Spatial Reasoning as Verbal Reasoning

Spatial Reasoning as Verbal Reasoning Antje Krumnack (antje.krumnack@psychol.uni-giessen.de) Experimental Psychology and Cognitive Science, Otto-Behaghel-Strasse 10F 35394 Giessen, Germany Leandra Bucher (leandra.bucher@psychol.uni-giessen.de) Experimental Psychology and Cognitive Science, Otto-Behaghel-Strasse 10F 35394 Giessen, Germany Jelica Nejasmic (jelica.nejasmic@psychol.uni-giessen.de) Experimental Psychology and Cognitive Science, Otto-Behaghel-Strasse 10F 35394 Giessen, Germany Markus Knauff (markus.knauff@psychol.uni-giessen.de) Experimental Psychology and Cognitive Science, Otto-Behaghel-Strasse 10F 35394 Giessen, Germany Abstract We introduce an approach for how spatial reasoning can be conceived as verbal reasoning. We describe a theory of how humans construct a mental representation given one- dimensional spatial relations. In this construction process ob- jects are inserted in a dynamic structure called a “queue” which provides an implicit direction. The spatial interpretation of this direction can be chosen freely. This implies that choices in the process of constructing a mental representation influence the result of deductive spatial reasoning. To derive the precise rules for the construction process we employ the assumption that humans try to minimize their cognitive effort, and a cost measure is introduced to judge the efficiency of the construc- tion process. From this we deduce how the queue should be constructed. We discuss empirical evidence for this approach as well as a computational implementation of the construction process. Keywords: Verbal Reasoning; Spatial Reasoning; Mental Models; Cost Function; Computational Framework Introduction One dimensional spatial relations like “right of”, “left of”, “in front”, “behind”, “north of” have in common that they are transitive and, thereby, allow us to create a linear order be- tween objects linked by one of these relations. Let us demon- strate this by an example. Consider the following two sen- tences, also called premises. 1. The apple is to the left of the mango. 2. The mango is to the left of the kiwi. These premises allow us to create a linear order of the ob- jects named in the premises, apple–mango–kiwi. This order enables us to draw conclusions about information not directly given in the premises: we can infer that the apple is to the left of the kiwi. The ability to infer information about relations between objects not explicitly yielded by the premises is the subject of theories about relational reasoning (cf. Johnson- Laird & Byrne,1991; chapter 5). The bases of such inferences are internal representations that reflect information conveyed verbally by the premises. There are several theories on how this is accomplished. Syntactic-based approaches (Braine & O’Brien, 1998; Rips, 1994; Hagert, 1984; Henst & Schaeken, 2005) suggest that the reasoning process is based on oper- ations similar to the syntactic rules of formal logic. A set of rules is applied to draw inferences from given premises in order to derive new information implicitly provided by the premises. Model-based approaches, such as the mental model theory (MMT) on the other hand, suggest that reasoners in- fer new information by inspecting a mental model, represent- ing the “state of affairs”, described by the premises (Johnson- Laird & Byrne, 1991). Polk and Newell (1995), however, point out that the deduc- tion process does not necessarily require deduction-specific mechanisms to operate on internal representations. Espe- cially in reasoners that are not specifically trained on de- ductive reasoning more general cognitive mechanisms might guide the reasoning process. They introduced a third ap- proach, called verbal reasoning, that assumes the cognitive processes in deductive reasoning to be based upon the same processes as language comprehension and generation. Verbal reasoning describes reasoning as transformation of verbal in- formation provided by the premises of an inference problem. Linguistic skills operate in order to encode and re-encode a reasoning problem until the conclusion becomes obvious or until the reasoner gives up. Polk and Newell (1995) hypothe- size that when task-relevant information is provided verbally, the crucial role in reasoning is played by the verbal processes of encoding and re-encoding accordingly and that inferences follow comparatively easily from the encoded information. In the following, we sketch how spatial reasoning can be conceived in Polk and Newell’s framework of verbal reason- ing, which covers reasoning about relations. In particular, we propose new theoretical assumptions for the special case of reasoning with spatial relations. The key assumption is that the process of constructing a mental representation – a mental model – from the premises influences deductive spa- tial reasoning. This implies that the process of encoding the