•

• °•

~,"

.. 4-

• ,"•

t

•"

."~

•

°,

~

•.

I

• g -.

*.

o•

~

I,.

i

0.;

6

:.

8

10 ]

F

I

[

.5

I

!

F

E

I

1.0

I

F

I

I

]

1.5

I

I

I

F

]

.5

I

I

I

I

I

1.0

I

I

I

I

I

15

I

I

I

I

I

,5

I

I

I

I

I

1.0

I

]

]

I

I

1.5

SEC.

Fig• 6• Frequency distributions of N1 (top panels) and 92 (bottom panels) peak amplitude at site I as a function of hit latency and misses (at 1.5 sec) to St stimuli• $1 equaled 1.0 mA for F and B; 0.7 mA for L. Dotted parallel lines represent median amplitude of SEPs in miss category. Note that the SEP amplitudes to hits tend to exceed the miss category median amplitude.

86

A.T. KULICS

-

s

i

_

[1

~

'/

I-

-,-<

fast

,--'~

o-ko.t

,,Y i:

-

I

--

I

i

-,-

o-

JM

k.. 2

-

..--

s2 ,.'.Y

~

//~',

0

slow hit

..2/``

\~

I.

hit

fast hit

....:7<

fast hit

.1

.2

.3

SEC.

Fig. 7. SEP amplitude associated with different behavioral response tendencies of subject F. $3, $2 and Sl equaled 1.4, 1.2 and 1.0 mA, respectively. The solid trace represents the mean SEP for all trials at that intensity.

observed, the N1 and P2 peak amplitude tends to exceed the median peak amplitude associated with the miss category. Chi-square tests were conducted to test whether the probability of larger amplitude responses to the hit category was greater than e x p e c t e d b y chance. In all cases, tests were significant b e y o n d the 0.001 level. Response distributions from F suggest that peak amplitude may

be inversely related to response latency. The behavioral response latency distributions from two sessions for each subject were subdivided into broad response classes representing hits of short, medium, and long latency and misses to each stimulus condition. An a t t e m p t was made to keep the number of behavioral events in each class to 25 + 5, which proved possible in most cases. The SEPs associated with each behavioral response class were averaged and scaled in S.D. units of their respective ECoG amplitude. Fig. 7 illustrates the change in SEP amplitude associated with the different response classes for one subject (F). Average amplitude for both N1 and P2 for fast hit categories exceeds the grand mean amplitude for all trials at that intensity {solid trace) and is less than the grand mean for slow hit and miss categories. In contrast, peak latency changes comparatively little across stimulus intensities and response classes. To illustrate the relationship between SEP amplitude and behavioral response latency more clearly, peak amplitude of N1 at site 1 was plotted as a function of the median response, latency of each behavioral class (Fig. 8). The miss groups were assigned the value > 1.5 sec, i.e., the end of trial interval. Filled symbols represent the grand mean peak amplitude and median response latency observed for a particular stimulus condition while open symbols of the same kind represent the group mean amplitude associated with hit or miss groups of a particular latency class at that intensity. For all subjects, average peak amplitude decreases as a function of increasing median behavioral response latency at all stimulus intensities. In some cases, particularly sessions represented in the top panel for F and the b o t t o m panels for L and B, a single monotonic function would closely fit most data points. P2 at site 1 together with N1 and P2 at surrounding sites behaved similarly to N1 at site 1. To demonstrate this essential similarity among sites, correlations were made between the site 1 sample mean amplitudes associated

SEP CORRELATES OF STIMULUS DETECTION IN MONKEY

F

87

B

L

8-

~-S

7-

%.

6-

3

m-s,

0\

,&-S 1

O

A\ "D\ \ \\\

5--

"~Z?

_ O

\\

4-

\\

3-

2-

1 --

¢5 v~

0

$3=1"4mA

$3:1.2mA

$2=1.2

$2=1.0,,

S~=1.2

S~:0.8,,

S~: 1.0 "

,'

S~ :1.0 ,,

I

I

I

I

I

T

1

[

I

I

[

I

I

I

I

I

I

I

I

i

I

]

r

I

!

I

I

I

l

,,

1

l

l

l

l

l

r

i

i

i

7-

6O\ \

5-

\O D\\

q

%

4

3-

m ~ :

~ m

2-

$3=1"2 rnA _

\\

S2= 1.0 ,,

1-

S~ =0.8

0

I

I

$ 2 = 0 . 8 ,, 8~0.6

,,

I

LJ

I

1 1 1 1 0 1 1 1 1 1 . 1 5 1 .

S3=1.2 m A

S3:I.0mA ~ ~ ~ ~

,

S2=1m0 ,,

- _ _ ~

~ ~

S~;0.8 ' '

- ~

I i j 1 1 1 1 1 1 [ I ] 1 1.0 1 . 5

_ _

O-~--~ ~ ~

15

I I I I I I I.5 I F 1 . 0

1 1 1 1 1 !

15

SEC.

Fig. 8. The change in N1 average peak amplitude as a function of the median latency of behavioral response groups in two sessions each (upper and lower panels) for the 3 m o n k e y subjects. Filled symbols represent grand mean SEP amplitude and median latency at a particular intensity while open symbols o f the same kind represent sample mean amplitude associated with hit or miss groups o f different response latency at that intensity. Miss groups are assigned a median latency value o f 1.5 sec. Dashed lines connecting open symbols with their respective closed ones express the degree o f departure o f sample means from the grand mean.

with each behavioral group and the respective mean amplitudes obtained simultaneously at the other sites. The number of samples from each site for 4 sessions equaled 38 for F and L, and 41 for B. Table I summarizes the results of the correlation analysis. The correlation coefficients (r) between site 1 and surrounding sites for N1 amplitude are high (left column section); in most cases, a linear function would account for 80% or more of the variance. The slope (a) and intercept (b) of the best-fitting linear function appear alongside r. The intercepts are within

+1 S.D. of zero. Thus, given N1 peak amplitude at site 1, N1 amplitude at surrounding sites may be accurately predicted on the basis of the slope of the relationship alone. Large r values were also observed between P2 amplitude at site 1 and surrounding sites (middle column section). Thus, the amplitude of analogous SEP components at different sites stand in constant ratio to one another. In addition, the successive N1 and P2 components at each site are strongly correlated (right column section). Earlier, it was suggested that the SEP at

. .

N1

1

B .

N1

N1 .

3

4

5 6

2

. . N1 N1 N1 N1

N1

1 2 3 4 5 6

L

. N1 N1 N1 N1 P60

N1

1 2 3 4 5 6

y

F

x

Site

Subj.

. .

.

0.9338

.

. .

0.8358 . .

. .

. . 0.8201 0.9280 0.9270 0.7002

. 0.9671 0.9436 0.9699 0.9898 --0.9500

r

.

. .

.

0.41 .

0.86

. .

0.46 0.55 0.50 0.22

0.57 0.55 0.71 1.20 0.68

a

+0.36 .

--0.08

.

+0.37 +0.78 +0.51 +0.72

--0.24 --0.45 --0.42 +0.16 --0.05

b

.

P2

P2

P2

x

P2

P2

--

-Nl14 P2 P2 P2 P2

. P2 P2 P2 P2 Nl16

y .

0.8652 0.8296

0.78 1.07

--

-1.30 0.95 1.18 0.79 0.58

---0.8861 0.7979 0.9214 0.9131 0.8165 --

0.43 0.64 0.78 1.12 1.10

a

. . 0.9196 0.9540 0.9769 0.9870 --0.9722

r

+0.20 --0.35

--

---0.22 +0.41 +0.34 +0.02 +1.25

+0.15 --0.43 --0.60 --0.02 +0.07

b

N1 N1

N1

N1 . N1 N1 N1 N1

N1 N1 N1 N1 N1 P60

x

.

P2 P2

P2

P2 P2 P2 P2

P2

P2 P2 P2 P2 P2 Nl16

y

.

0.86 0.61 0.92 0.91 0.79 1.25

a

--0.8100 --0.8106

--0.7252

0.48 1.29

0.44

--0.9083 0.62 . . --0.7304 1.07 --0.8157 1.22 --0.7681 0.84 --0.6159 0.95

--0.9608 --0.8532 --0.8969 --0.9293 --0.9540 --0.8854

r

+0.87 +0.16

+1.06

+0.36 --0.48 --0.19 +1.25

--0.20

+0.26 +0.53 +0.34 +0.11 +0.17 +0.58

b

T h e c o r r e l a t i o n c o e f f i c i e n t s (r) a n d b e s t - f i t t i n g l i n e a r f u n c t i o n s ( y = ax + b) d e s c r i b i n g t h e r e l a t i o n s h i p b e t w e e n SEP c o m p o n e n t a m p l i t u d e a t v a r i o u s r e c o r d i n g sites.

TABLE I

c~

>

00 0o

SEP CORRELATES OF STIMULUS DETECTION IN MONKEY site 6 for F, and site 2 for L are polarityinverted relative to recordings from other sites (Fig. 4). F o r F, P60 and N l 1 6 peak amplitude changed as a function of median behavioral response latency much like that observed in Fig. 8. This was also the case for N l 1 4 for L. In contrast, P34 amplitude did n o t decrease as a function of behavioral group latency.

Discussion

The present study demonstrates the feasibility of studying neural events in behaving monkey at stimulus intensities on a continuum from easily detectable to nearly undetectable. The sequence of cortical electrophysiologic events to such stimulation proved reliable and reproducible. At an intensity (1.2 mA) which was presumably clearly detectable, i.e., one which elicited consistent behavioral responses from all subjects (Fig. 2), a characteristic negative-positive complex was observed at 13 of a possible 18 cortical locations (Fig. 4). Twelve of the 13 electrodes were judged to be transcortical. In two subjects iF and L) at the center (no. 1) site of their electrode clusters, a small (1 S.D.) relative positivity (P1) was seen at approximately 12 msec (Fig. 4). This response temporally coincides with the familiar primary evoked response of somatosensory cortex which is usually recorded in a monopolar m o d e from the cortical surface, and which signals sensory input to the cortical region representing the portion of the b o d y stimulated (Woolsey et al. 1942; Amassian et al. 1964; Arezzo et al. 1981). In all likelihood, that which is seen here is the primary response recorded transcortically. The transcortical SEPs closely resemble the epidural recordings of Arezzo et al. (1981) from postcentral gyrus of the awake rhesus m o n k e y . Their P12, N45, and P l l 0 components which appear analogous to P1, N1 and P2 respectively, were shown to be generated within postcentral gyrus exclusively.

89

The absence of a larger primary response at specific cortical locations is probably n o t indicative of a failure to place electrodes within a region responsive to stimulation of the hand. When stimulus intensity was increased b e y o n d reported levels (i.e., to 2.5 mA) P1 responses became prominent at central sites (e.g., 4 and 9 S.D. above the ECoG at site 1 for F and L, respectively), and to a lesser degree, at surrounding sites. Except for site 5 of F, the shortest latency N1 response was observed at site 1 for each subject (Fig. 4); this difference in peak latency between site 1 and surrounding sites ranging from 4 to 9 msec. The large correlations between N1 peak amplitude at site 1 and surrounding sites (Table I) demonstrates their strong interdependency. The most likely explanation for both the latency difference and amplitude correlations is that electrophysiologic activity underlying N1 is propagated from some single source outward, resulting in the activation of a succession of concentric cortical regions more distantly removed from the focus. From the point of view of an external observer, this propagation would appear as a 'traveling wave' of activity over postcentral gyms. The average interelectrode distance of 4 mm and average latency difference of 6 msec suggest a propagation velocity of 0.67 m/sec. Such a relatively long conduction time leaves room for a number of possible explanations of the mechanism of propagation. For L, the shortest latency N1 appears to lie within the perimeter defined by the electrode cluster, suggesting that the generation of the earliest electrophysiologic events leading to N1 occurs within this region for this subject. For F, the shortest latency N1 occurs slightly more laterally on postcentral gyrus at site 5, which is indicative of a more lateral placement for the generator of earliest activity. The fact that the most prominent P1 response occurs in conjunction with the shorter latency N1 responses for F and L may have significance for understanding the generation of late responses. Some studies indicate

90 that the primary cortical response may initiate all subsequent evoked activity in m o n k e y (Eidelberg and Jenkins 1966) and man (Stohr and Goldring 1969; Williamson et al. 1970) although differences of opinion exist (e.g., Albe-Fessard and Besson 1973). The former interpretation might suggest that the most prominent P1 and the shortest latency N1 should be seen at contiguous loci, which appeared to be the case in the present study. If the significance of N1 for somatic sensation can be more firmly established, further study of these components simultaneously could lead to an understanding of where the earliest cortical somatosensory information processing, events occur in the behaving monkey. With regard to P2, its peak amplitude is strongly correlated with N1 peak amplitude as revealed in Table I. The work of Arezzo et al. (1981) suggests that the sources of N1 and P2 are within postcentral gyms and largely overlap. Beyond this, little is known concerning the mechanisms of generation of these late responses or whether they reflect the activity of a single neural generator. The presumed polarity reversal observed with two electrodes whose tips lay superficially rather than subcorticaUy is congruent with the above evidence and with the dipole model proposed to explain the location of cortical generators of evoked activity (e.g., Amassian et al. 1964). At one stage of the experiment simultaneous comparisons were made of the differential transcortical recording from a site with 'monopolar' recordings of the tip and barrel against a distant reference (i.e., a large screw at the occiput). These revealed that activity at the tip alone accounted for most of the activity observed in the transcortical recordings. The simplest interpretation o f N l 1 4 and N l 1 6 for L and F, respectively, is that they are analogous to P2, b u t because the electrode tips lie on the opposite side of a dipole source oriented from surface to depth in cerebral cortex, the activity recorded is reversed in polarity relative

A.T. KULICS to other P2 recordings. The same argument suggests that P60 of F at site 6 is analogous to N1. The fact that these 3 peaks correlate with behavioral response latency in much the same fashion as N1 and P2 further supports this interpretation. In contrast, the latency of P34 together with no observed correlation of this peak with behavior argues against it being analogous to N1. Regarding the correlation of neural and behavioral results, it was observed in Fig. 5 that the amplitude of N1 and P2 was roughly proportional to the index of sensory sensitivity, d', obtained to the same stimuli. However, since such a correspondence could be obtained even if the two response domains were completely independent of one another, such a comparison is interesting only if a stronger dependency can be established between the neural and behavioral response dimensions. The most convincing evidence for such a relationship was presented in Fig. 6, where peak amplitude and bar press latency were compared on a trial-by-trial basis. Although it is apparent that a significantly greater proportion of large amplitude N1 and P2 responses are associated with the hit group, there is a great deal of variability in the scatter plots. Much of this variability was anticipated; one source being the ongoing amplitude fluctuations of the ECoG upon which evoked activity is superimposed. Another source is a shifting response criterion (Cx) which would affect both response probability and response latency. Although response criterion is assumed to be constant, it is more likely that it undergoes slight fluctuations, at least, over trials. This points to one of the limitations of the present technique for the study of the neural correlates of sensation. At this stage of development, sources of variability attributable to both neural and behavioral response measurem e n t preclude rigorous trial-by-trial correlation of neural and sensory events. Presumably, control of extraneous noise sources (i.e., neural background activity and Cx variability)

SEP CORRELATES OF STIMULUSDETECTION IN MONKEY with refinements in b o t h neurophysiologic recording and psychophysical testing should produce closer approximations to the theoretical neural and sensory events. To minimize the influence of both sources of variability, measures of central tendency (i.e., average SEP amplitude and median hit latency) served as the basis for comparison of neural and behavioral responses (Figs. 7 and 8). Fig. 7 depicts many of the essential findings of the neural-behavioral correlational analysis. The change in average SEP amplitude in relation to psychophysical performance across intensities, in contrast to the constancy of peak latency, suggests that N1 and P2 amplitude may convey sensory information concerning sensation magnitude although peak latency, in all likelihood, does not. In contrast to N1 and P2, P1 changes as a function of stimulus intensity but not in relation to psychophysical performance (Kulics 1977; Kulics et al. 1977). Fig. 8 shows that the change in amplitude is inversely related to behavioral response latency at all intensities. In some cases, as indicated earlier, single monotonic functions would closely fit the data points. Such a relationship between SEPs and psychophysical performance would be expected if the specific groups of psychophysical reports of increasing latency reflect classes of sensations of decreasing magnitude, and N1 and P2 amplitude reflects the activity of neural events which give rise to these sensations (see Data interpretation). Plots of P1 as a function of behavior, however, would describe three separate linear functions, parallel to the abscissa, indicative of the independence of P1 amplitude with psychophysical performance and, by extension, subjective sensation. In support of this contention, Libet (1973) provides evidence that neural activity underlying the primary evoked response, although signalling the receipt of sensory information at the cortex, is unrelated to cerebral events leading to conscious somatosensory experience in man, although later components may

91

reflect activity necessary for sensation. Thus the evidence presented here is consistent with the hypothesis that N1 and P2 amplitude at postcentral gyms reflects the processing of sensory information pertinent to psychophysical performance of a somatosensory detection task in the monkey. If so, the presumed propagation of evoked activity could reflect the transmission of this somatic sensory information to surrounding cortical regions. Should the significance of late evoked activity for somatic sensation be established by convergent lines of evidence from other sources (Uttal 1973) further experiments such as the one described here may be important in demonstrating where and how sensory information transactions occur at the cortical level in monkey and in man.

Summary Rhesus monkeys were trained to respond to constant-current electrical pulse stimuli to the hand which are known to elicit touch sensation in man. Simultaneously, recordings of somatosensory evoked potentials (SEPs) were made from postcentral gyrus of the performing monkeys. The prominent features of the SEP at most recording sites were a negative (N1) component peaking at 50--65 msec followed by a positive wave (P2) peaking at 105--130 msec. Primary evoked activity (P1) was minimal or absent at most sites at the intensities employed. Differences in N1 peak latency ranging from 4 to 9 msec were observed between the central member of a cluster of recording sites and those surrounding it. These differences are thought to reflect the propagation of evoked activity from some unidentified focus in postcentral gyms to surrounding regions. N1 and P2 amplitude was found to decrease as a function of behavioral response latency at both the center and surrounding sites of the electrode clusters. The signal detection theoretical model, which provided

92

the interpretative framework for neurophysiologic and psychophysical responses, suggested that N1 and P2 peak amplitude may reflect somatosensory information processing events necessary for psychophysical performance of the monkey. The propagation of evoked activity to different sites on postcentral gyms could therefore signify the transmission of this sensory information to surrounding cortical regions. Since the psychophysical model is equally applicable to monkey or man, it is suggested that evidence presented here and in similar studies may be relevant to the question of the neural coding of conscious somatic sensory experiences of man.

Rdsumd Corr~lats neuroniques corticaux dvoquds de la d~tection du stimulus somato-sensoriel chez le singe Des singes rhdsus ont dtd entrainds rdpondre ~ des trains de stimulus de courant dlectrique constant donnds 0 la face palmaire de la main, stimulus semblables ~ ceux qui sont capables de produire la sensation du toucher chez lqlomme. Simultandment, des enregistrements de potentiels dvoquds somatosensoriels (PES) ont dtd effectuds au niveau du gyrus post-central chez les singes sujets ces expdriences. Les plus remarquables particularitds du 'PES' obtenues A la plupart des lieux d'enregistrements sont les suivantes: une composante ndgative (N1) culminant ~ 50--65 msec, suivie d'une composante positive (P2) culminant ~ 105--130 msec. L'activitd dvoqude primaire (P1) est minimale ou absente ~ la plupart des points enregistrds pour les intensitds utilisdes. Des diffdrences de faience du pic de NI, comprises entre 4 et 9 msec, ont dtd observdes entre le lieu gdomdtrique d'un groupe de points d'enregistrement et ces points euxm~mes. L'auteur interpr~te ces diffdrences comme ~tant le reflet de la propagation dhJne

A.T. KULICS

activitd dvoqude, d'une source non identifide dans le gyms postcentral aux rdgions environnantes. L'amplitude de N1 et P2 diminue en fonction de la latence de la rdponse comportementale aussi bien au point central qu'aux points environnants des faisceaux d'dlectrodes. Ce module thdorique apporte un cadre d'interprdtation des rdponses neurophysiologiques et psychophysiques. Le moddle sugg~re que l'amplitude des pics NI et P2 puisse refldter des activitds de transformation d'dvdnements somato~sensoriels, activitds ndcessaires ~ la production d'une performance psychophysique du singe. La propagation d'activitd dvoqude ~ diffdrents points du gyrus postdro-centrai correspond donc ~ la transmission de cette information sensorielle aux rdgions corticales voisines. Comme le moddle psychophysique est dgalement applicable au singe et ~ l'homme, il est donc suggdrd que les observations pr~sentdes dans cette publication et dans les prochaines du m ~ m e type peuvent ~tre pertinentes ~ l'dtude des phdnomdnes de codage neuronique d'expdriences conscientes somatosensorielles chez l~omme. The author thanks Dr. J. Roppolo for help in surgical procedures, Drs. T. Teyler, S. Fish and T. Voneida for review of this manuscript, Dr. Voneida for use of histological facilities,Dr. J. Gilloteaux for the French translation of the summary, and Mr. P. Langm a n for photographic servicesand aid in data analysis.

References Albe-Fessard, D. and Besson, J.M. Convergent thalamic and cortical projections--the non-specific system. In: A. Iggo (Ed.), Handbook of Sensory Physiology: Somatosensory System, Vol. 2. Springer, New York, 1973 : 489--560. Amassian, V.E., Waller, H.J. and Macy, Jr., J. Neural mechanism of the primary somatosensory evoked potential. Ann. N.Y. Acad. Sci., 1964, 112: 5--32. Arezzo, J.C., Vaughan, Jr., H.G. and Legatt, A.D. Topography and intracranial sources of somatosensory evoked potentials in the monkey. II. Cortical components. Electroenceph. clin. Neurophysiol., 1981, 28: 1--18.

SEP CORRELATES OF STIMULUS DETECTION IN MONKEY Carlson, K.R. A temporary restraint chair for monkeys. Physiol. Behav., 1972, 9: 493--494. Eidelberg, E. and Jenkins, I. Organization of somatic sensory mechanisms in the macaque cerebral cortex. Electroenceph: clin. Neurophysiol., 1966, 21: 437--459. Giblin, D.R. Somatosensory evoked potentials in healthy subjects and in patients with lesions of the nervous system. Ann. N.Y. Acad. Sci., 1964, 112: 93--142. Gibson, R. Electrical stimulation of pain and touch. In: D. Kenshalo (Ed.), The Skin Senses. Thomas, Springfield, Ill., 1968 : 223--261. Green, D. and Swets, J. Signal Detection Theory and Psychophysics. Wiley, New York, 1966:455 pp. Kulics, A.T. Sensory discriminability comparisons in human and monkey with implications for the study of central nervous correlates. Ann. N.Y. Acad. Sci., 1977, 299: 244--254. Kulics, A.T. and Lineberry, C.G. Sensory discriminability in monkeys and humans: direct comparisons by signal detection theory. Neuropsychologia, 1977, 15: 321--327. Kulics, A.T., Lineberry, C.G. and Roppolo, J.R. Neurophysiological correlates of sensory discrimination performance to electrical cutaneous stimuli in rhesus monkey. Brain Res., 1977, 136: 360--365. Libet, B., Electrical stimulation of cortex in human subjects, and conscious sensory aspects. In: A. Iggo (Ed.), Handbook of Sensory Physiology: Somatosensory System, Vol. 2. Springer, New York, 1973 : 743--790. Libet, B., Alberts, W.W., Wright, E.W. and Feinstein, B. Responses of human somatosensory cortex to stimuli below threshold for conscious sensation. Science, 1967, 158: 1597--1600. Lineberry, C.G. and Kulics, A.T. The effects of dia-

93

zepam, morphine and lidocaine on nociception in rhesus monkeys: a signal detection analysis. J. Pharmacol. exp. Ther., 1978, 205: 302--310. Mountcastle, V.B. The world around us: neural command functions for selective attention. Neurosci. Res. Frog. Bull., 1976, 14: 1--47. Stohr, P.E. and Goldring, S. Origin of somatosensory evoked scalp responses in man. J. Neurosurg., 1969, 31: 117--127. Sutton, S. The specification of psychological variables in an average evoked potential experiment. In: E. Donchin and D.B. Lindsley (Eds.), Average Evoked Potentials: Methods, Results, Evaluations. NASA SP-191, Washington, D.C., 1969: 237--262. Swets, J.A. The relative operating characteristic in psychology. Science, 1973, 182: 990--1000. Tsumoto, T., Hirose, N., Nonaka, S. and Takahashi, M. Cerebrovascular disease: changes in somatosensory evoked potentials associated with unilateral lesions. Electroenceph. clin. Neurophysiol., 1973, 35: 463--473. Uttal, W.R. The Psychobiology of Sensory Coding. Harper and Row, New York, 1973. Vaughan, Jr., H.G. and Ritter, W. Physiologic approaches to the analysis of attention and performance: tutorial review. In: S. Kornblum (Ed.), Attention and Performance, IV. Academic Press, New York, 1973: 129--154. Williarnson, P.D., Goff, W.R. and Allison, T. Somatosensory evoked response in patients with unilateral cerebral lesions. Electroenceph. clin. Neurophysiol., 1970, 28: 566--575. Woolsey, C.N., Marshall, L.L. and Bard, W.H. Representation of cutaneous tactile sensibility in the cerebral cortex of the monkey as indicated by evoked potentials. Bull. Johns Hopk. Hosp., 1942, 70: 399.