Proceedings of the Annual Meeting of the Cognitive Science Society

This paper describes the use of on-line processing of anaphoric reference to explore the structure of sentence representations. Of central concern is whether or not the representation of verbs is componential The differential amounts of processing required by subjects to resolve references to objects and simple or complex actions was used to provide insights into the nature of the internal representation of sentences. Two conditions were created; one in which componential effects occured, and another in which they did not. It was concluded that processing demands dictated the nature of the representation, and that either a wholistic or compnential mode of representation could be used on-line with equal facility

The analysis of perlocutionary acts is a fundamental component for a theory of human communication which accounts for the response of interactin gsubjects. We introduce two knowledge structures, namely games and contracts, which are used by the mental processes underlying perlocutionary acts. we present our mental processes underlying perlocutionary acts. We present our model through the discussion of an example of game; in particular we focus on the problem of shifting a from the wants of the actor to the wants of the partner.

In this paper we seek to trace the way in which psychologists' concepts of the mind have evolved over the roughly ninety years since the study of empirical psychology began in America. We examined metaphors used by psychologists to describe mental phenomena, based on a corpus of mental metaphors used in the journal Psychology Review from 1894 to the present. The cheif finding was that the nature of the mental metaphors changed over time. Spatial metaphors and animate-being metaphors predominate in the early stages, declining later in favor of systems metaphors predominate in the early stages, and artificial intelligence. A secondary finding was that the numbers of mental metaphors varied: They are more prevalent in the early and late stages of the century that in the mid-stages. These patterns are interpreted in terms of the evolution of psychologists' models of the mind.

Children with a year of Logo programming were asked to think-aloud about the functino of some Logo recursive programs, and then to predict by hand-stimulation of the programs what the graphics turtule will draw when the program is executed. If discrepancies arose, children were asked to explain them. A prevalent but misguided "looping" mental model of Logo recursion persisted even in the face of contradictions between program effects and the child's predicitions.

This paper is in two parts: I, a brief description of the identity theory of the development of the object concept and a computational model that describes the three stages of that theory and, II, the running of that model on two experiments important in the analysis of developnent of the object concept. Th emodel is designed to help determine which parameters in the development process might best explain changes whithin the object comprehension skills of infants. The discussion presents a set of perceptual invariants that can be used: a), to describe changes between developmental stages and, b) offer the potential for a cost/gain metric foor that development. The model is written in PROLOG, a very gigh level computer language. Detailed descriptions of PROLOG (6)(15) and further uses of PROLOG to model provlem solving skills (7) may be found elsewhere

The process of determining the temporal structure of a narrative text is extremely complex. In this paper, we examine a small but central part of this process: the roles of aspectual class and the progressive/non-progressive distinction. A standard set of aspectual calsses is presented and the temporal effects of each of these classes is discussed. Finally we briefly discuss an implementation of a system which can read a simple narrative text and construct a model of the temporal structure of that text.

Non-montonic logics have been used recently for a variety of A.I. purposes, including belief revision and default reasoning in question-answering and exper systems. This paper argues that by their nature, such systems dicard information which has a role in human belief systems. In particular, systems which use non-monotonic reasoning lose the distinction between fully justified inferences and reasonable presumptions, in the process losing the ability to record failed expectations as such, an ability which provides a useful measure of salience for A.I. systems.

Making a perveptual interpretation can be viewed as a computational process in which a plausible combination is chosen from a mong a large set of interdependent hypotheses. In a cooperative computation the hypotheses. In a cooperative computation the hypotheses are implemented by units that interact non-linearly and in parallel via excitatory and inhibitory links (Julesz 1971; Marr and Poggio, 1976;)

Successive tests pf the same memory often appear to change it. Either by consolodation of a correct response. Or an increased probability of repeating errors. This raises the question as to whether the act of attempted recall modifies existing memory traces or creates new ones. Experimental work in this area is complicated by difficulties (a) in isolating what the interations between successive trials actually reflect, and (b) determining exactly when such interactions have taken place. A methodology is described in which the difficulties can be overcome within the domain of a specific experimental paradigmZ: the repeated cueing (RTTT) method used by Jones (1974). Analysis provides three significant findings. Firstly. attempts at recall produce an engram which is normally unavailable after a small number of cases these traces may survive for at least the duration of the experiment. Finally. analysis of subject's confidence ratings indicates that subkects are able to distinguish between recalls of stimuli and previous answers implying an addition to. rather than a corruption of exisitng memory.