Author Notes: This research was supported by National Institute of Mental Health grant R01-MH42465 to F. K. Graham and R.F. Simons. Correspondence should be addressed to R.F. Simons, Department of Psychology, University of Delaware, Newark, DE 19716.
The general focus of the research described in this chapter is
on understanding the role in higher cognitive processes played
by two low-level reflexes-prepulse inhibition (PPI) and the orienting
response (OR). Specifically, this research had two primary aims:
1) To examine event-related potential (ERP) components under prepulse
conditions as part of a broader effort to determine how PPI effects
higher level perceptual processes, and 2) To determine whether
localized and generalized orienting are associated with attention-sensitive
Efforts to examine the inhibition of ERP components by prestimulation have been very successful, with two initial studies finding significant PPI beginning with themid-latency, positiveERP component at 50 ms (P50), but no evidence of PPI in the earlier P30 component. Because P30 and P50 are believed to reflect neural activity in two different auditory projection pathways, these effects were reproduced in a subsequent study which provided evidence that 1) PPI present in the nonspecific projection system (P50) is true PPI and is not due to local refractory effects, and 2) the absence of PPI characterizing P30 is a feature of other ERP components associated with the specific projection system. These experiments are described in the first section of the chapter.
The chapter's second section describes an independent series of studies that have utilized variations of the P300 component as a means of examining ERP characteristics associated with voluntary (active) and involuntary (passive) allocation of attention. Previous studies have agreed in finding parietal, posterior, P300s during active attention, but have been inconsistent in their findings of anterior P300s associated with passive attention, or orienting. Our own data now confirms the existence of a more anterior positive component in response to rare, nontarget events, and this component precedes, then overlaps in part, the parietal P300.
Prepulse inhibition is a robust phenomenon, present in both human
and nonhuman animals. It is elicited by any brief stimulus change
(S1) which shortly precedes a reflex-eliciting stimulus (S2) and
is defined as the automatic reduction of reflex size to S2 by
S1. In humans, the PPI effect has been demonstrated for the startle
blink, the most persistent component of startle (for reviews see
Anthony, 1985; Graham, 1975; Hoffman & Ison, 1980). The effect
is not due to learning (Krauter, Leonard, & Ison, 1973), frequency-dependent
auditory masking (Hoffman & Searle, 1968), the protective
reflex of the middle ear (Ison, Reiter, & Warren, 1979), or
S1-S2 response interference or muscle refractoriness (Graham &
Murray, 1977; Hammond, McAdam, & Ison, 1972).
Instead, PPI depends primarily on the transient characteristics of the prestimulus. An effective S1 can be any change in stimulation, such as an onset or offset (Lane, Ornitz, & Guthrie, 1991; Stitt, Hoffman, & Devido, 1980), a change in tone pitch (Cranney, Hoffman, & Cohen, 1984), or a brief change in the intensity of a continuous background signal (Reiter & Ison, 1977). Additionally, a change in stimulation lasting no longer than 6 ms is sufficient to produce an effect and a 20-ms S1 produces as great an effect as an S1 sustained until S2 onset (Giardina, 1989; Graham & Murray, 1977).
Neurophysiological (Davis & File, 1984), as well as behavioral (Graham, 1980) studies indicate that PPI is a form of active, extrinsic inhibition. Unlike refractoriness, which is produced by mechanisms that reduce transmission efficacy of the direct pathway mediating a response, PPI is mediated via a parallel indirect midbrain path, external to but synapsing with the direct hindbrain path which itself mediates startle.
Graham (1975) hypothesized that the mechanism underlying PPI protects preattentive stimulus processing, allowing a finer stimulus analysis to proceed with minimal interruption during the period critical for stimulus recognition. The mechanism, initiated by a rapidly-conducting transient detector, attenuates the response to a shortly-following intense S2, reducing the sensory impact of S2 and allowing the processing of S1 to proceed undisrupted through to recognition. "Without such a mechanism, temporally adjacent stimuli might more easily mask or merge with one another so that only a single stimulus could be perceived or two stimuli could be perceived but with altered characteristics" (Perlstein, Fiorito, Simons, & Graham, 1993, pp. 347-348).
The roughly 250 ms interval during which S1 processing moves from detection to recognition not only corresponds to the period of maximal PPI, but is also contemporaneous with the occurrence of early and mid-latency ERPs. If the mechanism underlying PPI reflects a sensory-gating or protective mechanism, then the PPI paradigm should have consequences for these exogenous ERPs and for the perception of both stimulus-pair members. We conducted a series of studies to explore these issues.
Our first two studies (Perlstein et al., 1993) evaluated PPI concomitants
of midline-recorded exogenous ERPs and perception using the standard
intramodal PPI paradigm. Single (75 db) or paired 800-Hz 40-ms
duration tones (S1, 75 db; S2, 110 db) were tested in separate
groups of subjects at stimulus onset asynchronies (SOAs) of 120
(Study 1) and 500 ms (Study 2). Stimuli were delivered at long
intertrial intervals (18 to 28 s) to allow for full recovery of
ERPs between trials. Blinks were recorded from the bipolar electromyographic
activity (EMG) of the orbicularis oculi; ERPs were measured from
midline frontal (Fz) and vertex (Cz) sites, referred to the mastoid.
Subjects provided unspeeded, unsignalled magnitude estimates of
the control stimuli and the second stimulus of each pair in the
short-SOA study and of control stimuli and both pair members in
the long-SOA study. Subjects' perceptual judgments were anchored
to a pair of 90-dB modulus tones that were presented prior to
each of four blocks of 20 trials. The magnitude estimation task
not only provided data regarding perceptual effects of PPI, but
also served to reduce habituation of blink.
ERPs were corrected for vertical electro-oculographic (EOG) activity using the method described by Gratton and colleagues (Gratton, Coles, and Donchin, 1983). Prior to measurement and analyses of baseline-to-peak amplitudes (mid- and later-latency EPs) and varimax-rotated PCA factor scores (midlatency EPs), paired-stimulus ERPs were subtracted from ERPs evoked by the low-intensity control stimuli to remove overlap of the response to S1 from response to S2.
Averaged ERPs are presented in Figure 1 and three of the main results are presented in Figure 2. First, in addition to significant reductions of blink at both SOAs, stimulus pairing markedly and significantly reduced P50 and later potentials, including N100 and P200; reductions were greater at the short than long SOA. Second, P30 was not affected by prestimulation at either SOA. Finally, consistent with the protection hypothesis, stimulus pairing also significantly reduced the perceived magnitude of 110-dB S2s, but magnitude estimates of the 75-dB S1s were increased relative to S1-alone levels.
Figure 1 Vertex (Cz) ERPs obtained under ipsimodal (acoustic) single- and paired-stimulus conditions when the SOA from S1 to S2 varied from 120 ms (top) to 500 ms (bottom).
Figure 2 Eyeblink, ERP and percepetual effects (loudness estimations) as a function of ipsimodal stimulus pairing at SOAs of 120 (top) and 500 ms (bottom).
These two studies demonstrate that it is possible to record mid-
and later-latency ERPs within the context of a PPI paradigm and
that these ERPs are differentially sensitive to
stimulus pairing. Furthermore, the finding that prestimulation
afforded some protection to perception of the prestimulus is consistent
with the protection hypothesis forwarded by Graham (1975).
Perlstein et al. (1993) suggested that prestimuli have different effects on P30 and P50 because the two potentials arise from functionally interacting thalamic-cortical pathways activated in parallel. Specifically, P30 reflects activity transmitted via the rapidly-conducting lemniscal or specific path which initiates the transient-dependent PPI effect in the more slowly-conducting, nonspecific or extralemniscal path leading to the P50 generator. This hypothesis is supported by findings of functional differences between P30 and P50 in human subjects (Buchwald, Rubenstein, Schwafel & Strandburg, 1991) and between analogous potentials in the cat (Erwin & Buchwald, 1987). Evidence from intracranial recordings suggests that the functionally distinct P30 and P50 are generated in primary and secondary auditory cortex, respectively (e.g., Liégeois-Chauvel, Musolino, & Chauvel, 1991; Scherg & von Cramon, 1986).
Thus, existing data suggest that P30 and P50 do not depend on serial transmission via the specific projection system but involve parallel asynchronous paths which may transmit information about transient and sustained stimulus characteristics, respectively. Furthermore, Perlstein et al. (1993) suggested that the prestimulus reductions of P50 and later potentials may be due, like blink PPI, to an active, extrinsic mechanism; that is, activity from the specific path feeds into and modulates activity in the nonspecific path. The suggestion of an extrinsic modulatory mechanism is in contrast to suggestions of others who have attributed ERP reductions in equal-intensity paired-stimulus paradigms to passive refractory-type processes (e.g., Roemer, Shagass, & Teyler, 1984).
These first two studies, however, could not directly evaluate the hypothesis of extrinsic, rather than intrinsic inhibition because the short SOA elicited near-maximal PPI of the blink and ERPs. Consequently, the SOA-response function for PPI of ERPs could not be distinguished from a refractory curve which also exhibits increased response size with increased SOA. Also, using ipsimodal pair members confounds refractoriness due to modality-specific effects at peripheral levels of the input path with other forms of modulation, such as PPI (Balaban, Anthony, & Graham, 1985; Rimpel, Geyer, & Hopf, 1982). Such confounding of intrinsic and extrinsic processes can be inferred from studies showing 50% greater reduction of N100 and P200 produced by ipsimodal than crossmodal pairs (e.g., Davis, Osterhammel, Wier, & Gjerdingen, 1972).
The third study in this series (Perlstein, Simons, & Graham,
in preparation) tested directly the hypothesis that an active
form of inhibition might account for prestimulus reduction of
P50 and later midline potentials. This study also evaluated the
nature of the pairing effects on specific and nonspecific subcomponents
of N100 described by Näätänen and Picton (1987)
and whether unimpaired perception of the S1 persisted in a crossmodal
Briefly, Näätänen and Picton (1987) described three distinct sources that contribute to the scalp-recorded N100. One source, generating Component 3, is modality nonspecific and can be elicited by auditory, visual, and tactile stimuli. Generator sources of this component appear to be widespread and involve a diffuse cortical system receiving projections from nonspecific thalamic nuclei via polysensory midbrain reticular formation (Velasco & Velasco, 1986; Velasco, Velasco, & Olvera, 1985) and projecting to the frontal, auditory association, and motor areas (e.g., Liégeois-Chauvel, Musolino, Badier, Marquis, & Chauvel, 1994; Giard, Perrin, Echallier, Thévenet, Froment, & Pernier, 1994).
The other two components are specific to the auditory system. Component 1 is best recorded electrically as a frontally-maximal negativity at 100 ms and, based on magnetic field (e.g., Pantev, Hoke, Lehnertz, Lütkenhöner, Anogianakis, & Wittkowski, 1988) and intracranial (Liégeois-Chauvel et al., 1994) recordings, reflects a source in or near primary auditory cortex. The second specific component, Component 2, can be identified over temporal regions (Wolpaw & Penry, 1975) as a biphasic "T-complex" consisting of a small P105 (Ta) and larger, later negativity (Tb) ranging in latency from about 120 to 165 ms in the literature. Generators of this component have been proposed in the secondary auditory cortex based on intracranial recordings (Celesia, 1976; Liégeois-Chauvel et al., 1994) and by scalp-distribution analyses (Scherg, Vajsar, & Picton, 1989; Scherg & von Cramon, 1985).
The hypothesis of a refractory effect on ERPs was tested in two ways. First, crossmodal (tactile-acoustic) rather than ipsimodal pairs served as stimuli. Crossmodal pairs provide an unconfounded test of the refractory hypothesis since they should eliminate the modality-specific repetition or refractory effect that occurs with intramodal pairs. Second, the effects on ERPs of an SOA too short to yield significant PPI of acoustic blink were compared with the effects of a longer SOA that should yield robust blink inhibition. It is known from studies of blink PPI that prestimulation with tactile S1s at short intervals (i.e. 25 to 50 ms) may produce facilitation of the reflex elicited by an acoustic S2 (e.g., Blumenthal & Gescheider, 1987). Thus, short-SOA prestimulus effects on ERPs might differentiate between intrinsic and extrinsic mechanisms.
In this study, S1 was a weak, 10-psi air puff, psychophysically matched to a 75-dB tone, delivered to the wrist; S2 was identical in quality to the 110-dB tone used in the first study. Stimuli were delivered at SOAs of 60 and 360 ms. ERPs were recorded from midline (Fz and Cz) and lateral (F8 and T4/6) sites. Vertical and horizontal EOG and EMG from three muscles-post-auricularis, medial frontalis, and masseter-were also recorded and ERPs were corrected to remove volume-conducted activity from these sites using the multiple-pass correction program of Miller, Gratton, and Yee (1988). Puff-alone ERP waveforms were subtracted from paired-stimulus waveforms and the data were analyzed as in the first study. Subjects performed a magnitude estimation task in which they judged the magnitude of all stimuli, relative to an infrequently-presented modulus.
Effects of pairing on ERPs and perception, illustrated in Figure 3, replicated our previous findings of differential effects on P30 and later midline ERPs and of reduced estimates of S2s and enhanced estimates of S1s. More interesting than replicating the effects, however, was the finding that prestimulus reductions of midline P50 and N100 were significantly greater at the long than short SOA. This effect, together with the crossmodal pairing, provides strong support for the hypothesis that PPI of exogenous ERPs, like blink PPI, results from an extrinsic rather than intrinsic mechanism. Specific and nonspecific N100 components, scaled to a common amplitude (Figure 4), were also differentially affected by prestimulation: The specific Component 1 showed recovery towards baseline levels at the long SOA; the nonspecific Component 3 and specific Component 2 showed parallel inhibitory effects which were greater at the long than at the short SOA.
Figure 3 Eyeblink, ERP and perceptual effects (loudness estimations) as a function of crossmodal (tactile-acoustic) stimulus pairing at SOAs of 60 and 360 ms.
Figure 4 Magnitude of the nonspecific N1 (N100) and two specific N1 components (C1 and Tb) recorded under single- and paired-stimulus conditions.
The findings of differential pairing effects on Components 1 and
3, which parallel effects on P30 and P50, respectively, support
previous work suggesting both their functional and anatomical
independence. For example, the magnetically-recorded N100 (N100m;
Component 1) exhibits ceiling effects at higher intensities and
is more dependent on stimulation rate than the electric N100/Component
3 (Hari, Kaila, Katila, Tuomisto, & Varpula, 1982; Reite,
Zimmerman, Edrich, & Zimmerman, 1982). Consistent with our
own data, N100m is unaffected by electrical prestimulation of
the median nerve at a 500-ms SOA (Huttunen, Hari, & Vanni,
1987), and is actually enhanced by pairing of moderate-intensity
acoustic stimuli at SOAs from 70 to 230 ms (Loveless & Hari,
1993; Loveless, Hari, Hämäläinen, & Tiihonen,
1989). This stimulation-rate sensitivity of N100m relative to
the nonspecific vertex N100 and its lack of sensitivity to attentional
manipulations (Arthur, Lewis, Medvick, & Flynn, 1991) is consistent
with the rapid recovery observed in the present study and with
it's localization to, at least, a region including the primary
Prestimulus reductions of Component 2, which paralleled reductions of Component 3, indicate that the pathway generating this response must receive input from a somatosensory-processing source. Findings of attention-related enhancements of Component 2 (Hackley, Woldorff, & Hillyard, 1990; Perrault & Picton, 1984), presumably including the nonspecific system since attention effects are not restricted to a single modality, also suggest overlapping functional aspects of Tb and Component 3. However, Loveless and Brunia's (1990) finding that effects of increased stimulus rise-time (increased latency and decreased amplitude) were restricted to the nonspecific component (3) indicates that Tb also exhibits functional characteristics of the specific system. Together, these data are consistent with Näätänen and Picton's (1987) suggestion that Tb receives both serial input from the primary auditory cortex and parallel input from the nonspecific thalamus; that is, the generator of Tb receives input from both specific and nonspecific projections.
The present research yielded four main findings regarding ERP
concomitants of PPI: (1) Specific and nonspecific ERPs are consistently
and differentially affected by both ipsi- and crossmodal prestimulation;
(2) prestimulus effects on midline ERPs are restricted to potentials
transmitted within the nonspecific path; (3) PPI of ERPs, like
blink PPI, is due to an active, extrinsic mechanism, and (4) the
differential effects of prestimulation on the various ERP components,
together with the literature, suggest three rather than two acoustic
projection systems. P30 and Component 1 of the N100 (N100m) is
associated with the specific, lemniscal, system, P50 and Component
3 of the N100 (vertex N100) with the nonspecific, extralemniscal,
system, and Component 2 of the N100 (Tb), exhibiting functional
characteristics of an intermediate system, perhaps reflects processing
in the "lemniscal adjunct" pathway described by Weinberger
and Diamond (1987).
The perceptual effects observed in this set of experiments revealed that perception of both tactile and acoustic prestimuli was undiminished by subsequent high-intensity acoustic stimuli and this effect has been observed as well by Filion and Cerani (1994). The dual findings of unimpaired S1 perception and reductions in S2 perceived magnitude provide preliminary support for the hypothesis that PPI reflects a protective mechanism and suggest, further, that use of the PPI paradigm to study both ERP and perceptual variables may clarify the mechanisms underlying the phenomenon as well as its adaptive function in information processing.
The issue of whether there is a unique positive ERP component
related to "novel" or "unpredictable" stimuli
has been under investigation in many laboratories but still remains
unresolved. This issue is important because orienting theory distinguishes
between a generalized OR to passively attended novel or deviant
stimuli and a localized OR to selected targets that are actively
attended. If a unique ERP response, specific to low probability
stimuli, can be reliably distinguished from an ERP component to
low-probability target stimuli, it would suggest that different
neural mechanisms are responsible for the productionof signal
and nonsignal ORs and would identify the two processes much earlier
in time than is possible with the slowly-developing autonomic
responses from which OR theory arose. This issue is also of interest
because it has been reported that the anterior, passively-attended
and posterior, actively attended P300 potentials show different
developmental courses (Courchesne, 1978). The posterior P300 can
be recorded from children as young as six years, while the anterior
P300 may not be present until adolescence.
Evidence for the existence of an anterior positivity specifically related to generalized orienting comes from two sources. The first source is a set of studies derived from procedures employed by N. Squires, Squires and Hillyard (1975) in which subjects were presented with two acoustic stimuli which differed only in frequency or intensity. In separate conditions, one of the two stimuli occurred with probabilities of 0.1, 0.5 and 0.9, and the other occurred with the complementary probability. At each probability, subjects either counted tones defined as targets (either the high or low probability tone), or ignored the tones while reading a book. With event-related potentials recorded from Fz, Cz and Pz, N. Squires et al. found a P300 deflection (P3b) related to both rare targets and rare non-targets, and noted another, earlier-onset P300 deflection (P3a) specifically related to all low probability stimuli (i.e. rare targets, rare non-targets and the ignored rare stimuli). P3b had a parietal scalp distribution, while P3a, in contrast, had a more anterior disctribution-centro-parietal in the rare target condition and centro-frontal in the ignore condition.
The second source of evidence suggesting that distinct P300 responses might be associated with passive and active attention derives from a series of studies by Courchesne and colleagues (e.g., Courchesne, 1978; Courchesne, Hillyard and Galambos, 1975; Courchesne, Kilman, Galambos and Lincoln, 1984; see also Knight, 1984). These studies employed a standard target-detection, or 'oddball' task, but in addition to rare target (oddball) and frequent standard stimuli, a variety of unannounced complex novel stimuli were embedded within the series of standards and targets. In each of these studies, a frontal or centro-frontal positive component was observed in response to the novel stimulus (the 'novels P3') while the parietal P300, as expected, was associated with the rare targets. It is unclear at this point whether the novels P3 described by Courchesne is the same potential as the P3a described by N. Squires et al. (1975).
As described briefly above, the first, and most compelling evidence
for the differentiation of a small, more anterior P3 response
from the much larger, later parietal P300 was provided by N. Squires
et al. (1975) and the subsequent replication by Snyder & Hillyard
(1976). Other studies, employing very similar procedures, have
been less successful, however, (e.g. Duncan-Johnson & Donchin,
1977; Sams, Ahlo & Naatanen, 1984; K. Squires, Donchin, Herning
& McCarthy, 1977) and have called into question the existence
of the P3a component or its independence from either the preceding
N200 or the subsequent P300 components, both commonly associated
with low-probability stimuli. Although it is frequently noted
in passing that two P300 peaks are seen in ERP waveforms, they
are rarely separately measured and tested statistically for differences
as a function of the experimental variables (e.g. probability,
electrode location, task). In fact, only N. Squires et al. (1975)
were able to show that the two deflections were associated with
different factors when the data were subjected to a Principal
Components Analysis (PCA).
The present P3a study was run as an 'exact' replication of N. Squires et al. (1975) in order to collect data relevant to three hypotheses. First, P3b and P3a, would show some similar and some different effects of experimental manipulations by virtue of their association with localized and generalized orienting. Second, P3b and P3a, would show different scalp distributions-the P3b characteristically parieto-central, while the P3a would be more anterior, either centro-parietal or centro-frontal as reported by N. Squires et al. (1975).
Third, explaining some of the discrepancies in the literature, the emergence and definition of P3a as a separate component in the PCA would be a function of the PCA methods employed. N. Squires et al. (1975) identified six components based on deflections in the ERP waveforms and measured their amplitudes with traditional 'peak-picking' algorithms. These component scores were then subjected to the PCA procedure. K. Squires et al. (1977), on the other hand, submitted the entire ERP waveform to PCA and allowed the factor analysis to identify components and derive component scores. It would not be surprising to learn that P3a, when submitted to PCA as one of six component scores might survive as an independent factor while P3a as a high frequency portion of a complex, multi-component waveform might not emerge as one of a small number of principal components.
In the present study (Chen, Simons & Graham, 1994), two 70 dB pure tones with frequencies of 1000 Hz and 1500 Hz and durations of 50 ms were used as stimuli and subjects were intructed to count 'high', count 'low' or ignore (read a magazine). Under each of these three instructional conditions, the probability of a 'high' (1500 Hz) tone was either 0.9, 0.5 or 0.1, for a total of nine experimental conditions, each delivered twice. The order of condition presentation was fixed, but balanced across two halves of the experimental session. When high-tone probability was either 0.9 or 0.1, the condition contained 200 trials. When high-tone probability was 0.5, the number of trials was 100. In each block, all single-trial ERPs associated with the 0.1 probability tone were stored, along with the same number of trials associated with the 0.9 and the 0.5 probability tones.
Monopolar EEG was recorded from midline frontal (Fz), vertex (Cz) and parietal (Pz) electrode sites. ERP waveforms were obtained by sampling each of the EEG channels at 500 cps and then corrected for vertical eye movements and eyeblinks (Miller et al., 1988).
Grand mean ERP waveforms for the low- and high-probability tones are presented in Figure 5. A computer algorithm was developed which identified and
Figure 5 Grand mean ERP waveform for the high (0.9)- and low (0.1)-probability stimuli across condition (attend rare, attend frequent, ignore) and electrode location (Fz, Cz, Pz). The labels identify the components that were measured as peaks. Slow wave (SW; not indicated) was an area measure (420-520 ms) subsequent to P3b.
Figure 6 Factor loadings obtained from the PCA of the six peak scores (PCA6) and the PCA using the entire 600 ms waveform (PCA600). The juxtaposition of the loadings was done by aligning the six discrete PCA6 loadings with the times in ms where the PCA factors were maximal.
scored the amplitude of N1, P2, N2, P3a, P3b and SW in each of
the constituent ERP waveforms. These component scores and the
entire 600 ms ERP waveform were then submitted to separate PCAs.
A comparison of the loadings produced by submitting the individual
component scores (solid lines) and the 600 ms waveform (dashed
lines) is provided graphically in Figure 6. The two methods of
component identification yielded similar, though not identical,
results. The six components identified from the peak amplitudes
(as per N. Squires et al., 1975) are comparable to six of the
first ten components extracted by PCA when the entire waveform
was factor analyzed (as per K. Squires et al., 1977). The most
noteworthy discrepancy between the two solutions occurs in the
P300 window. Though both PCAs contained evidence of two surface-positive
components in this window, the proportion of variance attributed
to each of the two factors was dependent upon the PCA strategy.
When submitted as one of six component scores, the P3a was the
first factor identified and accounted for 22% of the total variance.
P3b emerged as Factor 5 and accounted for 13% of the variance.
In the 600 ms, equal-interval, method, P3b was the first factor
extracted and accounted for 43% of the variance. P3a was identified
with Factor 8 and accounted for only 3% of the variance. Treating
the component scores as if they were spaced at equal intervals
(N. Squires et al., 1975), therefore, appears to overestimate
the contribution of P3a to the ERP waveform and, at the same time,
underestimate the contribution of P3b.
These two PCA solutions were virtually identical to those described by N. Squires et al. (1975) and K. Squires et al. (1977) and the use of the two different PCA techniques can easily account for the two different conclusions reached by the authors regarding the presence and significance of the P3a. Specifically, it appears that in the equal-interval, 600 ms PCA, the P3a, a relatively high-frequency component, is generated late and accounts for little variance when forced to compete with low frequency components, such as P3b and SW, that consist of many more data points. It would follow, then, that factor scores derived from factors which themselves account for little variance might also vary within a narrow range and show limited sensitivity to the experimental manipulations. It was this insensitivity that led K. Squires et al. (1977) to conclude that a positive factor observed at 250 ms (Factor 6) was not a P3a.
To examine this issue further, we computed an additional PCA on the 200 ms portion of the ERP waveform (220-420 ms) that contained the P300 complex. Again, both P3a and P3b emerged as independent factors, though in this case, P3a accounted for 20% of the variance-comparable to what it achieved when submitted as one of six component scores. Figure 7 depicts the P3a and P3b factor scores derived from the three PCAs as a function of electrode location and the target/nontarget status of the low-probability stimulus. The left-hand panels depict the centro-parietal scalp distribution of the early, P3a component. This topography was significantly quadratic when the P3a was a peak score submitted to PCA or when derived by PCA as a component in the narrow (220-420 ms) window. Note the lack of sensitivity in the P3a when extracted from the entire 600 ms data set in the manner
Figure 7 P3a and P3b factor scores plotted as a function of electrode location and the target/nontarget manipulation when the probability of the stimulus was low (0.1). Note how, relative to the other two conditions, the PCA600 poorly represents the P3a factor and the PCA6 poorly represents the P3b factor.
of K. Squires et al. (1977). Though the P3b effects are less dramatic,
it is nonetheless evident that underestimating the 'true' variance
accounted for by P3b is also reflected in the derived factor scores.
P3b was consistently larger in reponse to targets, but note how
the topography of P3b obtained from the peak-score factor analysis
appears less well defined then it is when the P3b measure is obtained
from the two equal-interval PCAs. In fact, there was not a significant
difference among the three scalp sites for the P3b factor in this
analysis. Interestingly, N. Squires et al. (1975) also reported
an "equipotentiality of the P3b factor across the
We believe that the results of this study replicate both N. Squires et al. (1975) and K. Squires et al. (1977). We believe further that these data confirm the existence of an anterior P3a component that is distinct from the later, though overlapping, P3b based on the four criteria for independence discussed by Fabiani, Gratton, Karis and Donchin (1987). In short, the two components 1) differed in morphology, 2)differed in scalp distribution, 3) were independent factors in each of three factor PCA analyses, and 4) responded differentially to the experimental manipulations.
Though this is only a first step in determining whether P3a and P3b can be associated with nonsignal and signal orienting, it is at least clear that the two ERP components are sensitive to probability. That is, both P3a and P3b were elicited by unexpected and unpredictable, low-probability events. Finding that P3b is more responsive to rare targets than to rare nontargets is compatiblewith P3b being associated with, or a component of the localized OR. Finding P3a, on the other hand, sensitive only to stimulus probability, is compatible with an association between P3a and generalized orienting (non-selective, passive attention). This conclusion is tentative, of course, since the present study differed in many ways with those traditionally designed to measure and investigate orienting. Perhaps most importantly, it was not designed to measure habituation, a key concept in orienting theory, and a criterion for discriminating components associated with generalized versus localized ORs. Nonetheless, this successful evocation of two independent P3 components can provide a platform from which a more targeted series of studies can be intitated.
In a parallel set of experiments, we have also been pursuing the
anterior P300 described originally by Courchesne et al. (1975).
This series of studies was designed to determine the conditions
under which unattended, novel, stimuli, embedded with targets
in a series of standards, would be associated with an anterior
P300 response that was distinct from the posterior P300 prompted
by attended targets. To examine more closely the possible association
between the P300 potentials and orienting, simultaneous measurement
of beat-by-beat changes in heart rate were obtained.
In our first three experiments (Simons, Balaban, Macy & Graham, 1986; Miles, Perlstein Simons & Graham, 1987; Simons, Graham, Miles & Balaban, in preparation; Simons, Graham & Rockstroh, in preparation), all stimuli were simple tones, white noise bursts, light flashes and air puffs. Novel stimuli were unannounced and of low probability. Targets were equal in probability to novels (p=.08), but required either overt or covert behavior from the subject when identified. For example, in one experiment, standard stimuli were a train of 1000 Hz pure tones, while the target was either a higher pitched (1500 Hz) pure tone or a white noise. In this case, when the high-pitched tone was the target, the white noise was novel; the high-pitched tone was novel when white noise was the target. The other two experiments were variations on this basic design across the three experiments (e.g. interstimulus intervals varied from 2 to 16 s; target/novel distinctions were ipsi- or cross-modal, target response requirements varied), the data contained in Figure 8 are representative.
Figure 8 ERP waveforms at each midline electrode site for target, simple novel, initial standards and standards matched with targets and novels on ordinal position. The heart-rate response to the same stimuli is presented on the right.
In short, target, novel and even the intitial six presentations
of the frequently occurring standard stimulus prompted only a
posterior P300 (P3b). There was no evidence at all of an anterior
P300 response associated with novelty. These conclusions held
whether the ERP waveform was scored by conventional 'peak picking'
or whether components were identified through factor analysis.
Despite this lack of sensitivity of P300, other measures consistently
differentiated the targets and novels. The slow wave (SW) component
was significantly more positive at Pz in response to targets than
it was to both novels and standards, and as illustrated in Figure
8, heart rate was also sensitive to the target/novel distinction.
Though the presentation of target, novel, and even intitial standard
stimuli was accompanied by a short-latency heart-rate deceleration,
the target stimulus prompted a subsequent and substantial acceleration
that was not present in the cardiac response to either novels
or standards. In these experiments, then, while target and novel
stimuli did prompt differential HR and ERP responses, the ERP
differences were subsequent to P300 and it was the target stimulus
that was unique; beyond the P300 window, the ERP response to novels
essentially matched the response associated with the initial and
remaining high-probability standards.
Our next experiment (Miles, 1992) assessed whether the emergence of an early, anterior P300 was a function of stimulus recognizability/complexity. In the three previously-described experiments, all stimuli were common, easily recognized and encoded, laboratory sights and sounds. These have not been the usual stimuli employed to study the novels P3. Courchesne (Courchesne et al., 1975; Courchesne, 1978) and others (e.g. Knight, 1984; Knight, Scabini, Woods & Clayworth, 1989; Friedman & Simpson, 1994) studying ERPs to 'novel' stimuli have normally employed stimuli that were essentially unrecognizable, such as computer generated sounds, synthesized dog barks and complex color geometrical designs. In other words, novel stimuli have been both unpredictable (rare) and unrecognizable.
As an attempt to explore the issue of stimulus specificity, we obtained a copy of the Courchesne et al. (1975) stimulus set. The target stimulus in this set was the synthesized word "you" (p=.10), the standard was the word "me" (p=.80) and the novels were complex mixtures of natural sounds, mechanical noises and digitally synthesized nonsense sounds (p=.10). These stimuli were presented at 75 dB, for 200 ms, to 24 undergraduate volunteers (12 female, 12 male) at SOAs of 1900 ms. Stimuli were presented in three blocks. The first block consisted only of standards and targets. Blocks 2 and 3 contained standards, targets and novels. In either Block 2 or Block 3, the novel stimuli varied from trial-to-trial; in the other block, balanced across subjects, one of the complex novels was used repetitively.
Monopolar EEG from three midline electrode sites (Fz, Cz, Pz), vertical EOG and heart rate were recorded. The EOG-corrected (Miller et al., 1988) ERPs and the heart-rate waveforms averaged across Blocks 2 and 3 are presented in Figure 9. The most striking aspect of the ERP data was the anterior
Figure 9 ERP waveforms at each midline electrode site for the target and complex novel stimuli. The heart-rate response to the same stimuli is presented on the right.
negative/positive complex (N2b/P3a) that preceded and merged with the more parietal P3b. To distinguish the early and later P300s and determine the independence of the N2b and P3a portions of the waveforms, the 750 ms ERP waveform was submitted to PCA. The factor loadings identified as the novels P3 and P3b are presented in the left-hand panel of Figure 10 (an independent N2 component was also identified), while the component scores as a function of Electrode Site and Stimulus (Target/Novel) are presented on the right. The P3b factor peaked at approximately 330 ms following stimulus onset and, as expected, was largest at Pz . It was equally large in response to targets and novels. In contrast, the factor identified as the novels P3 peaked at approximately 250 ms. It had a centrally maximal scalp distribution and was significantly larger in response to novels than to targets. Consistent with the full-waveform PCA in our P3a study, the P3b in this study was the first factor extracted and accounted for 37% of the total variance, while the novels P3 was extracted as Factor 5 accounted for only 5% of the variance.
Figure 10 The principal-component loadings identified as P3a and P3b (left) and their associated factor scores (right) plotted as a function of electrode for the target (solid line) and complex novel (dotted line) stimulus.
Once again, both the Slow Wave and the heart-rate response covaried
with the target/novel manipulation. Slow Wave, most positive at
Pz, was more positive in response to targets than novels. Consistent
with theoretical differences between passive and generalized orienting,
the heart-rate response to target stimuli consisted of a brief
deceleration and rapid return to baseline while the response to
novel stimuli was a deeper and more sustained heart-rate slowing.
The novels P3 and the P3b observed in the present study were strikingly similar to the P3a and P3b observed in the experiment previously described. Based on factor analyses of the entire ERP waveforms, P3b had a posterior scalp distibution in both studies and accounted for 43% and 37% of the total variance in the two studies respectively. P3a and the novels P3 both had centro-parietal scalp distributions and accounted for 3% and 5% of the variance. Although Couchesne (1978) has argued that P3a and the novels P3 are not the same component, the results of our factor analyses do not allow for such a differentiation.
The heart-rate data collected in this set of experiments provide some additional support for the hypothesized relationship between the two P300s and localized v. generalized ORs. Orienting theory (see Cook & Turpin, this volume) would predict that novelty (generalized orienting) would be associated with heart-rate deceleration while target detection (localized orienting) would prompt a more acceleratory response. This is precisely the pattern of results observed in the present set of studies. HR acceleration occurred in response to targets and was associated with a large amplitude P3b and a more positive slow wave. A pronounced deceleratory response occurred only in conjunction with the anterior P3 and only when the stimuli were highly novel. Interestingly, the P3b was evident in response to the complex novels as it had been in the ERP response to the simple novels that did not give rise to an
---------- Table 1 ----------
anterior P3. Table 1 summarizes these relationships and suggests the following scenario: All stimuli receive the same initial processing, indexed by equivalent ERP components through the N1-P2 vertex response. That is the extent of processing afforded to frequently occurring standard stimuli. Initial standard, simple novel, and target stimuli prompt a brief bradycardia and a P3b. This context updating response (Donchin, 1981) appears to be the terminus for the processing of initial standards and simple novels. Target stimuli go on to prompt SW and heart-rate acceleration that in this context most likely reflects the mobilization of response processes. Complex novels are associated with P3b, as were the simple novels, but, in addition, give rise to both the P3a and heart-rate deceleration. This association between the P3a and cardiac deceleration is further evidence of a link between the anterior P3 and generalized orienting and suggests that unlike targets, which seem to prompt response processes subsequent to P3b (SW, HR accel), these complex novels give rise to sensory enhancement processes (HR decel) that facilitate further stimulus intake.
The research described in this chapter was conducted, in part,
to examine the nature and information-processing effects of two
reflexive mechanisms -- 1) the prepulse inhibition of startle
and 2) the non-signal orienting response. PPI has been described
as a gating mechanism that protects the processing of information
in preattentive stores from interruption due to newly arriving
stimuli. The work described in this chapter is consistent with
this description. Lead stimuli not only inhibit the eyeblink and
other indices of startle, but also impact on higher-level processes
as reflected in the ERPs to subsequent stimuli and their perception.
The findings that these effects occur regardless of prepulse modality,
that ERP components are not modulated by prestimuli prior to 50
ms and that modulation differs among the various N100 components
suggest that this 'gating' occurs at a point somewhere beyond
the convergence of somatosensory and acoustic stimulus processing.
The finding that the apparent loudness of S1 is not reduced, even
when followed shortly by a powerful S2 is consistent with Graham's
(1975) hypothesized protective function of the PPI gate, but additional
studies will be necessary to more fully understand the complex
array of perceptual data we have collected to date and how these
perceptual effects relate to eyeblink and ERP component modulation.
Likewise, we continue to pursue the elusive frontal P300 component. Our studies to date have confirmed the existence of an independent anterior positivity under conditions at least superficially evocative of an OR interpretation. Furthermore, the PCA results suggest that the anterior P300s obtained using the Courchesne (Courchesne et al., 1975) and N. Squires (Squires et al., 1975) procedures are similar if not identical. Ongoing studies have been designed to address more directly the equivalence question (P3a v. novels P3) and to flesh out more completely the relationships among the P300 components and the varieties of attention/orienting with which they are presumably associated.