Cop\ right 1992 h\ the American Psychological Association. Inc. 0096-1523/92/S.VOO

Journal of Experimental Psychology Human Perception and Performance 1992. Vol. IS. No. 4. 1 1 2 1 - 1 1 3 8

Effects of Preliminary Perceptual Output on Neuronal Activity of the Primary Motor Cortex Jeff Miller

Alexa Riehle and Jean Requin Cognitive Neuroscience Laboratory, Marseille, France

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

University of California, San Diego

Observations of single neurons in the primary motor cortex of 1 monkey provided evidence that preliminary perceptual information reaches the motor system before perceptual analysis is complete. Neurons were recorded during a task in which 1 stimulus was assigned to a wrist flexion response and another was assigned to wrist extension. Two stimuli were assigned to a nogo response; each was visually similar to either the flexion or the extension stimulus. When a nogo stimulus was presented, neurons responded with weaker versions of the discharge patterns exhibited to the visually similar stimulus requiring a movement, suggesting that neurons receive partial perceptual information favoring that movement. Functionally separable neuronal populations were identified, and differences in the activations of these provide evidence about the functional effects of preliminary perceptual output on movement control processes.

A basic tenet of research into elementary, speeded decision making is that a choice reaction task requires the operation of a number of functionally distinct mental processes or stages such as perception, decision making, and response organization (e.g., Sanders, 1980; Smith, 1968). For logical and neurophysiological reasons, these stages are thought to exist within a sequential architecture, with the output of one stage serving as the input to the next (Kolb & Whishaw, 1985; Sternberg, 1969). The nature of the temporal relations among successive processes has been a subject of intense discussion in recent years, however. At one extreme are classic discrete models (e.g., Donders, 1868/1969; Sternberg, 1969), in which it is assumed that each stage must finish before the next can begin. In these models, each stage transmits to its successor a single output, containing all of its results, at the moment when it finishes. Hence, the stages operate in strict succession, with no temporal overlap. At the other extreme are recent, neurally inspired, continuous models (e.g., Eriksen & Schultz, 1979; McClelland, 1979), in which it is assumed that stages transmit arbitrarily

This work was begun while Miller was a visiting researcher in the Cognitive Neuroscience Unit of the Laboratory of Functional Neuroscience at the Centre Nationale Recherche Scientifique in Marseille. Riehle and Requin were supported by Grant N 00014-89-J-1557 from the Office of Naval Research. Portions of these data were presented in Riehle, Miller, and Requin (1991) and at Analytic Approaches to Human Cognition, an international conference in honor of Paul Bertelson, held on June 25-29, 1991, in Brussels. The data are discussed relative to models of the psychological refractory period in the report of the latter conference (Requin, Miller, & Riehle, in press). We thank Michael Coles, Steven Hackley, Gordon Logan, Allen Osman, and an anonymous reviewer for helpful comments on an earlier draft of the article. Correspondence concerning this article should be addressed to Jeff Miller, Department of Psychology, University of California, San Diego, La Jolla, California 92093-0109. or Alexa Riehle or Jean Requin, Cognitive Neuroscience Laboratory, 31 Chemin Joseph Aiguier, 13402 Marseille Cedex 9, France. 1121

small units of preliminary information. Using tiny amounts of preliminary output, a stage can begin very soon after its predecessor does, so there is almost complete temporal overlap of logically successive stages. Various intermediate possibilities also exist (see Miller, 1988, for a review). For example, Miller (e.g., 1982, 1983, 1988) has argued for an asynchronous discrete coding (ADC) model, in which stages transmit sizable chunks of information, with each chunk corresponding to a single, discretely coded stimulus attribute (e.g., color, size, shape). In this model, a stage begins as soon as it receives information about a single stimulus attribute, without waiting for information about all attributes. This would generally introduce a considerable delay between the onsets of successive processes (i.e., time enough to finish processing the fastest attribute), contrary to the most extreme continuous models, yet still allow some temporal overlap of successive processes, contrary to the most extreme discrete ones. Although there are also other important differences between discrete and continuous models as they are typically formulated (cf. Miller, 1988), the question of whether stages operate in strict sequence versus with some temporal overlap is probably the most fundamental with respect to the interpretation of reaction time (RT). Thus, a number of experimental paradigms have been developed in order to obtain empirical results relevant to the issue of discrete versus continuous transmission of information from stage to stage. Most have been directed toward the question of whether response preparation can begin before perceptual analysis of a stimulus finishes. (For reviews see Coles, 1989; Miller, 1988; and van der Molen, Bashore, Halliday, & Callaway, 1991; for recent examples see Miller & Hackley, 1992; Miller, Schaffer, & Hackley, 1991; Osman, Bashore, Coles, Donchin, & Meyer, 1988, 1992; Smid, Lamain, Hogeboom, Mulder, & Mulder, 1991; and Smid, Mulder, & Mulder, 1990). Results of studies using both RT and event-related potentials (ERPs) observed in average electroencephalographic (EEG) activity suggest that, under certain circumstances, the perceptual stage transmits early information about certain stimulus properties to decision and response preparation processes before analysis

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1122

J. MILLER, A. RIEHLE, AND J. REQUIN

of all relevant stimulus properties is complete. Specifically, in accordance with the ADC model, it appears that the perceptual system can transmit information about an easily discriminable and discretely codable stimulus attribute even while it continues to process a more difficult to discriminate attribute (e.g., Miller & Hackley, 1992; Osman et al., 1988, 1992). Some of the most direct and compelling evidence that response processes can begin before perceptual analysis is finished comes from a "left/right, go/no-go" paradigm developed independently by Miller and Hackley (1992; Hackley & Miller, 1990) and Osman et al. (1988, 1992). The minimal stimulus set for this paradigm consists of two pairs of stimuli defined by high visual similarity within pairs and low similarity between pairs (cf. Miller, 1982). For example, Miller and Hackley (1992, Experiments 2 and 3) presented subjects with four stimulus letters varying in name and size (i.e., S, S, T, and T). The size discrimination was adjusted to be much harder than the letter name discrimination, so that the perceptual analysis of a single stimulus letter would yield preliminary information about letter name well before complete information (i.e., name and size) was available. Three response alternatives are needed: left hand, right hand, and nogo (i.e., withhold the response), with one member of each visually similar stimulus pair assigned to each manual response and one member of each pair assigned to the no-go response. For example, Miller and Hackley (1992) required some subjects to respond with the left hand to S, to respond with the right hand to T, and to withhold the response to S or T. In this paradigm, the key predictions concern no-go trials. If response preparation begins before perceptual analysis is finished, consistent with continuous models, then the two response hands should be differentially activated. For example, when s is assigned to the left-hand response and S is assigned to the no-go response, the left hand should become more activated than the right hand when the S is presented. Early perceptual analysis should indicate that the stimulus is an "ess" before its size has been determined, simply because name is more easily discriminated than size. Under the assumption that response preparation begins as soon as any stimulus information is perceived, the left hand should become more activated than the right, because the left hand is a response alternative consistent with the early perceptual information and the right hand is not. If response preparation does not begin until perceptual analysis is finished, consistent with discrete models, then the left and right hands should remain equally activated. By the time stimulus analysis is finished, complete information is available to indicate that the correct response is to do nothing, so the response system never receives any differential information favoring the left versus the right response. The lateralized readiness potential (LRP) was used to determine whether no-go trials produced differential response preparation. In brief, the LRP is simply the difference in EEG activity detected by electrodes placed over the left and right sensorimotor cortex, and there is substantial evidence that it is a relatively direct measure of differential response preparation. (See Coles, Gratton, Bashore, Eriksen, & Donchin, 1985, or Coles, 1989, for reviews.) For example, when the right hand is more prepared for a response than the left hand, there

is more activity in the EEG over the left sensorimotor cortex than over the right, because of contralateral hand control. Miller and Hackley (1992) found evidence of differential response preparation on no-go trials, as did Osman et al. (1988, 1992) using a quite different stimulus set. Specifically, there was more EEG activity at the electrode contralateral to the responding hand associated with the preliminary information in the no-go stimulus (e.g., letter name) than at the other electrode. This contralateral excess, that is, the LRP, started approximately 250 ms after presentation of a no-go stimulus and lasted approximately 250 ms, as would be expected if preliminary information about an easy perceptual discrimination (e.g., letter name) caused some preliminary response preparation that dissipated when complete stimulus information (e.g., size) later became available to the perceptual system. The present article reports an attempt to extend the evidence of temporal overlap of perceptual and motor processes using single cell recording in primary motor cortex (MI). The results of the ERP studies reported by Miller and Hackley (1992) and Osman et al. (1988, 1992) mesh well with indirect behavioral evidence, but their interpretation is limited by the fact that the ERP is an aggregated measure of the activity of a large population of neurons (e.g., Meyer, Osman, Irwin, & Yantis, 1988). A demonstration of analogous phenomena in the behavior of single neurons would not only provide especially strong converging evidence but also provide an opportunity to gain new information through detailed analyses of the activities of individual members of this neuronal population. MI was chosen for study because it is intimately involved in the preparation (Evarts, Shinoda, & Wise, 1984; Lecas, Requin, Anger, & Vitton, 1986; Riehle & Requin, 1989) and execution (Evarts, 1981; Georgopoulos, 1991; Hepp-Reymond, 1988) of movements. For example, anticipation of a particular movement clearly produces substantial changes in the activity of MI neurons (e.g., Evarts et al., 1984), generally but not exclusively the specific neurons that are most active when that movement is carried out (Riehle, 1987; Riehle & Requin, 1989). The causal connection of this preparatory neuronal activity to movement execution seems clear from the finding that RT is negatively correlated with the magnitude of anticipatory change (Lecas et al., 1986). The study of single-neuron activity offers the opportunity to determine more precisely what types of movement-related activity can begin before stimulus analysis is finished, because there is evidence for three distinctly different types of units in MI (Lecas et al., 1986; Riehle, 1987; Riehle & Requin, 1989), as well as in other cortical structures involved in motor control (Seal, 1989; Seal & Requin, 1987). As reviewed by Requin, Riehle, and Seal (1988, in press), there appear to be input, or sensory, neurons that accept relevant stimulus information from other modules (e.g., perception); output, or motor, neurons that drive movement and perhaps send information to other brain areas as well; and interfacing, or sensorimotor, neurons that connect the two. Sensory neurons respond when the animal is given information indicating that a certain movement will be required (e.g., warning signals: Lecas et al., 1986; partial movement specifications: Riehle, 1987, and Riehle & Requin, 1989), even if the movement is not required

1123

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

PRELIMINARY PERCEPTUAL OUTPUT AND MOTOR CORTEX

immediately. In fact, after practice in a stimulus-response (SR) task, sensory neurons respond to presentation of the stimulus even if the task is changed so that the response is no longer permitted (Seal, 1989). Conversely, motor neurons only become active at a point just before overt movement begins, regardless of the task. Sensorimotor neurons appear to play the role of an intermediary, showing activity both at the presentation of a stimulus associated with a movement and in the interval just before and during overt movement, and may perhaps be responsible for implementing S-R associations. Because it is a measure of aggregate activity, of course, the LRP cannot indicate which type(s) of neurons in motor areas are sensitive to preliminary perceptual information in the left/ right, go/no-go paradigm used by Miller and Hackley (1992) and Osman et al. (1988, 1992). Given the apparent functional distinctions between these types of neurons, we must conclude that the functional meaning of the neuronal activation revealed by the LRP remains unclear. In view of this limitation, the present experiment was designed to directly examine these three types of neurons in a comparable task, in order to better identify the nature of the cortical processes carried out using preliminary perceptual information. The present study employed a variant of the left/right, go/ no-go paradigm used by Miller and Hackley (1992) and Osman et al. (1988, 1992). Stimuli were four lights varying in position, presented one at a time, with one pair of lights on each side of fixation. One light in each pair was associated with a pointing response (i.e., wrist flexion or extension to align a pointer with the light), and the other light was associated with a no-go response. The two lights in a given pair were very close to each other, so the within-pair discrimination was much more difficult than the between-pair discrimination. It was therefore anticipated that early perceptual information would indicate the possible direction of movement (i.e., flexion vs. extension) on the basis of the easy judgment of left versus right of fixation. The go/no-go decision should take much longer, because it requires localizing an individual light within a pair, which should take longer to determine. The question was what types of neuronal activity would take place in MI on no-go trials. If preliminary perceptual information is relayed to the motor areas before perception is complete, then there should be some electrophysiological evidence of preparation of the appropriate pointing response (flexion or extension). It is known that directionspecific preparatory neural discharges are observed when an explicit advance cue indicates the direction of an upcoming movement (Richie & Requin, 1989), although the parameters of the cuing paradigm do not permit the classification of the active neurons as sensory, sensorimotor, or motor. Nonetheless, this evidence strongly suggests that single-unit preparatory effects should be observed in this paradigm if the preliminary perceptual information can be used by the motor system in the same way as a cue is used. Method

Behavioral Task One monkey (Macaca fascicularis) was used and cared for in the manner described in the Guide for the Care and Use of Laboratory

Animals (U.S. Department of Health and Human Services, 1985). The animal was trained to grasp and rotate a vertical handle by performing wrist flexion and extension movements in the horizontal plane. A rod fixed on the handle served as a pointer moving in front of a semicircular, vertical panel on which two horizontal rows of light-emitting diodes (LEDs) were displayed, one above the other (see diagram inset in Figure 1). Two centrally located and continuously illuminated yellow LEDs, one in each row, indicated the starting position for all movements. The lower row also contained two green LEDs, at 40° angular rotation of the handle to the left and right of the starting position, that were jointly illuminated to provide a warning signal (WS). The upper row contained four red LEDs, two above the left WS and two above the right WS, that served as response signals (RS) and movement targets. On each side, the two red LEDs were separated by 4.5° of angular rotation (less than 1° of visual angle). To begin each trial, the monkey first had to rotate the handle so that the pointer was placed at the starting position and to maintain the pointer at this position for 1 s. After 1 s, both green LEDs were illuminated for 0.8 s as a WS. The WS was followed by a preparatory period whose duration (1 or .5 s) was constant within a session. At the end of the preparatory period, one of the four red LEDs was illuminated as an RS. When a distal LED was illuminated, the correct response was to make a rapid flexion or extension movement to align the pointer with the illuminated LED (go trials). When a proximal LED was illuminated, the correct response was to maintain the pointer at the starting position (no-go trials). The animal was tested in near darkness, so that the nonilluminated LEDs were not visible and could not be used as landmarks. The purpose of this was to make the discrimination between proximal and distal LEDs (i.e., go vs. nogo trials) rather difficult.

Reaction Times

200

300

400

500

600

reaction time (ms) Figure 1. Cumulative probability distributions of reaction time for flexion and extension movements. (Inset = experimental setup; empty circles = warning signals; filled circles = response signals; asterisks = starting position.)

A

' :.".\\;" " •• v. •* •. . . ij; —.-.."•.''.i .'.•'!»,'.._, .-.Iv •'• •.'.- «.»•' •'•• /"!T7-» '• ', .v*

Effects of preliminary perceptual output on neuronal activity of the primary motor cortex.

Observations of single neurons in the primary motor cortex of 1 monkey provided evidence that preliminary perceptual information reaches the motor sys...
2MB Sizes 0 Downloads 0 Views