How to describe how readers read

 The goal of this research, then, is to build a plausible theory which describes how reading can be accomplished, given the constraints which have been outlined in this chapter. This theory is cognitively inspired by appealing to information about human reading ability, although the theory is not intended to be a complete model of human reading. But, there is an additional issue to consider before embarking on the question of how readers read--namely, how should I, as a researcher, choose to describe how we humans read? By that I mean, what research   methodology should I follow as I attempt to answer the question?

  There are numerous styles of theories in the field of artificial intelligence, in particular, and cognitive science, in general. The approach that I take is a functional-computational-representational one. In this theoretical framework, it is assumed that a cognitive behavior can be explained in terms of a functional theory which can, in turn, be implemented as a computational model. While the process described in the theory and instantiated in the model is important, there is an equal importance attached to the representation underlying both theory and model.

Functional theories describe a cognitive behavior in terms of the input given to it and the output produced by it. Further, the behavior is broken down to the level of tasks, where each task is responsible for a smaller ``piece'' of the overall behavior. The tasks are described functionally in terms of input and output. They may be further broken down to the level of subtasks, which are again described in a functional fashion.

The computational aspect of this approach means that the theory will eventually be implemented as a working program on some computational system. This does not mean that this approach ties the researcher to a particular model instantiation; in fact, many different models could result from the same functional theory. It also does not mean that the computational model is the ``end'' of the research; rather, the model is another tool that the researcher may use in the theory refinement cycle. It forces the researcher to precisely specify all aspects of the theory, it provides a ``subject'' which can be used in empirical evaluation, and it acts as a sufficiency argument for the theory.

Finally, representation plays a key role in theories following this methodology. The cognitive process being described is expected to make use of extensive background knowledge in order to reach the desired level of expertise. The structure, content, and organization of this knowledge is a research issue at the same level as the process side of the theory. This forces another level of precision on the researcher--the researcher makes the commitment to provide their audience with specific details of the knowledge representation and organization which is often assumed to underlie a particular theory.