Cognitive mapping as a methodological problem

Basing on long-terms experiments with rats navigating mazes, Tolman suggested that “something like a field map of the environment gets established in the rat’s brain…And it is this tentative map, indicating routs and paths and environmental relationships, which finally determines what responses, if any, the animal will finally release” (Tolman, 1948, p. 192). This concept was based on Tolman’s general idea that animals do not merely base their actions on specific stimulus-response associations, but that they also internally reorganise acquired spatial information to form cognitive representations of the environment (Tolman et al., 1946 a, b). One important property of such representations is that they allow animals to react to stimuli that are not immediately present because the relationship of such stimuli to those actually perceived is maintained in a cognitive representation, in other words, a map. Once an animal has built a map, it can use the information to solve a variety of problems. If animals are introduced into an environment from a new starting point or if the previous path is blocked, they will be able to deduce the most direct trajectory from this place to a known goal.

Being “officially introduced” by Tolman in the forties, cognitive maps have come into fashion for behavioural neurobiologists and experimental psychologists in 80-th and, as Whehner and Menzel (1990) give this, it is not astonishing that recently even insects have been claimed to possess cognitive maps. As these authors note, any such claim depends crucially on how a “cognitive map” is defined in operational terms. Ants ran parallel ways in labyrinths with rats as early as in 30-th and these two groups of inhabitants of underground secret passages yielded to no one in this respect (Maier and Schneirla, 1935). Anyway, we are going to face some controversy when consider cognitive mapping in insects.

The main problem to be discussed is to what degree mapping is cognitive, or, in other words, is intelligence involved into a process of mapping. In order not to be lost in the labyrinth of wild mapping, let us start with several examples before entering into details of modern theories of cognitive maps.

In his pioneering experiments with map-like behaviour in chimpanzees E. Menzel (1973 a, b) began straight away with a problem of rational choosing of multi-destination routes. He carried young chimpanzees around while he hid pieces of fruit at 18 different sites in their enclosure. Animals could see where the experimenter hides the pieces of fruit. Then they were placed into a dark enclosure for 2 minutes. When the animals were released, they retrieved food from all 18 sites on a single rapid foraging expedition. Menzel supposed that the apes tended to use a “least-distance” strategy to retrieve the hidden food and that they were based on cognitive mapping.

Experiments with radial mazes that have been carried out in order to establish how much information animals can remember, also lead to discussion about cognitive mapping. Researchers have tried from 8 to 24 arms mazes to test rats which easily remembered position of the rewarded arm (Olton, 1978; Roberts, 1979). After a while the rat would be able to empty the maze of treats without going down the same arm twice. It did not follow a set pattern and the route it took could not be predicted. At first Olton (1978) suggested that the rat left an odour marker at the entrance to each arm, but the researcher conducting the experiment eliminated that possibility. The clue came when the maze was rotated through 90 degrees. The rat became disorientated, entering tunnels that it had already been down because their position in relation to the objects in the laboratory had changed. The test demonstrated that the rat could “picture” locations of the tunnels it had already entered and integrate them into cognitive map. It thus was able to reflect on its recent experience in order to find food with the minimum of effort.

Another set of maze experiments which has presented evidence of a cognitive map in rats has been the work of Morris (1981) using a swimming pool. He trained rats in a circular pool fool of opaque water, from which they could escape by climbing up onto a platform that was submerged 1 cm below the water level and thus could not be seen by the animals. The platform maintained a fixed position with respect to landmarks or objects in the experimental room. It turned out that rats can swim directly even towards a platform that is completely invisible. The results showed that the rats learned how to find objects that they could not see, smell or hear, locating their position based on their spatial relationship to landmarks in the room.

There is extensive evidence that rats and other laboratory animals such as pigeons and goldfish use complex relationship of landmarks to find a goal (Gould, J.L. and C.G. Gould 1994; Rodrí guez et al., 1994). Nutcrackers use an original form of space navigation when relocate the hidden food: the birds act as if they cover an open space by “mental triangles”. They seem to form representations of the locations of cache sites, each based upon the relationship between the goal and surrounding landmarks and thus the nutcrackers probably use cognitive maps (Kamil and Cheng, 2001).

Subsequently it has become an issue of great controversy whether or not animals construct “true cognitive” maps. Perhaps Noah faced the same problem when stuffed his Ark: the more claimants, the more problems. Even for a strong lobbyist of intelligence in animals it is somehow difficult to imagine that a goldfish is really able to mentally build up a global representation of space.

The cognitive map can be operationally defined as novel rout construction that cannot be explained by orienting either by pure path integration or by the use of beacons coincident with the goal (Jacobs and Schenk, 2003). O’Keefe and Nadel (1978) were the first to present a formal theory of the cognitive map. They define the cognitive map as “the representation of a group of places, some related to others by means of a set of rules of spatial transformation”. When an animal enters a new environment, it forms a representation of the spatial relationship maintained by landmarks or other cues that it perceives from its position. Starting from this initial place representation, and as the animal moves around the environment, it begins to incorporate new place representations into a cognitive map. This occurs because of the information coming from the sensory and motor systems with respect to the different distances that the animal perceives; these systems enable space to be represented in a relative way, with reference to the animal. O’Keefe and Nadel called this space relative or egocentric space. This theory explores the idea that navigation based on a cognitive map (“cartographic learning”) occurs in an all-or-nothing way and that hippocampus is the very structure that is responsible for cartographic learning.

In 1990s, Gallistel suggested more tolerant theory of cognitive map based on a claim that any orientation that includes implicitly distances and directions is evidence of a cognitive map (“metric map”). He defines cognitive map as “a record in the central nervous system of macroscopic geometric relations among surfaces in the environment used to plan movements through the environment” (Gallistel, 1990, p. 103). Since an animal can only perceive a part of its environment from any one vantage point, the construction of a cognitive map requires the integration of positional information derived from different views of the environment made at different times (Gallistel and Cramer, 1996). The construction of a map for navigation involves combining two sorts of position vectors: egocentric vectors which specify the locations of landmarks in a body-centred coordinate system and geocentric vectors which specifies the animal’s position in a coordinate system, defined with respect to animal’s body. The geocentric vectors are calculated by path integration and specify the animal’s position in a coordinate system based on the path’s starting point.

Recently Jacobs and Schenk (2003) “unpacked” cognitive map and proposed the parallel map theory (PMT) which is founded on the premise that there is not one map but more: the integrated (or cognitive) map and its two components, the bearing map and the sketch maps, which when integrated can create the novel shortcut. The bearing map is the multicoordinate grid map, derived from sources of distributed stimuli such as gradients. The sketch map encodes and stores fine-grained topographical data and is constructed from the memory of the positions of unique cues. In mammals the bearing and sketch maps are mediated by independent structures of the hippocampal formation, but the authors consider other structures in other taxa including invertebrates mediating these functions.

The bearing and the sketch maps thus represent two classes of maps. They are also constructed from two classes of cues, defined as directional and positional. Directional cues can include distributed cues such as gradients of odour, light or sound, as well as compass marks, distal visual landmarks which provide directional but not distant cues. Positional cues are discrete and unique objects, often near the goal, that can be used to estimate distance to the goal accurately. These maps differ in their stability and plasticity. The bearing map, once created, serves as a scaffold for the localisation of positional cues in absolute space or for global positioning. The sketch map, in contrast, is not a singular representation but a type of map. Sketch maps may coexist as a population of a topographic representation of local spaces, where each sketch map encodes the navigator’s position within a specific panoramic array. Such maps arise from disjoin exploration, where a navigator may experience several local areas without recognising their directional relationship to each other i.e. without forming an integrated map, and encodes them into separate sketch maps. Unlike the bearing map, where position can be deduced by estimating the rate of change in a gradient, a sketch map requires significant spatial memory to encode the features of individual landmarks in an array.

Jacobs and Schenk (2003) consider a set of examples when the bearing map overrides the sketch maps. This concerns navigation in clock-shifted birds. Because the sun’s movement is a distributed cue, the bearing map also incorporates data from the sun compass. With the clock shift, the pigeon shows a directional shift, even when the topographic information is not changed. The same phenomenon has been observed in small scale orientation in other contexts: despite the presence of well learned positional cues, clock-shifted scrub jays, nutcrackers, pigeons and black –capped chickadees persist in searching for hidden food in the predicted compass deviation.

With a concept of “unpacked” cognitive map in mind, we can consider the well known examples of map-like behaviour in insects and birds attributing them simpler than cognitive mapping means of collecting sketch maps. Wehner considers insects, like ants, bees, and wasps, possessing intricate navigation systems and remarkable memory store but lacking “true” cognitive maps (Wehner, 1999; Wehner and Menzel, 1990). Instead, mapping behaviour in insects can be attributed to snapshot matching. This strategy might be quite sophisticated in so far as landmarks positioned at different distances can be disentangled by motion parallax, and snapshots from the same scene might be taken from different points. According to Wehner (1997), insects do not incorporate their routes into a map-like system of reference. They cannot take a bird’s eye- or even bee’s eye-view of the terrain over which they travel; instead, they assemble the necessary information piecemeal over time. For insects, it might be a safer and more robust strategy to rely on sequentially organised, gazetteer-like memories rather than to encode the spatial relations among a multitude of similar sites and routes in a large-scale mental map.

It is still possible, however, that hymenopterans like ants and bees do possess a cumulative global path integration (PI) memory reflecting long-term experience of where abundant food is to be found (Collett et al., 2003).We should not to discard insects as lacking at least map-like organisation of spatial memory. Recently by using harmonic radar that monitors the flight path of an insect carrying the transponder antenna over a distance of up to 900 m, R. Menzel et al. (2005) reported the complete flight paths of displaced bees. A sequence of behavioral routines become apparent: (i) initial straight flights in which they fly the course that they were on when captured (foraging bees) or that they learned during dance communication (recruited bees); (ii) slow search flights with frequent changes of direction in which they attempt to " get their bearings"; and (iii) straight and rapid flights directed either to the hive or first to the feeding station and then to the hive. Two essential criteria of a map-like spatial memory are met by these results: bees can set course at any arbitrary location in their familiar area, and they can choose between at least two goals. This finding suggests a rich, map-like organization of spatial memory in navigating honey bees.

Recent experimental data on pigeon homing along highways in Europe (Lipp et al., 2004) demonstrated that in pigeons raised in an area characterized by navigationally relevant highway systems, mapping includes different stages. During early and middle stages of the flight, following large and distinct roads is likely to reflect stabilization of a compass course rather than the presence of a mental roadmap. A cognitive (roadmap) component manifested by repeated crossing of preferred topographical points, including highway exists, is more likely when pigeons approach the loft area.

Indeed, there is much of work to be done about map-like behaviour in different taxa. Theories and models of cognitive maps still cannot fairly explain the observed behavioural flexibility, for example, planning new trajectories, selecting a novel route to a goal and making a detour around an obstacle. It seems that at this stage adequate experimental method rather than additional models would be of great help. In the next issue we will consider several sets of examples linked to experimental methods which could be perspective for investigation of map-like behaviour.