Motor cortical encoding of serial order in a context-recall task

Science

Volumn 283, Issue 5408, 1999, Pages 1752-1757

  Carpenter, Adam F ^a   Georgopoulos, Apostolos P ^a   Pellizzer, Giuseppe ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

ANIMAL EXPERIMENT; ARTICLE; CONTROLLED STUDY; ELECTROMYOGRAM; EYE POSITION; MONKEY; MOTOR CORTEX; MOTOR PERFORMANCE; NERVE CELL; NONHUMAN; PRIORITY JOURNAL; RECALL;

ANALYSIS OF VARIANCE; ANIMALS; ELECTROPHYSIOLOGY; FIXATION, OCULAR; HAPLORHINI; MEMORY; MENTAL RECALL; MICROELECTRODES; MOTOR ACTIVITY; MOTOR CORTEX; NEURONS; PHOTIC STIMULATION; PSYCHOMOTOR PERFORMANCE;

EID: 0033548716     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.283.5408.1752     Document Type: Article

References (38)

1. K. S. Lashley, in Cerebral Mechanisms in Behavior, L. A. Jeffress, Ed. (Wiley, New York, 1951), pp. 112-136.
2. P. Barone and J.-P. Joseph, Exp. Brain Res. 78, 447 (1989); H. Mushiake, M. Inase, J. Tanji, J. Neurophysiol. 66, 705 (1991), S. Funahashi, M. Inoue, K. Kubota, Neurosci. Res. 18, 171 (1993); I. Kermadi and J.-P. Joseph, J. Neurophysiol. 74, 911 (1995); H. Mushiake and P. L. Strick, ibid., p. 2784; R. E. Kettner, J. K. Marcario, N. L. Port, Exp. Brain Res. 112, 347 (1996).
3. P. Barone and J.-P. Joseph, Exp. Brain Res. 78, 447 (1989); H. Mushiake, M. Inase, J. Tanji, J. Neurophysiol. 66, 705 (1991), S. Funahashi, M. Inoue, K. Kubota, Neurosci. Res. 18, 171 (1993); I. Kermadi and J.-P. Joseph, J. Neurophysiol. 74, 911 (1995); H. Mushiake and P. L. Strick, ibid., p. 2784; R. E. Kettner, J. K. Marcario, N. L. Port, Exp. Brain Res. 112, 347 (1996).
4. P. Barone and J.-P. Joseph, Exp. Brain Res. 78, 447 (1989); H. Mushiake, M. Inase, J. Tanji, J. Neurophysiol. 66, 705 (1991), S. Funahashi, M. Inoue, K. Kubota, Neurosci. Res. 18, 171 (1993); I. Kermadi and J.-P. Joseph, J. Neurophysiol. 74, 911 (1995); H. Mushiake and P. L. Strick, ibid., p. 2784; R. E. Kettner, J. K. Marcario, N. L. Port, Exp. Brain Res. 112, 347 (1996).
5. P. Barone and J.-P. Joseph, Exp. Brain Res. 78, 447 (1989); H. Mushiake, M. Inase, J. Tanji, J. Neurophysiol. 66, 705 (1991), S. Funahashi, M. Inoue, K. Kubota, Neurosci. Res. 18, 171 (1993); I. Kermadi and J.-P. Joseph, J. Neurophysiol. 74, 911 (1995); H. Mushiake and P. L. Strick, ibid., p. 2784; R. E. Kettner, J. K. Marcario, N. L. Port, Exp. Brain Res. 112, 347 (1996).
6. P. Barone and J.-P. Joseph, Exp. Brain Res. 78, 447 (1989); H. Mushiake, M. Inase, J. Tanji, J. Neurophysiol. 66, 705 (1991), S. Funahashi, M. Inoue, K. Kubota, Neurosci. Res. 18, 171 (1993); I. Kermadi and J.-P. Joseph, J. Neurophysiol. 74, 911 (1995); H. Mushiake and P. L. Strick, ibid., p. 2784; R. E. Kettner, J. K. Marcario, N. L. Port, Exp. Brain Res. 112, 347 (1996).
7. P. Barone and J.-P. Joseph, Exp. Brain Res. 78, 447 (1989); H. Mushiake, M. Inase, J. Tanji, J. Neurophysiol. 66, 705 (1991), S. Funahashi, M. Inoue, K. Kubota, Neurosci. Res. 18, 171 (1993); I. Kermadi and J.-P. Joseph, J. Neurophysiol. 74, 911 (1995); H. Mushiake and P. L. Strick, ibid., p. 2784; R. E. Kettner, J. K. Marcario, N. L. Port, Exp. Brain Res. 112, 347 (1996).
8. S. Sternberg, Psychon. Sci. 8, 55 (1967); Am. Sci. 57, 421 (1969); A. P. Georgopoulos and J. T. Lurito, Exp. Brain Res. 83, 453 (1991); G. Pellizzer and A. P. Georgopoulos, ibid. 93, 165 (1993).
9. S. Sternberg, Psychon. Sci. 8, 55 (1967); Am. Sci. 57, 421 (1969); A. P. Georgopoulos and J. T. Lurito, Exp. Brain Res. 83, 453 (1991); G. Pellizzer and A. P. Georgopoulos, ibid. 93, 165 (1993).
10. S. Sternberg, Psychon. Sci. 8, 55 (1967); Am. Sci. 57, 421 (1969); A. P. Georgopoulos and J. T. Lurito, Exp. Brain Res. 83, 453 (1991); G. Pellizzer and A. P. Georgopoulos, ibid. 93, 165 (1993).
11. S. Sternberg, Psychon. Sci. 8, 55 (1967); Am. Sci. 57, 421 (1969); A. P. Georgopoulos and J. T. Lurito, Exp. Brain Res. 83, 453 (1991); G. Pellizzer and A. P. Georgopoulos, ibid. 93, 165 (1993).
12. G. Pellizzer, P. Sargent, A. P. Georgopoulos, Science 269, 702 (1995).
13. The monkeys were trained to exert a force pulse on a two-dimensional semi-isometric handle in eight different directions (at 45° intervals). The manipulandum was a vertical rigid metal rod, with a disc attached to the top, which was placed in front of the animal in the midsagittal plane and which the animal grasped with the hand pronated. A net force feedback cursor was displayed on a monitor in front of the monkey. This cursor was deflected constantly downward to simulate a bias force of 54g and reflected, at any given moment, the net force (the vector sum of this simulated force and the force exerted by the animal on the manipulandum). At the start of the trial, a white stimulus appeared in the center of the screen and the monkey had to place the force feedback cursor or the center stimulus by exerting a force of 54g in the upward direction and then keep it there within a 11g-radius circular window. After 1 s, three to five yellow stimuli were presented on a 270g-radius circle in different directions. During the presentation of the stimuli, the force feedback cursor had to stay within the central window. The response was considered correct if (i) the threshold of 270g was exceeded, and (ii) the direction of the force pulse stayed within ±22.5° from the correct direction, from the center to the stimulus. The monkey received a liquid reward after each correct trial. Monkey 1 performed the context-recall task using sequences of three and four stimuli, with 400 ms elapsing between stimuli (epoch duration = 400 ms). Monkey 2 performed with sequences of three, four, and five stimuli, with an epoch duration of 650 ms. Sequences of different lengths (three, four, or five stimuli) were presented in a randomized block design for both monkeys. Given eight possible stimulus locations (one every 45°), there are 8!/(8 - 3)! = 336 unique sequences of three stimuli. For each of these sequences, the test stimulus could be either the first or second stimulus (serial positions 1 and 2); thus, a total of 672 unique trials are possible for sequences of three stimuli. Likewise, there are 8!/(8 - 4)! = 1680 sequences with three different serial positions for a total of 5040 unique trials for sequences of four stimuli, and 8!/(8 - 5)! = 6720 sequences with four different serial positions for a total of 26,880 unique trials for sequences of five stimuli. During training, sequences were selected at random from this set of all possible trials, with the constraint that an equal number of correct responses be performed for each serial position of the test stimulus. During neurophysiological recording, a subset of sequences was selected so that five correct trials (repetitions) could be obtained for each combination of sequence and serial position. Monkey 1 worked with a fixed set of sequences (16 sequences of three stimuli and four sequences of four stimuli) during the neural recordings. In the recording sessions with monkey 2, each block consisted of seven randomly generated sequences and one sequence drawn from a pool of two fixed sequences. (We incorporated fixed sequences in the experimental design for the purpose of pursuing population analyses in the future.) No consistent differences in behavior or neural activity were observed between the fixed and random sequences. For both monkeys, the sequences were chosen such that (i) each of the eight possible stimulus locations was presented an equal number of times, (ii) each stimulus location was presented equally often in each serial position, (iii) the test stimulus was presented in each of the eight locations equally often, and (iv) the direction of correct response was equally distributed among the eight directions. Monkey 1 performed correctly in 85% of the trials with sequences of three stimuli and 50% of the trials with sequences of four stimuli. Monkey 2 performed correctly in 79%, 75%, and 47% of the trials with sequences of three, four, and five stimuli, respectively. An incorrect response could be due to moving in the wrong direction or to initiating a motor response outside prespecified temporal limits.
14. The monkey was required to maintain the force feedback cursor within a small center window, and electromyographic (EMG) recordings revealed no significant change of muscle activity during this time (Fig. 1A). EMG activity was recorded with intramuscular, multistranded, Teflon-coated wire electrodes [A. B. Schwartz, R. E. Kettner, A. P. Georgopoulos, J. Neurosci. 8, 2913 (1988)] in separate sessions from the neural recordings. EMG activity was recorded in the following muscles: latissimus dorsi, rhomboids, paraspinal, infraspinatus, supraspinatus, trapezius (lower, middle, and upper), deltoid (anterior, middle, and posterior), pectoralis major, triceps (lateral and long heads), biceps, forearm flexor (unspecified), and forearm extensor (unspecified). Examination of EMG activity showed no change from the baseline level during the list presentation phase of the task; instead, when an increase of signal was observed, it began shortly before or during the motor response.
15. The electrical activity of single motor cortical neurons was recorded extracellularly with seven independently driven microelectrodes [V. B. Mountcastle, H. J. Reitboeck, G. F. Poggio, M. A. Steinmetz, J. Neurosci. Methods 36, 77 (1991); The placement of the recording chamber was done aseptically under general pentobarbital anesthesia. In both monkeys, the neurons were recorded from the bank and crown of the precentral gyrus (contralateral to the performing arm), medial to the genu of the arcuate sulcus and posterolateral to the precentral dimple. All isolated neurons were recorded regardless of their activity during the task. The eye position was monitored with an infrared oculometer (Dr. Bouis, Karlsruhe, Germany] or a scleral search coil (CNC Engineering, Seattle, WA) [A. F. Fuchs and D. A. Robinson, J. Appl. Physiol. 21, 1068 (1966); S. J. Judge, B. J. Richmond, F. C. Chu, Vision Res. 20, 535 (1980)]. Eye position was recorded simultaneously with neural recordings in monkey 2 and in separate recording sessions in monkey 1. A personal computer controlled the task display and data collection. Care and treatment of the animals during all stages of the experiments conformed to NIH's Principles of Laboratory Animal Care (revised 1995). All experimental protocols were approved by the appropriate institutional review boards.
16. The electrical activity of single motor cortical neurons was recorded extracellularly with seven independently driven microelectrodes [V. B. Mountcastle, H. J. Reitboeck, G. F. Poggio, M. A. Steinmetz, J. Neurosci. Methods 36, 77 (1991); D. Lee, N. P. Port, W. Kruse, A. P. Georgopoulos, in Neuronal Ensembles: Strategies for Recording and Decoding, H. Eichenbaum and J. Davis, Eds. (Wiley, New York, 1998), pp. 117-136]. D. Lee, N. P. Port, W. Kruse, A. P. Georgopoulos, in Neuronal Ensembles: Strategies for Recording and Decoding, H. Eichenbaum and J. Davis, Eds. (Wiley, New York, 1998), pp. 117-136]. S. J. Judge, B. J. Richmond, F. C. Chu, Vision Res. 20, 535 (1980)]. Eye position was recorded simultaneously with neural recordings in monkey 2 and in separate recording sessions in monkey 1. A personal computer controlled the task display and data collection. Care and treatment of the animals during all stages of the experiments conformed to NIH's Principles of Laboratory Animal Care (revised 1995). All experimental protocols were approved by the appropriate institutional review boards.
17. The electrical activity of single motor cortical neurons was recorded extracellularly with seven independently driven microelectrodes [V. B. Mountcastle, H. J. Reitboeck, G. F. Poggio, M. A. Steinmetz, J. Neurosci. Methods 36, 77 (1991); D. Lee, N. P. Port, W. Kruse, A. P. Georgopoulos, in Neuronal Ensembles: Strategies for Recording and Decoding, H. Eichenbaum and J. Davis, Eds. (Wiley, New York, 1998), pp. 117-136]. The placement of the recording chamber was done aseptically under general pentobarbital anesthesia. In both monkeys, the neurons were recorded from the bank and crown of the precentral gyrus (contralateral to the performing arm), medial to the genu of the arcuate sulcus and posterolateral to the precentral dimple. All isolated neurons were recorded regardless of their activity during the task. The eye position was monitored with an infrared oculometer (Dr. Bouis, Karlsruhe, Germany] or a scleral search coil (CNC Engineering, Seattle, WA) [A. F. Fuchs and D. A. Robinson, J. Appl. Physiol. 21, 1068 (1966); S. J. Judge, B. J. Richmond, F. C. Chu, Vision Res. 20, 535 (1980)]. Eye position was recorded simultaneously with neural recordings in monkey 2 and in separate recording sessions in monkey 1. A personal computer controlled the task display and data collection. Care and treatment of the animals during all stages of the experiments conformed to NIH's Principles of Laboratory Animal Care (revised 1995). All experimental protocols were approved by the appropriate institutional review boards. S. J. Judge, B. J. Richmond, F. C. Chu, Vision Res. 20, 535 (1980)]. Eye position was recorded simultaneously with neural recordings in monkey 2 and in separate recording sessions in monkey 1. A personal computer controlled the task display and data collection. Care and treatment of the animals during all stages of the experiments conformed to NIH's Principles of Laboratory Animal Care (revised 1995). All experimental protocols were approved by the appropriate institutional review boards. The square root transformation was applied to these firing rates to stabilize the variance [J. W. Tukey, Exploratory Data Analysis (Addison-Wesley, Reading, MA, 1977), p. 543; D. R. Cox and P. A. W. Lewis, The Statistical Analysis of Series of Events (Chapman & Hall, London, 1966); G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis.
18. The electrical activity of single motor cortical neurons was recorded extracellularly with seven independently driven microelectrodes [V. B. Mountcastle, H. J. Reitboeck, G. F. Poggio, M. A. Steinmetz, J. Neurosci. Methods 36, 77 (1991); D. Lee, N. P. Port, W. Kruse, A. P. Georgopoulos, in Neuronal Ensembles: Strategies for Recording and Decoding, H. Eichenbaum and J. Davis, Eds. (Wiley, New York, 1998), pp. 117-136]. The placement of the recording chamber was done aseptically under general pentobarbital anesthesia. In both monkeys, the neurons were recorded from the bank and crown of the precentral gyrus (contralateral to the performing arm), medial to the genu of the arcuate sulcus and posterolateral to the precentral dimple. All isolated neurons were recorded regardless of their activity during the task. The eye position was monitored with an infrared oculometer (Dr. Bouis, Karlsruhe, Germany] or a scleral search coil (CNC Engineering, Seattle, WA) [A. F. Fuchs and D. A. Robinson, J. Appl. Physiol. 21, 1068 (1966); S. J. Judge, B. J. Richmond, F. C. Chu, Vision Res. 20, 535 (1980)]. Eye position was recorded simultaneously with neural recordings in monkey 2 and in separate recording sessions in monkey 1. A personal computer controlled the task display and data collection. Care and treatment of the animals during all stages of the experiments conformed to NIH's Principles of Laboratory Animal Care (revised 1995). All experimental protocols were approved by the appropriate institutional review boards. S. J. Judge, B. J. Richmond, F. C. Chu, Vision Res. 20, 535 (1980)]. Eye position was recorded simultaneously with neural recordings in monkey 2 and in separate recording sessions in monkey 1. A personal computer controlled the task display and data collection. Care and treatment of the animals during all stages of the experiments conformed to NIH's Principles of Laboratory Animal Care (revised 1995). All experimental protocols were approved by the appropriate institutional review boards. The square root transformation was applied to these firing rates to stabilize the variance [J. W. Tukey, Exploratory Data Analysis (Addison-Wesley, Reading, MA, 1977), p. 543; D. R. Cox and P. A. W. Lewis, The Statistical Analysis of Series of Events (Chapman & Hall, London, 1966); G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis.
19. ANCOVAs were performed for each cell and each sequence size (three to five stimuli) as follows. The set of trials in a cell-sequence size combination was called 3 "case"; a total of 1812 cases were analyzed. For each period of interest (see below) in every correct trial, the firing rate was computed using fractional interspike intervals [M. Taira, J. Boline, N. Smyrnis, A. P. Georgopoulos, J. Ashe, Exp. Brain Res. 109, 367 (1996)]. The square root transformation was applied to these firing rates to stabilize the variance [J. W. Tukey, Exploratory Data Analysis (Addison-Wesley, Reading, MA, 1977), p. 543; D. R. Cox and P. A. W. Lewis, The Statistical Analysis of Series of Events (Chapman & Hall, London, 1966); G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis. G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis.
20. ANCOVAs were performed for each cell and each sequence size (three to five stimuli) as follows. The set of trials in a cell-sequence size combination was called 3 "case"; a total of 1812 cases were analyzed. For each period of interest (see below) in every correct trial, the firing rate was computed using fractional interspike intervals [M. Taira, J. Boline, N. Smyrnis, A. P. Georgopoulos, J. Ashe, Exp. Brain Res. 109, 367 (1996)]. The square root transformation was applied to these firing rates to stabilize the variance [J. W. Tukey, Exploratory Data Analysis (Addison-Wesley, Reading, MA, 1977), p. 543; D. R. Cox and P. A. W. Lewis, The Statistical Analysis of Series of Events (Chapman & Hall, London, 1966); G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis. G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis.
21. ANCOVAs were performed for each cell and each sequence size (three to five stimuli) as follows. The set of trials in a cell-sequence size combination was called 3 "case"; a total of 1812 cases were analyzed. For each period of interest (see below) in every correct trial, the firing rate was computed using fractional interspike intervals [M. Taira, J. Boline, N. Smyrnis, A. P. Georgopoulos, J. Ashe, Exp. Brain Res. 109, 367 (1996)]. The square root transformation was applied to these firing rates to stabilize the variance [J. W. Tukey, Exploratory Data Analysis (Addison-Wesley, Reading, MA, 1977), p. 543; D. R. Cox and P. A. W. Lewis, The Statistical Analysis of Series of Events (Chapman & Hall, London, 1966); G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis. G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis.
22. ANCOVAs were performed for each cell and each sequence size (three to five stimuli) as follows. The set of trials in a cell-sequence size combination was called 3 "case"; a total of 1812 cases were analyzed. For each period of interest (see below) in every correct trial, the firing rate was computed using fractional interspike intervals [M. Taira, J. Boline, N. Smyrnis, A. P. Georgopoulos, J. Ashe, Exp. Brain Res. 109, 367 (1996)]. The square root transformation was applied to these firing rates to stabilize the variance [J. W. Tukey, Exploratory Data Analysis (Addison-Wesley, Reading, MA, 1977), p. 543; D. R. Cox and P. A. W. Lewis, The Statistical Analysis of Series of Events (Chapman & Hall, London, 1966); G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis. G. W. Snedecor and W. G. Cochran, Statistical Methods (Iowa State Univ. Press, Ames, IA, ed. 8, 1989)]. Cells were included in the analyses if they had a mean firing rate of at least 0.1 impulses per second during the task. Data from two periods were analyzed, namely from the list presentation phase and the motor response period, both from the same trials. The list presentation phase comprised three to five epochs corresponding to the stimuli presented (5); the factors in this ANCOVA were Serial Position (number of levels = number of stimuli in the sequence). Location of the most recently presented stimulus (eight levels), and their interaction. In addition, the factor Repetition (five levels) was included to control for the effect of repeated measurement of the same cell, and the control period activity was used as a covariate to adjust for the baseline activity of the cell. The factors in the motor response ANCOVA included Motor Direction (eight levels) and Repetition (five levels); the control period activity was used as a covariate. Effects at a conservative probability level (P < 0.01) were considered statistically significant. Cells were classified as list presentation-related when any of the factors of Serial Position, Location, or their interaction was significant, and as motor-related when the factor Motor Direction was statistically significant in any of the sequence sizes analyzed. The program 4V of the BMDP/Dynamic statistical package (BMDP Statistical Software Inc., Los Angeles, 1992) was used to perform the ANCOVA analysis.
23. The percentages shown in Fig. 1D denote the relative frequency of an effect irrespective of the effects of other factors. On the other hand, exclusive effects can be calculated such that, for example, a "Serial Position" effect will mean that only this factor (and no other factor or interaction term) was statistically significant. This is the most stringent criterion for evaluating the importance of a specific factor on cell activity. These percentages are as follows: the main effect of Serial Position was the only statistically significant effect in 34.4 ± 6.71% of cells [mean ± SEM, N = 5 combinations of monkey and sequence size (5)]; the main effect of stimulus Location was the only significant effect in 0.6 ± 0.18% of cells; and the Serial Position × Location interaction effect was the only significant effect in 4.95 ± 0.54% of cells. Thus, Serial Position was the most frequent statistically significant effect in this analysis too. The cumulative frequency curve of the statistical significance of Serial Position in this case was also significantly higher than the Motor Direction curve (P < 0.0001, Kolmogorov-Smirnov test).
24. The number of simultaneously recorded neurons ranged from 2 to 11 for monkey 1 and from 2 to 16 for monkey 2. A quadratic discriminant analysis was performed [P. A. Lachenbruch, Discriminant Analysis (Hafner, New York, 1975)] because cells in a simultaneously recorded set commonly did not have a common covariance structure, hence a linear discriminant analysis was not appropriate. Percentages of correct classification were computed for each sequence within a list length and averaged across sequences.
25. The average percent of correct classification of serial position for a sequence of three stimuli was 69% (range 45 to 99%), for a sequence of four stimuli it was 64% (range 35 to 99.8%), and for a sequence of five stimuli it was 66% (range 32 to 94%). The chance levels of correct classification were 33.3%, 25%, and 20% for sequences of three, four, and five stimuli, respectively.
26. 2 = 0.995), where N is the number of cells in the ensemble. Extrapolating the power function indicates that an ensemble of as few as 16 motor cortical cells could perfectly classify all the items in a sequence of five stimuli.
27. Careful qualitative inspection of the data, using plots such as in Figs. 3 to 5, indicated that changes in cell activity could not, in general, be accounted for by eye movement, eye position, or stimulation of a specific part of the visual field. In addition, because the successively presented stimuli remained on the screen throughout the trial (Fig. 1A), there was a rich pattern of visual stimulation that differed from epoch to epoch and from sequence to sequence. Exhaustive quantitative analyses of these factors is beyond the scope of this report.
28. For reviews, see A. P. Georgopoulos, Annu. Rev. Neurosci. 14, 361 (1991); _, M. Taira, A. V. Lukashin, Science 260, 47 (1993). The changes in activity during the list presentation phase were not related to concomitant motor events because no motor response occurred during this period (6). Instead, these changes in activity occurred while the monkeys were viewing a sequence of stimuli, whose order they had to transiently remember. It is possible that this activity may reflect intended, but not executed, limb movements. However, if this were the case, we would expect cell activity to reflect the location of the currently presented stimulus, which would determine the direction of a hypothetical motor response. Instead, we found that cell activity was mostly related to serial position in the sequence rather than location. _, M. Taira, A. V. Lukashin, Science 260, 47 (1993). The changes in activity during the list presentation phase were not related to concomitant motor events because no motor response occurred during this period (6). Instead, these changes in activity occurred while the monkeys were viewing a sequence of stimuli, whose order they had to transiently remember. It is possible that this activity may reflect intended, but not executed, limb movements. However, if this were the case, we would expect cell activity to reflect the location of the currently presented stimulus, which would determine the direction of a hypothetical motor response. Instead, we found that cell activity was mostly related to serial position in the sequence rather than location.
29. For reviews, see A. P. Georgopoulos, Annu. Rev. Neurosci. 14, 361 (1991); _, M. Taira, A. V. Lukashin, Science 260, 47 (1993). The changes in activity during the list presentation phase were not related to concomitant motor events because no motor response occurred during this period (6). Instead, these changes in activity occurred while the monkeys were viewing a sequence of stimuli, whose order they had to transiently remember. It is possible that this activity may reflect intended, but not executed, limb movements. However, if this were the case, we would expect cell activity to reflect the location of the currently presented stimulus, which would determine the direction of a hypothetical motor response. Instead, we found that cell activity was mostly related to serial position in the sequence rather than location. _, M. Taira, A. V. Lukashin, Science 260, 47 (1993). The changes in activity during the list presentation phase were not related to concomitant motor events because no motor response occurred during this period (6). Instead, these changes in activity occurred while the monkeys were viewing a sequence of stimuli, whose order they had to transiently remember. It is possible that this activity may reflect intended, but not executed, limb movements. However, if this were the case, we would expect cell activity to reflect the location of the currently presented stimulus, which would determine the direction of a hypothetical motor response. Instead, we found that cell activity was mostly related to serial position in the sequence rather than location.
30. A similarly phasic encoding response has been described for inferotemporal neurons during an object recognition memory task [E. K. Miller, L. Lin, R. Desimone, J. Neurosci. 13, 1460 (1993)].
31. P. S. Goldman-Rakic, J. F. Bates, M. V. Chafee, Curr. Opin. Neurobiol. 2, 830 (1992); D. G. Beiser and J. C. Houk, J. Neurophysiol. 79, 3168 (1998).
32. P. S. Goldman-Rakic, J. F. Bates, M. V. Chafee, Curr. Opin. Neurobiol. 2, 830 (1992); J. C. Houk and S. P. Wise, Cereb. Cortex 5, 95 (1995); J. C. Houk and S. P. Wise, Cereb. Cortex 5, 95 (1995);
33. P. S. Goldman-Rakic, J. F. Bates, M. V. Chafee, Curr. Opin. Neurobiol. 2, 830 (1992); J. C. Houk and S. P. Wise, Cereb. Cortex 5, 95 (1995); D. G. Beiser and J. C. Houk, J. Neurophysiol. 79, 3168 (1998). D. G. Beiser and J. C. Houk, J. Neurophysiol. 79, 3168 (1998). M. Petrides, Proc. R. Soc. London Ser. B 246, 299 (1991); J. Neurosci. 15, 359 (1995).
34. B. Milner, M. Petrides, M. L. Smith, Hum. Neurobiol. 4, 137 (1985); J. Neurosci. 15, 359 (1995).
35. B. Milner, M. Petrides, M. L. Smith, Hum. Neurobiol. 4, 137 (1985); M. Petrides, Proc. R. Soc. London Ser. B 246, 299 (1991); M. Petrides, Proc. R. Soc. London Ser. B 246, 299 (1991);
36. B. Milner, M. Petrides, M. L. Smith, Hum. Neurobiol. 4, 137 (1985); M. Petrides, Proc. R. Soc. London Ser. B 246, 299 (1991); J. Neurosci. 15, 359 (1995). J. Neurosci. 15, 359 (1995).
37. This study was not designed to address issues of neural coding of serial order per se, as an abstract entity (that is, the first-ness, second-ness, third-ness, ..., k-ness of a series of stimuli). Rather, we focused on a specific motor task whose correct performance depended on identifying the serial order of stimuli. The motor cortex (this report) as well as premotor and prefrontal areas [A. F. Carpenter, G. Pellizzer, A. P. Georgopoulos, Soc. Neurosci. Abstr. 24, 1426 (1998)] are involved in this task.
38. We thank P. Sargent for participating in part of the experiment. Supported by a VA Merit Review Award (G.P.), NIH grant NS17413 (A.P.G.), NSF training fellowship GER9454163 (A.F.C.), the U.S. Department of Veterans Affairs, and the American Legion Brain Sciences Chair.