In the theory of mind an intelligent supervisor in the brain that reads information and commands action

Homunculi (Latin, diminutive of *homo, man) have a rich history in science and philosophy, but not all homunculi were created equal. As a term in biology, 'homunculus' is a close relative of'animalculus' ('little animal'). The latter was a generic term used by the early microscopists of the seventeenth century to refer to the microorganisms discovered under the magnifying lens (Wilson 1995). Guided by a combination of naive observations and wishful thinking, animalculi were occasionally depicted as miniature editions of full-sized animals. Probably the most famous one was invented by the Dutch microscopist Nicolaas Hartsoeker. While investigating sperm, he portrayed the first 'homuncu-lus': a curled-up tiny human being enclosed inside a spermatozoon like a passenger on an aeroplane diving for a crash. Hartsoeker actually wrote that what he drew was not what he saw but what one might hope to see, yet this remark was rapidly forgotten, whereas the homunculus was not (ibid.). The mythical homunculi in sperms are currently of interest primarily to historians of biology. In this discussion we will deal with another species of homunculus, created in philosophy to account for the operation of the mind. Basically, so goes this version of the homunculus story, there is a little man inside our head, that sees, hears, smells and tastes, feels, contemplates and plans, pulls pulleys, presses levers, and makes us think what we think and do what we do. Admittedly, such an homunculus does have an intuitive appeal. How nice would it be to be able to explain the "enigmas of the brain by "reducing them to a search for a little fellow that hides somewhere there and is responsible for it all. The problem is that the solution simply postpones the difficulty, is clearly too "anthropomorphic, and any smart homunculus would have rejected it.

Rephrasing Eliot on cats (Eliot 1939), the naming of homunculi is a difficult matter. In psychology they become variants of 'central executives' (see "working memory). In physiology they resurface disguised as 'grandmother cells', 'command neurons', and more (see below). In philosophy, homunculi have been discussed many years before they were so named. Descartes' proposal that the pineal gland is the site of the soul (Descartes 1649) is a version of the homunculus with assigned neuroanatomical coordinates. Other thinkers were much less enthusiastic; Leibniz, for example, opposed those searching for explanations in 'little demons or imps which can without ado perform whatever is wanted' (Leibniz 1704). More than two centuries later, Homunculi found their way into the debates on "behaviourism in psychology. Homunculi-based explanations of behaviour were inherently problematic for behaviourism. On the one hand, homunculi could relieve psychology from the need to explain inner mental processes; on the other, themselves they were postulated inner mental structures. 'The function of the inner man is to provide an explanation which will not be explained in turn', so Skinner (1971),'... Explanation stops with him'. But he also added: 'Science does not dehumanize man; it de-homunculizes him' (ibid.).

Another window through which the homunculus stares at us is artificial intelligence. In this discipline, the assumption that intelligent systems must ultimately harbour an intelligent executive agent, that is more-or-less in the image of its host, is dubbed 'the homunculus fallacy' (Kenny 1971). Some authors argue that depiction of 'homunculi' does not necessarily implies adapting this fallacy, because homunculi can be discharged with a hierarchy of progressively sillier homunculi: 'Homunculi are bogeymen only if they duplicate entire the talents they are rung to explain... If one can get a team or committee of relatively ignorant, narrow-minded, blind homunculi to produce the intelligent behavior of the whole, this is progress' (Dennett 1978; see also Minsky 1985). Note, however, that if the homunculus is multiplied extensively and made really stupid, one disposes of the homunculus rather than of the fallacy.

Most neuroscientists are familiar with another, more tangible use of the term 'homunculus', in neu-roanatomy and neurology. Resting on the shoulders of the great European neuroanatomists of the nineteenth century, several teams of investigators have mapped the representation of body surface in the brain (Marshall et al. 1941; Penfield and Rasmussen 1950). They came up with the finding that somatic sensations arising from the body surface map on to specific areas of the primary somatosensory cortex, although the "map is distorted: different surfaces of the body occupy areas that are disproportional to their actual physical size. This cortical topographical map is termed 'sensory homunculus'. An analogous 'motor homunculus' exists in the primary motor cortex. Despite the graphic representation of bodily figures on cortical areas in neurology books, the use of the term 'homunculus' in this context does not imply that a supervisor resides in those areas, even if some newcomers to brain research might erroneously suppose that it does.

From the point of view of memory research, the concept of 'homunculus' is pertinent to two critical albeit related issues: the encoding of "internal representations, and the localization of the "engram. For if there were homunculi, one possibility would have been that learning involves changes in the way they read and govern. There are two extreme views on the nature of representational codes in the brain. One is that complex representations are realized in single neurons ('unitary code', e.g. Barlow 1972). Such units are dubbed 'grandmother', 'gnostic', or 'pontifical' cells or units (Konorski 1967; Baum et al. 1988).1 A related concept exists for motor programmes: the 'command neuron', a neuron responsible for a certain behaviour and critical in generating it (Wiersma and Ikeda 1964; Kupfermann and Weiss 1978; Edwards et al. 1999). Now, single units that encode complex representations do carry a connotation of 'homunculi': if there are cells that encode the grandmother, why not cells that encode complete autobiographical narratives, and yet others that read the whole brain, encode, and navigate our consciousness? The opposing view considers internal representations as distributed over many neurons, while none of the individual units encodes a significant part of the representation ('population code', "cell assembly). In between these views one may envisage neuronal populations of various sizes, the members of which respond to complex stimuli and possibly even represent meaningful chunks of complex representations, e.g. cells that respond to hands or faces (Gross et al. 1972; Desimone 1991). Such cells are detected in circuits in multiple locations in the brain (Desimone 1991; 0 "Scalaidhe et al. 1997). Similarly, even 'command neurons' in simple systems are commonly described as groups of cells (Kupfermann and Weiss 1978). Internal representations seem thus distributed, even though the critical number of units that encode a meaningful chunk of the representation might be small (e.g. Young and Yamane 1992; Shadlen et al. 1996; "cell assembly). At least in the mammalian brain, the search for the engram should therefore focus on distributed neuronal populations, not on a single cell in a fixed address or on an elusive "plastic homunculus.

The main virtue of the 'homunculus' is that as a concept it forces us to think about how the brain understands and controls itself. All this notwithstanding, it is tempting to assume that the homunculus per se is still popular, more than we tend to concede, even in the mind of serious neuroscientists. We must overcome some very basic intuitions in order to drive him, or her, out.

Selected associations: Anthropomorphism, Reduction, Subject

1The important element in the concept is the convergence on a single processor. Gnostic units,for example, were not depicted necessarily as single cells, but were always contrasted with a representation that is distributed over many cells (Konorski 1967).

Was this article helpful?

0 0

Post a comment