<<Back to Cognition Class Home Page
---------------------------------------------------------------------------------------
CLASS 1 - Friday April 3rdMETHODS IN COGNITIVE PSYCHOLOGY
Metaphor lecture
What do you think? What was not clear? Did it make you think to something new?
What the lecture made me think was how metaphors may be useful and dangerous in teaching. On the one hand, metaphors and examples are essential in communicating a new concept or abstract knowledge. On the other hand, if I focus on a particular metaphor, it may bias your idea on the concept that I'm teaching toward this metaphor. Gestalt psychologists showed how sometimes we create very rigid concepts that prevent us from finding new solutions.
A possible solution is to give you different metaphors that capture different aspects of the phenomenon we are studying.
Levels of Analysis : David Marr
David Marr made a big contribution to the foundation of cognitive science when he described three different levels of analysis of cognitive operations:
1. The Computational levelWhat is the input? What is the output? What is the type of computation that is performed and what are its properties? WHAT 2. The representational or Algorithmic level What are the possible sequences of logic operations that we can use to carry out this computation? HOW 3. The Hardware Implementation Which algorithm is actually used in the device we are studying? What is the hardware structure of the device? How is the algorithm implemented?
Cognitive psychology basically focuses on the first two levels, whereas cognitive neuroscience focuses on the third.
More recently Churchland and Sejnowski (1991) criticized Marr's claim that these three levels of analysis are independent in cognitive neuroscience. They wrote:
"There are indefinitely many computation models one might dream up, and it might be that none are even close to how the brain in fact achieves solutions to difficult computational problems." (p. 16)Neurobiological constraints can be use to limit the possible algorithms actually used in the brain to carried out certain computations.
METHODS
Which are the experimental methods used in cognitive psychology? We will talk about two main methods, RT and study of error patterns.
In the next class professor Posner will talk about some of the methodologies used in cognitive neuroscience, and in particular Event related potentials, PET and fMRI.
Here, I will focus on how we can get some information regarding the mind using these techniques. We will talk about how we use these techniques to learn something about our mental processes and which assumptions we have to make in order to do so.
This is very important: any time a certain method is used, there are always some assumptions that are made, in order to extract conclusions from the data collected with that method and to make generalizations.
REACTION TIMES
How do you use reaction times to learn something about the mind?
HELMHOLTZ
The first and most intuitive way to use reaction time is to measure the speed of cognitive processes. The first to use RT was Helmholtz. He stimulated the nerve in a frog and measured how long did it take to move the leg. Using this information, an estimate of he nerve length and assuming a constant speed, he computed the speed of nerve transmission as 100 m/sec.
We very rarely use reaction times in this way now. Usually we compare different conditions and we compute DIFFERENCES among reaction times or differences between different populations (for example yound and older adults).
DONDERS
A "mental processes" translation of this methodology was deviced by Donders. He developed a method called subtractive method.
Description (picture from page 20)
Task A -> simple reaction times
Task B -> choice reaction times
Task C -> go/no go task
C - A gives the stimulus selection time
D - C gives the response selection time
Assumptions:
(1) Mental operations occur in real time
(2) The nature of the mental operations does not change when the task is made more complicated (pure insertion assumption).
Interestingly, a similar methods has been used in neuroimaging studies, for examples in PET studies. The first step is to divide the task in simpler components. In a study by Peterson et al (1988) the process of reading a word is decomposed in 4 parts.
A word:
* has visual features (lines, angles, etc)
* is composed by the letters
* respond to orthographic rules
* has meaning
In this study, each participant saw 4 different stimuli (false fonts, letter strings, pseudowords and words) and a PET scan was taken when he or she was looking at the stimuli. The assumption is that each stimulus required a particular set of mental operations. Using the subtracting method was possible to isolate certain regions of the brain that were highlighted when that operation was carried out.
These studies are really important, but sometimes they received the same kind of criticisms that Donder's method received. Can we really believe that we can isolate the location of precise mental operations in this way?
STERNBERG
One way to solve the problem created by the pure insertion assumption was proposed by Sternberg. Instead of eliminating one stage, he used an experimental design in which the same stage was repeated several times. For example, we can measure the times that a person takes to decide whether a particular letter is the same as a letter that has been previously presented and that he or she memorized. This method will allow us to know how long it takes to decide whether a letter was presented before, but also may allow us to understand something about the process. For example, we can investigate whether this process is parallel or serial
---------------------------------------------------------------------------------------