Incremental Model Construction: Eye-movements reflect mental representations and operations–even if there is nothing to look at
Marco Ragni, Thomas Fangmeier, Andreas Bittner, Lars Konieczny · Proceedings of the Annual Meeting of the Cognitive Science Society · 2009
Incremental Model Construction: Eye-movements reflect mental representations and operations – even if there is nothing to look at. Marco Ragni, Thomas Fangmeier, Andreas Bittner, Lars Konieczny ({ragni, fangmeier, bittner, lars}@cognition.uni-freiburg.de) Center for Cognitive Science Friedrichstr. 50 79098 Freiburg, Germany Abstract In the following we present a cognitive model and a visual world experiment to test a fundamental hypothesis derived from the mental model approach: the assumption that special relational reasoning relies on a mental model manipulation device responsible for model construction, inspection and variation. We will provide evidence for a direct linking hypothesis of eye-movements, demonstrating that the eye reflects the fundamental operations of model construction, even if there are no objects to look at. Keywords: Spatial reasoning; eye-movement study; preferred mental models; visual world experiment; attention focus Introduction There are a number of different cognitive theories which try to explain the human reasoning process. Proponents of theories based on formal rules (Braine & O’Brien, 1998; Rips, 1994) claim that people solve reasoning problems by applying formal rules (e.g. transitivity rules) to abstract representations of the premises. In contrast, proponents of mental models claim that the main strategy employed in reasoning is the successive construction of a so-called mental model of the state of the affairs. This model contains all the information given in the premises. New information, such as a reasoning problem’s conclusion, is generated or evaluated by inspecting possible models (Johnson-Laird & Byrne, 1991). To provide an example: (1) The plate is to the left of the knife. The fork is to the left of the knife. The glass is in front of the knife. The spoon is in front of the plate. This describes the following two possible models: fork spoon plate glass knife spoon plate fork glass knife The structure of such tasks is typically identical in everyday life and in psychological laboratories: there is always some given spatial information (the so-called premises) and the participants have to generate a (putative) conclusion based on these premises. The premise information can be discerned by small-scale spatial objects (e.g. fruit or cutlery which have to be arranged) or large- scale objects (e.g. prominent buildings such as church towers). In the domain of reasoning with spatial relations, the theory of mental models (MMT) has received substantial empirical support (Byrne & Johnson, Laird, 1989; Johnson- Laird & Byrne, 1991; Jahn, Knauff, Johnson-Laird, 2007; Ragni, 2008). MMT assumes a three stage process consisting of a comprehension, a description, and a validation phase. In the comprehension phase, reasoners construct a mental model that reflects the information from the premises. If new information is encountered during the reading of the premises it is immediately used in the construction of the model. During the description phase, this model is inspected to find new information that is not explicitly given in the premises. Finally, in the validation phase the reasoner tries to construct alternative models that refute this putative conclusion. However, some questions remain open with regards to how people deal with multi- model problems. For example, which model is constructed first, and does this model construction adhere to certain principles? Why do reasoners neglect some models? The preferred mental model theory (PMMT) claims that humans generally tend to construct a preferred mental model (PMM). The PMM is the starting point for deriving a putative conclusion. In the model variation phase the participants tend to make local and continuous transformations starting from the PMM to search counter- examples (Rauh et al., 2005). Several predictions of the PMMT about insertion principles as well as transformation strategies in spatial relational reasoning can be shown (Ragni et al., 2006). For the second premise in example 1 “the fork is to the left of the knife” there are two possible arrangements. Humans typically tend to process these premises sequentially, i.e. first, a mental model of “plate - knife” is generated and then the new information, the fork, is inserted into the existing model. The second premise allows the insertion of the fork at two different places: inbetween the plate and the knife (first fit principle), or to the left of the plate (first free fit principle). It had been shown empirically that PMMs are generally constructed by using the fff-principle (Ragni et al., 2006). One very important point, however, is: is it possible to decide the classic question about the way we reason based on evidence from eye-movements? MMT has a necessary assumption–that mental models are constructed