The role of locus coeruleus in the regulation of cognitive performance

Science

Volumn 283, Issue 5401, 1999, Pages 549-554

  Usher, Marius ^a   Cohen, Jonathan D ^a   Servan Schreiber, David ^a   Rajkowski, Janusz ^a   Aston Jones, Gary ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

ARTICLE; COGNITION; COUPLING FACTOR; EXPLORATORY BEHAVIOR; LOCUS CERULEUS; MONKEY; NEUROMODULATION; NONHUMAN; PRIORITY JOURNAL; STIMULUS RESPONSE; TASK PERFORMANCE; VISUAL DISCRIMINATION;

EID: 0033593588     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.283.5401.549     Document Type: Article

References (55)

1. W. Schultz, P. Dayan, P. R. Montague, Science 275, 1593 (1997).
2. G. Aston-Jones, J, Rajkowski, T. Alexinsky, J. Neurosci. 14, 4467 (1994); J. Rajkowski, P. Kubiak, G. Aston- Jones, Brain Res. Bull. 35, 607 (1994).
3. G. Aston-Jones, J, Rajkowski, T. Alexinsky, J. Neurosci. 14, 4467 (1994); J. Rajkowski, P. Kubiak, G. Aston-Jones, Brain Res. Bull. 35, 607 (1994).
4. Training and experimental recording sessions took place in an acoustically insulated, electrically shielded metal chamber (IAC, Bronx, NY). Monkeys were trained to depress a lever and to stably foveate a fixation stimulus on a video monitor, at which point this stimulus was replaced by a target or nontarget stimulus (horizontally or vertically oriented rectangle). The animal was required to selectively release the lever in response to the target stimulus (20% of trials). Responses to the other stimulus were not reinforced but instead generated a 3-s time-out. Training continued until animals performed at a level of at least 85% correct. See (2) for more details.
5. Typically, epochs of poor performance contained more than seven times the frequency of FA errors as epochs of good performance. The hit rates varied only slightly between these periods, remaining either constant or declining slightly during poor performance intervals. For the three monkeys analyzed, the d' values in poor compared with good periods increased from 2.9 to 5.1,3.7 to 4.7, and 3.7 to 5.1. The response criterion b also increased during the good periods from 0.23 to 0.82, 0.36 to 2.92, and 0.06 to 1.11, respectively. For these monkeys, the standard deviations of RTs were 58, 55, and 46 ms, respectively, during poor intervals and 35, 33, and 35 ms, respectively, during epochs of good performance (P < 0.001; Levene test of variances).
6. S. L. Foote, G. Aston-Jones, F. E. Bloom, Proc. Natl. Acad. Sci. U.S.A. 77, 3033 (1980); S. J. Grant, G. Aston-Jones, D. E. J. Redmond, Brain Res. Bull. 21, 401 (1988).
7. S. L. Foote, G. Aston-Jones, F. E. Bloom, Proc. Natl. Acad. Sci. U.S.A. 77, 3033 (1980); S. J. Grant, G. Aston-Jones, D. E. J. Redmond, Brain Res. Bull. 21, 401 (1988).
8. B. Knight, J. Gen. Physiol. 59, 734 (1972).
9. th.
10. O. Bernander, C. Koch, M. Usher, Neural Comput. 6, 622 (1994).
11. C. K. Aghajanian, J. M. Cedarbaum, R. Y. Wang, Brain Res. 136, 570 (1977); T. M. Egan, C. Henderson, R. A. North, J. T. Williams, J. Physiol. 345, 477 (1983); M. Ennis and G. Aston-Jones, Brain Res. 374, 299 (1986).
12. C. K. Aghajanian, J. M. Cedarbaum, R. Y. Wang, Brain Res. 136, 570 (1977); T. M. Egan, C. Henderson, R. A. North, J. T. Williams, J. Physiol. 345, 477 (1983); M. Ennis and G. Aston-Jones, Brain Res. 374, 299 (1986).
13. C. K. Aghajanian, J. M. Cedarbaum, R. Y. Wang, Brain Res. 136, 570 (1977); T. M. Egan, C. Henderson, R. A. North, J. T. Williams, J. Physiol. 345, 477 (1983); M. Ennis and G. Aston-Jones, Brain Res. 374, 299 (1986).
14. Lateral inhibition occurs with a rise time of about 25 ms after LC cell firing and a decay of 250 ms [S. L. Foote, F. E. Bloom, G. Aston-Jones, Physiol. Rev. 63, 844 (1983)]. This collateral NE release regulates the firing rate of the LC population: After each target-evoked, synchronized response of the population (see below), a slightly delayed inhibitory effect appears (as reflected in the PSTH histograms; Figs. 1C and 3A).
15. Electrotonic coupling is consistent with observations of gap junctions among LC neurons in neonatal rats (21) and with recent evidence for coupling among LC neurons in the adult rat (22, 23). We assume that coupling produces a weak ohmic conductance between pairs of cells, which reaches a maximum of about 2.5% of the input current received by the cell, corresponding to the amount of current found in gap junctions identified in neonatal LC neurons (21).
16. This network is not intended to be a detailed simulation of specific neuronal circuits at the cellular level. Rather, it is intended to simulate task performance, with mechanisms that are consistent with those of biological information processing [see, for example, J. L. McClelland, in Attention and Performance, vol. XIV (MIT Press, Cambridge, MA, 1993), pp. 655-688; D. E. Rumelhart and J. L. McClelland, Parallel Distributed Processing (MIT Press, Cambridge MA, 1986)]. For example, the behavior of cell assemblies thought to represent task-relevant stimuli and responses in the cortex is simulated as single processing units with continuous-valued activation levels, on the assumption that information is represented in the cortex as the average spike rate of cell populations [D. J. Amit, Modeling Brain Function (Cambridge Univ. Press, Cambridge, 1989)]. Recurrent self-connections simulate mutual excitatory synapses between cells that belong to a particular assembly.
17. This network is not intended to be a detailed simulation of specific neuronal circuits at the cellular level. Rather, it is intended to simulate task performance, with mechanisms that are consistent with those of biological information processing [see, for example, J. L. McClelland, in Attention and Performance, vol. XIV (MIT Press, Cambridge, MA, 1993), pp. 655-688; D. E. Rumelhart and J. L. McClelland, Parallel Distributed Processing (MIT Press, Cambridge MA, 1986)]. For example, the behavior of cell assemblies thought to represent task-relevant stimuli and responses in the cortex is simulated as single processing units with continuous-valued activation levels, on the assumption that information is represented in the cortex as the average spike rate of cell populations [D. J. Amit, Modeling Brain Function (Cambridge Univ. Press, Cambridge, 1989)]. Recurrent self-connections simulate mutual excitatory synapses between cells that belong to a particular assembly.
18. This network is not intended to be a detailed simulation of specific neuronal circuits at the cellular level. Rather, it is intended to simulate task performance, with mechanisms that are consistent with those of biological information processing [see, for example, J. L. McClelland, in Attention and Performance, vol. XIV (MIT Press, Cambridge, MA, 1993), pp. 655-688; D. E. Rumelhart and J. L. McClelland, Parallel Distributed Processing (MIT Press, Cambridge MA, 1986)]. For example, the behavior of cell assemblies thought to represent task-relevant stimuli and responses in the cortex is simulated as single processing units with continuous-valued activation levels, on the assumption that information is represented in the cortex as the average spike rate of cell populations [D. J. Amit, Modeling Brain Function (Cambridge Univ. Press, Cambridge, 1989)]. Recurrent self-connections simulate mutual excitatory synapses between cells that belong to a particular assembly.
19. The weak weight from each input unit to the opposite decision unit captures our assumption that the stimuli used in the task have overlapping features and therefore each partially activates the representation of the other. The weights from the distractor decision unit to the response unit and LC network are zero (and therefore not implemented). This value corresponds to our assumption that, because the animal has been overtrained to respond to the target but not the distractor, there has been selective strengthening of projections from the target decision unit to the response unit and LC module, but not for the distractor unit.
20. ij and processing rate T. The gain parameter is determined by the summed output of units in the LC network, with a lag time of about 55 to 90 ms between a change in mean LC unit activity and the consequent change in the gain parameter of units in the behavioral network (consistent with physiological data concerning the time constants governing LC activity and cortical release of NE) (10). Note that the effect of NE on the behavioral network (gain modulation) is different than its local effects within the LC (inhibitory), consistent with empirical data regarding its effects in each of these areas (10). The gain effect in the behavioral network is consistent with more detailed hypotheses about the effects of NE on cortical circuits [M. E. Hasselmo, Behav. Brain Res. 67, 1 (1995)].
21. D. Servan-Schreiber, H. W. Printz, J. D. Cohen, Science 249, 892 (1990).
22. Even the weak degree of coupling implemented (11) produces significant synchronization of the LC population, as predicted by theoretical models of coupled phase oscillators [Y. Kuramoto, Chemical Oscillators, Waves and Turbulence (Springer-Verlag, Berlin, 1984); A. Sherman and J. Rinzel, Proc. Natl. Acad. Sci. U.S.A. 89, 2471 (1992); S. Strogatz and R. Mirollo, J. Stat. Phys. 63, 613 (1991)], and decreases spontaneous firing rate. The relative decrease in the spontaneous rate of discharge observed in our simulations (from 0.90 to 0.67 Hz) is similar to that observed empirically in LC cells during good behavioral epochs (2.95 to 2.01 Hz). To assess the specificity of this effect, we explored other schemes for simulating variations in LC firing, none of which reproduced the patterns in the empirical data. For example, increasing lateral inhibition [for example, C. Harris, Z. Hausken, J. Williams, Neuroscience 50, 253 (1992)] or decreasing noise among LC cells both reduced tonic discharge and increased synchronization of activity [C. van Vreeswijk, L. F. Abbott, B. C. Ermentrout, J. Comput. Neurosci. 1, 303 (1994)] but did not produce the amplification of stimulus-evoked responses observed empirically during epochs of good behavioral performance (Fig. 1, C compared with D). Conversely, other schemes such as induction of a low-threshold calcium conductance [H. Jahnsen and R. Llinas, J. Physiol. 349, 227 (1984)] (not reported to exist in LC neurons) or simultaneously increasing inhibitory and target-evoked excitatory inputs can produce the amplification effect but do not lead to the reduction of tonic activity or increased synchrony.
23. Even the weak degree of coupling implemented (11) produces significant synchronization of the LC population, as predicted by theoretical models of coupled phase oscillators [Y. Kuramoto, Chemical Oscillators, Waves and Turbulence (Springer-Verlag, Berlin, 1984); A. Sherman and J. Rinzel, Proc. Natl. Acad. Sci. U.S.A. 89, 2471 (1992); S. Strogatz and R. Mirollo, J. Stat. Phys. 63, 613 (1991)], and decreases spontaneous firing rate. The relative decrease in the spontaneous rate of discharge observed in our simulations (from 0.90 to 0.67 Hz) is similar to that observed empirically in LC cells during good behavioral epochs (2.95 to 2.01 Hz). To assess the specificity of this effect, we explored other schemes for simulating variations in LC firing, none of which reproduced the patterns in the empirical data. For example, increasing lateral inhibition [for example, C. Harris, Z. Hausken, J. Williams, Neuroscience 50, 253 (1992)] or decreasing noise among LC cells both reduced tonic discharge and increased synchronization of activity [C. van Vreeswijk, L. F. Abbott, B. C. Ermentrout, J. Comput. Neurosci. 1, 303 (1994)] but did not produce the amplification of stimulus-evoked responses observed empirically during epochs of good behavioral performance (Fig. 1, C compared with D). Conversely, other schemes such as induction of a low-threshold calcium conductance [H. Jahnsen and R. Llinas, J. Physiol. 349, 227 (1984)] (not reported to exist in LC neurons) or simultaneously increasing inhibitory and target-evoked excitatory inputs can produce the amplification effect but do not lead to the reduction of tonic activity or increased synchrony.
24. Even the weak degree of coupling implemented (11) produces significant synchronization of the LC population, as predicted by theoretical models of coupled phase oscillators [Y. Kuramoto, Chemical Oscillators, Waves and Turbulence (Springer-Verlag, Berlin, 1984); A. Sherman and J. Rinzel, Proc. Natl. Acad. Sci. U.S.A. 89, 2471 (1992); S. Strogatz and R. Mirollo, J. Stat. Phys. 63, 613 (1991)], and decreases spontaneous firing rate. The relative decrease in the spontaneous rate of discharge observed in our simulations (from 0.90 to 0.67 Hz) is similar to that observed empirically in LC cells during good behavioral epochs (2.95 to 2.01 Hz). To assess the specificity of this effect, we explored other schemes for simulating variations in LC firing, none of which reproduced the patterns in the empirical data. For example, increasing lateral inhibition [for example, C. Harris, Z. Hausken, J. Williams, Neuroscience 50, 253 (1992)] or decreasing noise among LC cells both reduced tonic discharge and increased synchronization of activity [C. van Vreeswijk, L. F. Abbott, B. C. Ermentrout, J. Comput. Neurosci. 1, 303 (1994)] but did not produce the amplification of stimulus-evoked responses observed empirically during epochs of good behavioral performance (Fig. 1, C compared with D). Conversely, other schemes such as induction of a low-threshold calcium conductance [H. Jahnsen and R. Llinas, J. Physiol. 349, 227 (1984)] (not reported to exist in LC neurons) or simultaneously increasing inhibitory and target-evoked excitatory inputs can produce the amplification effect but do not lead to the reduction of tonic activity or increased synchrony.
25. Even the weak degree of coupling implemented (11) produces significant synchronization of the LC population, as predicted by theoretical models of coupled phase oscillators [Y. Kuramoto, Chemical Oscillators, Waves and Turbulence (Springer-Verlag, Berlin, 1984); A. Sherman and J. Rinzel, Proc. Natl. Acad. Sci. U.S.A. 89, 2471 (1992); S. Strogatz and R. Mirollo, J. Stat. Phys. 63, 613 (1991)], and decreases spontaneous firing rate. The relative decrease in the spontaneous rate of discharge observed in our simulations (from 0.90 to 0.67 Hz) is similar to that observed empirically in LC cells during good behavioral epochs (2.95 to 2.01 Hz). To assess the specificity of this effect, we explored other schemes for simulating variations in LC firing, none of which reproduced the patterns in the empirical data. For example, increasing lateral inhibition [for example, C. Harris, Z. Hausken, J. Williams, Neuroscience 50, 253 (1992)] or decreasing noise among LC cells both reduced tonic discharge and increased synchronization of activity [C. van Vreeswijk, L. F. Abbott, B. C. Ermentrout, J. Comput. Neurosci. 1, 303 (1994)] but did not produce the amplification of stimulus-evoked responses observed empirically during epochs of good behavioral performance (Fig. 1, C compared with D). Conversely, other schemes such as induction of a low-threshold calcium conductance [H. Jahnsen and R. Llinas, J. Physiol. 349, 227 (1984)] (not reported to exist in LC neurons) or simultaneously increasing inhibitory and target-evoked excitatory inputs can produce the amplification effect but do not lead to the reduction of tonic activity or increased synchrony.
26. Even the weak degree of coupling implemented (11) produces significant synchronization of the LC population, as predicted by theoretical models of coupled phase oscillators [Y. Kuramoto, Chemical Oscillators, Waves and Turbulence (Springer-Verlag, Berlin, 1984); A. Sherman and J. Rinzel, Proc. Natl. Acad. Sci. U.S.A. 89, 2471 (1992); S. Strogatz and R. Mirollo, J. Stat. Phys. 63, 613 (1991)], and decreases spontaneous firing rate. The relative decrease in the spontaneous rate of discharge observed in our simulations (from 0.90 to 0.67 Hz) is similar to that observed empirically in LC cells during good behavioral epochs (2.95 to 2.01 Hz). To assess the specificity of this effect, we explored other schemes for simulating variations in LC firing, none of which reproduced the patterns in the empirical data. For example, increasing lateral inhibition [for example, C. Harris, Z. Hausken, J. Williams, Neuroscience 50, 253 (1992)] or decreasing noise among LC cells both reduced tonic discharge and increased synchronization of activity [C. van Vreeswijk, L. F. Abbott, B. C. Ermentrout, J. Comput. Neurosci. 1, 303 (1994)] but did not produce the amplification of stimulus-evoked responses observed empirically during epochs of good behavioral performance (Fig. 1, C compared with D). Conversely, other schemes such as induction of a low-threshold calcium conductance [H. Jahnsen and R. Llinas, J. Physiol. 349, 227 (1984)] (not reported to exist in LC neurons) or simultaneously increasing inhibitory and target-evoked excitatory inputs can produce the amplification effect but do not lead to the reduction of tonic activity or increased synchrony.
27. Even the weak degree of coupling implemented (11) produces significant synchronization of the LC population, as predicted by theoretical models of coupled phase oscillators [Y. Kuramoto, Chemical Oscillators, Waves and Turbulence (Springer-Verlag, Berlin, 1984); A. Sherman and J. Rinzel, Proc. Natl. Acad. Sci. U.S.A. 89, 2471 (1992); S. Strogatz and R. Mirollo, J. Stat. Phys. 63, 613 (1991)], and decreases spontaneous firing rate. The relative decrease in the spontaneous rate of discharge observed in our simulations (from 0.90 to 0.67 Hz) is similar to that observed empirically in LC cells during good behavioral epochs (2.95 to 2.01 Hz). To assess the specificity of this effect, we explored other schemes for simulating variations in LC firing, none of which reproduced the patterns in the empirical data. For example, increasing lateral inhibition [for example, C. Harris, Z. Hausken, J. Williams, Neuroscience 50, 253 (1992)] or decreasing noise among LC cells both reduced tonic discharge and increased synchronization of activity [C. van Vreeswijk, L. F. Abbott, B. C. Ermentrout, J. Comput. Neurosci. 1, 303 (1994)] but did not produce the amplification of stimulus-evoked responses observed empirically during epochs of good behavioral performance (Fig. 1, C compared with D). Conversely, other schemes such as induction of a low-threshold calcium conductance [H. Jahnsen and R. Llinas, J. Physiol. 349, 227 (1984)] (not reported to exist in LC neurons) or simultaneously increasing inhibitory and target-evoked excitatory inputs can produce the amplification effect but do not lead to the reduction of tonic activity or increased synchrony.
28. Note, however, that if it were to occur immediately, it would also potentiate processing in the distractor unit, which is transiently activated by the target stimulus (see Fig. 2B). This would lead to an increase in misses (through competition with the target unit) as well as an increase in FAs. However, the target-evoked LC response occurs about 100 ms after target presentation, which is after the time interval of transient activation of the distractor unit.
29. Studies with rapid serial visual presentation methods [E. Weichselgartner and G. Sperling, Science 238, 778 (1987); A. Reeves and G. Sperling, Psychol. Rev. 93, 180 (1986); J. E. Raymond, K. L. Shapiro, K. M. Arnell, J. Exp. Psychol. Hum. Percept. Perform. 21, 653 (1992) ] have demonstrated that targets are most efficiently accessed during an attentional window of about 180 ms [for review, see H. E. Egeth and S. Yantis, Annu. Rev. Psychol. 48, 269 (1997)].
30. Studies with rapid serial visual presentation methods [E. Weichselgartner and G. Sperling, Science 238, 778 (1987); A. Reeves and G. Sperling, Psychol. Rev. 93, 180 (1986); J. E. Raymond, K. L. Shapiro, K. M. Arnell, J. Exp. Psychol. Hum. Percept. Perform. 21, 653 (1992) ] have demonstrated that targets are most efficiently accessed during an attentional window of about 180 ms [for review, see H. E. Egeth and S. Yantis, Annu. Rev. Psychol. 48, 269 (1997)].
31. Studies with rapid serial visual presentation methods [E. Weichselgartner and G. Sperling, Science 238, 778 (1987); A. Reeves and G. Sperling, Psychol. Rev. 93, 180 (1986); J. E. Raymond, K. L. Shapiro, K. M. Arnell, J. Exp. Psychol. Hum. Percept. Perform. 21, 653 (1992) ] have demonstrated that targets are most efficiently accessed during an attentional window of about 180 ms [for review, see H. E. Egeth and S. Yantis, Annu. Rev. Psychol. 48, 269 (1997)].
32. Studies with rapid serial visual presentation methods [E. Weichselgartner and G. Sperling, Science 238, 778 (1987); A. Reeves and G. Sperling, Psychol. Rev. 93, 180 (1986); J. E. Raymond, K. L. Shapiro, K. M. Arnell, J. Exp. Psychol. Hum. Percept. Perform. 21, 653 (1992) ] have demonstrated that targets are most efficiently accessed during an attentional window of about 180 ms [for review, see H. E. Egeth and S. Yantis, Annu. Rev. Psychol. 48, 269 (1997)].
33. LC phasic responses to targets (which modulate the gain of decision and response units in the model) are characterized by a concentration of spikes 80 to 150 ms after stimulus presentation, which is compensated by postactivation inactivity, so that only the temporal alignment, but not the total number of spikes, differs after target stimuli. Such synchronous activity in one part of a system can enhance transmission into a subsequent processing stage [M. Abeles, Isr. J. Med. Sci. 18, 83 (1982)], resulting in attentional modulation [E. Niebur and C. Koch, J. Comput. Neurosci. 1, 141 (1994)]. Our results extend this computational principle from cortico-cortical interactions to interactions between subcortical structures and the cortex.
34. LC phasic responses to targets (which modulate the gain of decision and response units in the model) are characterized by a concentration of spikes 80 to 150 ms after stimulus presentation, which is compensated by postactivation inactivity, so that only the temporal alignment, but not the total number of spikes, differs after target stimuli. Such synchronous activity in one part of a system can enhance transmission into a subsequent processing stage [M. Abeles, Isr. J. Med. Sci. 18, 83 (1982)], resulting in attentional modulation [E. Niebur and C. Koch, J. Comput. Neurosci. 1, 141 (1994)]. Our results extend this computational principle from cortico-cortical interactions to interactions between subcortical structures and the cortex.
35. Previous experimental studies have linked electrotonic coupling with an increase in synchrony of firing [for example, R. Llinas, R. Baker, C. J. Sotelo, J. Neurophysiol. 37, 560 (1974); C P. Taylor and F. E. Dudek, Science 218, 810 (1982); A. Draguhn, R. Traub, D. Schmitz, J. G. R. Jefferys, Nature 394, 189 (1998)].
36. Previous experimental studies have linked electrotonic coupling with an increase in synchrony of firing [for example, R. Llinas, R. Baker, C. J. Sotelo, J. Neurophysiol. 37, 560 (1974); C P. Taylor and F. E. Dudek, Science 218, 810 (1982); A. Draguhn, R. Traub, D. Schmitz, J. G. R. Jefferys, Nature 394, 189 (1998)].
37. Previous experimental studies have linked electrotonic coupling with an increase in synchrony of firing [for example, R. Llinas, R. Baker, C. J. Sotelo, J. Neurophysiol. 37, 560 (1974); C P. Taylor and F. E. Dudek, Science 218, 810 (1982); A. Draguhn, R. Traub, D. Schmitz, J. G. R. Jefferys, Nature 394, 189 (1998)].
38. M. J. Christie, J. T. Williams, R. A. North, J. Neurosci. 9, 3584 (1989); M. J. Christie and H. F. Jelinek, Neuroscience 56, 129 (1993).
39. M. J. Christie, J. T. Williams, R. A. North, J. Neurosci. 9, 3584 (1989); M. J. Christie and H. F. Jelinek, Neuroscience 56, 129 (1993).
40. M. Ishimatsu and J. T. Williams, J. Neurosci. 16, 5196 (1996); R. A. Travagli, M. Wessendorf, T. V. Dunwiddie, J. T. Williams, Soc. Neurosci. Abstr 20, 1734 (1994); R. A. Travagli, T. V. Dunwiddie, J. T. Williams, J. Neurophysiol. 74, 519 (1994).
41. M. Ishimatsu and J. T. Williams, J. Neurosci. 16, 5196 (1996); R. A. Travagli, M. Wessendorf, T. V. Dunwiddie, J. T. Williams, Soc. Neurosci. Abstr 20, 1734 (1994); R. A. Travagli, T. V. Dunwiddie, J. T. Williams, J. Neurophysiol. 74, 519 (1994).
42. M. Ishimatsu and J. T. Williams, J. Neurosci. 16, 5196 (1996); R. A. Travagli, M. Wessendorf, T. V. Dunwiddie, J. T. Williams, Soc. Neurosci. Abstr 20, 1734 (1994); R. A. Travagli, T. V. Dunwiddie, J. T. Williams, J. Neurophysiol. 74, 519 (1994).
43. P. B. Osborne and J. T. Williams, J. Neurophysiol. 76, 1559 (1996).
44. Previous electrophysiological studies reveal that the adenylyl cyclase stimulator forskolin increases coupling among adult LC neurons, suggesting that coupling among these cells may be modulated by a cyclic adenosine monophosphate (cAMP)-dependent mechanism in the rat (23). A cAMP mechanism has also been described for modulated coupling in other central neurons, and such coupling has been found to be modulated within a time frame of seconds, consistent with the data and model presented here [D. C. McMahon and D. R. Brown, J. Neurophysiol. 72, 2257 (1994); D. C. McMahon and M. P. Mattson, Brain Res. 718, 89 (1996)].
45. Previous electrophysiological studies reveal that the adenylyl cyclase stimulator forskolin increases coupling among adult LC neurons, suggesting that coupling among these cells may be modulated by a cyclic adenosine monophosphate (cAMP)-dependent mechanism in the rat (23). A cAMP mechanism has also been described for modulated coupling in other central neurons, and such coupling has been found to be modulated within a time frame of seconds, consistent with the data and model presented here [D. C. McMahon and D. R. Brown, J. Neurophysiol. 72, 2257 (1994); D. C. McMahon and M. P. Mattson, Brain Res. 718, 89 (1996)].
46. R. S. Siegler, Cognit. Psychol. 28, 225 (1995).
47. L. P. Kaelbling, M. L. Littman, A. W. Moore, J. Artif. Intel. Res. 4, 237 (1996).
48. Our model provides an explicit mechanism that can account for the involvement of the LC system in selective attention [for example, N. R. Selden, T. W. Robbins, B. J. Everitt, J. Neurosci. 10, 531 (1990); N. R. Selden, B. J. Cole, B. J. Everitt, T. W. Robbins, Behav. Brain Res. 39, 29 (1990)] and in exploratory behavior and responsiveness to novelty (for example, S. J. Sara, C. Dyon-Laurent, A. Herve, Cognit. Brain Res. 2, 181(1995); S. O. Ogren, T. Archer, S. B. Ross, in Catecholamines: Neuropharmacology and Central Nervous System Theoretical Aspects, M. Sandler, Ed. (Liss, New York, 1984), pp. 285-292)], by suggesting that each of these is associated with a different mode of LC function and that the LC may mediate shifts between them.
49. Our model provides an explicit mechanism that can account for the involvement of the LC system in selective attention [for example, N. R. Selden, T. W. Robbins, B. J. Everitt, J. Neurosci. 10, 531 (1990); N. R. Selden, B. J. Cole, B. J. Everitt, T. W. Robbins, Behav. Brain Res. 39, 29 (1990)] and in exploratory behavior and responsiveness to novelty (for example, S. J. Sara, C. Dyon-Laurent, A. Herve, Cognit. Brain Res. 2, 181(1995); S. O. Ogren, T. Archer, S. B. Ross, in Catecholamines: Neuropharmacology and Central Nervous System Theoretical Aspects, M. Sandler, Ed. (Liss, New York, 1984), pp. 285-292)], by suggesting that each of these is associated with a different mode of LC function and that the LC may mediate shifts between them.
50. Our model provides an explicit mechanism that can account for the involvement of the LC system in selective attention [for example, N. R. Selden, T. W. Robbins, B. J. Everitt, J. Neurosci. 10, 531 (1990); N. R. Selden, B. J. Cole, B. J. Everitt, T. W. Robbins, Behav. Brain Res. 39, 29 (1990)] and in exploratory behavior and responsiveness to novelty (for example, S. J. Sara, C. Dyon-Laurent, A. Herve, Cognit. Brain Res. 2, 181(1995); S. O. Ogren, T. Archer, S. B. Ross, in Catecholamines: Neuropharmacology and Central Nervous System Theoretical Aspects, M. Sandler, Ed. (Liss, New York, 1984), pp. 285-292)], by suggesting that each of these is associated with a different mode of LC function and that the LC may mediate shifts between them.
51. Our model provides an explicit mechanism that can account for the involvement of the LC system in selective attention [for example, N. R. Selden, T. W. Robbins, B. J. Everitt, J. Neurosci. 10, 531 (1990); N. R. Selden, B. J. Cole, B. J. Everitt, T. W. Robbins, Behav. Brain Res. 39, 29 (1990)] and in exploratory behavior and responsiveness to novelty (for example, S. J. Sara, C. Dyon-Laurent, A. Herve, Cognit. Brain Res. 2, 181(1995); S. O. Ogren, T. Archer, S. B. Ross, in Catecholamines: Neuropharmacology and Central Nervous System Theoretical Aspects, M. Sandler, Ed. (Liss, New York, 1984), pp. 285-292)], by suggesting that each of these is associated with a different mode of LC function and that the LC may mediate shifts between them.
52. We have found in preliminary studies that direct manipulation of LC activity by local microinjections produces effects in task performance predicted by our model [S. Ivanova, J. Rajkowski, V. Silakov, T. Watanabe, G. Aston-Jones, Soc. Neurosci. Abstr. 23, 1587 (1997)].
53. P. R. Montague, P Dayan, T. J. Sejnowski, J. Neurosci. 16, 1936 (1996).
54. I. Nelken and E. A. Vaadia, Biol. Cybern. 64, 51 (1990).
55. We thank M. Stemmler, E. Niebur, B. Waterhouse, and R. Zemmel for comments on the manuscript and S. Aston-Jones for illustrations. Supported by the Human Frontiers Science Program, Air Force Office of Scientific Research grant F49620- 93-1-0099, and National Institute of Mental Health grants MH47566, MH45156, MH 55309, and MH 58480.