Neural rhythmicity, feature binding, and serotonin: A hypothesis
Abstract
Natural selection favors animals that make successful predictive theories about the world. The first step in the formation of these theories is the construction of complex, multi-feature percepts. This process requires resolution of the binding problem, possibly via rhythmic cortical oscillations as suggested by von der Malsburg, Singer, Koch & Crick, and others. If the binding process were made rewarding, animals might enjoy theory-making and spontaneously become "smarter." I argue that the serotonergic raphe may have been used by evolution to link cortical binding with limbic reward centers and so serve as a neural substrate for the enjoyment of successful theory-making. I present evidence from the study of disorders like OCD and autism and drugs like LSD and MDMA suggesting that rhythmicity, reward, and pattern recognition are causally linked. I also propose that the genus Homo has tied powerful symbol manipulation hardware ("language") to the binding/theory-making circuits, allowing the construction, rehearsal, and communication of sophisticated models of the world. I suggest that many interesting phenomena, such as music-induced euphoria, deja vu, and the so-called "temporal lobe personality" can be explained by the interactions between these systems.
Introduction
Why do brains look for patterns? What is the hardware that causes us to theorize about events we experience? I use the word "theory" simply to mean the identification of percepts as belonging to some already understood class. For example, the classification of a particular Poodle as a dog is a "static" theory, predicting that the attributes of this particular object can be predicted by the attributes of the class Dog. Each assignment of an object into a class is in fact an experiment. We experience a few features of an object, and based on this, we tentatively place the object into a previously known class. There are also "dynamic" theories. If a ball is thrown, we use our in-house model of ballistics to predict the ball’s trajectory. We could be wrong; perhaps the ball has a spin, generating a force causing the ball to deviate from a normal path. Scientific theories are of both sorts, static and dynamic: The Linnean system of biological classification is a static theory, Newtonian mechanics is a dynamical theory.
That we try to make sense of the world in these ways is undeniable; that it is advantageous is intuitive. If we have good predictive ability, we can choose behaviors leading to the most desirable states of affairs. Even simple successes in theory-making are rewarding. We are all familiar with the satisfaction of remembering a location, fixing a broken device, or generally just being right. Can we explain why, neurobiologically, our brains cause us to behave and feel this way? What neural mechanisms underlie the search for order?
Feature binding and rhythmicity
The richness of our environment requires any would-be theory maker to segment the sensory world into discrete objects. However, there is a problem: If each complex object in the environment activates large sets of cells simultaneously, how will the brain know where the representation for one object ends and another begins? If we are presented with a gray square and a gray circle, three feature detectors activate: gray, square, and circle. In this case, downstream neurons observing this obtain a more-or-less clear interpretation of the world. Suppose instead that we see a gray square and a black circle. Now, four feature detectors are activated: gray, square, black, circle. If these detectors are all simultaneously active, the downstream interpretation is ambiguous. Which shape is which color? This is the so-called "binding problem" (Fig. 1).
Figure 1. The binding problem. In world #1, the presentation of a red square and a red circle excites three feature detectors: red, square, and circle. Downstream neurons can obtain a more-or-less clear interpretation of the objects in the world based on this activity. In world #2, the red square and blue circle excite four feature detectors: red, blue, square, circle. If all feature detectors fire simultaneously, the downstream interpretation cannot resolve which shape is which color. One possibility is to use rhythmic oscillations of two sets of feature detectors, so that at any given time, there is an unambiguous and accurate (though incomplete) representation of objects in the world.
As first suggested by Milner (1), and later clarified by von der Malsburg (2), a possible solution to the binding problem is to have sets of feature detectors rhythmically firing together in correlated bursts, i.e. (gray + square), then (black + circle), then (gray + square), etc. At any moment, the set of active feature detectors represents a single percept. This scheme predicts that the perception of complex objects (required for useful theory-building) will be accompanied by cortical oscillations, as subpopulations of feature detectors go on and off in their allotted time windows. Such oscillations have been observed (3-11), and indeed, there is a popular opinion that the oscillation solution to the binding problem may be correct (12-16). Crick and Koch (12,15) have even argued that such oscillations are the neural basis of consciousness itself.
Symbolic representations
Imagine an animal theorizing about lions. If the area of the brain housing the model activates the area of the brain which is activated when a lion is actually seen, this will be indistinguishable from actually seeing a lion in the world. Using the primary representation for model-building will cause hallucinations. To prevent this, some intermediary placeholder, or token, is required which will be used in all reasoning, predicting, all model building in general. Token use will, in fact, permit model building in the first place and will be a major evolutionary advantage. Using models, an animal could internally rehearse proposed actions and anticipate the consequences without taking any real risks. The system of tokens and token manipulation that evolution and associative learning have produced constitute what I will call "language."
Note that linguistic tokens are not necessarily things that can be said. It is quite common to use non-verbal tokens for mental objects, as when comparing two tastes. You can think of them as "this taste" and "that taste" – though you probably do not go to the actual trouble of doing so – but there is a richness to each idea beyond its "thisness" or "thatness." Somehow, an idea without a name still has a mentally graspable handle, its token. This general ability to juggle named and unnamed concepts is what I mean by "language." That is, it refers to all symbolic manipulation, not just that which can be verbalized, gestured, or written.
As the richness of the internal representational scheme grows, tokens attach themselves to clouds of sense-data and ideas of causal flow are extracted from the dynamics of the world. As this happens, tokens ("words") can be combined into strings ("sentences") representing experiences, both actual and imagined, and used in communication. In this way, the lexicon of a language is the collection of our static theories, and the grammar is the skeleton of our dynamical theories, both extracted from sensory experience. Although simplifying the world by regarding objects as members of a class entails information loss, throwing away complexity and reducing the entropy of experience is what enhances survival and makes language useful.
On the road to symbolic theory-making, then, there are three crucial steps. First, mental objects must be bound together and assembled from their constituent perceptual features. Second, each object must linked to a memory, so that the learned characteristics of that class of objects can be reconstructed; some of these characteristics may not be superficially apparent. Finally, the compound sense-memory object must be replaced by a token. Token labeling is usually learned, as each instance of experience is described or given a name. Eventually, a token’s "meaning" is the center of mass of its associated perceptual cloud (Fig. 2).
Figure 2. Percepts that strongly resemble prototypical memories result in more binding and require less information encoding than non-prototypical percepts. Life experience activates cell assemblies for each percept. These percepts are collections of features bound together by rhythmically oscillating neural ensembles. Repeated presentations of similar objects excite overlapping but non-identical neural ensembles; these ensembles form a cloud in neural activation space. At the center of the cloud is a fictional object, the prototype, which has probably never been exactly experienced. When a new percept excites ensemble activity very similar to that of the prototype, strong binding results and little additional information needs to be represented; the prototype, which is just the mean of the stored neural activity patterns, is by definition the expectation of any new stimulus of this class. When, instead, a new percept is very different than the prototype, surprise is greater and weak binding results. One goal of the animal is to experience as little surprise as possible; thus prototypical percepts, via their high levels of induced cortical rhythmicity, are most attractive.
Serotonin
A simple way to produce smart animals is to make the process of getting smarter enjoyable. If theory-making requires cortical binding and, in turn, rhythmic neural oscillations, then evolution could potentially have produced smart animals by (1) creating a neural subsystem which has broad sensory-cortical connectivity, which is (2) specifically activated by rhythmic input, and which (3) modulates the areas of the limbic system involved in reward. The serotonergic system of the raphe nucleus is a prime candidate.
The broad connectivity of serotonergic cells is almost unmatched, each cell contacting 500,000 cortical targets and each cortical cell receiving approximately 200 serotonergic varicosities (17). Interestingly, this cortical projection is selectively larger in primates than in rats (18). The densest collections of serotonergic fibers are those associated with sensory information processing and the limbic system (17,19,20). The limbic projections allow serotonin to control reward, which it appears to do via modulation of dopamine’s effects in the nucleus accumbens (19,21,22).
A major proposal for the preferred stimulus of serotonergic cells is rhythmic activity (17,23,24), from the observation that while raphe activity increases with behavioral arousal, spiking frequencies are highest during rhythmic motor activities such as licking, chewing, and running on a treadmill. Because of serotonin’s dual roles, Jacobs (23) suggested that "normal" people engage in repetitive activities such as knee-bouncing, chair-rocking, and head-bobbing in order to self-medicate with serotonin and obtain the associated reward. This bold proposal has many implications for various disorders, as described below, and is a central element in the construction of the present hypothesis, described next.
The hypothesis
The hypothesis is as follows. Rhythmic neural activity is induced by the binding of a set of features into a unitary percept (Figs. 1, 2). This percept is compared to and possibly bound to an object stored in long-term memory. The percept-memory binding activates the serotonergic raphe, which then activates, via the thalamus, three areas (Fig. 3). First is the frontal cortex, possibly the lateral orbitofrontal region, which uses the percept to update its model of the state of the world. Second is the temporal cortex, possibly Wernicke’s area, which retrieves the corresponding linguistic token and brings its representation into the binding circuit. The more the current percept resembles the prototypical object, the stronger this involvement will be. Third, serotonin released by the raphe enhances the reward associated with dopamine release in the nucleus accumbens (NACC). This makes the recognition process pleasurable. Finally, I argue that positive feedback from the cortex, through the raphe and thalamus and back to cortex, allows serotonin release to facilitate oscillatory binding as well as be facilitated by oscillatory binding.
Figure 3. Possible circuit diagram for the rhythmicity hypothesis. In sensory cortex, collections of feature detectors are bound together into unitary percepts via cortical or cortico-thalamic oscillations. Each percept is compared to long-term memory, searching for a match. If a match is found, the ensemble representation for the memory is brought into the binding circuit with the current percept, solving the secondary binding problem. This binding to memory induces rhythmic activity sufficient to activate serotonergic cells of the raphe nucleus, which then activates the thalamus. The thalamus is the main branch point for subsequent effects: three pathways lead to neocortex. A pathway to lateral orbitofrontal cortex (OFC) updates the internal model for the present state-of-affairs. A pathway to the temporal cortex (TCTX) activates a linguistic token for the memory-bound percept, if such a token exists. A pathway to the nucleus accumbens (NACC) could also come directly from the raphe; this projection modulates the reward associated with dopamine release in the NACC and is responsible for the putatively rewarding effects of effective theory formation and use. A final thalamo-cortical pathway to the prefrontal or premotor cortex (PFC) controls ongoing behavioral output; this circuit is not intimately related to the hypothesis of this paper but is shown for context. The feedback loops with the basal ganglia (which may go directly to cortex rather than return to thalamus) allow the generation of dynamical theories, expressions, and plans (25,26). In such cases each cortical area forwards a series of candidate projections for the next time step, activated according the probabilities based on previous experience. The basal ganglia (BG) imposes a winner-take-all competition and chooses the most likely next state.
I propose that the release of serotonin during the cortical binding associated with theory-making was used by evolution to make animals smarter. When an animal’s brain releases serotonin during successful theorizing, the animal will enjoy this process and will, over time, learn to understand and predict the world. This advantage will enhance its survival and will be favored by natural selection. I believe that there was at least one sharp increase in the quality of this system during the evolution of genus Homo, and that this increased quality resulted in our ability to create and use language and other forms of symbolic representation. At the same time, the relationship between rhythmicity and reward has allowed us to commandeer the underlying neural hardware and obtain pleasure from evolutionarily neutral repetitive stimulation, as proposed by Jacobs (23).
In this hypothesis, the frontal cortex houses acquired knowledge concerning projections of the current world state into the future. When several possible futures are available, a feedback loop with the basal ganglia resolves conflicting predictions (25,26). Because the tokens for each world-state are activated this process, the loop from frontal and temporal cortex to basal ganglia and back houses a generative grammar used in the production of sentences.
Drugs of abuse
One set of evidence for the rhythmicity hypothesis comes from the study of LSD and MDMA (Ecstasy), both serotonergic agonists (27-29). LSD users report hallucinations reminiscent of synesthesia – cross-modal binding of colors or textures with sounds, for example. They also report feelings of deja vu, intense insight, and enhanced pleasure with rhythmic activity, especially music. The subjective effects of MDMA are less hallucinatory and more euphorigenic, but there are similar effects regarding reactions to music.
Reconsider the model: serotonin is the link between oscillations underlying feature binding and reward, and there is positive feedback between serotonin and binding (Fig. 3). Because LSD and MDMA facilitate serotonin, binding should become enhanced, and features which do not actually belong to the same object get spuriously bound in cortex. Because higher serotonin levels are associated with mood elevation, this should be pleasurable. Finally, the positive feedback to oscillatory neural activity and the enhancements of neural reward centers would make rhythmic activity more rewarding than normal, hence the elevated appreciation for music. The recent emergence of the musical/drug phenomenon known as "raves" highlights this effect.
If the process which binds percepts to memories became artificially enhanced, what would result? If binding were sufficiently diffuse, the subjective impression would perhaps be familiarity with the overall state of affairs. It would resemble, I believe, a perception that the entire perceptual snapshot had been experienced before. Electrophysiologically, certain parts of the cortex would be in a state of abnormally high rhythmicity, i.e. a seizure. I argue that this is in fact the basis for déjà vu, which explains why this unusual mnemonic/perceptual experience is often reported by patients with certain seizure disorders, and is also induced by LSD.
Neuropsychiatric disorders
If this hypothesis is correct, then there should be correlations between various phenomena connected to serotonin, neural rhythmicity, reward, and pattern classification. Some data along these lines come from the study of obsessive-compulsive disorder (OCD), Gilles de la Tourette Syndrome (GTS), autism, depression, and epilepsy.
A common symptom of OCD, GTS, and autism is rhythmic motor activity. Patients with OCD and GTS report having to repeat an action over and over until they get it "just right" (perhaps until it exactly matches an action prototype?) and patients with autism spend hours alone, rocking or flapping their hands. Excessive rhythmicity actually extends beyond the behavioral scale into the neurological: there is high comorbidity of seizure disorders with each of these pathologies, and in some cases (GTS) symptoms are reminiscent of temporal lobe epilepsy (30-32).
A second point is that these disorders are often treated with serotonergic agonists (33-38). This is noteworthy because this is the same class of drugs used to treat patients with chronic depression. In fact, unmedicated OCD and GTS patients prevented from engaging in rhythmic or compelling behaviors become depressed. This might be interpreted as suggesting that they suffer from abnormally low levels of serotonin and reward, which they return to normal by engaging in rhythmic activities. Compulsions might then represent emotional pressures to make the sensory environment conform to learned prototypes, a subjective experience reminiscent of reports from the autism literature. Notably, serotonergic agonists are most effective in specifically reducing the repetitive behaviors and stereotypy pressures while not noticeably improving, for example, panic/anxiety symptoms seen in autism (36,39-42).
Depression is also treated with rapid eye movement (REM) sleep deprivation (43) and electroconvulsive therapy (ECT) (44,45). The therapeutic effect that has been reported with REM deprivation is presumably due to serotonergic quiescence during REM. By reducing REM, quiescence may be prevented so that serotonin levels rise [but see (46)]. ECT, on the other hand, is the paragon of rhythmic neural activity, and must strongly activate the raphe. A generalized seizure would resemble a state of intense cortical binding, raphe cells would fire vigorously, and serotonin levels would be predicted to rise.
A popularized though empirically rare aspect of autism [and GTS - see (47)] is the savant syndrome, characterized by exceptional ability in one narrow domain (47-49). A speculative explanation is that elevated serotonin levels (40,42,50,51) facilitate binding and pattern recognition skills. A corollary prediction, stemming from the ostensible connection between the elevated 5-HT levels of hyperserotonemia and cortical rhythmicity, is that there should be high comorbidity of autism and seizure disorders, which is the case (30-32).
Beauty and the religious experience
As mentioned earlier, the more a percept resembles a category prototype, the stronger the binding. This implies that oscillatory activity, and consequently reward, should be more pronounced as things tend more toward averageness or genericity. Thus, this hypothesis accounts for the finding that average faces are judged more attractive (52,53). If the phenomenon of averagenes preference is a general one – as this hypothesis would predict – then the most average head of broccoli, the most average rose, and the most average country farmhouse will also be judged as aesthetically superior. From another perspective, genericity implies low entropy and high predictability, which would be expected to make the world simple and safe.
Clinical observations have suggested an association between temporal lobe seizures on one hand and hyperreligiosity and hypergraphia on the other (54-57). While organized/ institutional religion is a complex social and psychological phenomenon, it is possible that the canonical transcendental religious experience, at least in part, has a simple underlying neural foundation. Consider the subjective experience that would be associated with large-scale cortical rhythmicity. The corresponding percept would be an object which possessed all features, universal in character. It would be impossible to describe adequately, as many of its features would be mutually contradictory, but nevertheless the neural circuitry for language would be strongly activated. Awareness of this concept would be predicted to be highly pleasurable, because strong binding would facilitate the serotonergic modulation of reward and pleasure centers in the NACC. I propose that something like the rhythmicity hypothesis explains the connection between seizure disorders (excessive rhythmicity), hyperreligiosity (global feature binding), and hypergraphia (excessive linguistic token activation). In crude terms, God is a seizure. Consistent with this proposal is the fact that in some societies, religious rituals involve the ingestion of a hallucinogen and the performance of rhythmic drumming or chanting. I suggest that because both of these stimuli would be expected to enhance cortical rhythmicity, they would indeed facilitate the religious experience.
Discussion
Obviously, this is a controversial hypothesis with many possible objections. One of the most immediate is the disparity between the frequency of oscillations thought to be involved in sensory binding, i.e. the 40 Hz gamma-band, and that involved in the behaviors known to activate the raphe, i.e. 1-5 Hz. On one hand, this objection might be investigated empirically by examining raphe response to 40 Hz cortical oscillations. On the other hand, there are defenses available in the absence of this evidence. One possibility is that the raphe responds only to every n’th peak in the binding oscillations. This would simply be a kind of frequency division, where the raphe is tuned to the power in lower harmonics of the gamma band. A second possibility is that there is another piece of neural hardware that is specifically designed to do frequency division on modules of activated cortex. Evidence for such a system is given by the hippocampal model of Jensen and Lisman (58,59), in which working memory scans through its current contents by packing multiple gamma-band peaks into each theta cycle in the hippocampus (58,59) and allowing each gamma peak to excite a population of cells. Thus, it could be hippocampal theta-oscillations and not cortical gamma-oscillations that excite the raphe.
I have greatly simplified serotonergic physiology in order to make a compact point in this paper. Before closing, several additional points should be raised. First is a finding that serotonergic cells fall silent abruptly when an unexpected event occurs, as when an experimenter suddenly enters the room of a grooming cat (23). Gradually, unit activity recovers to normal. When the experimenter leaves the room, there is a brief lull, less pronounced than that seen on the way in. This finding is at least circumstantially supportive of the rhythmicity hypothesis. When the cat is surprised by the opening door, the world is transformed into unpredictability. Binding is correspondingly low, and serotonergic unit activity crashes. As the world becomes comprehensible again, unit activity recovers, and because surprise is less pronounced when the experimenter leaves, the lull is less pronounced.
A second point regarding serotonin concerns dreaming. As mentioned earlier, raphe activity is reduced during REM sleep. If it is true that serotonin facilitates the binding of percepts to tokens and modulates reward, its quiescence during REM might imply that sensory-cortical experiences during REM never become "token-ized" or reinforced by NACC DA release. Because no language-based narrative would then occur on-line, it is possible that these experiences would therefore be difficult to reconstruct after the fact, accounting for the subjective effect that dreams are hard to remember. This is especially interesting in light of the hippocampus’s role in memory and the work of Jensen and Lisman (58,59).
Conclusion
The desire to make the world comprehensible is one of humanity’s deepest idiosyncrasies. It seems reasonable to assume that evolution would impel animals with the drive not only to survive, but to make survival easier; and to do this it would have to create neural hardware for the purpose. What is the nature of that hardware? What are the consequences of its failure? Even when the system is functioning perfectly, there are opportunities to take advantage of the brain’s rhythmic solution to the binding problem and the reward that results. Astonishingly, the pursuit of these opportunities has resulted in some of our greatest achievements.
"Music is a higher revelation than philosophy." – Ludwig von Beethoven
"Beauty is truth, truth beauty – that is all ye know on earth, and all ye need to know." – John Keats
"In every true searcher of nature there is a kind of religious reverence, for he finds it impossible to imagine that he is the first to have thought out the exceedingly delicate threads that connect his perceptions." – Albert Einstein
"You all, healthy people, can’t imagine the happiness which we epileptics feel during the moments before our fit…I don’t know if this felicity lasts for seconds, hours, or months, but believe me, I would not exchange it for all the joys that life may bring." – Fyodor Dostoyevsky
Acknowledgements
Barry Jacobs introduced me to the idea that rhythmic activity might be rewarding because it releases serotonin. For helpful discussions and comments I also wish to thank Brian Craft, Ole Jensen, Larry Abbott, John Lisman, Paul Mineiro, Jon Horvitz, and Kamal Sen. Through the course of this work, I was supported by an Office of Naval Research predoctoral fellowship, a Princeton University Honorific Fellowship, an NIH postdoctoral training grant, and a Sloan Fellowship for Theoretical Neuroscience.
References
[1] Milner, P. A model for visual shape recognition. Psychol. Rev. 1974;81:521-535.
[2] von der Malsburg, C and Schneider, W. A neural cocktail party processor. Biol. Cybernetics 1986;54:29-40.
[3] Gray, C, Konig, P, Engel, A and Singer, W. Oscillatory responses in cat visual cortex exhibit inter-columnar synchronization which reflects global stimulus properties. Nature 1989;338:334-337.
[4] Gray, C, Engel, A, Konig, P and Singer, W. Stimulus-dependent neuronal oscillations in cat visual cortex: receptive field properties and feature dependence. Euro. J. Neurosci. 1990;2:607.
[5] Gray, C and Viana di Prisco, G. Properties of stimulus-dependent rhythmic activity of visual cortical neurons in the adult cat. Soc. Neurosci. Abstr. 1993;19:868.
[6] Livingstone, M. Oscillatory firing and interneuronal correlations in squirrel monkey striate cortex. J. Neurophysiol. 1996;75:2467-2485.
[7] Tallon-Baudry, C, Bertrand, O, Delpuech, C and Pernier, J. Stimulus-specificity of phase-locked and non-phase-locked 40 Hz visual responses in human. J. Neurosci. 1996;16:4240-4249.
[8] Nicolelis, M, Baccala, L, Lin, R and Chapin, J. Sensorimotor encoding by synchronous neural ensemble activity at multiple levels of the somatosensory system. Science 1995;268:5215.
[9] Engel, A, Kreiter, A, Konig, P and Singer, W. Synchronization of oscillatory neuronal responses between striate and extrastriate visual areas of the cat. Proc. Natl. Acad. Sci. 1991;88:6048-6052.
[10] Engel, A, Konig, P, Kreiter, A, Schillen, T and Singer, W. Temporal coding in the visual cortex: new vistas on integration in the nervous system. Trends Neurosci. 1992;15:218-226.
[11] Murthy, V and Fetz, E. Synchronized 25-35 Hz oscillations in sensorimotor cortex of awake monkeys. Soc. Neurosci. Abstr. 1991;17:310.
[12] Crick, F and Koch, C. Towards a neurobiological theory of consciousness. Sem. Neurosci. 1990;2:263-275.
[13] Eckhorn, R, Bauer, R, Jordan, W, et al. Coherent oscillations: a mechanism of feature linking in the visual cortex? Biol. Cybernetics 1988;60:121-130.
[14] Singer, W. Putative functions of temporal correlations in neocortical processing. Large Scale Neuronal Theories of the Brain. C. Koch and J. Davis. Cambridge, MIT Press 1994.
[15] Koch, C and Crick, F. Some further ideas regarding the neuronal basis of awareness. Large-Scale Neuronal Theories of the Brain. C. Koch and J. Davis. Cambridge, MIT Press 1994.
[16] Koch, C and Davis, J, Eds. Large-Scale Neuronal Theories of the Brain. Cambridge, MIT Press 1994.
[17] Jacobs, B and Azmitia, E. Structure and function of the brain serotonin system. Physiol. Rev. 1992;72:165-229.
[18] Azmitia, E and Gannon, P. Anatomy of the serotonergic system in the primate and sub-primate brain. Advances in Neurology 1986;43:407-468.
[19] Broderick, P and Phelix, C. Serotonin (5-HT) within dopaminergic reward circuits signals open-field behavior: II. Basis for 5-HT–DA interaction in cocaine dysfunctional behavior. Neuroscience and Biobehavioral Reviews 1997;21(3):227-260.
[20] Jacobs, B. Serotonin and behavior: emphasis on motor control. J. Clin. Psychiatr. 1991;52:17-23.
[21] Brodie, M and Bunney, E. Serotonin potentiates dopamine inhibition of ventral tegmental area neurons in vitro. J. Neurophysiol. 1996;76(3):2077-2082.
[22] Boulenguez, P, Rawlins, J, Chauveau, J, Joseph, M, Mitchell, S and Gray, J. Modulation of dopamine release in the nucleus accumbens by 5-HT1B agonists: involvement of the hippocampo-accumbens pathway. Neuropharmacology 1996;35(11):1521-1529.
[23] Jacobs, B. Serotonin, motor activity, and depression-related disorders. Amer. Sci. 1994;82:456-463.
[24] Ribeiro-Do-Valle, L, Fornal, C, Litto, W and Jacobs, B. Serotonergic dorsal raphe unit activity related to feeding/grooming behaviors in cats. Soc. Neurosci. Abstr. 1989;15:1283.
[25] Mink, J. The basal ganglia: focused selection and inhibition of competing motor programs. Prog. Neurobiol. 1996;50:381-425.
[26] Redgrave, P, Prescott, T and Gurney, K. The basal ganglia: a neural architecture for conflict resolution. Soc. Neurosci. Abstr. 1997;23:192.
[27] Rudnick, G and Wall, S. The molecular mechanism of "ecstasy" [3,4-methylenedioxymethamphetamine (MDMA)]: serotonin transporters are targets for MDMA-induced serotonin release. Proc. Natl. Acad. Sci. 1992;89:1817-1821.
[28] Teitler, M, Leonhardt, S, Appel, N, De Souza, E and Glennon, R. Receptor pharmacology of MDMA and related hallucinogens. Ann. NY Acad. Sci. 1990;600:626-638.
[29] White, S, Obradovic, R, Imel, K and Wheaton, M. The effects of methlyenedioxymethamphetamine (MDMA, "ecstasy") on monoaminergic neurotransmission in the central nervous system. Prog. Neurobiol. 1996;49(5):455-479.
[30] Volkmar, F and Nelson, D. Seizure disorders in autism. J. Amer. Acad. Child and Adolescent Psychiatr. 1990;29(1):127-129.
[31] Wong, V. Epilepsy in children with autistic spectrum disorder. J. Child Neurology 1993;8(4):316-322.
[32] Rossi, P, Parmaegianni, A, Bach, V, Santalucci, M and Visconti, P. EEG features and epilepsy in patients with autism. Brain Devel. 1995;17(3):169-174.
[33] Devaugh-Geiss, J. Pharmacologic therapy of obsessive compulsive disorder. Advances in Pharmacology 1994;30:35-52.
[34] Baumgarten, H and Grozdanovic, Z. The role of serotonin in behavior modulation. Fortschr. Neurologie-Psychiatrie 1995;63(Suppl. 1):3-8.
[35] Pigott, T. OCD: where the serotonin selectivity story begins. J. Clin. Psychiatr. 1996;57:11-20.
[36] Levine, J, Aviram, A, Holan, A, Ring, A, Barak, Y and Belmaker, R. Inositol treatment of autism. J. Neural Transmission 1997;104(2):307-310.
[37] Silvestri, R, Raffaele, M, De Domenico, P, Tisano, A, Lagana, A and Di Perri, R. Serotonergic agents in the treatment of Gilles de la Tourette’s syndrome. Acta Neurologica (Napoli) 1994;16(1):58-63.
[38] Rapoport, J, Inoff-Germain, G, Imel, K and Wheaton, M. Tourette syndrome: medical and surgical treatment of obsessive-compulsive disorder. Neurologic Clinics 1997;15(2):421-428.
[39] Cook, E and Leventhal, B. The serotonin system in autism. Current Opinion in Pediatrics 1996;8(4):348-354.
[40] Cook, E, Courchesne, R, Lore, C, et al. Evidence of linkage between the serotonin transporter and autistic disorder. Molec. Psychiatry 1997;2(3):247-250.
[41] Lincoln, A, Courchesne, E, Harms, L and Allen, M. Sensory modulation of auditory stimuli in children with autism and receptive developmental language disorder: event-related potential evidence. J. Autism and Developmental Dis. 1995;25(5):521-539.
[42] McDougle, C, Naylor, S, Cohen, D, Aghajanian, G, Heninger, G and Price, L. Effects of trytophan depletion in drug-free adults with autism. Arch. Gen. Psychiatry 1996;53.
[43] Riemann, D, Schnitzler, M, Hohagen, F and Berger, M. Depression and sleep — the status of current research. Fortschr. Neurologie-Psychiatrie 1994;62(12):458-478.
[44] Figiel, G, Zorumski, C, Doraiswarmy, P, Mattingly, G and Jarvis, M. Simultaneous major depression and panic disorder: treatment with electroconvulsive therapy. J. Clin. Psychiatr. 1992;53(1):12-15.
[45] Ghaziuddin, N, King, C, Naylor, M, et al. Electroconvulsive treatment in adolescents with pharmacotherapy-refractory depression. J. Child. Adolesc. Psychopharmacol. 1996;6(4):259-271.
[46] Farooqui, S, Brock, J and Zhou, J. Changes in monoamines and their metabolite concentrations in REM sleep-deprived rat forebrain nuclei. Pharmacology, Biochemistry, and Behavior 1996;54(2):385-391.
[47] Moriarty, J, Ring, H and Robertson, M. An idiot savant calendrical calculator with Gilles de la Tourette syndrome: implications for an understanding of the savant syndrome. Psychological Med. 1993;23(4):1019-1021.
[48] Young, R and Nettelbeck, T. The abilities of a musical savant and his family. J. Autism and Developmental Dis. 1995;25(3):231-248.
[49] Kehrer, H. Savant capabilities of autistic persons. Acta Paedopsychiatrica 1992;55(3):151-155.
[50] Martineau, J, Barthelemy, C, Jouve, J, Muh, J and Lelord, G. Monoamines (serotonin and catecholamines) and their derivatives in infantile autism: age-related changes and drug effects. Develop. Med. and Child Neurol. 1992;34(7):593-603. [51] Malenka, R, Hamblin, M and Barchas, J. Biochemical hypotheses of affective disorders and anxiety. Basic Neurochemistry: Molecular, Cellular, and Medical Aspects. S. GJ. New York, Raven Press 1989.
[52] Langlois, J and Roggman, L. Attractive faces are only average. Psychological Science 1990;1:115-121.
[53] Rhodes, G and Tremewan, T. Averageness, exaggeration, and facial attractiveness. Psychological Science 1996;7(2):105-110.
[54] Waxman, S and Geschwind, N. The interictal behavior syndrome of temporal lobe epilepsy. Arch. Gen. Psychiatry 1975;32(12):1580-1586.
[55] Csernansky, J, Leiderman, D, Mandabach, M and Moses, J. Psychopathology and limbic epilepsy: relationship to seizure variables and neuropsychological function. Epilepsia 1990;31(3):275-280.
[56] Tucker, D, Novelly, R and Walker, P. Hyperreligiosity in temporal lobe epilepsy: redefining the relationship. J. Nervous and Mental Dis. 1987;175(3):181-184.
[57] Ramachandran, V, Hirstein, W, Armel, K, Tecoma, E and Iragui, V. The neural basis of religious experience. Soc. Neurosci. Abstr. 1997;23(2):1316.
[58] Jensen, O and Lisman, J. Theta/gamma networks with slow NMDA channels learn sequences and encode episodic memory: role of NMDA channels in recall. Learning and Memory 1996;3:264-278.
[59] Jensen, O and Lisman, J. Novel lists of 7 +/- 2 known items can be reliably stored in an oscillatory short-term memory network: interaction with long-term memory. Learning and Memory 1996;3:257-263.