Brain regions responsive to novelty in the absence of awareness

Science

Volumn 276, Issue 5316, 1997, Pages 1272-1275

  Berns, Gregory S ^a   Cohen, Jonathan D ^b,c   Mintun, Mark A ^d  

^a UNIVERSITY OF PITTSBURGH   (United States)

^b CARNEGIE MELLON UNIVERSITY   (United States)

^c UNIVERSITY OF PITTSBURGH MEDICAL CENTER   (United States)

^d WASHINGTON UNIVERSITY SCHOOL OF MEDICINE   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

ARTICLE; AWARENESS; BRAIN MAPPING; CINGULATE GYRUS; CORPUS STRIATUM; HUMAN; PREMOTOR CORTEX; PRIORITY JOURNAL; REACTION TIME; SELECTIVE ATTENTION; TASK PERFORMANCE;

ADULT; AWARENESS; BRAIN MAPPING; CEREBROVASCULAR CIRCULATION; CORPUS STRIATUM; FEMALE; GYRUS CINGULI; HUMANS; LEARNING; LINGUISTICS; MALE; MOTOR CORTEX; PERCEPTION; PSYCHOMOTOR PERFORMANCE; REACTION TIME; TOMOGRAPHY, EMISSION-COMPUTED;

EID: 0030950252     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.276.5316.1272     Document Type: Article

References (44)

1. C. E. Shannon and W. Weaver, The Mathematical Theory of Communication (Univ. of Illinois Press, Chicago, 1949).
2. E. Tulving, H. J. Markowitsch, S. Kapur, R. Habib, S. Houle, NeuroReport 5, 2525 (1994); E. Tulving et al., Proc. Natl. Acad. Sci. U.S.A. 91, 2012 (1994); E. Tulving, H. J. Markowitsch, F. I. M. Craik, R. Habib, S. Houle, Cereb. Cortex 6, 71 (1996); N. C. Andreasen et al., NeuroImage 2, 296 (1995).
3. E. Tulving, H. J. Markowitsch, S. Kapur, R. Habib, S. Houle, NeuroReport 5, 2525 (1994); E. Tulving et al., Proc. Natl. Acad. Sci. U.S.A. 91, 2012 (1994); E. Tulving, H. J. Markowitsch, F. I. M. Craik, R. Habib, S. Houle, Cereb. Cortex 6, 71 (1996); N. C. Andreasen et al., NeuroImage 2, 296 (1995).
4. E. Tulving, H. J. Markowitsch, S. Kapur, R. Habib, S. Houle, NeuroReport 5, 2525 (1994); E. Tulving et al., Proc. Natl. Acad. Sci. U.S.A. 91, 2012 (1994); E. Tulving, H. J. Markowitsch, F. I. M. Craik, R. Habib, S. Houle, Cereb. Cortex 6, 71 (1996); N. C. Andreasen et al., NeuroImage 2, 296 (1995).
5. E. Tulving, H. J. Markowitsch, S. Kapur, R. Habib, S. Houle, NeuroReport 5, 2525 (1994); E. Tulving et al., Proc. Natl. Acad. Sci. U.S.A. 91, 2012 (1994); E. Tulving, H. J. Markowitsch, F. I. M. Craik, R. Habib, S. Houle, Cereb. Cortex 6, 71 (1996); N. C. Andreasen et al., NeuroImage 2, 296 (1995).
6. A. S. Reber, Implicit Learning and Tacit Knowledge. An Essay on the Cognitive Unconscious (Oxford Psychology Series No. 19, Oxford Univ. Press, Oxford, 1993).
7. M. Hartman, D. S. Knopman, M. J. Nissen, J. Exp. Psychol. Learn. Mem. Cognition 15, 1070 (1989); R. C. Mathews et al., ibid., p. 1083; M. A. Stadler, ibid. 21, 674 (1995); A. Cohen, R. I. Ivry, S. W. Keele, ibid. 16, 17 (1990); T. Curran and S. W. Keele, ibid. 19, 189 (1993).
8. M. Hartman, D. S. Knopman, M. J. Nissen, J. Exp. Psychol. Learn. Mem. Cognition 15, 1070 (1989); R. C. Mathews et al., ibid., p. 1083; M. A. Stadler, ibid. 21, 674 (1995); A. Cohen, R. I. Ivry, S. W. Keele, ibid. 16, 17 (1990); T. Curran and S. W. Keele, ibid. 19, 189 (1993).
9. M. Hartman, D. S. Knopman, M. J. Nissen, J. Exp. Psychol. Learn. Mem. Cognition 15, 1070 (1989); R. C. Mathews et al., ibid., p. 1083; M. A. Stadler, ibid. 21, 674 (1995); A. Cohen, R. I. Ivry, S. W. Keele, ibid. 16, 17 (1990); T. Curran and S. W. Keele, ibid. 19, 189 (1993).
10. M. Hartman, D. S. Knopman, M. J. Nissen, J. Exp. Psychol. Learn. Mem. Cognition 15, 1070 (1989); R. C. Mathews et al., ibid., p. 1083; M. A. Stadler, ibid. 21, 674 (1995); A. Cohen, R. I. Ivry, S. W. Keele, ibid. 16, 17 (1990); T. Curran and S. W. Keele, ibid. 19, 189 (1993).
11. M. Hartman, D. S. Knopman, M. J. Nissen, J. Exp. Psychol. Learn. Mem. Cognition 15, 1070 (1989); R. C. Mathews et al., ibid., p. 1083; M. A. Stadler, ibid. 21, 674 (1995); A. Cohen, R. I. Ivry, S. W. Keele, ibid. 16, 17 (1990); T. Curran and S. W. Keele, ibid. 19, 189 (1993).
12. Finite-state grammars consist of nodes connected by allowable transitions, with each transition corresponding to one of the stimuli (3). With complex grammars, individuals learn the dependencies after ∼30,000 trials (as evidenced by decreased reaction times compared with trials with random stimuli) without developing awareness of the grammar's presence [A. Cleeremans and J. L. McClelland, J. Exp. Psychol. Gen. 120, 235 (1991)]. In pilot studies, we found that simpler grammars can be learned in shorter periods of time without individuals developing awareness of the grammar.
13. There were five male and five female participants, average age = 26.0 years (SD = 6.4, range 20 to 39). Informed consent was obtained from all participants after the nature and possible risks of the experiment were explained. Before the study the participants current and past medical histories were reviewed to exclude those with neurologic, psychiatric, or active medical disorders. The protocol was approved by the University of Pittsburgh Medical Center institutional review board.
14. Stimuli appeared every 700 ms regardless of whether the response was correct. Stimuli were determined by a five-node finite-state grammar similar to those of previous studies [A. S. Reber, J. Verb. Learn. Verb. Behav. 6, 317 (1967)]. From a given node, one of the available transitions was randomly chosen and the corresponding stimulus was displayed. This was then repeated from the new node. The final node in the grammar was the same as the first. Stimuli associated with the transitions were set up so that the overall probability of each stimulus was 1/3. Thus, any learning that occurred used information about second-order or higher probabilities. To maintain a general level of motivation, the participants were informed at the beginning of the task that they would earn monetary bonuses for better performance, and accuracy was more important than speed.
15. The architecture of grammar B was the same as that of grammar A except that the stimuli associated with the node transitions were permuted so that 1 became 2, 2 became 3, and 3 became 1.
16. 15O-water activity that are closely proportional to rCBF [M. E. Raichle, W. R. W. Martin, P. Herscovitch, M. A. Mintun, J. Markham, J. Nucl. Med. 24, 790 (1983); P. Herscovitch, J. Markham, M. E. Raichle, ibid., p. 782].
17. 15O-water activity that are closely proportional to rCBF [M. E. Raichle, W. R. W. Martin, P. Herscovitch, M. A. Mintun, J. Markham, J. Nucl. Med. 24, 790 (1983); P. Herscovitch, J. Markham, M. E. Raichle, ibid., p. 782].
18. 15O-water activity that are closely proportional to rCBF [M. E. Raichle, W. R. W. Martin, P. Herscovitch, M. A. Mintun, J. Markham, J. Nucl. Med. 24, 790 (1983); P. Herscovitch, J. Markham, M. E. Raichle, ibid., p. 782].
19. Scans were realigned to each individual's structural magnetic resonance image [R. P. Woods, J. C. Mazziotta, S. R. Cherry, J. Comput. Assisted Tomogr. 17, 536 (1993)] and subsequently transformed to a standard atlas [J. Talairach and P. Tournoux, Coplanar Stereotaxic Atlas of the Human Brain (Thieme, Stuttgart, 1988)]. With SPM95 [K. J. Friston, C. D. Frith, P. F. Liddle, R. S. J. Frackowiak, J. Cereb. Blood Flow Metab. 11, 690 (1991)], analyses of variance (ANOVAs) were performed at each voxel in the PET space. To control for global differences in blood flow both within and between individuals, we first normalized activity counts using analysis of covariance (ANCOVA) to a mean of 50 ml per 100 ml per minute. Activity maps were smoothed with a 20 mm by 20 mm by 12 mm Gaussian kernel to increase the signal-to-noise ratio and account for interparticipant anatomic variability. The time required for learning each grammar precluded a balanced experimental design, so effects confounded with time may be significant (such as fatigue). We therefore included scan number as a covariate of noninterest in the ANOVA model, and only the remaining voxels with significant changes across the whole study at P < 0.05 were used for further specific t tests. To identify rCBF changes due to the grammar change, we performed contrasts between the end of grammar A (scans 5 through 8) and the beginning of grammar B (scans 9 through 12). Two one-tailed paired t tests identified those voxels with significant increases or decreases, respectively. A conservative threshold of significance was used (P < 0.001).
20. Scans were realigned to each individual's structural magnetic resonance image [R. P. Woods, J. C. Mazziotta, S. R. Cherry, J. Comput. Assisted Tomogr. 17, 536 (1993)] and subsequently transformed to a standard atlas [J. Talairach and P. Tournoux, Coplanar Stereotaxic Atlas of the Human Brain (Thieme, Stuttgart, 1988)]. With SPM95 [K. J. Friston, C. D. Frith, P. F. Liddle, R. S. J. Frackowiak, J. Cereb. Blood Flow Metab. 11, 690 (1991)], analyses of variance (ANOVAs) were performed at each voxel in the PET space. To control for global differences in blood flow both within and between individuals, we first normalized activity counts using analysis of covariance (ANCOVA) to a mean of 50 ml per 100 ml per minute. Activity maps were smoothed with a 20 mm by 20 mm by 12 mm Gaussian kernel to increase the signal-to-noise ratio and account for interparticipant anatomic variability. The time required for learning each grammar precluded a balanced experimental design, so effects confounded with time may be significant (such as fatigue). We therefore included scan number as a covariate of noninterest in the ANOVA model, and only the remaining voxels with significant changes across the whole study at P < 0.05 were used for further specific t tests. To identify rCBF changes due to the grammar change, we performed contrasts between the end of grammar A (scans 5 through 8) and the beginning of grammar B (scans 9 through 12). Two one-tailed paired t tests identified those voxels with significant increases or decreases, respectively. A conservative threshold of significance was used (P < 0.001).
21. Scans were realigned to each individual's structural magnetic resonance image [R. P. Woods, J. C. Mazziotta, S. R. Cherry, J. Comput. Assisted Tomogr. 17, 536 (1993)] and subsequently transformed to a standard atlas [J. Talairach and P. Tournoux, Coplanar Stereotaxic Atlas of the Human Brain (Thieme, Stuttgart, 1988)]. With SPM95 [K. J. Friston, C. D. Frith, P. F. Liddle, R. S. J. Frackowiak, J. Cereb. Blood Flow Metab. 11, 690 (1991)], analyses of variance (ANOVAs) were performed at each voxel in the PET space. To control for global differences in blood flow both within and between individuals, we first normalized activity counts using analysis of covariance (ANCOVA) to a mean of 50 ml per 100 ml per minute. Activity maps were smoothed with a 20 mm by 20 mm by 12 mm Gaussian kernel to increase the signal-to-noise ratio and account for interparticipant anatomic variability. The time required for learning each grammar precluded a balanced experimental design, so effects confounded with time may be significant (such as fatigue). We therefore included scan number as a covariate of noninterest in the ANOVA model, and only the remaining voxels with significant changes across the whole study at P < 0.05 were used for further specific t tests. To identify rCBF changes due to the grammar change, we performed contrasts between the end of grammar A (scans 5 through 8) and the beginning of grammar B (scans 9 through 12). Two one-tailed paired t tests identified those voxels with significant increases or decreases, respectively. A conservative threshold of significance was used (P < 0.001).
22. Participants were formally debriefed after completion of all PET scans with a series of general ("Did you notice anything about the sequence of numbers?") to specific questions ("Halfway through the task, the sequences changed. Did you notice this?"). No one reported awareness of sequential structure to the stimuli. One of the 10 participants reported that something may have changed during the course of the task, but the other nine denied being aware of the grammar change.
23. When compared with the scans taken at rest, several other regions (particularly the right cerebellum and left sensorimotor areas) had significant increases in blood flow during the entire task; however, these areas did not show a statistically significant difference in blood flow when the grammar was changed.
24. Reaction time is a surrogate measure of learning, so any regions that show either positively or negatively correlated blood flow are potentially involved in the learning of the contextual information. Significant positive correlations were limited to the right ventral striatum and left parahippocampal region.
25. S. T. Grafton, E. Hazeltine, R. Ivry, J. Cognitive Neurosci. 7, 497 (1995); S. L. Rauch et al., Hum. Brain Map. 3, 271 (1995). Other studies of sequential or procedural learning (with awareness) have shown activation of some basal ganglia regions, as well as other areas [S. T. Grafton et al., J. Neurosci. 12, 2542 (1992); R. J. Seitz and P. E. Roland, Eur. J. Neurosci. 4, 154 (1992); I. H. Jenkins, D. J. Brooks, P. D. Nixon, R. S. J. Frackowiak, R. E. Passingham, J. Neurosci. 14, 3775 (1994)].
26. S. T. Grafton, E. Hazeltine, R. Ivry, J. Cognitive Neurosci. 7, 497 (1995); S. L. Rauch et al., Hum. Brain Map. 3, 271 (1995). Other studies of sequential or procedural learning (with awareness) have shown activation of some basal ganglia regions, as well as other areas [S. T. Grafton et al., J. Neurosci. 12, 2542 (1992); R. J. Seitz and P. E. Roland, Eur. J. Neurosci. 4, 154 (1992); I. H. Jenkins, D. J. Brooks, P. D. Nixon, R. S. J. Frackowiak, R. E. Passingham, J. Neurosci. 14, 3775 (1994)].
27. S. T. Grafton, E. Hazeltine, R. Ivry, J. Cognitive Neurosci. 7, 497 (1995); S. L. Rauch et al., Hum. Brain Map. 3, 271 (1995). Other studies of sequential or procedural learning (with awareness) have shown activation of some basal ganglia regions, as well as other areas [S. T. Grafton et al., J. Neurosci. 12, 2542 (1992); R. J. Seitz and P. E. Roland, Eur. J. Neurosci. 4, 154 (1992); I. H. Jenkins, D. J. Brooks, P. D. Nixon, R. S. J. Frackowiak, R. E. Passingham, J. Neurosci. 14, 3775 (1994)].
28. S. T. Grafton, E. Hazeltine, R. Ivry, J. Cognitive Neurosci. 7, 497 (1995); S. L. Rauch et al., Hum. Brain Map. 3, 271 (1995). Other studies of sequential or procedural learning (with awareness) have shown activation of some basal ganglia regions, as well as other areas [S. T. Grafton et al., J. Neurosci. 12, 2542 (1992); R. J. Seitz and P. E. Roland, Eur. J. Neurosci. 4, 154 (1992); I. H. Jenkins, D. J. Brooks, P. D. Nixon, R. S. J. Frackowiak, R. E. Passingham, J. Neurosci. 14, 3775 (1994)].
29. S. T. Grafton, E. Hazeltine, R. Ivry, J. Cognitive Neurosci. 7, 497 (1995); S. L. Rauch et al., Hum. Brain Map. 3, 271 (1995). Other studies of sequential or procedural learning (with awareness) have shown activation of some basal ganglia regions, as well as other areas [S. T. Grafton et al., J. Neurosci. 12, 2542 (1992); R. J. Seitz and P. E. Roland, Eur. J. Neurosci. 4, 154 (1992); I. H. Jenkins, D. J. Brooks, P. D. Nixon, R. S. J. Frackowiak, R. E. Passingham, J. Neurosci. 14, 3775 (1994)].
30. Other cognitive processes occurring at the grammar change may also contribute to the pattern of rCBF changes. Overriding the response tendencies from the first grammar may be significant, even if the individual is not conscious of it. Our study does not allow for the separation of the novelty from the process of overriding responses. However, the ventral striatal activation, being more transient, may be more related to novelty sensitivity, and the left anterior cingulate activation, because it peaks after the grammar change and not during the first grammar, may be more related to overriding prepotent responses.
31. P. R. Montague, P. Dayan, T. J. Sejnowski, J. Neurosci. 16, 1936 (1996).
32. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
33. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
34. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
35. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
36. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
37. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
38. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
39. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
40. P. Barone and J. P. Joseph, Exp. Brain Res. 78, 447 (1989); J. D. Cohen et al., Nature 386, 604 (1997); S. Courtney, L. G. Ungerlieder, K. Kell, J. V. Haxby, ibid., p. 608; J. M. Fuster and G. E. Alexander, Science 173, 652 (1971); P. S. Goldman-Rakic, in Handbook of Physiology, Section 1, The Nervous System, V. B. Mountcastle, Ed. (American Physiological Society, Bethesda, MD, 1987), vol. 5, pp. 373-417; P. M. Grasby et al., Brain 116, 1 (1993); J. Jonides et al., Nature 363, 623 (1993); E. K. Miller, C. A. Erickson, R. Desimone, J. Neurosci. 16, 5154 (1996); H. Niki, Brain Res. 68, 185 (1974).
41. J. D. Cohen, T. S. Braver, R. O'Reilly, Philos. Trans. R. Soc. London Series B 351, 1515 (1996).
42. Another study of implicit learning also found increasing right DLPFC activation as a sequence was learned [S. T. Grafton, E. Hazeltine, R. Ivry, J. Cognitive Neurosci. 7, 497 (1995)]. Because the sequence was simpler, several individuals developed awareness of it, so it was not clear whether DLPFC activity increased because of the increases in maintenance requirements or in the awareness of the grammar.
43. T. S. Braver et al., Neuroimage 5, 49 (1997).
44. We would like to thank M. B. Wiseman for assistance with computer processing and C. S. Carter and J. L. McClelland for comments on the manuscript. Supported by the Functional Brain Imaging Center (MH 49815).