Psy 416: Reasoning and Problem Solving

This chapter presents a series of studies on different types of problems where the primary focus is on how long different processes take. Implicit in the studies is the belief that the mind is basically a information processing (computational) device. Information processing is a set of sequential or simultaneous processes, each of which take time to complete. If we can figure out how much time different tasks take and the consequences of limiting available time, perhaps we can get some insights into the algorithms used by the problem solver to solve the problems. In addition creative experimentation gives (tentative) answers to how some of these algorithms may be implemented.

The two primary methods of measuring how long processes take to complete are Reaction Time experiments (RT) and Tachistoscopic exposure time (ET) experiments. By measuring how long it takes to do different tasks (RT), or by measuring how good performance is for different tasks after a limited amount of time to work on it (ET), a researcher can generate theories about what underlying processes are involved. A large amount of cognitive research uses variants of these methods.

Donders and the subtractive method

1. Simple RT                  RT=t_o
2. Go--no go RT            RT= t_o+a           a=stimulus identification time
3. choice RT                  RT= t_o+a+b       b=response selection time

1. Posner and physical, name matching and category matching: t_o + recognition time+(naming time) + (categorization time) + comparison time + (negation time)
2. Sternberg and memory sampling: t_o + time to match (each letter)  + time to check matching box + (change response time)
3. Shepard and mental rotation: number of degrees of rotation, matches or mismatches of templates. Time to form an image+time to rotate at approximately a constant rate + time to compare images + (negation time). Pictures

Pylyshyn's criticism of rotation, argument for underlying propositional format
4. Gough and sentence verification: time to evaluate truth of a sentence against a picture; active sentence vs passive sentence, true sentences vs false sentences, affirmative sentences vs negative sentences, and combinations of these variables.
5. Clark and Chase and sentence verification:  comparison times of true and false, affirmative and negative, with different relation terms. Variables are combined here also.
6. Greenspan and Segal and sentence evaluation: comparisons based on the evaluation process and what is active in memory from previous trial; variable search times, matching times, negation times, and others.
7. Lawson and sentence evaluation: using analogical representations. How does one search a complex picture or a set of sentences, for whether a new sentence is true or false in respect to that data. Designed to evaluate both form of representation and search process.
8. Tanenhaus and sentence evaluation: measuring eye movements and testing for interaction between linguistic and perceptual phenomena.
9. Olson & Filby, actives, passives and focus: Tests for how one relates a picture to a sentence being evaluated for its truth value. Pictures

The models are conceptually compatible with the general information processing models and thus with computational models. One can simulate these processes. Such modeling is one of the major pastimes of many AI researchers.  For tasks that require fairly complex processing steps, certain researchers have introduced a modification of old introspective reports. Subjects are asked to 'think aloud' while they are attempting to solve a problem. Subjects thus report their experience, what they are attending to, and what their intentions are 'on-line', that is, while the process is going on. This generates a record of the sequence of processes which can then be analyzed and converted into an effective procedure representing the heuristics used during the problem-solving activity.

 Back to Psy 416 Syllabus