There are cerebellum-like structures, including the valvula and the electroreceptive lateral line lobe (ELL), that are found in certain species of fish, as well as the mammalian dorsal cochlear nucleus, the architecture and molecular biology of which in some respects strikingly resemble that of the cerebellum of vertebrates. These merit separate discussion from the topics in Chapter 2 on comparative anatomy of the cerebellum.

The Valvula

In addition to a true cerebellum (corpus cerebelli), there are several structures in electroreceptive fish (particularly the mormyrids), including the valvula, that are recognizably cerebellar or cerebellar-like. This resemblance suggests that a comparative study of them could yield some general rules concerning the cerebellum (Bell and Szabo 1986). Thus, among the bony fishes (teleosts), the mormyrids, which comprise certain African fresh-water fish, possess a cerebellum that is divided into the usual parts, that is, the body (corpus cerebelli), and a part arranged quite differently, the valvula. This highly unusual structure is considered in some detail.

The valvula, which is actually unique to the mormyrids, can grow to relatively large dimensions, covering all other parts of the brain (Fig. 12.1). This enlargement has been considered to be related to the high degree of differentiation of the lateral line system in these fish, which have a weak electric organ in the tail. They also have electroreceptors, which to a large extent are constituted of lateral line organs distributed over the body, and which have become modified to be sensitive to electric fields in the water (Nieuwenhuys and Nicholson 1969a).

History

In the introductory chapter of his magnificent three-volume monograph, The Comparative Anatomy and Histology of the Cerebellum, Olof Larsell (1967; Larsell and Jansen, 1972) provides a history of the cerebellum in its gross aspects, which reads in part as follows (additional historical details can be found in Clarke and O'Malley 1968):

In view of the importance attached to the climbing fibers and the inferior olivary nucleus (also termed the inferior olivary complex, inferior olive, or simply IO) in a number of theories of cerebellar function (Chapters 13, 14, 16, and 17), its anatomy and physiology is surveyed in some detail (see also Chapter 6).

The sequence of topics in this chapter is as follows: organization of the inferior olivary nucleus itself; its afferent input from the periphery; projections to it from other structures (e.g., the cerebellar nuclei); and finally, the climbing fibers (i.e., the output system from the inferior olive to the cerebellar cortex with their collaterals to the cerebellar nuclei). The chapter closes with a consideration of theories of function of the inferior olivary nucleus and a discussion of modulation of sensory (peripheral) input to the inferior olive. For purposes of differentiation, arbitrarily, afferent refers to input derived from the periphery, whereas projection refers to input derived from or directed to another brain structure (e.g., cerebellar cortex or nucleus).

The Inferior Olivary System

Organization of Inferior Olivary Nucleus

The inferior olive is a folded gray mass in the medulla, lying just above and slightly lateral to the pyramidal tracts (Fig. 5.1), and consists of a principal olive and dorsal and medial accessory olives. The principal olive (Fig. 5.2, bottom, 4) is a folded narrow band of cells in which a ventral and a dorsal lamella can be distinguished.

In view of the limited scope of this book and the correspondingly limited treatments of cerebellar anatomy and physiology, the following books and papers may be cited as supplementary sources: Brodal (1981); Brodal (1998); Eccles (1973, 1977b); Gilman, Bloedel, and Lechtenberg (1981); Ito (1984); Kandel, Schwartz, and Jessell (1991); Lechtenberg (1981); and Rothwell (1994).

The anatomical aspects of this chapter are based primarily on Brodal (1981); Eccles, Ito, and Szentágothai (1967); Ghez (1991); Ito (1984); Paley and Chan-Palay (1974); Palkovits, Magyar, and Szentágothai (1972); Parent (1996); and Ramón y Cajal (1995), supplemented in many instances from more recent publications. A synopsis of cerebellar anatomy appears in Voogd and Glickstein (1998). Further aspects of cerebellar system anatomy that are particularly relevant to the topic of adaptive control are considered in subsequent chapters: the mossy fibers (Chapter 4), the climbing fibers and their source, the inferior olive (Chapter 5), and the cerebellar nuclei and their efferent pathways (Chapter 6).

Synopsis of Principal Findings in Human Cerebellar Disease

To provide the elements of a clinical perspective, this chapter includes limited information about symptoms and signs of cerebellar disease in humans. From the wealth of available sources on clinical neurology, the material here has been adapted primarily from Victor and Ropper (2001), with additional material from the classical paper by Holmes (1939).

Lesions of the cerebellum in humans give rise to basically three types of abnormalities: (1) decrease of muscle tone (hypotonia), (2) incoordination (ataxia) of volitional movement, and (3) disorders of equilibrium and gait. Extensive lesions of one cerebellar hemisphere can give rise to all three types of abnormalities, on the same side as the lesion, as can lesions of the cerebellar nuclei and/or cerebellar peduncles. Lesions of the dentate nucleus or of the superior cerebellar peduncle result in the most severe and lasting symptoms.

Hypotonia, which refers to a decrease in passive resistance of muscles to movement (e.g., extension of a limb) is ascribed to a depression of gamma and alpha motor neuron activity, and tends to disappear with time.

The most prominent manifestation of cerebellar disease (i.e., abnormalities of volitional or intended movement) are included under the general heading of cerebellar incoordination or ataxia. These terms include dyssynergia, dysmetria, and dysdiadochokinesis (impaired or slowed repetitive reversals motion [e.g., of alternation or pronation-supination of the forearm; Fig. 11.1], or successive touching of each finger to the thumb). Complete impairment of such alternation maneuvers is termed adiadochokinesis.

The Cerebellar Nuclei

This chapter, which is largely based on Brodal (1981) and Chan-Palay (1977), is concerned mainly with the lateral (dentate) nucleus, as representative of the cerebellar nuclei, which also include the interpositi (emboliform and globose) and the medial (fastigial) nuclei, the terms in parentheses indicating those for the human cerebellum.

Anatomical Aspects

In humans, there are four distinct cellular masses or nuclei in the white matter of each half of the cerebellum (also termed the deep cerebellar nuclei). Most medial is the fastigial nucleus, followed more laterally by the small globose and emboliform nuclei and, most laterally, the dentate. The dentate nucleus appears in sections as a wrinkled band of gray matter (not unlike the inferior olive) with a medioanteriorly directed hilus. In the rat, cat, monkey, and most mammals, the usually accepted counterparts are the nucleus medialis, nucleus interpositus anterior and posterior, and nucleus lateralis, respectively.

The dentate nucleus is enormous in humans, both in comparison with the other nuclei and in comparison with other species; in fact, it has been estimated to contain some 284,000 cells (see also Heidary and Tomasch 1969). The principal afferent fibers to the cerebellar nuclei are the Purkinje cell axons from the cerebellar cortex, which are inhibitory. Other afferents include collaterals of the climbing and the mossy fibers, which are excitatory. Almost all efferent fibers from the cerebellum are axons of cells in the cerebellar nuclei.

Introduction

In this final chapter, a brief retrospective is presented on the history of the cerebellum as an adaptive controller, a term used as an inclusive term also for adaptive signal processor (ASP) and adaptive filter, to refer to a system having the capability to adjust or optimize its own parameters automatically. An alternative but equivalent view of the cerebellum as a three-layer neural net follows. Then, the significance of the basic uniformity of the structure of the cerebellar system is briefly reviewed, as a preliminary to several other topics. (As previously in this book, the term cerebellar system refers to the cerebellar cortex, cerebellar nuclei, and the inferior olive [the olivary nucleus], together with their interconnections.) The significance of the two principal afferent fiber systems of the cerebellum is then considered in relation to their features and to their counterparts in adaptive controllers.

Next, the vestibulo-ocular reflex (VOR), which is based on the vestibulocerebellum, is viewed as the simplest and most straightforward manifestation of an adaptive controller model of the cerebellum. The VOR is then used as a model for the functions of the remaining parts of the cerebellum, the spinocerebellum and the cerebrocerebellum, including the controversial question of the cerebellum and cognition. Then, some remaining problems are considered, such as the functions of the cerebellar nuclei and especially the inferior olive, and the function of negative and positive feedback circuits.

The implementation of the adaptive signal processors (ASPs) depicted in Figures 15.11, 15.12, and 15.15, with some modifications, was as follows. In the case of the linear predictor of Figure 15.11 (in which the training signal is identical with the input signal), the processor consists basically of a 16-tap delay line (Fig. A.1), the outputs of taps 9–16 of which are fed, respectively, to eight amplitude-matching units, each of the type shown in Figure 15.1A. (If the 16 taps of such a delay line are summed by means of precision (1%) resistors, linear interpolation results [compare traces 1 and 2 of Fig. A.2]. If two such units are cascaded, a sigmoid [actually a double parabolic] curve can be generated from a step function [compare traces 1 and 3 of Fig. A.2].)

As indicated in Figure A.1, the analog tapped delay line itself (which has a DC response) consists of two synchronously driven 16-channel analogue multiplexers, which serially transfer, via their interconnected common terminals, the stored voltages between adjacent condenser-based storage (sample-and-hold) units (Barlow 1993, pp. 313–315). The outputs are available both in individual form and in multiplexed form, the latter from the interconnected common terminals of the multiplexers (Fig. A.1). In essence, the unit operates as a “bucket brigade,” as does a charge-transfer device. In the present unit, however, it is voltage rather than charge that is transferred. Figure 15.10 shows a sine wave read out from the tapped delay line at four successively increasing delays.

This chapter is devoted to a relatively technical discussion of adaptive control and adaptive signal processors (ASPs; as mentioned previously, the two terms are used more or less interchangeably), as a preliminary to the continuation of the consideration of adaptive control models of the cerebellum in Chapter 16 and to Chapter 17.

The discussion in Chapters 13 and 14 suggests that two of the leading and more promising approaches to modeling the cerebellum have been the closely interrelated approaches of neural networks, on the one hand, and adaptive controllers, on the other hand. These two approaches share the feature of self-adjustments of weights to adapt themselves automatically to a given training signal or to a desired output, or to a succession of desired outputs.

In this chapter, a closer look is taken at the process of adaptive control in the form of a specific ASP of the author's own design and construction, which functions on-line and in real time. (As previously indicated, the term adaptive controller is used in reference to a system that adaptively controls a specific object [e.g., a biological or robotic limb], and adaptive signal processor is used to indicate basically the same process, but without controlling a specific object.) This exercise begins with a simple task and progresses to a series of more complex tasks. Sample results (i.e., waveforms) are given from which a better impression of the operation of these devices can be obtained than from descriptions and block diagrams alone.

On a theoretical basis (as summarized in Chapter 13), both Marr (1969) and Albus (1971) invoked memory storage at the parallel fiber–Purkinje cell dendritic tree interface if the parallel fibers and the climbing fibers (both excitatory) were conjointly active. In Marr's theory, the efficacy of these synapses was postulated to become enhanced (facilitated) with such conjoint activation, whereas in Albus's theory, synaptic efficacy became diminished (depressed). Ekerot and Oscarsson (1981) found that impulses in climbing fibers resulted in a depolarization of Purkinje cell dendrites lasting about 100 milliseconds. The authors conjectured that this effect might induce plastic changes in the parallel fiber synapses onto Purkinje cell dendrites, as envisaged in theories of motor learning by the cerebellum.

The diminished synaptic efficacy predicted by Albus was soon discovered (Ito and Kano 1982; Ito, Sakurai, and Tongroach 1982) in the form of long-term depression (LTD; i.e., a significant diminution of the parallel fiber postsynaptic potentials induced in Purkinje cells and lasting for at least 1 hour).

That there is indeed a capability for plasticity is illustrated for a subject (Fig. 7.1) throwing darts at a target (Fig. 7.2) before wearing prism spectacles, while wearing the spectacles, and after removing them. It is the direction of throw of the dart (which is normally determined from the direction of gaze) that undergoes change followed by adaptation or recalibration, which is then followed by a rebound and reverse adaptation upon removal of the prism spectacles.

This book assembles evidence that the requirements of a model to meet the unique anatomical and functional features that characterize the cerebellum are currently best met by adaptive control models (or their neural net equivalents), the signal feature of which is their ability to adjust (i.e., to optimize) their own parameters automatically.

In the survey of models in Chapter 13, several different types of cerebellar models, including some adaptive control and some neural network models, were treated more or less equally. In this chapter, by contrast, emphasis is placed on general features of (the closely interrelated) adaptive control and neural network models, especially concerning the question of plasticity (memory; i.e., changeability of weights), as a model for changes of efficacy of mossy fiber–Purkinje cell dendrite synapses under the influence of the climbing fibers. This chapter paves the way for Chapter 15, in which the features of adaptive controllers (as well as, in principle, neural network models) are illustrated with the aid of a specific implementation. Thus, a look is taken inside the “black box” or schematic diagram. In turn, more advanced and recent adaptive control and neural network models are considered in Chapter 16.

A brief word should be mentioned about terminology and synonyms, a topic that will also arise later. Neural nets, nerve nets, or (artificial) neural networks, are also known as connectionist models (theories) of computation (Rumelhart 1990). Adaptive controllers are also known as adaptive filters, adaptive signal processors, state estimators, and Kalman filters or Kalman–Bucy filters. These terms are used more or less interchangeably in this text. (The term, adaptive controller, has a slightly different meaning from that of adaptive signal processor; the latter implies that there is no specifically controlled object; see also Chapter 16.)

The Flocculonodular Lobe (the Vestibulocerebellum)

The flocculonodular lobe is the oldest part of the cerebellum (i.e., the archicerebellum). It occupies the major portion of the primitive cerebellum (e.g., in the lamprey and urodele amphibia; Nieuwenhuys 1967), as indicated in Chapter 2. Further, it is with the flocculonodular (posterior) lobe of the cerebellum that the vestibular nuclei of the brain stem are most closely associated (Brodal and Jansen 1954). Correspondingly, among the afferent fibers to the flocculonodular lobe, the vestibular ones are the most significant. Primary or direct vestibular fibers (i.e., from the end-organ, the vestibular organ) reach the flocculus, nodulus, and the adjoining part of the uvula, as well as the fastigial nucleus and the lingula. Of these, the nodulus is a later phylogenetic development than the flocculus. It appears that, whereas primary vestibular fibers decrease in phylogenesis, secondary fibers undergo an increase.

In a study of the climbing fiber projection to the rat flocculus and adjacent ventral paraflocculus, Ruigrok, Osse, and Voogd (1992) found that two parts of the inferior olive (the dorsal cap of Kooy and the ventrolateral outgrowth) are both connected with a set of two alternating zones of floccular/ventral parafloccular Purkinje cells, suggesting that these zones reflect functionally distinct and discrete units related to specific aspects of visuomotor control.

Early in this century it was shown that aging is accompanied by a decrease in intranasal chemosensory sensitivity to camphor (Vaschide, 1904). Numerous studies have confirmed such findings for various odorants (Venstrom and Amoore, 1968; Schiffman, Moss, and Erickson, 1976; Stevens and Cain, 1987). Others have reported decreased ability to identify odorants with increasing age (Doty et al., 1984; Wood and Harkins, 1987; Cain and Gent, 1991), as well as greater tendencies for olfactory adaptation and slower recovery of threshold sensitivity (Stevens et al., 1989). In contrast, few investigators have reported stable olfactory function over a life span (e.g., Rovee, Cohen, and Shlapack, 1975).

What Is the Anatomical Substrate for Age-related Loss of Chemosensory Function?

The olfactory system is part of a living organism that undergoes significant changes related to aging. Both peripheral sensory and central processing units are affected in that process. In the following we shall try to summarize some of those effects concerning the aging olfactory epithelium, as well as changes in the central nervous system that may contribute to the deterioration of the aging sense of smell.

Central Nervous System and Olfactory Bulb

Olfactory receptor neurons (ORNs) may be subject to alterations similar to those seen in neurons in the aging brain. Alterations in the aging central nervous system (CNS) include, for example, deficits in the regulation of intracellular calcium levels, increased leakage of synaptic transmitters, and changes in neuronal arborization (Smith, 1988).

During the past decade, scientists have identified a large number of genes coding for olfactory receptor proteins in vertebrates, including humans, and in insects and nematodes (Buck and Axel, 1991; Selbie et al., 1992; Sengupta, Colbert, and Bargmann, 1994; Gao and Chess, 1999; Clyne et al., 1999). Much earlier, such entities had been hypothesized to exist, probably on the intuition that the nature of an odor was not “ethereal” but rather a material part of the odor source (Lucretius, De rerum natura, IV) and thus could interact directly with the detecting organism. During the twentieth century, that concept was commonly used by physiologists to discuss the coding of odors (Zwaardemaker, 1925; Guillot, 1948) and by pioneer chemists who postulated receptive sites for odor molecules (Amoore, 1967; Beets, 1982). Its recent implementation in identifying receptor proteins has emitted a “strong scent of success” (Lancet, 1991). However, when the data from molecular biology and the physiological properties of the olfactory system are compared, it becomes clear that the final word has not yet been spoken on olfactory coding.

When olfactory signals are detected and differentiated, they gain behavioral significance when recognized as representing particular odor sources. In terms of neurophysiology, such processes require highly organized neuronal circuitry, and an important finding in recent studies of receptor proteins is that the receptors themselves are involved in determining the neural space devoted to representation of the chemical environment.

Whether or not human pheromones exist and how they might influence human physiology and behavior have been debated for decades; for a review, see Preti and Wysocki (1999). In this chapter, we present the concepts and data that are shaping our understanding of pheromones and pheromone-like effects in humans and other species. This will include the semantic issues associated with use of the term “pheromone” and descriptions of experiments designed to determine the psychological effects of two putative human pheromones. The expectation that human pheromones can consistently elicit stereotyped behavior is unrealistic. We argue that it is more likely that airborne signals are context-dependent and have more general, modulatory effects that can best be captured conceptually in terms of “modulatory pheromones” or “social chemosignals” (McClintock, 2000). Research with humans also presents a unique opportunity to consider the functional role of an awareness or conscious experience of pheromones and social chemosignals, whereas animal research is limited to studies of overt behavior.

Definition of “Pheromone”

According to the classic definition (Karlson and Lüscher, 1959), pheromones are airborne chemical signals emitted by an individual that trigger specific neuroendocrine, behavioral, or developmental responses in other individuals of the same species. Early pheromone research began with insects (Karlson and Butenandt, 1959), and the term “pheromone” was coined to designate a group of externally released “active substances” that triggered specific behavioral responses and were similar to, but could not be called, hormones. An example is bombykol, a substance emitted by female silkworm moths.