Chapter 11 Discussion
2.4 Spatial knowledge
2.4.3 Survey knowledge: the cognitive map?
Yet another metaphor in the spatial cognition community which generates
controversy is the idea of the ‘cognitive map’, a phrase first used in 1948 by Edward Tolman. Tolman found that rats were able to take novel shortcuts to return to origin and proposed the term ‘cognitive map’ to reflect a form of mental cartographic knowledge that could be accessed and used to inform navigation. O'Keefe and Nadel (1978) in their seminal work proposed that the neural substrate of the cognitive map lay within the hippocampus in the brain. More recently, structural changes have been found in the hippocampi of London taxi drivers who display excellent survey level knowledge of the London area (Maguire et al., 2000). The continuing research developments in the neuropsychological literature are not matched in the cognitive science literature. Regardless of the neuroanatomy of spatial knowledge, the structure, content and even existence of the ‘cognitive map’ is debated. Mackintosh (2002) makes the distinction between knowledge that is
explicitly spatial and knowledge organised into ‘temporal sequences’. For example, a route consisting of paths interspersed with landmarks is not inherently spatial in nature, more an organised sequence of memories. For Mackintosh, the cognitive map metaphor holds only if there are explicit spatial relationships between different locations, of the sort which tend to be reported more by males than females (Bell and Saucier, 2004). The cognitive map is a representation of the external
environment in memory, closest to the survey knowledge level described in the LRS model. The levels of knowledge that are described in the LRS model are contained and structured within the cognitive map providing access to different levels of knowledge required for different tasks. Bennett(1996) reviews and critiques three original conceptions of the cognitive map found in the literature. The conception of the cognitive map by Tolman (1948) and later O'Keefe and Nadel (1978) embody the concept of distorted and incomplete information being stored in memory that enables novel shortcuts to be taken in the environment. This conception of distorted but usable information both pre-dates and supports the fluid LRS model proposed by Montello (1998). The cognitive map within this conception can support navigation under uncertainty, when certain parts of a route are missing or incomplete; other information contained within the map supports decision making.
Thinus-Blanc (1987) suggests a very different view of the cognitive map. Her conception is far more ‘map-like’ in nature; the individual has an allocentric representation of the environment in their heads from which to navigate with. Thinus-Blanc suggests that the use of this map is fundamentally at the procedural knowledge level so may not be highly accessible to reason and as such, not reported accurately. Gallistel (1993) takes this explanation one step further and suggests that a cognitive map is a representation of any space with the central nervous system of the animal and is concerned directly with the geometry of the
maps arguing that novel shortcuts could be taken while adjusting the individual’s position with respect to seen landmarks or other knowledge of the area, rather than a definite representation of the environment itself. Essentially, Bennett suggests that simpler explanations may exist for navigation not requiring a complex, allocentric representation in memory with which to work from. These simpler strategies
described in section 2.3.2 would incur less cognitive demand and therefore would be preferable to complex access and retrieval processes involving a cognitive map. Since navigation may take place with incomplete, simplified knowledge about an area, there is little reason to suspect that complex, survey like representations are formed for everyday tasks. The cognitive map may well be a case of analogical reasoning rather than having any explanatory power. Recent research indicates that again, individuals are likely to engage strategies that minimise cognitive demand. Foo et al. (2005) show that individuals tended to use route knowledge, correcting errors by using landmark knowledge rather than an allocentric cognitive map of the area.
More recently, Newman et al. (2007) provide further evidence to support this position finding that individuals tend to move down the LRS hierarchy rather up it when correcting for error, even if they have a degree of survey knowledge about an environment. In their study, individuals drove a taxi around a VE. Survey information and landmark information was disrupted and the study showed more disruption in navigation when landmarks were disrupted rather than the overall layout. Although this may represent the way the environment is learned, it does not support the case for a detailed, veridical, cognitive map.
2.4.4 Section summary
Spatial knowledge may be divided into declarative knowledge, which is accessible to reason, and procedural knowledge. Procedural knowledge is likely to play a role in automatic navigation of familiar routes and may be developed through experience of
navigation or sensory experiences while navigating. Declarative knowledge about space can be gained from direct experience or through geocentric, external
representations of space. Either source of spatial knowledge can inform navigation. The LRS model is the dominant framework for understanding how spatial knowledge is organised. The model proposes three levels of knowledge in a nested hierarchy: landmark, route and survey knowledge. Recent research indicates that landmark knowledge plays an especially important part in navigation and route knowledge may even be a connected series of landmarks. Some features in space are more likely to be encoded and used in navigation than others and these features may constitute landmarks. Examples include route intersections or structural features in buildings such as staircases. When using external representations of space,
recognition is more likely to take place if features in space that are more likely to be encoded by individuals are represented on the plan or map. The existence of survey knowledge or ‘the cognitive map’ is controversial however, it has been shown that humans are able to report map-like knowledge of space and this knowledge may well inform navigation. However, the implied level of detail from a cognitive map is not required for successful navigation. The knowledge may be significantly distorted, generalised or otherwise incomplete but still support navigation effectively.