Cognitive map Mental model of physical or imaginary space

Mappa is a cloth or napkin in Latin; yet the first maps were probably drawn not on cloth but rather on clay, by the Babylonians. In the neurosciences, the term is occasionally used in its everyday connotation (definition 2), to denote functional architecture and spatiotemporal activity patterns in discrete brain areas (e.g. in "functional neuroimaging), without necessarily making assumptions about the code and the fine properties of the map. This usage needs no further elaboration. But 'map' is also used more specifically in two different contexts or "levels of analysis. One refers to the mode of representation ofsensorimotor space in neuronal space (definitions 3 and 4), the other to cognitive representation of the world (definition 5). The first requires "reductive analysis of brains, the latter may even satisfy psychologists and philosophers who prefer not to map the world in terms ofcircuits and spikes.

Let us navigate our way in the neuroscience map-world top-down, from behaviour to neurons. Cognitive maps are mental models of the world. In its most common usage, a 'cognitive map' is meant to be a mental analogue of a topographic map. The idea dates back much before scientific psychology was born, and is reflected, for example, in ancient methods of "mnemonics. It was Tolman's neo"behaviourism that has endowed the concept with a stronger scientific flavour. Tolman trusted that mental maps are instrumental in enabling the organism to locate goals and get from one place to another (Tolman 1948). The spatial map concept was later supported and much expanded in a series of elegant experiments on the role of "hippocampus in guiding "rats in "mazes (O'Keefe and Nadel 1978). One can entertain multiple versions of such maps, differing, for example, in their resolution and behavioural control—from detailed plans that exploit salient landmarks to permit navigational plasticity, to less-detailed global representations of space (ibid; Gallistel 1989). Analysis of experience-dependent modifications in so-called hippocampal 'place maps' and their candidate components, 'place cells', is a fruitful branch of the molecular neurobiology and "neurogenetics of learning and memory (e.g. Wilson and Tone-gawa 1997).Whether the major role of the hippocampus is to encode topography is still debated. Some authors propose that the mapping functions of the hippocampus much exceed the spatial domain (e.g. Wood et al. 1999). Few, though, will dispute that spatial coordinates do fulfil a major role in our cognition. The need to map space in order to navigate to food and away from predators may have provided many species, including invertebrates, with the capacity for cognitive maps. However, the risk of "anthropomorphizing in construing animal behaviour should always be considered; complex navigational abilities in "'simple' species may not necessarily require off-line 'cognitive maps' (e.g. Wehner and Menzel 1990; Bennett 1996). Note that the concept of 'cognitive map' is not limited to representation of physical space, and "generalizes to faculties that do not rely directly on sensory attributes, such as mental depiction of "taxonomies, or of social status (Aronson 1995; Laszlo et al. 1996). Such nonspatial mapping may still involve imagery of imaginary or "metaphorical spatial coordinates. There was even a proposal that spatial mapping has provided a phylo-genetic platform for the emergence of human language (O'Keefe 1996).

Another context in which mapping is employed in the neurosciences is sensory or motor encoding in brain circuits. When the neurons in such circuits are arranged so that their spatial relationship conserve those in the peripheral sensory epithelium, the map so formed is termed 'projectional', or 'topographical' (Konishi 1986; Kaas 1997) (Figure 44). A "classic example is the

Classical Space

Fig.44 A central topographical map of sensory space and its modification by experience. (A) A simplified lateral view of the right neo*cortex in the owl monkey.Areas 1—3b in the primary somatosensory cortex contain a somatotopic representation of the body surfaces (*homunculus).The location of the hand representation is marked by hatching. (B) The hand surface of an adult monkey. Numbers 1-5 denote the digits (1 is the thumb); distal (d), middle (m), and proximal (p) phalanges; Pi—4, the palmar pads at the base of the digits. (C) A map of the representation of hand surfaces indicated in B, in area 3b of the somatosensory cortex.The map is rotated 90° counterclockwise with respect to A. Grey areas, dorsal (hairy) skin on each digit. (D) The behavioural apparatus used for studying the effect of differential stimulation of restricted skin surfaces of the hand on the representation of these surfaces in area 3b. The monkey was trained to maintain contact with a rotating disk in order to get banana pellets. Only the distal aspect of the distal segment of digits 2, 3, and occasionally 4, contacted the disk. (E) The cortical hand representation of the same monkey as in C, after about 20 weeks of daily training (1.5 h/day) in the apparatus depicted in D. Note the remodelling of the map and the marked expansion of the representations of the distal aspects of digits 2,3, and, to a lesser degree, 4. (Adapted from Jenkins and Merzenich 1987.)

/.in. Banana A"A pellet

Fig.44 A central topographical map of sensory space and its modification by experience. (A) A simplified lateral view of the right neo*cortex in the owl monkey.Areas 1—3b in the primary somatosensory cortex contain a somatotopic representation of the body surfaces (*homunculus).The location of the hand representation is marked by hatching. (B) The hand surface of an adult monkey. Numbers 1-5 denote the digits (1 is the thumb); distal (d), middle (m), and proximal (p) phalanges; Pi—4, the palmar pads at the base of the digits. (C) A map of the representation of hand surfaces indicated in B, in area 3b of the somatosensory cortex.The map is rotated 90° counterclockwise with respect to A. Grey areas, dorsal (hairy) skin on each digit. (D) The behavioural apparatus used for studying the effect of differential stimulation of restricted skin surfaces of the hand on the representation of these surfaces in area 3b. The monkey was trained to maintain contact with a rotating disk in order to get banana pellets. Only the distal aspect of the distal segment of digits 2, 3, and occasionally 4, contacted the disk. (E) The cortical hand representation of the same monkey as in C, after about 20 weeks of daily training (1.5 h/day) in the apparatus depicted in D. Note the remodelling of the map and the marked expansion of the representations of the distal aspects of digits 2,3, and, to a lesser degree, 4. (Adapted from Jenkins and Merzenich 1987.)

somatotopic "homunculus in the primary sensorimotor "cortex (Penfield and Rasmussen 1968). When the map does not conserve the topography of the corresponding sensory epithelium, it is 'centrally synthesized' or 'computational' (Knudsen et al. 1987). The distinction is methodologically convenient but should be taken with a pinch of salt; 'projectional' mapping involves central computations as well. Examples for 'computational maps' are many (e.g. Knudsen et al. 1987; Knierim and Van Essen 1992; Schreiner 1995), although in most cases the contribution to the overall "internal representation is still not fully appreciated. A "classic example in which the contribution of a centrally synthesized map to an internal representation is evident, is that of the map that encodes interaural time differences of the barn owl: here spike time-code is transformed into place code by a brain-stem circuit that uses delay lines and "coincidence detectors (Konishi 1986) (Figure 45).

Some general issues concerning sensory maps are noteworthy:

1. Traditionally, 'central maps' are expected to display systematic variation in the value of at least one sensory attribute across at least one dimension of neural structure. The lack of apparent systematicity was actually taken to indicate that mapping is not indispensable for neural computation (Knudsen et al. 1987). However, in some brain circuits we cannot yet conclude whether there is an orderly, systematic variation in the encoding of an important attribute across a dimension of neural structure. In other cases, miniature ordered maps are discovered dispersed in a seemingly irregular mosaic (e.g. pinwheel-like patterns of orientation selectivity in the mammalian visual cortex; Bonhoeffer and Grinvald 1993). Declaring a representation 'non-map' by the criteria of definition 3 may therefore prove premature. It might be useful to relax the constraints and extend the concept of'brain map' to all cases in which world attributes are represented in a confined, dedicated brain area (definitions 1 and 2). In that case, 'map' becomes more of a generic term for localized candidate internal representations (that can of course map into each other)—but see 6 below.

2. Mapping involves transformation of codes, for example from time code (see above, Figure 45) or chemical code (e.g. Rubin and Katz 1999) into place code.

3. Maps recombine to produce higher-order maps. In the process, topographical and computational maps of different modalities interact and align,

Fig. 45 A schematic *model of a computational map of acoustic representation in the brainstem of the barn owl. The map converts spike time code into place code. The circuit uses delay lines and 'coincidence detection for measuring and encoding interaural time differences. Neurons A-E are arranged in an array and fire maximally only when signals from the left and from the right arrive simultaneously. Temporal information about the acoustic signal is encoded by spike timing. The axonal path increases in opposite directions for the two sources, thus creating a left-right asymmetry in transmission delays. When binaural disparities in the acoustic signals exactly compensate for this asymmetry, the neurons fire maximally. The output of the neurons does not use spike timing to encode time, but rather the position of the neuron in the array signals the interaural time differences for which the neuron responds maximally. (Adapted from Konishi 1986.)

Fig. 45 A schematic *model of a computational map of acoustic representation in the brainstem of the barn owl. The map converts spike time code into place code. The circuit uses delay lines and 'coincidence detection for measuring and encoding interaural time differences. Neurons A-E are arranged in an array and fire maximally only when signals from the left and from the right arrive simultaneously. Temporal information about the acoustic signal is encoded by spike timing. The axonal path increases in opposite directions for the two sources, thus creating a left-right asymmetry in transmission delays. When binaural disparities in the acoustic signals exactly compensate for this asymmetry, the neurons fire maximally. The output of the neurons does not use spike timing to encode time, but rather the position of the neuron in the array signals the interaural time differences for which the neuron responds maximally. (Adapted from Konishi 1986.)

presumably to permit coherent "perception and action (e.g. Wallace et al. 1996; Feldman and Knudsen 1997). This indicates that the brain uses general strategies to process codes regardless of the original modality or of whether the map was topographical or computational to begin with. The shape of maps may reflect developmental constraints, or phylogenetic pressures to optimize wiring or facilitate computations (Konishi 1986; Schreiner 1995; Kaas 1997; Van Essen 1997). The possibility should, however, be considered that in some maps topography has little to do with the ultimate representational meaning (see 6 below).

5. Suppose the spatiotemporal states of brain maps could be recorded in physiologically meaningful time windows—will we be able to read in individual maps only types of computations, or also tokens, namely statements with a specific representational 'semantics' which encode a unique physiological and behavioural situation?1

6. Ample evidence now indicates that central maps can be altered by experience (e.g. Weinberger 1995; Kilgard and Merzenich 1998; Faber et al. 1999). What is the relevance of such changes to learning and memory? Does the observed experience-dependent modifications in the map reflect lasting representational alterations, or solely changes auxiliary to the representational change, for example, expansion of computational space (Dudai 1999)? With the impressive advances in "methods such as functional neuroimaging and molecular neurobiol-ogy, more and more experience-dependent alterations are bound to be detected in brain maps. The temptation to declare each of these a manifestation of learning should be better quenched until we understand what the map really charts.

Selected associations: Internal representation, Model,

Plasticity, System

1For more on type vs. token, see *system.

Was this article helpful?

0 0

Post a comment