Langston, Cognitive
Psychology, Notes 1 -- Introduction/History
I. Goals.
A. What is cognitive psychology?
B. The information processing paradigm.
C. Classic cognitive psychology architecture.
II. What is cognitive psychology?
Text: "Cognitive Psychology refers to all processes by which
sensory input is transformed, reduced, elaborated, stored, recovered,
and used." Alternatively, "The experimental study of human information
processing
in its many manifestations."
Experimental study: Based on the experimental method, empirical,
scientific.
Human information processing: People sometimes operate as
information
processors.
Many manifestations: Information comes from the environment,
is stored briefly, some is selected for additional processing,
something
is done to it, it may result in some additional behavior.
That's the "transformed, reduced, elaborated, stored, recovered, and
used" part.
Some examples of questions of interest:
What is the capacity of short-term memory?
How is short-term memory searched?
How long can memories last?
How do people understand language?
What is attention?
Cognitive psychology covers everything that happens between the
presentation of a stimulus and making a response to it.
Top
III. The information processing paradigm. (A lot
of this was derived from Lachman, Lachman, and Butterfield, 1979.
If you're interested in Cognitive Psychology, this is the best
introduction
to the paradigm there is.)
A. From the philosophy of science we get the concept of a
paradigm.
It's the overarching world view shared by a group of researchers.
A paradigm has six major parts:
1. Intellectual antecedents: The areas of science that
have come before and form the foundation.
2. Pretheoretical ideas: Assumptions and beliefs underlying
a theory.
3. Subject matter: What we study (the choice here will
determine the nature of the science).
4. Analogies: Analogies borrowed from more well-understood
systems. These can have a large influence on the questions asked.
5. Concepts and language: How we talk about the science
(reflects pretheoretical ideas).
6. Methodology: How we do our work.
Cognitive psychology operates under the information processing
paradigm.
Let's go over it in detail. The point here is that if you don't
know
where you come from, you won't know where you're going.
B. Intellectual antecedents. Why are these important to
the paradigm? Because a lot of the new paradigm is designed as a
response to inadequacies in the old paradigm. If you don't know
where
you come from, it's hard to appreciate why you're doing what you're
doing.
A bunch to consider.
1. Experimental roots:
a. Structuralists: Wundt (the first psychological
laboratory
was established by Wundt in 1879). Experience is important, but
what
is it (what is the structure of experience, and how are the pieces
combined)?
The technique was to have a trained observer introspect about
experience
to uncover the structure. This had its problems, primarily:
1) Arbitrary: There's no systematic way to get at the
structure.
2) Introspecting changes what you're introspecting about.
b. Behaviorists (neobehaviorists): Watson, Pavlov,
Skinner.
American response to structuralism, especially introspection.
What
goes on in the head isn't observable, if we want to progress in
understanding,
we must focus only on observables (behavior). The radical version
of this doesn't sound very psychological today. For example, if I
don't get any water for three days, I'm likely to drink a lot when I
get
my chance. You might be tempted to say I was "thirsty." A
behaviorist
would not agree. For them, a mental construct, like thirst, is
unnecessary
to understand the situation. A simple model,
Longer without water -> More organism drinks
is sufficient. We don't need mediating concepts,
Longer without water -> More thirst -> More organism drinks.
This attitude developed from a belief in logical positivism.
This was a set of rules that determined which questions were capable of
being answered with empirical means (a.k.a. scientific) and which
weren't
(a.k.a. unscientific). Questions about mind, thinking,
consciousness,
etc. were unscientific, and couldn't be answered. So, these areas
were neglected. Part of the development of cognitive psychology
was
a reaction to this attitude. What makes people interesting to
study
was explicitly ruled out by logical positivism.
They had two approaches. Loosely speaking, classical conditioning
is the development of associations between novel stimuli and
biologically
relevant events. Operant conditioning is the development of
associations
between stimuli and responses. This S-R psychology was the big
player
before the cognitive revolution.
While we explicitly allow mental representation in cognitive
psychology,
some of the behaviorist program has been retained. Two examples:
1) Empiricism. We still focus on observables, most of what
we can observe is behavior. But, we use those observations to
infer
something about what's going on mentally. Most of our observation
takes place in controlled, laboratory settings. We'll see an
example
of that in a moment.
2) Nomothetic explanation. Instead of understanding
individuals,
we're interested in general rules that would apply to anyone (as
opposed
to ideographic explanation).
Some other areas of explicit disagreement:
1) Not totally committed to learning. We'll pay attention
to potentially innate contributions to behavior if there appears to be
evidence for them.
2) Totally committed to humans. There may be similarities
in the ways human and animal nervous systems work, and there may be
similarities
in learning. But, we're only interested in humans. That's
who
we study.
A related research area was verbal learning. The materials were
sort of verbal (mostly nonsense syllables), and the participants were
humans,
but verbal learning was more behaviorist than information
processing.
Their focus was on how associations were formed and what factors
influenced
the formation of associations. the same concepts (like mind) that
were taboo for neobehaviorists were also taboo for most verbal
learners.
The information processing paradigm did borrow some things from the
verbal
learners:
1) Focus on memory
2) Laboratory procedures (we'll see some of this in the
methodology
section).
2. Human engineering: This was primarily motivated by World
War II. Why do people make mistakes (like shooting friendly
forces)?
Incorporating the human element into the design of machines caused
humans
to be viewed as processors of information. This leads to the
question:
How do people process information? That takes us to the kinds of
questions cognitive psychologists ask.
3. Information processing roots: Communication theory has
to do with transmitting a signal (information) along a channel to a
receiver.
This notion inspired early cognitive psychologists, who regarded humans
as information processors. Information is transmitted along a
channel,
received, and processed. The processing results in
behavior.
In other words, the mental state dictates behavior, not the physical
event.
a. Mental chronometry: Donders, Sternberg. Mental
events take time. We can estimate how much time each activity
takes
through the use of clever tasks. Once we know this, we can see
what
happens to information inside. Donders’ techniques were
problematic.
Sternberg's updating of this technique in the 60’s was a major step for
the field.
b. Information: Shannon. Shannon et al. figured out
how to quantify information. His work inspired psychologists to
investigate
the effect of information on behavior. For example, the more
information
a stimulus conveys, the harder it is to detect the stimulus in a noisy
environment.
4. Linguistic roots: Chomsky. Chomsky pointed out
in 1959 that behaviorist accounts of language could never work.
Some
of his arguments had to do with whether or not language rules could be
learned. Chomsky's conclusions were that:
a. Some aspects of language are innate (the basic grammar).
b. Language is rule-governed.
c. Language rules are stored in the head and are used to produce
and understand.
Note how these ideas are outside the realm of behaviorist
thinking.
These points aren't universally accepted, but early cognitive
psychologists
were interested in the representation of language.
5. Synthesis: Cognitive psychology took:
a. Focus on observable behavior, but behavior can reflect mental
events.
b. The notion of information.
c. The idea of internally represented rules.
The result is an experimental science of mental life. As long
as the questions are framed correctly, and observations are properly
made,
we can understand what happens in people's minds.
C. Pretheoretical ideas.
1. Symbol manipulation: Humans encode, compare, store,
etc. What they use for this are symbols. Thinking is the
manipulation
of symbols.
2. Representation: Some aspects of the world are
represented
in the mind. This representation may be analogical (what's in the
head resembles what's being represented). The representation is
symbolic
(made of stuff that stands for other stuff). This is contrasted
with
direct perception where no intermediate representation is required
(Gibson).
3. Some things may be innate. They interact with learned
stuff to produce cognition.
4. Cognitive processing takes time, and we can measure that time
and use those measurements to understand subparts of the mind.
D. Subject matter. Overall, mental processes. Broken
down, we get attention, perception, problem solving, learning, memory,
language, knowledge representation, meaning, and comprehension.
Note
similarities between this list and the table of contents for the book.
E. Analogies.
1. People are information processing devices. You'll see
a lot of words like input and output that might also be used to
describe
machines. However, the idea that people are an information
processing
channel with limited capacity is losing force because it's not very
productive.
2. The computer metaphor. Minds work like computers.
This is more popular, but it may be losing some ground (possibly to a
brain
analogy, but that's iffy).
F. Concepts and language. It reflects what we've looked
at so far.
G. Methodology. Borrow an analogy from the biology of the
cell. We're spies looking at a factory. We see trucks with
supplies come in one end, and other trucks come and pick up the
finished
product. We have to figure out what goes on in the factory.
The only tool we have is to occasionally send in a shipment of supplies
and see what comes out. We can't go in the factory
ourselves.
The mind is like the factory for us. We want to know about the
mental
stuff, but can't go in and see it. We can see what goes in and
what
comes out, and sometimes we can put stuff in. Keep in mind that
everything
mental is going to be inferred. We say "If it took this long to
do
task X, then Y and Z (mental processes) must have intervened. A
lot
of our methodology is about making safe inferences.
1. Convergent techniques. We have some process we think
exists. We attack with several types of investigation coming from
different directions, and we see if we get similar results. A
classic
example of this is 7±2. This is the capacity of short-term
memory. It kept showing up in a variety of situations, and
eventually
George Miller realized it was a fundamental capacity. Along with
convergent techniques, there is also a lot of strong inference.
This
involves listing several possible explanations for a phenomenon, then
conducting
experiments to rule out all but one. By doing this, we are able
to
eliminate incorrect theories.
2. Computer simulation. One problem with inferring
processes
is that you can infer anything. We can't take brains apart and
put
them together again to check if we're right. Instead, we make
computer
simulations that have just the parts we claim operate, and then
manipulate
those to see if they do what people do. If the simulations work,
it's more evidence that we've identified the right processes.
Note,
we could still be wrong.
3. Reaction time. This is our favorite technique.
Many experiments are based on timing people doing various tasks and
inferring
processes based on the differences in time. The Stroop task is an
example.
Demonstration: Color Stroop task and how we can infer
mental
activities.
Here are some other examples of reaction time being used to make
inferences
about behavior.
a. A task: Inferences when reading. Do you
automatically
think of all of the inferences when you read a sentence? Try this
sentence:
"John pounded in the nail"
You might make something called an instrument inference, which is
inferring
the instrument used to do the action. In this case, that would be
a hammer. How could we tell if you made this inference?
1) Naming:
Description: I present you with a stimulus, usually a word, and
you simply name it (out loud) as fast as you can. So, if I show
you
the word "phone" you say "phone."
Example: You read "John pounded in the nail" and then name either
"hammer" (related to the sentence) or "knife" (unrelated). If
hammer
is part of your representation (you formed the inference), then you
should
name hammer faster than knife.
2) Lexical decision:
Description: I present you with a string of letters ("kpaln"
vs. "plank") and you tell me if that string is a word.
Example: You read our sentence and I show you "hammer" or
"knife".
If you formed the inference, then you should say hammer is a word
faster
than you say knife is a word. (The answer is "yes" in both cases,
knife is just less likely for pounding nails.)
3) Item recognition:
Description: You read a segment of text, and at some point
(during
or after) I present an item (a word or a phrase) and ask you if it came
from that text.
Example: You read "John pounded in the nail" and I ask if
"hammer"
was in the text. If you drew the inference, then it should be
harder
to say "no" to hammer (and it should take longer) than it would be to
say
no to "knife."
There are more, but that gets the idea across.
Top
IV. Classic cognitive psychology architecture. Now,
you know something about the background of cognitive psychology.
How do we put that stuff to work? First, we need an outline of
what
the mind might look like. Note, this is just an outline.
We'll
end up with something different.
You have three main sets of activities: Stuff comes in, something
happens to it, and stuff goes out. Here's a rendition of that.
A. Getting stuff in: The basic process is to convert from
a physical stimulus to mental energy. Based on work in
neuroscience,
most people think this part is wired into the brain (sensation).
After that, a lot of complex cognitive stuff happens
(perception).
It looks like every aspect of a visual scene is broken down into its
elements
by the time it reaches the higher cortex. So, the problem is how
it gets put back together. One of the biggest concerns here has
to
do with how much "thinking" can influence perception.
Note that there is filtering, not everything gets in. After that,
there is selection. Some stuff gets more attention than other
stuff.
B. The middle (something happens to it): For classic
cognitive
psychology, there are only two things to study: Representation
and
process. Representation is concerned with what is stored and how
it is stored. Process is concerned with what you do with that
stuff.
Our goal for this lecture is to determine what would be required for
the
human cognitive architecture. I'll maintain two lists, one for
representation
and one for process.
1. Representation:
a. The mind is a physical symbol system. Physical:
Instantiated in a physical thing (brain) as opposed to floating in the
ether. Symbol: What it stores are discrete symbols that
each
stand for (represent) some entity (usually in the world, but possibly
something
that can only exist in your mind). Each symbol is stored in a
particular
location. This buys into the computer metaphor where computers
manipulate
symbolic information that's stored at particular addresses in memory.
b. There's some medium for storage (which we won't attempt to
specify for now). Whatever this medium is, it stores both the
facts
and operations that you can perform on these facts. In other
words,
whatever representational stuff is combined to make a symbol will also
combine into a representation of a rule.
Note: a and b, and some of the others below, take place in the
box labeled long-term memory in the picture.
c. The language of thought is propositions. Your intuition
might be that you think in words, but it should be pretty easy to
demonstrate
that that isn't true. Propositions are abstract entities that
encode
single ideas. These can be arranged into networks that capture
the
relationships between ideas.
Demonstration: Tip of the tongue task.
d. Your cognitive architecture needs some way to represent mental
images. You can give yourself the phenomenological experience of
an image if you try to answer the question "How many windows are there
in the house in which you grew up?" You'll probably imagine
yourself
walking through the house. But, is this an actual image like real
vision, or is it just a byproduct of thinking in propositions?
For
now, we'll admit an image storage and viewing area to our architecture.
e. Your architecture stores processes. In particular, rules
for performing automatic processes will be stored. An automatic
process
has three basic properties:
1) It can be done in the absence of focused attention.
2) You usually can't verbalize how you're doing it.
3) You usually can't stop it once you start it, or prevent
yourself
from starting it when the situation in which you need it arises.
Processes start out as controlled processes (the opposite of
automatic),
but with repeated practice they become automated. We'll think of
them as compiled computer programs, and store them in the architecture.
f. The architecture has a working memory. It's a place
for processing information. It's in representation because it's a
representational medium. The rest of the architecture is a
potential
information storage, meaning that you could get at it if you wanted it,
but it isn't presently doing anything. Working memory is the
active
part of that. It's useful to have a limit on what you can think
about
simultaneously to prevent information overload during processing.
If you carefully considered every possibility you could think of before
acting, or brought together everything you know about a topic before
deciding
about it, you'd spend most of your life in a catatonic, immobile
state.
This working memory is the short-term memory box in the picture.
2. Processes:
a. Some way to satisfy goals. We would like a general
process
to solve any problem with which the organism is faced. This is
means-ends
analysis. Basically, you put some end (goal) into a register, and
then retrieve the means required to achieve that goal. You can
nest
means and ends as deeply as required to get to a solution. For
example,
here's something that happened recently: I want a sandwich for
lunch,
but have no mustard. I go to the store and get some. I
realize
I have nothing but water to drink. I put down the mustard and
return
to the store to buy some Pepsi. I come home and the dog has
stashed
the mustard somewhere (apparently thinking it's a toy). I search
for the mustard. I don't find it. I use mayonnaise
instead.
The main goal is to have a sandwich, but en route to achieving it, a
succession
of subgoals had to be solved. Goals can come from three sources:
1) Memory (as in you remember you were supposed to read for
class).
2) Cognition (as in mustard is required to kill the meaty taste
of luncheon meat).
3) The environment (as in the mustard has been hidden by the
dog).
Means can also come from these sources. For example, if I have
a goal to reach something high, the presence of a chair in the
immediate
environment might provide a means of achieving that goal. This
highlights
the value of working memory. With all of the stuff coming in, you
need a limit to keep you on the main path. Otherwise, all of the
possibilities would overwhelm action.
b. Some image manipulation processes. There's evidence
that you can scan and manipulate mental images. So, we'll have
some
processes for it.
c. A way to automate control processes and a way to execute
processes.
Someone has to do the actual work.
d. Some interpretation processes. All of that information
is useless without some way to make sense of it.
e. Attention. This is like the processes’ working
memory.
It's a limit on the amount of information that you can process at one
time.
Top
Cognitive Psychology Notes 1
Will Langston
Back to Langston's Cognitive Psychology
Page