The task description of reading

The high level claim contains two key ideas--one relating to reading in general and one relating to the understanding of unknown concepts. Each of these will be decomposed into smaller claims, until I arrive at a level capable of being tested. On the reading side, I propose that   THE BEHAVIOR KNOWN AS READING IS MADE UP OF A COLLECTION OF TASKS. During the course of this dissertation, I will describe the specific task set which I have found to be sufficient, as well as my methods for implementing those tasks in the ISAAC computer model. To illustrate, consider what happens as you read the first two paragraphs of the following short story, Zoo ([#!story:zoo!#]):

   

The children were always good during the month of August, especially when it began to get near the twenty-third. It was on this day that the great silver spaceship carrying Professor Hugo's Interplanetary Zoo settled down for its annual six-hour visit to the Chicago area.

Before daybreak the crowds would form, long lines of children and adults both, each one clutching his or her dollar, and waiting with wonderment to see what race of strange creatures the Professor had brought this year.

  In order to comprehend these first two paragraphs, a reader is performing a number of related cognitive tasks. First, the reader is acting to retrieve concepts based on the words in the text. This information, along with knowledge of syntactic categories and rules, allows the reader to parse the sentences. While doing this, the reader is undergoing shifts of attention and focus as the text is transformed into internal representations. These are just a few of the processes going on in the mind of the reader. To have a theory of reading, then, I need to present what the tasks are and how they relate to one another.

This presentation of the specific tasks is the aspect of this claim which should be focused on. Many researchers in the history of the study of reading have suggested that a set of tasks is necessary in order to explain the process (see Chapter 3 for some examples). At one level, my own work simply reiterates this need.   At another level, however, studies of the ISAAC model has enabled me to enumerate a set of tasks which permits creative reading to occur. Other task sets will also work; my own set is probably not the set of necessary tasks, it is a set of sufficient tasks. Moreso, it represents a set of tasks which has been shown to be sufficient for the reading of published short stories, which required creativity in order to be comprehended.

  In addition to providing a set of tasks, my work on reading supports the claim that THE REQUIRED TASKS WORK LARGELY INDEPENDENTLY BUT ARE ORGANIZED WITH A HIGHER-LEVEL STRUCTURE. Prior work on reading in the fields of reading education, artificial intelligence, and psychology has proposed that a wide-range of tasks are required, ranging from specific reading tasks to more general reasoning tasks. For example, some of these include syntactic processing, memory retrieval, time management, and character identification.   However, the task level of description does not capture the organization which the claim suggests. By themselves, the tasks contain no organization, they merely exist. A theory which appeals to the task level can describe the connections between various tasks, of course, but the claim I am making is stronger than that. My research has indicated that tasks with similar functions should be grouped together into larger processing entities--these are known as   supertasks and are part of the claim that   ORGANIZING THE TASK SET INTO SUPERTASKS ALLOWS TASKS WITH SIMILAR PURPOSES TO COORDINATE THEIR EFFORTS. Education literature suggests that these larger divisions are one correct way to view and model the overall reading process; in addition to that, there are strong functional arguments to be made for their existence. In no particular order, these supertasks are             memory, reasoning, sentence processing, scenario comprehension, story structure comprehension, and control. Briefly (see Chapter 5 for the in-depth analysis of these supertasks):

 
• The memory supertask handles the need to retrieve concepts from and store concepts in a memory system. Notice that this supertask is not dedicated to reading; in a general purpose entity, this supertask would be responsible for all memory operations. Memory is associative and built around the idea of experiential storage.

   

• reasoning is the supertask acting as the heart of generalized reasoning. This supertask is responsible for coming up with explanation, making predictions, and drawing conclusions about concepts presented to it. As with memory, it is a largely general purpose facility. Of course, it makes heavy use of the memory supertask to fulfill its own goals.

   

• sentence processing handles sentence-level phenomena. Parsing, initial concept retrieval, and disambiguation are example tasks from this supertask. While the purpose of this one is more reading-oriented, note that it is also used in auditory comprehension.

   

• scenario comprehension provides the ability to comprehend events happening around an agent. This supertask functions in day-to-day interactions with the world, as well as when the reasoner is reading a story describing some set of events.

   

• story structure comprehension allows the reader to comprehend the elements that make up a particular type of text, the story. Character identification, setting identification, and plot development are example tasks which fall here.

     

• Finally, control acts as a metareasoning component which serves to coordinate the other supertasks. While each supertask has the ability to perform some actions autonomously, the ``big'' cognitive decisions filter through the control supertask. In addition to controlling the interactions between the other supertasks, this one is also responsible for changing the depth of reading, abandoning a failing understanding attempt, and so forth.