Acta Psychologica 46 (1980) 129-140

0 North-Holland Publishing Company

EVOKED POTENTIAL CORRELATES OF VISUAL SELECTIVE

ATTENTION *

D .G. WASTELL h4RC Applied Psychology Unit, Cambridge, England

and

D. KLEINMAN Dept. of Psychology, University of Durham, Durham, England

Accepted January 1980

Two experiments are reported in which the relationship is explored between the Nl and P2

components of the visual evoked potential and visual selective attention. In both experiments,

channels were defined in terms of stimulus attributes (experiment 1 - intensity; experiment 2 -

shape) with all stimuli being presented at a common spatial locus in order to preclude fixation

shifts. Under these circumstances, P2 was found in both experiments to correlate with visual

selective attention whereas Nl did not. It is concluded from the null result for Nl that a pre-

cortical gating mechanism for visual inputs within a single spatial channel is not a component of

the visual system. The P2 correlate of visual selective attention, because of its long latency, is

attributed to the differential post-perceptual processing inevitably received by task-relevant and

task-irrelevant inputs.

Introduction

The initial negative wave (Nl) of the cortical evoked potential (EP), in reflecting the initial concerted mass activation of the cortex in response to sensory stimulation, is clearly well-placed to index the operation of pre-cortical selective processes. Using EPs in combination with adequate controls over peripheral mechanisms (inner ear muscle contractions,

* This research was conducted as part of the fist author’s doctoral research programme at Dur-

ham University and the support of an SRC grant is acknowledged.

Requests for reprints may be sent to D.G. Wastell, MCR Applied Psychology Unit, 15 Chau-

cer Road, Cambridge, U.K.

129

130 D.G. Wastell, D. Kleinmar~ /EP correlates of visual selectivity

head position etc.) the work of Hillyard’s group in particular (e.g.Hill- yard et al. 1973; Schwent and Hillyard 1975) has demonstrated an enhancement of Nl in relation to the direction of auditory selective attention. On the basis of this evidence, a pre-cortical gating mechanism ‘filtering’ cortical input (located by Picton et al. 1978 in the non- specific thalamus) has been inferred for the auditory modality. How- ever, in the visual system the evidence for such a central selection mechanism is not so convincing. Much of the earlier work, for instance, (e.g. Donchin and Cohen 1967) failed to ensure the complete unpredictability of the relevant and irrelevant stimuli, and thereby con- founded the operation of the selective machinery with possible general state changes developing in advance of the relevant material (the so- called ‘differential preparation’ artifact discussed in depth by Naatanen 1975).

Van Voorhis and Hillyard (1977) using a suitable paradigm, have, however, more recently been able to demonstrate an Nl enhancement of the visual EP to attended material in experiments using spatially separated channels. In order to control for the possibility of fixation shifts mediating their effects van Voorhis and Hillyard monitored the electro-oculogram (EOG). However careful reading of their method reveals that subject’s heads were apparently not clamped in their study. Furthermore the EOG results in their investigation, which are clearly crucial to their argument, are only reported in the briefest qualitative terms.

In the work reported here, control over the peripheral mechanism of visual fixation was achieved by presenting both attended and unattended stimuli within a small visual angle around a single point of fixation: i.e. in such a way that shifts of fixation would bring no advantage. Harter and Salmon (1972), in a methodologically sound paper with respect to the differential preparation artifact, also presented stimuli in this way. Their EP analysis is based on difference waveforms and does not indicate attention-related effects in evoked activity before a relative negative shift in the EP to attended stimuli which peaks at 235 ms post-stimulus, i.e. substantially later than the usual visual-N1 latency of around 130 ms or so. Although Harter and Salmon imply this shift to be evidence that visual selective attention is subserved by the early modulation of sensory input (1972: 6 1 l), the long latency of their effects makes this conclusion tenuous, and indeed it receives no mention in a subsequent paper (Harter and Previc 1978).

D.G. Wastell, D. Kleinman IEP correlates of visual selectivity 131

The purpose of the present investigation was to employ experimental procedures more closely resembling the established visual selective attention paradigms of cognitive psychology than those typically used in previous EP work in an attempt (using a conventional EP analysis) to demonstrate an attention-related enhancement of evoked activity at the latency of Nl . Such an enhancement could then be adduced, as per the auditory-N1 data, as evidencing the existence of precortical selective machinery in the visual system.

Two experiments are reported whose procedures, in turn, draw upon the selective reading paradigm (Neisser 1969) and the visual search para- digm respectively, crudely adapting them for EP work. In both experi- ments attended and nonattended stimuli are presented at a single point of fixation in order to preclude the involvement of fixation shifts to facilitate selection. Furthermore, following NZttinens critique (1975) of experimental methodology in relation to EP’s and selective atten- tion, the attended and nonattended stimuli were delivered according to a random temporal sequence in both experiments, thereby avoiding the ‘differential preparation’ artifact. The requirement of placing a high processing load on subjects in selective attention experiments, first pointed out in the EP connection by Wilkinson and Lee (1972) and Hartley (1970) and subsequently consolidated in the work of Schwent et al. (1976a, b), was also recognised and satisfied in both the experi- ments reported here, although different measures were implemented in each case.

Experiment 1

Introduction

In the first experiment a high processing load was achieved with the combination of a difficult task and the minimal IS1 for its successful performance. More specifi- cally, the attended and unattended channels were defined (in an analogous way to the use of colour in the selective reading paradigm, Neisser 1969) in terms of stimulus intensity, with Ss being required to attend to stimuli at one intensity level (the attended channel), whilst ignoring those at the other (the unattended channel). The stimuli within both channels were identical, comprising in each case a sequence of ones and zeros, and selective attention to one channel was achieved by instruct- ing Ss to count the number of 1s and OS in that channel. This first task thus required the subject to discriminate between two stimuli in the attended stream and accordingly update the appropriate one of two ongoing totals. The demands of the

132 D.G. Wastell, D. Kleinman /.EP correlates of visual selectivity

task were thus exacting and, indeed, preliminary work indicated that the fastest stimulus presentation rate at which the S could successfully manage it was 2 per sec.

Method

EEG was recorded monopolarly (ear lobe reference, preamplifier time constant = 0.1) from the vertex of five S’s and was subsequently analysed off-line by an IBM 370 computer. A 5 X 3 matrix of red light emitting diodes (LEDs) was mounted 1 metre from the S at the far end of a dimly illuminated tunnel and subtended an angle at the eye of approximately 1 degree.

The LED matrix was controlled by the transistor outputs of an IBM 1130 com- puter and was used to present a sequence of ‘ones’ and ‘zeros’ of either a high or low brightness level. Luminances of the LEDs were measured using an SE1 spot photometer with values of 17 candela/m* and 1.5 candela/m’ being obtained for individual elements for the high and low intensities respectively. The S’s task was to count the number of 1s and the number of OS of the intensity specified for the current block of trials. Eight blocks of trials, each block consisting of the random-

ised presentation of 20 of each type of stimulus (i.e. 80 trials per block), con- stituted the experimental session. On four of the blocks the S was instructed to count high intensity Is and OS (attend hi condition) and on the remaining blocks to count the low intensity stimuli (attend lo condition), the order of the conditions being counter-balanced within Ss. A constant IS1 of 500 ms was used as already indicated, and during this interval the central LED of the matrix remained illu- minated. Ss were required to fixate this point throughout. Before each block began Ss were given a run of practice trials which they terminated when they felt capable of performing the task; and at the end of each block the S was asked to report the number of Is and OS counted.

Given the importance of achieving a high processing load in this work, the possi- bility must be considered of S’s in this experiment simply ‘catching on’ to the fact that there were 20 of each ‘target’ stimulus per block, and thus responding 20-20 without actually attempting to count. Ss were questioned to this effect at the end of the experiment and indicated that they had not performed in this way. They unanimously indicated that they had consistently found the task difficult, this dif- ficulty finding expression in consistent miscounting which resulted in an average discrepancy of 2 betihreen each reported total and the actual number of targets (i.e. 20).

Results

Eight average EPs were computed for each S representing the brain response to each of the four types of stimulation (high intensity 1, hi 0, low intensity 1, lo 0) under each attention condition. As the experimental session comprised 640 trials, each of these average EPs was thus based on 80 trials. Four measures were derived from these averages: the amplitude and latency of the negative wave peaking at an aver- age latency of 122 ms (Nl), and the amplitude and latency of the subsequent posi-

D.G. Wastell, D. Kleinrnan /EP correlates of visual selectivity 133

50 . ...____* *--____ 5 -.PZ

=i 10

: 3.0

z 20 -Nl 7 a xc

10

< ATT IGN HI LO 1 0

3 OA ATT HI LO 1 0

ATTENTION INTENSITY STRUCTURE

Fig. 1. Experiment 1: Nl and P2 amplitude and latency are shown plotted as a function of attention, intensity and stimulus structure.

tive deflection that peaked on average 78 ms later (P2). Amplitudes were measured with respect to a baseline representing average activity in the first 20 msec post- stimulus (Walter ( 1964) indicates a latency in excess of 20 msec for the activation of the cortex by a visual stimulus). A long pre-stimulus baseline was not used due to possible contamination by late potentials associated with the preceding stimulus. A three factor repeated measures analysis of variance (attend vs nonattend, hi vs lo intensity, ones vs zeros being the three factors) was conducted for each measure, with the means expressing the main effects of attention, intensity and spatial struc-

ture upon each measure being shown in fig. 1. Considering Nl amplitude and latency first, neither measure was found to be

sensitive to the main effect of attention with F ratios of less than one (i.e. 2 p > 0.5) in both cases. It did, however, appear that the Nls evoked by less intense stimuli were both smaller in amplitude (F( 1,28) = 10.6, p < 0.01) and longer in latency (F( 1,28,) = 24.57, p < 0.001) and there was also evidence that 1s evoked a shorter latency Nl than did OS (F( 1,28) = 4.33, p < 0.05). All the remaining effects and interactions in both analyses were non-significant.

A significant main effect of attention (F( 1,28) = 8.9, p < 0.01) upon the ampli- tude of the P2 component was revealed in the ANOVA for P2 amplitude, with attention producing a 1.1 PV enhancement in the amplitude of this component. No other main effects nor any of the two-way interactions were significant.

Turning finally to P2 latency, it was noted that no attention-related latency changes, accompanying the amplitude enhancement, were present (F( 1,28) = 1.78 ns). Indeed the latency of this component behaved more in the manner of both N 1 amplitude and latency in being primarily sensitive to the effect of stimulus intensity (F( 1,28) = 41.26, p < 0.001). Neither the remaining main effect nor any of the interactions were found to be significant.

134 D.G. Wastell, D. Kleinman IEP correlates of visual selectivit]

The results of this experiment, although failing to reveal a significant main effect of selective attention upon N 1 amplitude, demonstrate an interesting dissociation between this component and the subsequent positive deflection of P2. Whereas N 1 amplitude and latency are influenced by stimulus intensity, with less intense stimuli

evoking smaller and later potentials (a result in accord with the literature; Regan 1972), P2 amplitude is not. Conversely P2 amplitude in this experiment, unlike N 1, does appear to index selective attention. The most reasonable interpretation of this dissociation is that N 1, being earlier, reflects the earlier mental events of perceptual processing and would differentiate between intense and less intense stimuli but not necessarily between attended and unattended ones. Post-perceptual processing would, on the other hand, be expected to reflect more of the relevance of stimuli rather than their perceptual content and ipso f&to so would any concurrent post- perceptual brain activity, which we are arguing, includes P2. The observation that the latency of P2 is influenced by stimulus intensity does not contradict this hypothesis for the reason that any delay in the earlier perceptual stages of proces- sing would inevitably be expected to delay all subsequent stages. This reasoning remains valid providing that no significant additional delay in the latency of P2 can be demonstrated, a stipulation that the data appears to support with the delay of 30.8 ms in the latency of P2 being only 4.6 ms longer (t(df = 4) = 1.38 ns) than the delay of 26.2 ms in the occurrence of N 1 to the low intensity stimuli.

Further discussion of these results will be deferred until the discussion section of the next experiment wherein the implications of the results of both experiments for models of visual selective attention will be discussed together.

Experiment 2

Introduction

Although no particular features of experiment 1 were identified as central to its failure to demonstrate an Nl-amplitude correlate of visual selective attention, a second experiment was conducted which exhibited the following modifications:

(a) The attended and unattended stimuli were defined in terms of their patterning rather than their intensity. Considering the general psychological literature, selective processes in vision have very largely been investigated in relation to the spatial structure of stimuli, that is in terms of searching for particular letters or forms (Haber and Hershenson 1973). Although colour has also been employed by some investigators (most notably, Neisser 1969; and Willows and McKinnon 1973) as the basis of selection, the use of stimulus intensity, as in

D.G. Wastell, D. Kleinman /EP correlates of visual selectivity 135

experiment 1, is a somewhat unusual procedure. The present study, therefore, investigates visual selective attention in a context where selection has been thoroughly investigated, i.e. in relation to character recognition. In short, Ss were required to detect the presence of a single target letter in a stream of letters presented successively in the visual field using the apparatus of experi- ment 1. As such this experiment is directly related to a long tradition of psychological experimentation, and in particular to the work of Neisser and his various co-workers (e.g. Neisser 1963; Neisser and Beller 1965; Neisser and Stoper 1965). One disadvantage, however, in this adaptation of the visual search paradigm as against the procedure of experiment 1 is that, with the S ‘looking for’ different patterns, his fixation point might vary across conditions. It was considered that with the visual display subtending only 1 degree at the eye, and with a central fixation point remaining on throughout the experiment, any such variations would be unimportant.

(b) With the S being simply required to detect and count a single target item in an

ongoing stream, a faster presentation rate than that employed in experiment 1 was necessary in order to induce the high information load that is a necessary condition for producing an effect of selective attention upon Nl amplitude (Schwent et al. 1976a, b). The achievement of an effective processing load by this device is a more proven method than the use of the demanding task in experiment 1, thus representing a further potential improvement over this earlier design.

Method

Six Ss participated in the experiment and their EEG was recorded monopolarly (ear lobe reference, preamplifier time constant = 0.1) from two electrode placements, one at the vertex (Cz) and the second positioned at a point midway between the two occipital lobes (0~). The apparatus of experiment 1 was used to present a random sequence of the eight letters D, E, F, H, I, 0, T and U at a rate of 3/set. The experiment was comprised of two such sequences of 256 stimuli each (yielding on average 32 relevant stimuli per sequence) with Ss being required to count the number of Is in one sequence and the number of Es in the other. The choice of Is and Es, although somewhat arbitrary, was because, with the apparatus used (that is, the 5 X 3 LED display) they were easily recognised and were rarely confused with other letters. The order of presentation of the two sequences was balanced across Ss and practice trials given in the manner of the previous experiment. It should be noted that unlike experiment 1, where the sequences of stimuli were random permutations of 20 of each of the 4 stimulus types, the sequences in this experiment were completely random. Although this inevitably meant that varying numbers of attended and unattended stimuli would occur in each sequence, this procedure was adopted as it removed a possible basis for the S to predict the stimuli above chance level [ 11.

[l] This statistical point, that the predictability of stimuli in random permutations is above chance level, is probably worth emphasising in the EPs and selective attention connection, as it

136 D. G. Wastell, D. Kleinman /EP correlates oj’visual selectivity

__ ATTEND ------ IGNORE

VERTEX OCClPUl

Fig. 2. Experiment 2: Composite vertex and occipital average EPs associated with attended and

non-attended stimuli.

At the end of each sequence, Ss were asked to indicate the number of relevant stimuli they had counted. Upon questioning, no reliable preference for either the attend-I or the attend-E condition was expressed by the Ss, a finding corroborated by the identical mean error scores of 3.4 (11%) for the two conditions (error being indexed by the absolute discrepancy between the total reported by the S and. the actual number of target stimuli).

Results

Eight digitally smoothed (Waste11 1979) average EPs (based on, on average, 32 stimuli each) were calculated for each S, representing the brain activity evoked by the Is and Es under each attention condition at each electrode placement. Compo- site EPs, illustrating brain act,ivity collapsed across subjects for attended and non- attended stimuli, are shown in fig, 2. For both occipital and vertex average EPs the amplitude of the negative peak (Nl) and the subsequent positive trough(P2) were measured as per experiment 1.

Remembering that the experiment comprised two blocks of stimuli, with Is being attended in one block and Es in the other, the following index was adopted in order to express the effect of selective attention upon EP amplitude (AMPL.) for each S’s data:

A = (AMPL. attended I + AMPL. attended E)/2- (AMPL. nonattended I + AMPL. nonattended E)/2

draws attention to an inherent contradiction between the concern for equal numbers of each

stimulus type in a sequence and the necessity to achieve complete randomness. It need not be felt that this difficulty only applies significantly to short permutations: basing its prediction on

a memory of previous stimulation, for instance, the author’s IBM 1130 computer was able to

predict the occurrence of ‘relevant’ events in a random permutation containing 100 such events

randomly intermixed with 100 irrelevant ones with success rates of 57%, 58%, 5.5%, 57% and

58% for the first five sequences in a run of 50 such simulations, where the overall average success rate was 54.5% (A.P. White, personal communication).

D.G. Wastell, D. Kleinman fEP correlates of visual selectivity 137

Not only is the analysis simplified by combining the EP data for Is and Es in this way, but the effect of any general state change between blocks (which might add a constant amount to the EP amplitude for all stimuli in a particular block) is also eliminated.

The effect of attention was thus calculated for both EP componentsiN 1, P2) at each recording site (C,, 0,) and evaluated using Student’s t-tests (elf= 5 through- out). No effect of attention upon the Nl component at either location was found (C,: t = -1.0 ns; 0,: t = -0.4 ns), but, following experiment 1, P2 is enhanced with attention at the vertex (A = 2.0 yV; t = 4.6, p < 0.01). There is also a correspond- ing, but diminished, effect of attention upon the occipital P2 (A = 0.8 /.LV; t = 7.4, p < 0.001). A further t-test indicated this interaction of the effect of attention with electrode location to be reliable (t = 3.2, p < 0.01).

One problem in the design of this experiment and the use of the above index of selective attention is that the experimental model is assuming no order effects inter- acting with the effect of attention between block 1 and block 2. However, such effects are intuitively likely: once S has been attending to, say, Is in block 1, for instance, it is not reasonable to assume that they then become just as ignorable in block 2 as the Es were in block 1. In order to assess such effects the data were reanalysed, restricting the analysis to the first block of each S’s data. The effect of attention was thus reestimated by:

A’ = AMPL. attended stimulus block 1 - AMPL. nonattended stimulus block 1

Although this meant that any effects of attention were necessarily confounded with EP amplitude differences due to the different physical characteristics of the attended and nonattended stimuli in any individual S’s data, these latter differences would be cancelled out in the group mean. Statistical analysis of these data revealed essentially the same results as above, with no effect of attention upon N 1 amplitude at either the vertex (t = -0.8 ns) or the occiput (t = -1.3 ns). A significant

enhancement of P2 amplitude at the vertex (A’ = 2.1 /JV; f = 2.1, p < 0.05) was again obtained, although the smaller occipital effect disappeared (t = 0.3 ns). Thus, although the attention-related enhancement of P2 at Cz and the null results for N 1 appear to be robust with respect to order effects, the small enhancement of P2 at 0, should be treated with some caution.

Discussion

The results of this experiment confirm those of experiment 1 with selective atten-

tion to a visual information channel being indexed by the amplitude of the P2 com- ponent of the visual EP, but not by the amplitude of the Nl component. It is further shown that this effect of attention is essentially restricted to the brain activity recorded from a vertex as against an occipital placement. It is generally held (see Picton et al. 1976: 362) that vertex potentials reflect the activity of the frontal association cortex and thus our topographic effect is seen to be consistent with Pribram’s (Pribram and McGuinness 1975) functional dichotomy between frontal and posterior cortex which associates these areas with ‘attentional’ (the orienting

138 D.G. Wastell, D. Kleinman /EP correlates of visual selectivity

reaction) and ‘sensorial’ (visual discriminatory processes) aspects of processing respectively.

It was indicated in the Introduction that the amplitude of the Nl wave of the auditory EP is generally held to index the operation of a central selection mecha- nism, probably located at the level of the thalamus, which preferentially transmits information on an attended channel through to the cortex, Hillyard et al. (1973), drawing on Broadbent (1970), describe this mode of attention as ‘stimulus set’ connoting its concern with the early perceptual stages of analysis and distinguish- ing it from the subsequent ‘response set’ mode in which specific signals within the attended stream are recognised. However, although the evidence of experiment 1 indicates the visual N 1 to be sensitive to perceptual variables, neither experiment 1 nor experiment 2 demonstrate any relationship between Nl and visual selective attention; and the suggestion therefore is that, in the paradigms used in this study, central gating of the visual sensory array does not occur. Whilst this null result does call into some question the generality of the concept of a precortical gating mechanism subserving the ‘stimulus set’ mode of attention, an important caveat should be considered before generalising the present findings too far. In the audi- tory system the Nl-correlate of selective attention has generally been investigated using analogues of the dichotic listening paradigm, i.e. in paradigms where attended and unattended streams are clearly differentiated in terms of spatial location. In the present visual paradigms, in order to preclude the use of peripheral mechanisms, all stimuli were presented at a common fixation point. Channels were defined in terms of stimulus intensity and shape rather than spatial locus. Most parsimoniously then, all one can conclude from our data is that selective attention to visual inputs from a single spatial source is not reflected in the amplitude of Nl. The present data therefore do not preclude a link between visual Nl amplitude and selectivity where spatially separate channels are used, and are therefore not inconsistent with the results of van Voorhis and Hillyard (1977) which provide some evidence for such a link.

Whether or not the organism is operating a visual ‘stimulus set’ in the experi- ments of this paper, relevant and irrelevant stimuli certainly receive differential processing at some stage. It is proposed that the P2 correlate of selective attention indexes this differential processing, with the enhancement of P2 to stimuli in the attended stream reflecting the increased post-perceptual processing (i.e. counting) required by task-relevant inputs. In other words, given the long latency of P2 (peak- ing at around 200 ms or so, i.e. at a latency comparable with the appearance of simple motor reactions) and the fact that it is preceded by an earlier cortical wave (N 1) which is apparently related to the perceptual qualities of stimulation, the labil- ity of its amplitude with respect to attention is regarded, in contrast to N 1, as being more related to a post-perceptual ‘response set’ mode of selective attention. It is pertinent to note here that Hillyard (e.g. Hillyard et ul. 1973: 179) makes this same distinction between N 1 and subsequent positivity, although in his case he implicates P3 rather than P2 as indexing ‘response set’.

D.G. Wastell, D. Kleinman IEP correlates of visual selectivity 139

References

Broadbent, D.E., 1970. ‘Stimulus set and response set: two kinds of selective attention’. In: D.I. Mostofsky (ed.), Attention: contemporary theory and analysis. New York: Appleton-

CenturyCrofts.

Donchin, E. and L. Cohen, 1967. Averaged evoked potentials and intramodality selective atten- tion. Electroencephalography and Clinical Neurophysiology 22,5 37-546.

Haber, R.N. and M. Herchenson, 1973. The psychology of visual perception. New York: Holt,

Rinehart and Winston.

Harter, M.R. and F.H. Previc, 1978. Size-specific information channels and selective attention:

visual evoked potential and behavioural measures. Electroencephalography and Clinical

Neurophysiology 45,628-640.

Harter, M.R. and L.E. Salmon, 1972. Intra-modality selective attention and evoked cortical

potentials to randomly presented patterns. Electroencephalography and Clinical Neurophy-

siology 32, 6055613.

Hartley, L.R., 1970. The effect of stimulus relevance on cortical evoked potentials. Quarterly

Journal of Experimental Psychology 22,531-546.

Hillyard, S.A.. R.F. Hink, V.L. Schwent and T.W. Picton, 1973. Electrical signs of attention in

the human brain. Science 182, 177-180.

Naatanen, R., 1975. Selective attention and evoked potentials in humans - a critical review. Biological Psychology 2, 237-307.

Neisser, U., 1963. Decision-time without reaction time: experiments in visual scanning. Ameri- can Journal of Psychology 76, 3766385.

Neisser, U., 1969. Selective reading: a method for the study of visual attention. Paper presented at the symposium on Attention: some growing points in recent research. XIX International Congress of Psychology, London.

Neisser, U. and H.K. Beller, 1965. Searching through word lists. British Journal of Psychology 56, 349-358.

Neisser, U. and A. Stoper, 1965. Redirecting the search process. British Journal of Psychology

56,359-368.

Picton, T.W., K.B. Campbell, J. Baribeau-Braun and G.B. Proulx, 1978. ‘The neurophysiology of human attention: a tutorial review’. In: J. Requin (ed.), Attention and performance VII.

New Jersey: Lawrence Erlbaum. pp. 429-467.

Picton, T.W., S.A. Hillyard and R. Galambos, 1976. ‘Habituation and attention in the auditory system’. In: W.D. Keidel and W.D. Neff (eds.), Handbook of sensory physiology V/3. Berlin:

Springer-Verlag. pp. 343-389.

Pribram, K.H. and D. McGuinness, 1975. Arousal, activation and effort in the control of atten- tion. Psychological Review 82, 116-149.

Regan, D., 1972. Evoked potentials in psychology, sensory physiology and clinical medicine. London: Chapman and Hall.

Schwent, V.L. and S.A. Hillyard, 1975. Evoked potential correlates of selective attention with multi-channel auditory inputs. Electroencephalography and Clinical Neurophysiology 38,

131-138.

Schwent, V.L., S.A. Hillyard and R. Galambos, 1976a. Selective attention and the auditory

vertex potential. I. Effects of stimulus delivery rate. Electroencephalography and Clinical

Neurophysiology 40,604-614.

Schwent, V.L., S.A. Hillyard and R. Galambos, 1976b. Selective attention and the auditory vertex potential. II. Effects of signal intensity and masking noise. Electroencephalography and Clinical Neurophysiology 40,615-622.

140 D. G. Wastell, D. Kleinman /EP correlates of visual selectivity

van Voorhis, S. and S.A. Hillyard, 1977. Visual evoked potentials and selective attention to points in space. Perception and Psychophysics 22, 54-62.

Walter, W.G., 1964. The convergence and interaction of visual, auditory and tactile responses in

the human non-specific cortex. Annals of the New York Academy of Sciences 112, 320&

361.

Wastell, D.G., 1979. The application of low-pass linear filters to evoked potential data: filtering

without phase distortion. Electroencephalography and Clinical Neurophysiology 46, 355-

356. Wilkinson, R.T. and M.V. Lee, 1972. Auditory evoked potentials and selective attention.

Electroencephalography and Clinical Neurophysiology 33,411&418.

Willows, D.M. and G.E. Mackinnon, 1973, Selective reading: attention to unattended lines.

Canadian Journal of Psychology 27, 2922304.