Advancing the Understanding of Spatial Cognition by Considering Control

The ability to process spatial information is crucial for various tasks as diverse as, navigation, planning layouts, and managing abstract concepts. Accordingly, considerable effort has been spent on understanding how spatial cognition is realized in the human mind. These endeavors have, however, so far virtually neglected one important aspect of spatial cognition, namely control. In this contribution we show that neglecting control constitutes a serious lack in understanding spatial cognition. Moreover, we propose conceptions for computational cognitive models including control for two particular spatial cognition tasks. Besides constituting first approaches to integrating control and spatial cognition these models potentially allow giving a more detailed account of the respective spatial cognition phenomena than currently available theories. Control in Spatial Cognition The ability to process spatial information, reason about space, and communicate about it is crucial in various domains of human endeavor. Without this ability people would not be able to, for instance, navigate their environment, exchange knowledge about dangerous or attractive places, plan vacations, execute directed movements, etc. Apart from these domains spatial cognition has furthermore been shown to be essential for managing abstract concepts (e.g., Torralbo, Santiago, & Lupianez, 2006) and gaining scientific insight (Machamer, Darden, & Carver, 2000). Thus, spatial cognition plays an important role in virtually all human activity. According to this importance there has been considerable effort to unravel how spatial information is processed in the human mind. For example, abundant experimental data exists on human performance in tasks like perspective taking (e.g., Avraamides & Kelly, 2005), navigation (e.g., Golledge, 1995), and architectural design (e.g., Verstijnen, Leeuwen, Goldschmidt, Hamel, & Hennessey, 1998). Such data has usually been analyzed and interpreted with respect to the types of representations and / or operations defined on these representations used during spatial cognition. By taking such a focus on representations and (corresponding) operations, researchers, so far, mostly (see Allen, 1999, for an exception) seem to have neglected a third aspect essential for understanding spatial cognition, namely control. The necessity for taking control into account is nicely exhibited by work done in the philosophy of science. There it is a general observation that proper explanations of scientific phenomena need to include—though sometimes named differently—control aspects. For instance, in terms of the mechanisms approach to scientific explanation proposed by Machamer et al. (2000), to be able to understand a phenomenon it is not enough to consider only the entities and activities involved in the scientific phenomenon. As Bechtel (2005) remarks, “The secret (. . . ) is to organize components appropriately so that their operations are orchestrated to produce something beyond what the components can do.” In addition to illustrating that control should be part of a proper explanatory description, Bechtel’s remark highlights the fact that control (which he terms “organize”, and “orchestrated”) might in some cases even be an indispensable aspect for arriving at a satisfactory explanation of some phenomenon. Consequently, neglecting control in trying to understand spatial cognition and, thus, an essential part of human behavior, seems to be insufficient. It is not alone the (type of) representations and their associated processes which constitute the basis for human spatial cognition, but also how these representations and operations are organized. To the best of our knowledge, however, detailed accounts of control in spatial cognition are missing so far. This is not to say that existing models of human spatial knowledge processing (e.g., Gunzelmann, Anderson, & Douglas, 2004; Barkowsky, 2001) are realized without involving control mechanisms. Any computational model necessarily has some form of control mechanisms implemented, because otherwise it would not lead to reasonable results. However, these control mechanisms have been realized more as the result of this necessity than as the result of a careful consideration of the specifics of human control mechanisms in the respective tasks. Put differently, available control conceptions are rather unelaborate byproducts of developing models which mainly focus on representations and operations. In this contribution we will show that explicitly considering control in spatial cognition is directly relevant to gain a satisfactory understanding of spatial cognition and, thus, the available approaches are insufficient. Furthermore, we will propose conceptions for two computational models including control for two spatial cognition tasks. Since control has been virtually neglected so far in spatial cognition research, these models are the first to integrate aspects of both cognitive control and spatial cognition. In addition, the models potentially allow giving more detailed accounts of the modeled spatial cognition phenomena than previously proposed theories for these phenomena. The remainder of this article is structured as follows: In the next section the phenomena observed in human imaginal perspective taking, existing explanatory approaches for this task, how considering control might improve the explanatory power of existing approaches, and first steps towards a computational cognitive model for imaginal perspective taking will be detailed. The subsequent section will comprise the same aspects, but for a different spatial cognition task, namely the apprehension of spatial terms. Finally, issues for future work will be highlighted in the conclusions. Control in Imaginal Perspective Taking As Rieser (1989) remarks, planning and executing actions when moving through an environment requires judging spatial relations in this environment from certain locations and / or orientations before the moving body actually is at the corresponding locations / orientations. Likewise, for example tele-operating a vehicle, giving and understanding route directions, or considering whether changing one’s location will improve one’s view on some audio-visual display may call for judging spatial relations from certain perspectives without moving the body into this perspective. Common to all these situations is that (a) one is inside the environment for which the spatial relations have to be identified and (b) the sensory information about the environment available is just the egocentric visual and auditory impression (i.e., in particular, a map-like bird’s eye view is not available). The task of judging spatial relations in such situations from a different perspective than the bodily one has been termed imaginal perspective taking (IPT see May, 2004). As illustrated by the above examples, IPT is essential for everyday life. Thus, IPT mirrors the importance of spatial cognition in general for human behavior (see above) and, accordingly, in order to understand how this ability is realized in the human mind numerous experiments have been conducted. The general setup of such experiments and their main results will be described in the next section. Investigation of Imaginal Perspective Taking Growing interest in understanding IPT has resulted in an abundant number of experiments (see e.g., Avraamides & Kelly, 2005; May, 2004; Farrell & Robertson, 1998; Easton & Sholl, 1995; Presson & Montello, 1994; Rieser, 1989). Although they differ with respect to the precise factors they are investigating, the general design is usually the same in these experiments. A typical IPT experiment consists of two phases. In the first phase subjects are placed at a certain location with a certain orientation in an environment. Besides the participants there are a number of objects in the environment which surround the subject (see Figure 1; the location of the person is at the origin of the two arrows, her orientation is indicated by the solid arrow). The participants are told to memorize the locations of the objects. After the subjects have sufficiently learned the spatial arrangement the second phase begins1. In the second phase they are placed in the same environment and usually at the same location and orientation as in the first phase. This time, however, they are deprived of any visual or auditory information (e.g., by blindfolding and putting on headphones). The subjects are then asked to judge a number of orientation relations between themselves and the surrounding objects. Importantly, they often have to judge this relation Sometimes the participants are tested in environments they already know from everyday life. The first phase is obsolete then. 1