Evoked Potentials, Neural Efficiency and I.Q.

Dr. John Ertl, PhD

In order to avoid the Psychologists dilemma in defining intelligence as a score on an I.Q. test, the concept of neural efficiency is introduced and the concept of intelligence is eliminated entirely from this paper. I.Q. test scores are regarded as useful measures of the behavioral efficiency of the organism and some degree of relationship with neurological efficiency is predicted.

Neurological efficiency can be defined independently on a speculative basis, but at least measurable quantities are involved and testable hypotheses are produced with some experimental evidence in support.

Thus the correlations presented in this paper between the latency of components of the averaged evoked potentials (AEP) and I.Q. test scores are regarded simply as evidence in support of the neurological efficiency hypothesis and do not imply a cause effect relationship between AEP latencies and intelligence.

Past attempts to correlate electrophysiological variables with behavioral indices of intelligence have been inconclusive (Vogel and Broverman, 1964). The hypothesis of a relation between AEP latency and psychometric intelligence is supported by animal evidence (Bradley, et al., 1964) and exploratory work with humans (Chalke and Ertl, 1965).

The possibility of relating microscopic and macroscopic electrophysiology of the central nervous system is so important that any reasonable hypothesis, even an oversimplified and naive hypothesis such as I am developing, may be worth considering.

As MacKay has said, “Science knows three kinds of frontier; those of the very large, the very small and the very complex.” Most of us at this Symposium are concerned with the very complex. I agree with MacKay that our conceptual resources are in danger of being swamped with data of embarrassing richness, so let me take a small step from the more complex to the less complex. Neural efficiency can be operationally defined at two levels;

1) based on the latency of components of the AEP detected at the scalp of human subjects, and 2) at the neuronal level.

1) Research to date indicates that the late components of the AEP are most sensitive to changes in stimulus parameters involving decision making (Sutton, et al., 1965), pattern recognition, attention and problem solving (Beinhocker, et al., 1966, Callaway, 1966, Chapman and Bragdon, 1964, Uttal, 1965)• drugs inducing changes in levels of alertness (Allison, et al., 1963, Garcia-Austt, 1963, Brazier, 1963) and generally the informational content of the stimulus.

Only a few of dozens of studies are cited, but there can be little doubt that the AEP with all its technical problems is a meaningful and most fruitful variable in the study of the central nervous system.

It is true that most studies have concerned themselves with the amplitude domain of AEPs, however it appears that the time domain is more relevant to the present investigation (Chalke and Ertl, 1965, Ertl, 1967). Neurological efficiency therefore, at this level is operationally defined as the latency of sequential AEP components.

2) In an important and controversial paper arising from the work of Gerstein (1960, 1961, 1964), Fox and O’Brien (1965) have shown that the post stimulus histogram (curve of probability) of single cell discharges duplicates the AEP wave form detected from the same area with gross electrodes in cats.

The importance of this work is discussed by Uttal (1965) and others, and is ignored by Creutzfeldt, et al., (1966) in a major paper dealing with “Relations between EEG phenomena and potentials of single cortical cells”. Creutzfeldt found close relations between evoked cortical EEG phenomena and transmembrane potential changes of cortical cells, mostly slow dendritic potentials and synchronized afferent and efferent fiber activity.

This difference of opinion may seem trivial in relation to the present paper, but is in fact central to the interpretation of these findings and also other evoked potential studies and of course it is central in the definition of neural efficiency. This whole problem has recently been brought into sharp focus by Kaufman and Price (1967) in a paper in which they describe “The detection of cortical spike activity at the human scalp”.

In order to operationally define neural efficiency at this level the ideas that information is transmitted as a spike interval code (MacKay and McCulloch, 1952, Braitenberg, 1967), that the AEP detected at the scalp is a description of the probability of firing of single cells or groups of cells firing in synchrony (Fox, 1965) are accepted.

Furthermore, most evoked potential studies cited above indicate that various components of the AEP can be influenced selectively by manipulation of the stimulus or the biochemical environment of the organism. A paper in Science by Libet a few weeks ago is in my opinion conclusive in demonstrating that the late components of the evoked potential are related to high level conscious processing; the abstract reads as follows:

Averaged evoked responses of somatosensory cortex, recorded subdurally, appeared with stimuli (skin, ventral posterolateral nucleus cortex) which were subthreshold for sensation. Such responses were deficient in late components. Subthreshold stimuli could elicit sensation with suitable repetition. The primary evoked response was not sufficient for sensation.

These facts bear on the problems of neurophysiological correlates of conscious and unconscious experience, and of “subliminal perception”. Thus the entire response from the onset of the stimulus to the cessation of time locked activity may be considered as a description of a specific pattern of spike sequences. In view of this, neurological efficiency is defined as the latency of the most probable times of occurrence of a sequence of spike discharges following a stimulus.

Method

The experimental sample of 170 male and 130 female subjects was randomly drawn from the total population of 4,170 pupils attending grades two, four, and seven of the 39 primary schools in the Ottawa Separate School system. The mean age of the subjects in the sample was 124 months with a range from 86 to 185 months.

Each subject was given three widely used psychometric tests of intelligence, (The Otis Quick-Scoring Test of Mental Ability, The Primary Mental Abilities battery 1962 revision, and the individually administered Wechsler Intelligence Scale for Children).

The EEG of each subject was recorded from bipolar scalp contact electrodes 6 cm. apart over the right sensori-motor area, parallel with the midline and astride in the 10-20 system with ground to the right ear lobe (upward deflection in reported data indicates negativity of anterior electrode with respect to the posterior). Electrode impedances were held below 5K ohms.

The raw EEG was amplified to the required voltage in a bandwidth 3db down at 5 and 100 Hz and recorded on multi-channel magnetic tape together with trigger pulses corresponding to the onset of flash stimuli. Subjects sat with eyes open in a darkened shielded room fixating a spot on a reflecting screen 7 feet away. A stimulator lamp was located above and behind the subject outside the room so that clicks from the lamp’s gas discharge tube were not audible to subjects.

Bright, microsecond duration photic stimuli were delivered according to a gaussian stimulus interval distribution with mean of 1.12 seconds and standard deviation of .75 seconds. Average evoked potentials in response to 400 stimuli in a 625 msec interval following stimulation were extracted and analyzed by two methods - conventional amplitude summation and zero-crossing analysis - using the Enhancetron ND-801 digital computer.

Using the first method, 400 cortical responses of each subject were averaged in the conventional amplitude summation manner by the Enhancetron (1024 channels) in sets of 200 and read out by Moseley X-Y recorder. Short-term reliability assessments of the average evoked potential and identification of its sequential components were made by visual cross-correlation from these graphs.

To eliminate much of the subjectivity inherent in this methodof AEP component identification and to statistically determine the presence or absence of AEP components a second zero-crossing analysis technique was employed (Ertl, 1965). This method facilitated identification of low amplitude, high frequency and high synchrony components which are suppressed by the conventional amplitude summation method (Ertl, 1967).

The EEG of each subject was converted to pulses corresponding to baseline crossings where the waveform passed from positive to negative voltage. A distribution of zero crossing occurrences against time following 400 stimuli (Fig. IB) was made by the Enhancetron using its multichannel scaling mode across 64 channels.

A control run was also generated using the same EEG data as in the stimulated run, but by triggering the multiscaling process 400 times randomly, i.e., with no time relation to experimental stimulation. The mean count of both the Computer studies indicated that the best compromise between resolution and signal detection is reached when channel dwell time is 1 to 2% of the analysis period.

Ttimulated and control runs and the standard deviation of the control run were computed. If the number of zero crossings in any channel of the stimulated run was greater than two standard deviations from the mean of the control run a statistically significant event (evoked potential component) was considered to be present. These events were labeled in sequence El, E2, E3, and E4.

Each event so identified corresponded in time to the zero crossing of the average evoked potential obtained by conventional amplitude summation (Fig. 1 A). To gain time resolution (625 msec/1024 compared to 625 msec/64), the amplitude summated peak preceding each significant zero crossing event was labelled and its latency measured with an error of measurement estimated to be plus or minus 3 msec.

The latency of the first four sequential evoked potential components detected were inter-correlated with all the psychometric measures of intelligence obtained using the Pearson r correlation coefficient.

Results and Discussion

Experimental results are presented in Tables 1,2 and illustrated in Figure 2. Short-term reliability of the average evoked potential waveforms particularly in the first 250 msec after stimulation were found by visual inspection to be very high. This observation is supported by other work on the short and long-term stability of AEPs (Dustman and Beck, 1963,Werre and Smith, 1964).

The long-term stability of AEP component latencies was also demonstrated by data from one subject on 101 days across a ten month period yielding the following standard deviations for the first four AEP components: El - 4.1 msec; E2 - 3.9 msec; E3 - 4.7 msec; E4 - 9.7 msec. No significant sex differences were noted in either the physiological or psychometric measures. Analysis of variance revealed that neither age nor the interaction between age and psychometric intelligence made a significant contribution to the variability in the latency of AEP components.

Consequently data was pooled for determination of correlations across the entire sample of 300 subjects.

Substantial inverse correlations between the latency of all AEP components and psychometric I.Q. (Otis, WISC, PMA Total) were obtained (Table 1). It would appear that the electrophysiological measures are related to some common factor tapped by all the highly intercorrelated psychometric tests.

It is interesting to note that sub tests of the PMA (Table 1) which is a factor analyzed test, all correlate significantly with the E3 component latency while many of the WISC sub tests do not (Table 2). It would be premature to attempt a discussion of the entire correlation matrix from the psychological point of view, there are, however, a few points of interest and anomalies in terms of the neural efficiency hypothesis.

1) The correlations of every psychological test and its sub-tests are higher with the late components of the AEP (E3, E4) than with the early components of the AEP (El, E2). In view of the large sample and the large number of sub test scores, it is highly improbable that such a trend could be due to chance. This agrees well with the generally accepted observation that the late components of the AEP are involved with higher mental activity.

Since information processing time, as measured by the latency of AEP components, is the key definition of neurological efficiency, one would expect that the late components would differentiate best between people of varying neurological efficiency. In this connection it must be noted that the inter-individual variability of the latency of El and E2 is less than half the variability of E3 and E4 (Table 1). 2) The intercorrelations of the latencies of sequentially identified AEP components are quite high ranging from .30 to .50 for non-adjacent components and .71 to .77 for adjacent components (Table 1). However, these correlations are not so high as to preclude the possibility that each component represents a distinct phase in the processing of information.

This also agrees well with the findings that the AEP components can be selectively influenced by stimulus variables. 3) An anomaly in the findings is that digit span does not correlate with AEP latencies. Intuitively one would expect this test to be a good measure of neurological efficiency; however it is known to correlate poorly with I.Q. (.27 in this sample) which intuitively is also to be expected. It is known that a large number of variables affect the waveform of the AEP particularly in the amplitude domain (Beinhocker, et al., 1966).

The inter-individual latency differences noted in this sample are greater than any latency changes attributable to variables such as attention, arousal level, pupil diameter, etc. It was impossible to control all potentially interfering variables, but it is unlikely that the observed relationship between AEP latency and I.Q. can be attributed to any of these uncontrolled variables.

Summary

1. A concept of neurological efficiency is proposed and defined at two levels; in terms of averaged evoked potentials at the scalp and at the neuronal level. The relationships between these two levels of definition are discussed.

2. Experimental evidence is presented indicating substantial inverse correlations (.33 - 5l) between evoked potential component latencies and three standardized I.Q. tests in a sample of 300 children, these findings are viewed as supportive evidence for the proposed neurological efficiency hypothesis.

References

Allison, T., Goff, W.R., Abrahamian, H.A., and Rosner, B.S., “The effects of barbiturate anesthesia upon human somatosensory evoked responses”, Electroenceph. clin. Neurophysiol., 1963, Suppl. 24, 68-75*

Beinhocker, G.D., Brooks, P.R., Anfenger, E., and Copenhaver, R.M.,

“Electroperimetry”, IEEE Trans. on Bio-Med. Eng., 1966, vol. BME-13,11-18.

Bradley, P.B., Eayrs, J.T., and Richards, N.M., “Factors influencing potentials in normal and cretinous rats”, Electroenceph. Clin. Neurophysiol., 1964, vol 17, 308-313-

Braitenberg, V., “On the use of theories, models and cybernetical toys in brain research”, Brain Research. 1967, vol. 6, 201-216.

Brazier, M.A.B., “Information carrying characteristics of brain responses”, Electroenceph. clin. Neurophysiol., 1963, Suppl. 24, 55-67.

Callaway, E., “Averaged evoked responses in psychiatry”, Journal Nervous and Mental Pis.. 1966, vol. 143, 80-92.

Chalke, F.C.R., and Ertl, J.P., “Evoked potentials and intelligence”, Life Sciences. 1965, vol. 4, 1319-1322.

Chapman, R.M., and Bragdon, H.R., “Evoked responses to numerical and non-numerical visual stimuli while problem solving”, Nature, 1964, vol. 203, 1155-1157.

Creutzfeldt, O.D., Watanabe, S., and Lux, H.D., “Relations between EEG phenomena and potentials of single cortical cells. I. Evoked responses after thalamic and epicortical stimulation”, Electroenceph.

clin. Neurophysiol., 1966, vol. 20, 1-18. Dustman, R.E., Beck, E.C., “Long-term stability of visually evoked potentials in man”, Science, 1963, vol. 142, 1480-1481. Ertl, J., “Detection of evoked potentials by zero crossing analysis”,

Electroenceph. clin. Neurophysiol., 1965, vol. 18, 630-631. Ertl, J., “Evoked potential recovery from tape recorded zero crossings of the EEG”, Electroenceph. clin. Neurophysiol., 1967, vol. 22, 387-388.

Fox, S.S., and O’Brien, J.H., “Duplication of evoked potential waveform by curve of probability of firing of a single cell”, Science, 1965, vol. 147, 888-890. Garcia-Austt, E., “Influence of the states of awareness upon sensory

evoked potentials”, Electroenceph. clin. Neurophysiol., 1963, Suppl. 24, 76-89.

Gerstein, G.L., and Kiang, N.Y.S., “An approach to the quantitative analysis of electrophysiological data from single neurons”, Biophys. J.,

I960, vol. 1, 15-28. Gerstein, G.L., and Kiang, N.Y.S., “Responses of single units in the

auditory cortex”, Exptl. Neurol., 1964, vol. 10, 1-18. Gerstein, G.L., “Neuron firing patterns and the slow potentials”, Electroenceph.

clin. Neurophysiol., 1961, Suppl. 20, 68-71. Kaufman, L., and Price, R., “The detection of cortical spike activity at the

human scalp”, IEEE Trans, on Bio-Med. Eng., 1967, vol. BME-14, 84-90. MacKay, D.M., and McCulloch, W.S., “The limiting information capacity of a

neuronal link”, Bui, of Math. Biophysics, 1952, vol. 14, 127-135.

Sutton, S., Braren, M., Zubin, J., and John, E.R., “Evoked potential

correlates of stimulus uncertainty.”, Science, 1965, vol. 150, 1187-1188.

Uttal, W.R., “Do compound evoked potentials reflect psychological codes?”

Psychological Bulletin, 1965, vol. 64, 377-392. Vogel, W., and Broverman, D.M., “Relationship between EEG and test intelligence: A critical review”, Psychological Bulletin, 1964, vol. 62, 132-144.

Werre, P.F., and Smith, C.J., “Variability of responses evoked by flashes in man”, Electroenceph. clin. Neurophysiol., 1964, vol. 17, 644-652.