Tip:
Highlight text to annotate it
X
Lecture 17: Short-term memory, working memory, and attention.
In the last lecture we saw one of the ways in which perception is related to memory,
which is that, in the act of perception, the perceiver draws on information stored in memory.
Perception doesn't just extract information from the stimulus. Perception, as the active
construction of a mental representation of the environment, as a problem to be solved
by the perceiver, draws on the perceiver's permanent repository of world knowledge stored
in memory, as well as the perceiver's more momentary expectations about what he or she
is going to encounter. These expectations are also stored in memory.
But that's not the end of it. Perception doesn't just draw on memory: it also changes memory,
because perception leaves a trace of itself in memory, a mental representation of the
stimulus event, which persists after the stimulus itself has been terminated. Memory frees our
experience, thought and action from control by the immediately present physical stimulus
environment, and allows us to perceive, contemplate, and respond to events in the past as well
as events in the present. Viewed from this perspective, memory is critical for what we
ordinarily construe as intelligent behavior, and certainly it's a necessary component for
any system that is to learn from experience. We talk about memory as if it's one thing,
but a little reflection shows that there are lots of different kinds of memory. One distinction
among memories is temporal - how long the memory lasts. This distinction is based on
the intuition that remembering an unfamiliar telephone number that you've just looked up
is somehow different from remembering your own telephone number, or the telephone number
of your boyfriend or your girlfriend. One memory is permanent. The other is gone almost
instantly. This intuition is captured in a classic view
of memory known as the Multi-Store model of memory, proposed by a number of investigators
in the 1960s. The Multi-Store model of memory is based on a computer model of the mind,
and distinguishes among three different types of memory, three different storage structures
that hold information either temporarily or permanently. Information moves among these
structures by means of a set of control processes. At the earliest stage of memory processing
is a set of sensory registers, one for each sensory modality, that hold a complete, veridical
representation of sensory input. They hold all the information that's presented to them.
The sensory registers draw on information held in long-term memory to recognize patterns
of features in the manner I described earlier in the lectures on perception.
In the model, some subset of the information held in the sensory registers is then transferred
to a second store known as short-term memory, by virtue of a control process known as attention.
By paying attention to some things as opposed to others those things we pay attention to
get transferred into short-term memory. Items in short-term memory can be maintained there
indefinitely by means of another control process known as rehearsal. Again, think of the telephone
number example. You look up a telephone number, and then you repeat it to yourself over and
over again until you get a chance to write it down or dial it. That's what rehearsal
is like. You stop rehearsing the number, and pretty
soon it disappears from memory. But by means of yet another control process, known as encoding,
information can be transferred from the relatively temporary short-term memory, to the permanent
repository of stored knowledge known as long-term memory. And finally, by means of yet another
control process, known as retrieval, information can be transferred from long-term memory to
short-term memory. If I ask you what your telephone number is you pull that information
out of long-term memory and hold it in short-term memory while you answer my question.
The multi-store model of memory became very popular in psychology -- so popular, in fact,
that it was known as the modal model of memory, that is, the model of memory that was embraced
by the largest number of investigators. They all had the same idea, that memory was a collection
of storage structures linked by a set of control processes, though the particular names that
they gave to these structures differed from one model to another. Sometimes the sensory
registers were called sensory memories or sensory stores. Short-term memory was sometimes
called primary memory; it's now often called working memory. Long-term memory was sometimes
called secondary memory. And the models differed among themselves in terms of various details.
But the general idea that there was a set of storage structures linked by a set of control
processes is what the multi-store model of memory is all about.
Let's look for a moment at the properties of each of these storage structures. As indicated
before, the sensory registers contain a complete, veridical -- that is, accurate -- representation
of sensory input. They're of unlimited capacity so they're able to hold all the information
that's presented to them at one or another of the sensory surfaces. According to the
model, there's one sensory register for each sensory modality. But two of the sensory registers
have been studied most intensively and they've been given special names: the icon, the sensory
register for vision, and the echo, the sensory register for audition. According to most versions
of the multi-store model, the sensory registers store information in pre-categorical form
-- that is, the input is not yet processed for meaning. These sensory registers have
unlimited capacity, but information is held in the sensory registers for only a brief
period of time, perhaps less than a second. Unless some information is transferred from
the sensory registers to short-term memory, it's gone. Information is lost from the sensory
registers either through a decay process or by displacement by newly arriving information.
It can't be maintained in the sensory registers by any kind of cognitive activity.
We know about the properties of the sensory registers from experiments employing a paradigm
initially devised by George Sperling, in which subjects were briefly presented -- that is,
for just a few hundred milliseconds, less than a second, with a three by four visual
array of letters. After the array disappeared, it was followed by a retention interval of
up to one second -- just a second -- after which the subjects were asked to report the
contents of the array. There were two conditions under which they were asked to make these
reports. In the whole-report condition, the subjects were asked to report the entire array,
all 12 letters. In the partial-report condition, the subjects were cued by a tone of low, medium,
or high pitch to report the contents of only the bottom, middle, or top rows.
Here are the results of the experiment. In the whole-report condition, the subjects were
able to report relatively few of the items, only about four or five on average. But in
the partial-report condition, the subjects performed considerately better. If the tone
was presented before the array was shown, or precisely when the array was shown, the
subjects were able to report virtually its entire contents, nine or 10 of the letters.
The implication is that the entire array was actually represented in iconic memory, because
the subjects could report accurately the contents of any randomly selected row; so the contents
of all the rows were available to them. However, even in the partial report condition, memory
dropped off rapidly over the retention interval. With retention intervals of just one second,
the performance equaled that of the subjects in the whole report condition. In the whole
report condition, apparently, subjects couldn't extract all the information in the array before
it decayed. So this simple experiment tells us that all the information in the visual
array is available in the sensory register, so long as you can get it out. You've got to
get it out very fast, because it decays very rapidly.
Something like the sensory registers is almost logically necessary for the sensory-perceptual
system to make contact with memory. But some investigators have wondered just how useful
it is. After all, in the real world stimuli usually remain present for longer than just
a few hundred milliseconds; and this ecological fact, this fact about the environment, may
obviate any need for the sensory registers. As the psychologist Ralph Haber once put it
in a famous paper, iconic memory may only be useful for reading at night in a lightning
storm. Still, the sensory registers are memory-storage
structures -- the point where sensory information first makes contact with the cognitive system.
And they do give an organism the opportunity to react to very brief events. It's been proposed
that the icon was probably especially useful in the evolutionary scheme of things, because
it enables organisms to catch the movements of predators and prey that are very rapid.
And the echo seems even more useful because many sounds are normally brief -- in particular,
the sounds that make up the phonemes of spoken language.
Short-term memory is probably of clearer relevance to everyday life, and not just for remembering
telephone numbers either. Information is transferred from the sensory registers to short-term memory
after it's received some degree of processing; after it's been subject to feature detection,
pattern recognition and directed attention. In some versions of the multi-store model,
primary memory or short-term memory is thought of as storing an acoustic representation of
the stimulus -- that is, what the name of the stimulus is, or a description of what
the object or event sounds like. In any event, short-term memory has a limited capacity. It can contain
only a small number of items at one particular time -- a number that's been estimated in
a famous paper at about seven plus or minus two items. Think about telephone numbers:
there's the three-digit exchange and there's the four digit number. In fact, it was psychological
experiments at Bell Laboratories that determined just what the limits of a usable telephone
number might be. Information in short-term memory can be retained
there indefinitely by means of rehearsal -- repeating it over and over, either out loud, or to yourself.
With enough rehearsal -- especially, as we'll see later, enough rehearsal of a particular
kind -- information can be transferred from the temporary store of short-term memory to
the permanent store of long-term memory. If it's not rehearsed, the information in short-term
memory either decays over time, or more likely is displaced by new incoming information.
Again, the capacity of short-term memory is limited to seven plus or minus two items.
So if you're going to put anything new in, something's got to come out.
In order to demonstrate the limited capacity of short-term memory let's do an experiment
known as the digit span test. I'm going to read to you a list of digits, and then after
I stop I want you to just write them down. Here's the first list, 5, 9, 0. Write it down.
Here's the next list. 4, 8, 6, 1. Write it down.
Here's the next list. 7, 3, 0, 9, 4. Write it down.
Here's the next list. 2, 4, 9, 6, 5, 8. Write it down.
Next list, 1, 4, 6, 8, 2, 4, 5. Write it down. Next list. 3, 9, 2, 1, 5, 7, 6, 0. Write it
down. Next list. 6, 2, 5, 7, 3, 9, 1, 8, 4. Write
it down. One more list. 0, 6, 3, 8, 9, 4, 1, 7, 2,
5. Write it down. Okay, now, turn to the next slide and check
your lists against the lists I just read. Here are the lists. You probably got the first
three or four lists with no trouble at all. Then you started having a little bit of trouble
remembering all the items, when we got to lists of five or six items in length. Seven
items, even harder. Eight items: probably too hard. The capacity of short-term memory
is seven plus or minus two items. It doesn't have to be digits. You can also
do this with letters. Here's an alphabetical digit span test. I'm going to read you a list
of letters, and I want you to write them down after I stop.
Here's the first list. Y, S, P, B, C, U, J, B, L, D, S, L, B, G,
K, A, I, C, I, B, F. Write them down. Okay, here's another one:
F, B, I, C, I, A, K, G, B, L, S, D, L, B, J, U, C, B, P, S, Y.
Write them down. Now, turn to the next slide and check your list.
Okay, here they are. For the first list you probably got the first 5, 6, or 7 items, but
then it was all over. But in the second list, which had the same number of letters, you
might even have gotten all of them. That's because in the second list, as opposed to
the first list, you could break the letters up into meaningful chunks.
FBI; CIA; KGB, the old Soviet secret police; LSD, LBJ, the initials of an American president;
UCB, for UC Berkeley; PSY for psychology. Chunking items together in this way is a means
of increasing the effective capacity of short-term memory. So the capacity of short-term memory
isn't seven plus or minus two items, it's seven plus or minus two chunks, where the
chunks can actually be pretty large. If you can chunk the items together, grouping them
together into meaningful units, all you have to do is remember the chunks. So, to return
to our example of the telephone numbers, these days, telephone numbers are pretty long. There's
a traditional seven-digit telephone number, consisting of a three-digit exchange, and
a four-digit number. But then there's a three-digit area code, and maybe a two- or three-digit
country code. Put those all together and you've got a string of digits that vastly exceeds
the capacity of short-term memory. But by virtue of chunking you can remember quite
a bit. So, for example, all telephone numbers at the University of California, Berkeley
begin with either 642 or 643. So you don't have to know those three numbers. All you
have to do is remember whether it's a two or a three. Berkeley is in area code 510.
Most of the east part of San Francisco Bay is in 510. San Francisco is 415; New York,
212; Washington, 202; and so on. The country code for the United States is 1; for England,
44; for Italy 39; and so on. You don't have to remember all these numbers; all you have
to do is remember the chunks. And that brings us to long-term memory, the
permanent repository of stored knowledge in the mind. In some sense, long-term memory
is a passive store of knowledge. Whereas we're immediately aware of the contents of short-term
memory, we're not immediately aware of all the contents of long-term memory. We have
to retrieve that information and bring it in to short-term memory. The capacity of long-term
memory is essentially unlimited. So, whereas the capacity of short-term memory is roughly
seven plus or minus two items, or chunks, there seems to be no limit to the amount of
stuff that you can get into long-term memory. And long-term memory is also apparently permanent.
Information might be lost from the sensory registers, or from short-term memory, through
decay or displacement. But there's essentially no forgetting from long-term memory. Now obviously
we do forget things from long-term memory. But, as we'll see later, there are reasons
for thinking that those items are not permanently lost. Forgetting from the sensory registers,
or from short-term memory, is permanent; but forgetting from long-term memory appears to
be a temporary thing. Support for a distinction between short-term
and long-term memory is provided by a phenomenon known as the serial-position effect. Consider
a form of memory experiment known as single-trial free recall. The experimenter presents a list
of items for a single study trial, and then the subject simply must recall the items that
were presented to him or her. If we plot the probability of recalling each item against
its position in the study list, we typically observe a bowed curve. Items in the early
or late portions of the list are more likely to be recalled than those in the middle.
These are known as the primacy and recency effects in memory. In the primacy effect,
memory is better for items that occurred early in the list, compared to items that were in
the middle. In the recency effect, memory is better for the last items in the list compared
to the middle. The primacy effect appears to reflect retrieval from long-term memory.
The recency effect appears to reflect retrieval from short-term memory. But how do we know
this? How do we know that primacy reflects retrieval from long-term memory, and recency
reflects retrieval from short-term memory? It turns out that the primacy and recency
effects are affected by different sorts of variables.
For example, slowing down the rate of presentation, increasing the interval between adjacent items,
and thus increasing the amount of rehearsal each item can receive, increases the primacy
effect, but has no effect on recency. The idea is that by giving the item more opportunity
for rehearsal, we increase the likelihood that it will be transferred to long-term memory.
Similarly, increasing the retention interval, the period of time after the list has been
presented, but before the subject has been asked to recall the items, affects recency
but not primacy. Even with a relatively short list, increasing the retention interval to
as little as 30 seconds virtually abolishes the recency effect. Of course, the trick in
this experiment is that the retention interval is filled by a distracting task to prevent
the subject from overtly rehearsing the items. Still, slowing the rate of presentation increases
primacy, but has no effect on recency, and increasing the retention interval has a big
effect on recency, but no effect on primacy. This tells us that the primacy and recency
effects are due to two quite different kinds of memory.
Other evidence supporting a distinction between short-term and long-term memory comes from
neuropsychological studies from patients like H.M. who had amnesia. Testing revealed that
patient H.M. had a normal digit span, seven plus or minus two items just like everybody
else. Apparently, his short-term memory was unimpaired. But no matter how slowly you presented
a list of words for free recall, he just couldn't remember any of it, after even just a short
period of distraction. He seemed to have no capacity for long-term memory.
Interestingly, there is another patient, known as patient K.F., who has damage in quite a
different area of the brain: the left occipito-parietal area, who has exactly the opposite pattern
of memory impairment. Patient K.F. has severely impaired digit span, apparently no short-term
memory, but normal free recall of lists even as long as 10 items -- apparently normal long-term
memory. So here we have two patients: Patient H.M., who has normal short-term memory but
impaired long-term memory, and patient K.F., who has impaired short-term memory, but normal
long-term memory. This pattern certainly suggests that short-term
and long-term memory are structurally distinct, but there's a problem. Remember that the multi-store
model of memory says that short-term memory is the pathway to long-term memory. It's by
virtue of rehearsal that items move from the short-term store to the long-term store. But
if KF doesn't have any short-term memory, how does KF get long-term memory to begin
with? Findings like this led some investigators
to abandon an important aspect of the multi-store model of memory, which held that short-term
memory was a pathway to long-term memory -- that it was somehow necessary for long-term memory
to occur. We no longer think that that's the case. That's not to say that we don't have
something like short-term memory. We obviously do. But recent theorists have begun talking
about a working memory instead of a short-term store. Working memory is so-named because
its function goes way beyond the function attributed to short-term memory, which is
simply to keep an item in an active state. Working memory does keep items in an active
state, but it keeps them in an active state while work is being performed on them, while
the information in short-term memory is being used in the service of some task. So working
memory is not simply a route to long-term memory. Rather, it's like a cognitive workspace,
where information is actively processed. Here's one view of what working memory might
look like, and as you can see it consists of a number of different elements. There is,
for example, a structure known as a phonological loop for auditory rehearsal, like repeating
a telephone number over and over again. And there is a visuo-spatial sketchpad that recycles
visual images of a stimulus in the same way. And there's a central executive that controls
conscious information processing, that is actually operating on the information that's
in working memory. And then there's a buffer memory that connects the phonological loop
and the visuo-spatial sketchpad to long-term storage.
You could think of working memory as just another name for short-term memory. But there's
a big difference, which is that much more goes on in working memory than the simple
rehearsal that was the primary function of short-term memory. If you continue in your
studies of psychology, you're going to see a lot more references to working memory than
you do to short-term memory these days. The distinction between short-term memory
and long-term memory has a long history in psychology. It goes back at least to the time
of William James in the late 19th century, who distinguished between primary memory and
secondary memory in terms of attention. Primary memory, which is what James meant by short-term
memory, or working memory, is the memory we have of an object while we are still paying
attention to our image of it - our mental representation of it; while we're rehearsing
or thinking about it, or doing something else to it, in short-term or working memory. Secondary
memory, or what we would now call long-term memory, is the memory we have of an object
once we've stopped paying attention to it. Let's go back to the telephone number example.
Someone gives you her telephone number, and you rehearse it to yourself, or you think
about it, or you divide it up into chunks so you can remember it better later. You perform
some work on it. You're paying attention to that number all the time. Then you get interrupted,
you start talking to somebody else, and you're no longer paying attention to that telephone
number. You're paying attention to the new thing. So while you're paying attention to
something it's in short-term, or working, memory -- what James called primary memory.
After you've turned your attention to something else, then the representation of that telephone
number resides in long-term memory, or what James called secondary memory, or memory proper.
Which raises the question: if attention is what links perception and memory, then what's
attention? Well, James had an answer. He said everyone knows what attention is: it's the
taking possession by the mind in clear and vivid form of one out of what seems several
simultaneously possible objects or trains of thought. Vocalization, concentration of
consciousness, are of its essence. It implies withdrawal from some things in order to deal
effectively with others. James' verbal description of attention has never been bettered, but
cognitive psychologists have now achieved a much more detailed understanding of the
mechanics of attention. Some of the earliest studies of attention
made use of a paradigm known as dichotic listening, based on what has been called the cocktail
party phenomenon. When you're at a cocktail party, there are lots of conversations going
on, but you can only pay attention to one of these at a time. And attentional selection
is accomplished in some respects by virtue of spatial and visual processing. You look
at the person you're talking to, and if you should happen to look away for a moment -- for
example, to ask somebody to refresh your drink -- you maintain the conversation by staying
focused on the sound of the other person's voice. Of course, the moment you look away,
your attention is distracted from the conversation, and you're likely to miss something that's
been said. Attention seems to be drawn and focused based on physical grounds.
Colin Cherry, a British psychologist, simulated the cocktail party phenomenon in an experimental
paradigm known as dichotic listening, or shadowing. In a dichotic listening experiment, the subject
is asked to repeat, or shadow, a message presented over one of a pair of earphones, or speakers,
while ignoring a competing message presented over the other device. Normal subjects can
do this successfully, but their ability to repeat the target passage comes at some expense.
While they're able to remember pretty much of what was presented over the attended channel,
they pretty much forget whatever was presented over the unattended channel. Moreover, while
they generally noticed when the voice on the unattended channel switched from say, male
to female, they failed to do so when the voice switched from one language to another or from
forward to backward speech. The dichotic listening experiment simultaneously
reveals the limited capacity of attention and the basis on which attentional selection
occurs. You can pay attention to only one conversation at a time; we process information
from only one channel at a time. And just as we pay attention to our companions by looking
at them, so we discriminate between the attended and unattended channels on the basis of their
spatial location or other physical features. Physical analysis comes first, analysis of
meaning comes later -- or, at least, that's the way it seemed from these early experiments.
Based on experiments of this sort, Donald Broadbent, another British psychologist, proposed
a filter model of attention which looks a little bit like the multi-store model of memory
that we examined earlier. In Broadbent's model, information arriving at the sensory receptors
is first held in the sensory registers, or something very much like them, from which
it passes through a selective filter into a limited capacity processor, like short-term
memory, that compares sensory information with information already present in long-term
memory. Depending on the results of this comparison, the newly arrived sensory information may
itself be deposited in long-term memory, and may be used to generate some response executed
through bodily systems like the muscles and the glands.
In Broadbent's system, the limited-capacity processor, like short-term memory, is tantamount
to consciousness. Thus, attention is the pathway to awareness. Pre-attentive processing is
unconscious processing. Post-attentive processing or attentive processing is conscious processing.
Which raises the further question of how much pre-attentive processing there is, and according
to Broadbent's theory, there's not very much. According to Broadbent, pre-attentive processing
is limited to a very narrow range of physical properties such as the spatial location and
the physical features of the stimulus. Anything else, like the analysis of the stimulus for
meaning, has to be done attentively, in the limited-capacity processor, like short-term
or working memory. That's why, in the dichotic listening experiment, subjects were able to
focus on one ear, and ignore the information in the other ear. Attention is selecting one
channel, as opposed to another, according to the physical features of that channel.
The filter theory was a good start, but it turns out to have some problems. For example,
Neville Moray, yet another English psychologist, found that in the dichotic listening experiment
subjects were distracted from the attended channel toward the unattended channel when
their own name was presented over the unattended channel. And Anne Treisman, still another English
psychologist, found that subjects would follow the shadowed message if the presentation was
shifted from one ear to the other. The fact that subjects can pick up on their own names
presented over the unattended channel, or that they follow a message when it's shifted
from one ear to the other, suggests that some attentional selection goes beyond the
physical features of the stimulus. Subjects can pay attention to things based on their
meaning. But if subjects can pay attention to something
based on its meaning, to shift attention from one channel to another based on meaning, then
there has to be some pre-attentive processing of the meaning of an event, too, not just
its spatial and physical features. Findings such as these led some theorists to propose
late-selection theories of attention, as opposed to theories like Broadbent's, which are early-selection
theories. According to early selection theories, like
Broadbent's original proposal, attentional selection occurs relatively early in the sequence
of information processing before meaning analysis can occur. In early-selection theories, attentional
selection is based on an analysis of the physical and spatial properties of the stimulus. After
attention has selected some objects based on their physical properties, only then are
those attended objects given any semantic analysis, or analysis of meaning. Physical
analysis occurs pre-attentively, pre-consciously. Semantic analysis occurs post-attentively,
or consciously. In late-selection theories, both physical
and semantic analyses occur early in the information processing sequence, before attention is directed
to them, so stimuli are analyzed for both their physical, and their semantic features.
And then on the basis of this analysis, some are attended to, and play a role in ongoing
thought and behavior. Again, according to early selection theories, pre-attentive processing
is limited to developing a physical description of the stimulus. All available stimuli are
processed at this stage, but things like identification, semantic description, categorization, and
response are limited to the single stimulus selected on the basis of the physical descriptions
composed at the early stages. But according to late selection theories, all available
stimuli are also identified and processed for meaning pre-attentively, outside of consciousness.
Attention and consciousness are therefore required only for response. In late selection
theories, you direct attention to some stimulus, not just based on its physical features, but
also based on its meaning, on its relevance, or its pertinence for the task at hand.
The controversy between early selection and late selection theories of attention continues
right down to this day, and can sometimes get very, very vigorous. It boils down to
a question of the extent of pre-attentive processing. How much can you analyze a stimulus
without paying conscious attention to it? Is preattentive processing limited to analyses
of perceptual structure, as implied by the early selection theories? Or can it extend
to semantic meaning as well, as implied by the late selection theories of attention?
This issue bears on the problem of unconscious perception, raised in an earlier lecture. It seems obvious
that you can't pay attention to an object that you can't detect, which is what a subliminal
stimulus is, just by definition. So the question of pre-attentive processing boils down to,
how much can you process a subliminal stimulus? Can a subliminal stimulus be analyzed at all
perceptually? Or if it can, is the perceptual analysis limited to analyzing the physical
features of the stimulus? Or can it extend to semantic features, like the meaning of
the object? The seemingly endless debate between early-
and late-selection theories of attention gave rise to a complete reformulation of the idea
of attention in what are now known as capacity theories of attention. In this view, attention
is not identified with any kind of filter but rather with mental capacity, with a person's
ability to deploy his or her cognitive resources in various directions. So we can equate attention
not with some kind of filter, but rather with the amount of cognitive effort that the person
is devoting to some particular task. These capacity theories assume that an individual's
cognitive resources are limited. There's only so much cognitive capacity to be devoted to
any particular task. But that we don't devote the same amount of cognitive capacity to every
task that comes our way. Rather, the amount of attention required depends on the task
to be performed, the demands of that task. Some tasks are very demanding: they require
the allocation of considerable attentional resources in what is known as controlled processing.
But other tasks are undemanding, they require little or no attentive effort, and they can
be performed automatically, in what is known as automatic processing.
The interesting feature of this capacity theory of attention is that we can reduce the amount
of attention that a task requires by becoming very good at it. As a result of extensive
practice, processes that were once performed in a controlled manner, drawing on attentional
resources, can now be performed in an automatic manner, drawing on little or no resources.
The distinction between controlled and automatic processes, and the idea that controlled processes
can be automatized through extensive practice, sheds new light on the question of the extent
of pre-attentive processing. In the traditional view, elementary analyses, say of the spatial
or physical properties of a stimulus, are performed preattentively, before attention
has been devoted to the stimulus. But complex processes, such as those involved in identifying
an object, analyzing its meaning, categorizing the object -- they can only be performed after
attention has been directed to them. Meaning analysis can't be performed preattentively.
The revisionist capacity view agrees that elementary processes are typically performed
before attention is directed to the object. It asserts that these elementary processes
are just performed automatically, but it insists that even complex processes can be performed
pre-attentively as well: identification, meaning analysis, categorization -- all that semantic
stuff can be performed pre-attentively, provided that the process has been automatized through
practice. The fact that a task that was once very difficult
can now be performed automatically is vividly illustrated by the Stroop interference experiment.
In this experiment, subjects are asked to name the color of the ink in which words are
printed ignoring the words themselves. And subjects find this very difficult to do. They're
trying simply to name the color, which ought to be a very simple task, based on just the
physical analysis of the stimulus. But they can't help reading the word, and this gets
in the way of the color-naming task. That's known as Stroop interference, and it occurs
because for skilled readers of English, which we all are -- those of us who are taking this
course, or teaching it -- reading just occurs automatically, whether we intend to or not,
whether we're paying attention to the words or not. It just happens. But it didn't happen
when you were five or six years old and just learning to read, that was hard work. But
now that you're a skilled reader, reading doesn't take that kind of effort anymore.
It's as easy as walking. Before you turn to the next slide, just demonstrate
the Stroop interference effect for yourself. Beginning with the leftmost column, simply
name the color of the ink in which each word is printed. Ignore the words themselves, just
name the color of the ink, and see how it goes.
Whether a process is innately automatic, like a reflex, or whether that process has been
automatized through extensive practice, automatic processes seem to have a set of features that
they share in common. First is inevitable evocation. Automatic processes are inevitably
evoked by the appearance of specific environmental stimuli, regardless of the person's conscious
intentions. In the Stroop interference experiment, you can't help but read the word, even though
you are not intending to do so. Second, incorrigible completion: Once engaged, automatic processes
proceed inevitably to their conclusion. It's a little bit like a ballistic missile: once
it's launched, there is nothing you can do about it, it's going to fly until it hits
the ground. By efficient execution, we mean the execution of an automatic process concerns
no attentional resources -- or very, very little by way of attentional resources. This
permits parallel processing: because they consume no attentional resources, automatic
processes do not interfere with other ongoing cognitive processes. You can multitask, you
can do some things in parallel, so long as they're all automatized.
Automatic processes are unconscious processes in the strict sense of the term, because they
operate outside of our phenomenal awareness. We have no idea that we're doing what we're
doing automatically. And they operate outside voluntary control. We don't control their
initiation and we don't control their completion. Controlled processing, by contrast, is conscious
processing. We initiate and terminate it at will. Conscious processing, controlled processing,
consumes cognitive resources. And when it comes to controlled processing, we're pretty
much confined to doing one thing at a time. As we'll see in the next lecture, the amount
of attention devoted to an event and the kind of attention devoted to an event is critical
for its fate in long-term memory.