Information holding mechanisms
Environmental stimuli
generally pass by too quickly for us to select and interpret them
satisfactorily. To slow things down we need some types of holding mechanisms — hypothesized
storage processes that hold information long enough so that it can be properly
processed. This chapter discusses some such holding mechanisms.
A complication in
the learning process is that not everything we wish to associate together
occurs together in time. Two stimuli to be associated might be separated by
several seconds, or the reward for a behavior might come after the behavior
has stopped. Learning such delayed associations may also involve some type
of holding mechanism.
Stimuli that affect
the sense receptors produce neural activity in various parts of the central nervous
system. It appears that only part of the information that
reaches the sensory areas is later attended to and perceived. The rest
dissipates in a short period of time, probably less than a second. The holding
mechanism that holds all the sensory input for a short time is called sensory storage (also known as short term sensory storage,
sensory memory, and pre-perceptual store).
In a classic experiment
on sensory storage, Sperling (1960) presented to subjects an array of 12 letters,
three rows of four letters each. This array was shown to the subjects for
only 50 milliseconds. Immediately after the presentation of the array, the
subjects heard one of three tones which told them which row of letters to
try to remember. Since the tone came on after the letters were presented, the subjects could not know in advance which
row they would have to remember. The subjects were good at this task, remembering
a high percentage of the letters of the signaled row. However, once remembering
one row they had great difficulty trying to remember either other row, for
the image of the 12 letters quickly dies as the subject recites the letters
of the required row. Similarly if the signal tone, instead of coming on immediately
after the array of letters, is delayed by one second, memory of the signaled
row drops to less than one half of what it is with no delay.
This suggests that
sensory storage very briefly holds more information than can be perceived
and processed. That part of the information that is quickly chosen (attended
to?) continues in perceptual processing, while the rest of the information
is lost from the system. Although not as well researched as vision, auditory
perception seems to have similar storage phenomena (Massaro, 1972). .
It is assumed that
the information that is “chosen” from sensory storage goes into another holding
mechanism, short term memory. Some of the information from short term memory
is then assumed to enter long term memory. This trichotomy of sensory storage,
short term memory, and long term memory, however, is an explanatory fiction.
One can argue that there are fewer or more separate stages. Also, it is possible
that none of the proposed stages may correspond to any real physiological
processes, although we will suggest some. Finally, the following discussion
is based on a linear model which shows information going from sensory storage
to short term memory to long term memory. However, some theorists argue for
a parallel model where information from sensory storage may go simultaneously
into short term memory and long term memory.
A person looks up a
phone number and remembers it just long enough to dial it. If he gets a busy
signal and decides to dial again, he might find he has forgotten the number
already. This appears to be an example of short term memory (STM),
or primary memory, a short duration memory-holding mechanism. STM is generally
considered a store of limited capacity in which information dissipates with
time and/or is easily displaced by newer information. It may be somewhat
contradictory to describe this holding mechanism in terms of short term memories, since memories are usually defined as being relatively permanent (see
Chapter 1). Broadbent (1963) suggests that forgetting in STM is due more to the
deterioration of information with time than to actual stimulus properties of
the information. That is, information that enters STM weakens over time, and
this effect is largely independent of any interfering effects between different
pieces of information in memory storage. Waugh and Norman (1965), on the other
hand, suggest that information loss in STM (they call it primary memory) is not
due to a dissipation with time but is due to the information being displaced by
newer information. They suggest that primary memory has a limited capacity, and
new information, if not redundant, will displace the old.
After looking up
the phone number we can make sure we remember it until dialed by running it
over and over again in our minds or by repeating it out loud. This process
of rehearsal is a common way to keep information in STM. It is as if information keeps
being taken out of STM and put back in to be sure it is never lost from STM.
Rehearsal is a key part of many models of STM. For example, in the Waugh and
Norman (1965) model rehearsal is seen as a way to keep information in primary
memory as well as a way to facilitate the entry of the information into more
permanent memory storage systems (i.e., secondary or long term memory).
Rehearsal often involves
the subject’s coding the stimulus into verbal symbols, such as words, which
can then be rehearsed. If shown a string of letters to remember for a short
time, a person probably won’t rehearse a visual image of the letters as much
as a verbal reading of the letters. However, rehearsal does not have to be verbal or even conscious.
If STM does have
limited capacity, what is its storage capacity? Miller (1956) suggested that
the answer is seven, plus or minus two, units of information. That is, STM
usually holds between five and nine units of information, depending on some
aspect of the person’s intelligence and the nature of the material. Thus a
person might be able to remember a string of nine random digits or a string
of seven random letters. Miller uses the word chunks to refer to the
units of information that can be remembered. He argues that although a person
can remember only about 7 chunks of information, people differ in how
much information they code into any one chunk. The process of chunking, then, is a coding procedure for converting environmental information into
chunks of information.
The following example
illustrates how Miller’s theory works. If we read the following series of
digits—011100101011001010101100—to a person and ask him to repeat back, in
order, as many as he can, he might remember only the first eight digits, for
each 0 or 1 is coded as a single chunk and he remembers 8 chunks. If, however,
we teach the subject a simple binary code in which two digits are chunked
together and coded as one digit (00=0, 0 1=1, 10=2, 11=3), then the first
eight digits would be coded as follows: 1302. After learning well this binary
chunking code, our subject might still remember only eight chunks, but this
would now correspond to 16 digits. Miller reports examples of several engineers
who learned to remember a string of 20 lights in terms of the specific sequence
of on and off. They did this by grouping the lights in sets of three, chunking
the on — off information for each group into a single
chunk, and then remembering about seven chunks (on—off—on might be considered
101 and coded as the number 5). Chunking of more meaningful material often
utilizes codes where considerable information can be carried by one meaningful
word.
A number of distinctions
have been made between the nature and processes of STM and those of long term
memory (LTM):
|
1.
|
STM
is generally considered to be of limited capacity whereas LTM, for all
practical purposes, is unlimited. |
|
2.
|
Information in STM dissipates with time and/or is displaced by new information.
Storage in LTM, on the other hand, is considered fairly permanent, and
forgetting is explained in terms of interference from other learned material.
(This will be covered in more detail in the next chapter.) |
|
3.
|
Less
processing of information takes place in STM than in LTM. |
|
4.
|
STM
is affected more by acoustic similarity of the material than by semantic
similarity, while the opposite is true for LTM. That is, interference
and confusion between material in STM primarily depends on how similar
the different materials sound to each other, whereas in LTM the meaning of the material
plays a greater role. |
|
5.
|
Some
theories, such as Hebb’s (discussed later), suggest that STM involves
an active, ongoing physiological process, while LTM involves some structural
change in the CNS. |
Other theorists,
such as Melton (1963), argue against the dichotomy between STM and LTM. They
suggest that the characteristics of STM storage are basically the same as those
of LTM. Current research is showing that many of the principles of forgetting
that are known to be true in LTM also apply to STM. Apparent differences
between STM and LTM, then, do not indicate different processes, but are
artifacts of the amount of time the subject has to do his tasks and the type of
task required of him. STM and LTM are just different points on the same
continuum that differ quantitatively, but not qualitatively.
For example, consider
the distinction that STM is primarily acoustic whereas LTM is primarily semantic.
Shulman (1971), who was one of Melton’s students, has argued that semantic
encoding can be demonstrated in STM tasks if the task requires it or if the
rate of incoming information is slow. However, in most STM tasks the subject
does not have time, particularly if rehearsal is taking place, to encode stimuli
semantically, and therefore the acoustic properties of the stimuli are more
important. With more time, semantic encoding is possible, and becomes dominant.
Thus the degree of semantic encoding is assumed to lie along a continuum of
time available for the task, rather than to reflect distinct processes of
STM and LTM. However, Baddeley (1972) argues that semantic coding, being more
complex and slower than acoustic coding, results in a more durable memory
trace. He suggests that if LTM is defined in terms of trace durability, then
semantic coding takes place in only LTM, and not in STM. According to Baddeley,
effects of apparent semantic coding in STM reflect semantically coded retrieval
rules from LTM being used to
interpret acoustically coded material in STM.
Even if many of the
mechanisms of STM and LTM are the same, there might still be some qualitative
differences. Although Peterson (1966) agrees that many mechanisms affect STM
and LTM similarly, he suggests that there still is a recency factor in STM,
the effectiveness of which decreases with time and which interacts with learning
mechanisms.
Hebb (1961) did an
experiment in which he disproved to his satisfaction his own assumption of
the independence of STM and LTM. He gave his subjects the task of learning
an unordered sequence of nine digits, presented at the rate of one per second.
The subjects then had to immediately repeat them back. This was done for 24
trials. However, on every third trial the same sequence of digits was presented,
while the other 16 sequences were all different. Now since the subject is
asked only to remember a sequence long enough to repeat it back, the sequence
of each trial should replace in
STM the sequence of the previous trial. So,
Hebb argued, if STM and LTM are distinct, recall of the repeated sequence
should not be better than that for any other sequence. However, the experiment
showed that recall of the repeated sequence improved over trials. Melton (1963)
repeated and extended Hebb’s experiment, and achieved similar results. Hebb
and Melton interpreted these results as suggesting a continuum of STM and
LTM. It could, however, also be argued that STM and LTM are distinct and parallel
processes, and that the information during the Hebb experiment entered both
STM and LTM simultaneously.
A general problem
in memory experiments, except perhaps with those carried out in long time
intervals, is that we often can’t be sure whether the recall came from STM
or LTM. Just because we choose a short recall interval and call our experiment
a study of STM doesn’t mean that the recall couldn’t really have come from
an LTM system which is qualitatively different than an STM system. Perhaps
many of the similarities found between STM and LTM merely show that the experimenter
was really measuring LTM in both cases. In this sense the terms “short term
memory” and “long term memory” may be misleading since they imply a time distinction
which may or may not exist. There is no reason why the processes of STM and
LTM can’t overlap in time. Thus the question of whether there are two storages
or one remains unanswered.
LONG TERM MEMORY
With long recall intervals,
theorists agree that the information comes from an LTM storage. What are the
properties of this storage? How is information stored and how is it found again
(retrieved) once stored? Underwood (1969) suggests that memories can be
conceptualized as involving a collection of attributes. The attributes of a
memory are a result of the process by which information is encoded into LTM.
They are the distinguishing characteristics of the memories by which the
encoding process separates one memory from another. Underwood gives a number of
possible attributes, including the following:
1. A spatial attribute occurs when the memory can be associated with some spatial coordinates.
In trying to remember a chemistry formula the student might remember that
it was at the bottom of one of the pages in the text, and this spatial cue
may then facilitate his remembering the formula.
2. A temporal attribute occurs when the memory can be placed in time relative to some
other event. You remember that you went to Dunham’s Flower Shop before you
went to Dunn’s Antique House.
3. A modality attribute is based on the sense mode of the incoming information. Information
might be coded according to whether it was written or verbal, or there might
even be different storage systems for visual and auditory memory.
Attributes, then,
are the labels and hooks that serve to discriminate memories and facilitate
later retrieval. In reviewing studies of animal memory, Spear (1973) argues
that “the attributes of a memory represent those events that were noticed
by the organism during learning.” Spear suggests that events that trigger
memory attributes may reactivate inactive memories and improve retrieval of
these memories.
Norman and others
(Bower, 1973; Norman, 1969, Chap. 6) have investigated the properties of LTM
in terms of systems that people use to improve their memory. These memory
devices are called mnemonics,
after Mnemosyne, the Greek goddess of memory.
Norman included in his studies the following mnemonic devices: rhymes, method
of loci, and analytic substitutions.
Rhymes. “Thirty days
hath September...” is a common rhyme that people use to help remember the
numbers of days in each month, for it is easier to remember the rhyme than to
memorize the number of days for each month. Another common rhyme is the
spelling rule “I before e except after c.”
Method of Loci. Early Greek orators who had to remember long speeches from memory used
a simple device now called the method of loci. They would memorize each part of
the speech in association with a particular part of their home, i.e., by
practicing that part while looking at or thinking about the corresponding part
of the home. Then when giving the speech their minds would systematically move
through their home in the same order, letting the images of the different loci
facilitate the memory of the associated part of the speech.
Analytic
Substitutions. The
method of analytic substitutions consists of translating the material to be
remembered into a form that is easier to remember. This translation may involve
numbers, sounds, words, and so forth. For example, to remember number values
for the transcendental number pi (the ratio of the circumference of a circle to
its diameter) many people have devised sentences where the number of letters in
successive words give the sequence of digits for pi. One such sentence devised
by Sir James Jeans goes, “How I want a drink, alcoholic of course, after the
heavy chapters involving quantum mechanics.” A common way of remembering
numbers is by coding each digit into a specific word: one—bun, two—shoe,
three—tree, four—door, five — hive, six — sticks, seven — heaven,
eight — gate, nine —line, ten — hen.
First the person learns the code well. Then when he wishes to remember a number
he forms a bizarre image based on the words corresponding to the numbers. For
example, to remember that a person was born in ‘52 you might imagine him
walking over to a bee hive and putting his shoe in it.
Analytic substitution
schemes can get quite complicated. It may take a long time to learn and to
develop any skill at using some of the schemes, but when mastered they impressively
improve memory. Without going into the systems it might be interesting to
look at a complex example from Norman (p. 122). Consider having to learn the
following sequence of digits: 001100001001100001101010001111111100000. This
sequence could first be encoded into the octal digits 1411415217740, where
each octal digit represents (chunking) three of the original digits. Using
a number-consonant transformation on the octal digits results in the letters
trttrtlntkkrs. These letters are then changed into the equivalent (by this
system) sequence of trd drt ln tng grs which is put into the word-picture
“tired dirty lion eating grass.” Thus 39 original digits were translated into
one easy to remember word- picture.
Norman suggests that
mnemonic devices work simply by reducing long, unrelated strings of material
into short, related lists according to a previously learned scheme. Norman
also suggests that mnemonic devices are effective for retrieval rather than
learning, as they usually add more to learn. Paivio (1969) suggested that
some of these devices may work by increasing the amount of imagery associated
with the material to be learned and recalled.
Norman (p. 123)
offers the following maxim for improving memory: “If you wish to learn something,
rather than plunge blindly ahead reciting the material endlessly, it would
be best to first summarize briefly its overall meaning and structure, second,
to decide how it relates to what you already know and, finally, to divide
the material into a small set of logical subdivisions.”
When forgetting does
take place in LTM it appears to be a retrieval problem rather than a storage
problem. Although some information may be lost from LTM storage, as with advanced
age, forgetting is generally due to not being able to retrieve information that is in the storage. The major cause of this retrieval
failure, as will be discussed in the next chapter, is interference resulting
from other learned material. In trying to retrieve one piece of information
you instead retrieve another piece of information.
Figure 4-1 (an extension
of Figure 3—1 from the preceding chapter) shows a possible set of relationships
between the processes discussed so far.
Lorente de No (1938)
was one of the first to provide evidence for networks of neurons in the brain
that close back on themselves and thus could be potentially self-exciting.
Neural impulses that start in such a network, as the result of stimulation from
neurons outside the network, might travel through the network, which closes
back on itself, and thus restimulate the same neurons again. This neural activity
might keep circling through the same network over and over many times. Such
closed networks of neurons that can keep restimulating themselves are called reverberatory circuits.
_dir%5Ctem91D1seg411.jpg)
Physiological
evidence related to the functioning of such circuits is still highly debatable.
Some evidence was supplied by Burns (1954) working with cortical slabs isolated
from the rest of the brain but with an intact blood supply. Burns found that
electrical stimulation of parts of these slabs could result in bursts of
electrical activity lasting for 30 minutes. Such activity may involve the
activity of reverberatory circuits. However, there is some question as to how
representative such a surgical preparation is of what goes on in the cortex of
an intact animal.
Figure 4—2 shows
a highly simplified schema of one reverberatory circuit. It should be remembered
that the only criterion for a reverberatory circuit is that it close back
on itself and re-excite the network. It might be composed of only two neurons
or of hundreds of neurons. It could be in the form of a simple circle or it
could be a complex network with many different routes and side-paths. The
neural impulses might or might not utilize exactly the same neurons through
the network each time. But to be a reverberatory circuit the network must close on itself and re-excite itself.
The reverberatory
circuit is a possible candidate for a holding mechanism. Information could
go into such a circuit and stay there during the duration of the reverberation.
Many psychological theories have drawn on such a construct, the most influential
theory being that of Hebb.
_dir%5Ctem9324seg421.jpg)
Hebb’s original
theory (1949) explained STM in terms of reverberatory circuits, meaning, for
example, that remembering a phone number just long enough to dial it involves
holding the information in reverberatory circuits. Continued activity of a
neural network was assumed to produce structural changes in the neuron,
probably at the synapse. (Hebb suggested the possibility of enlargement of the
synaptic knobs.) This structural change was thought to be the basis for LTM. At
first, then, Hebb had a sharp dichotomy: STM involved an activity trace, LTM
involved structural change. More recently, as was discussed earlier in this
chapter, Hebb has been questioning this distinction.
According to Hebb’s
model, continued use of a neural pathway results in facilitation of transmission
in the involved synapses. This facilitation makes it more probable that the
next neural impulse will take the same path which then can lead to reverberatory
activity of a given neural network. Hebb calls such a network a cell-assembly. With more complex learning, cell-assemblies
become associated so that the firing of one cell- assembly tends to also fire
related cell-assemblies. Hebb calls such a group of associated cell-assemblies
a phase sequence.
There were a number
of problems with the early Hebbian model, many of which were related to the
predicted spread of excitatory effects from cell-assemblies and phase sequences:
What keeps impulses from spreading to inappropriate pathways? What keeps too
many neurons from being tied up with single memories? Why doesn’t the whole
brain eventually become one giant phase sequence? These problems arose because
Hebb emphasized the excitatory effects of neurons on each other, but did not
account for any checks on these excitatory effects. To deal with these problems
Milner (1957) introduced the concept of inhibition into the Hebbian system.
That is, in addition to excitatory effects, neurons also
have inhibitory effects on other neurons, and thus input into a nerve network
might be inhibitory or excitatory. Repeated activity of a neural network was
assumed to inhibit neighboring neurons, thus restricting excitatory neural
activity to the appropriate neurons. The activity of a cell-assembly might
be stopped by inhibitory influences from outside the network. Now the Hebbian
model, like almost all current psychological models of brain functioning,
explains behavior and brain processes as being the result of a complex interplay
of excitatory and inhibitory effects.
Using Hebb’s theories,
a tremendous number of psychological phenomena can be explained in terms of
contructs such as cell-assemblies. For example, Hebb (1972, Chap. 5) uses
cell-assemblies to explain the mediating processes between stimuli and responses
in situations which involve a sequence of associations between stimulus and
response as opposed to cases where the stimulus elicits a reflex or simple
conditioned response. Simple holding of information for a period of time is
accomplished by cell-assemblies. Thoughts are mediating processes involving
one or more cell-assemblies, with thinking occurring when a number of such
mediating processes stimulate each other. Attention and set involve the selective
activation of one group of cell-assemblies over other groups.
An interesting prediction
from reverberatory circuit models of learning such as Hebb’s is that if immediately
after an animal learns something you disrupt the reverberatory circuits, then
the structural change necessary for LTM will be stopped, as well as the activity
of STM. This is indicated by studies in which an animal is given a disruptive
agent right after learning (e.g., a drug that produces convulsions), after
which he appears to not be able to remember what he learned. This type of
finding led to the research on consolidation discussed below. However, although
reverberatory circuit models were a major factor leading to consolidation
research and theories, current theories and explanations of consolidation
do not depend on any particular assumptions about reverberatory circuits.
CONSOLIDATION
Whatever
physiological change might underlie LTM, it is highly improbable that this
change occurs simultaneously with learning. Rather, after a learning experience
there is probably a period of time during which some process, called consolidation, converts the effects of the experience into
LTM. The assumption of consolidation is only an assumption that there is such a
non-instantaneous process, and not an explanation of what the process is or how
long it will take. From this assumption has come a massive number of
experiments and theories (Grossman, 1967, Chap. 14; Lewis, 1969; McGaugh, 1966;
McGaugh & Herz, 1972; Spevack & Suboski, 1969). First we will consider
some indirect evidence for consolidation — phenomena that are
supportive of consolidation, but are equally well explained without appealing
to consolidation processes.
In conditioning eyelid
blinking in humans, Spence and Norris (1950) found that the percentage of
conditioned responses increased as the mean inter-trial interval (ITI, the
amount of time between successive conditioning trials) increased from 9 to
90 seconds. A possible interpretation of this is that with short ITI one trial
impairs the consolidation of the previous trial, thus retarding learning.
Similarly Kettlewell and Papsdorf (1967) conditioned an eyelid response in
rabbits while the rabbits were in a darkened chamber. They found that 10 seconds
of illumination during the ITI impaired learning. Although there are other
possible explanations (see Ost, 1969), it may be that the ITI illumination
impaired the consolidation of the previous trial.
Another phenomenon
suggested as being supportive of consolidation is the Kamin effect. This refers to the observation that retention
is often poorest at some intermediate time after learning. Kamin (1963) trained
rats in a shuttle box, where they had to run from one side to the other on
cue to escape or avoid shock, to a criterion of three correct avoidance responses.
The rats were tested for retention 1 minute, 30 minutes, 1 hour, 6 hours,
24 hours, and 20 days after the last training trial. Kamin found little deficit
in all but the 1-hour and 6-hour groups. The deficit at 1 and 6 hours might
be because the learning is being consolidated and thus less available for
retrieval, although other research suggests that consolidation is completed
much sooner than this. A popular alternative explanation of the Kamin effect
(see Barrett et al., 1971) is that the shocks during training induce some
type of behavioral inhibition or suppression which impairs performance of
the avoidance response during retention. This inhibition is assumed to have
its maximum effect at intermediate retention intervals. Its effect gradually
increases and then decreases.
If a person is in
an accident in which he is hit on the head or receives some other type of
traumatic brain injury, retrograde
amnesia often occurs (Russell
& Nathan, 1946). In retrograde amnesia the person cannot remember what
happened during some period of time just prior to the accident. With time,
many of the memories come back, the temporally most distant memories returning
first. But there is often a period of time just prior to the accident that
is never recovered in memory. (Similar results occur following cerebral anoxia,
carbon monoxide poisoning, and Korsakoff psychosis.) Although these results
can be interpreted as being due to the accident’s disruption of consolidation
of the memories just prior to the accident, we will need a somewhat different
explanation for those memories that eventually recover.
To investigate this
retrograde amnesia with animals in the laboratory researchers have utilized
a variety of agents that are presumed to disrupt the consolidation process.
The most frequently used disruptive agent is electroconvulsive shock (ECS).
ECS is usually administered by attaching electrodes to both ears of the animal
and inducing a strong enough current to produce convulsions, although some
researchers (e.g., Jarvik & Kopp, 1967) have reported ESC-induced amnesia
with stimulation below the level necessary for convulsions. The direct physiological
effects of ECS are unknown, as it affects a number of variables such as amount
of neural firing and neural metabolism. A human parallel of ECS is electroshock
therapy, discussed later.
Other disruptive
agents include drugs such as metrazol (pentylenetetrazol) which produce convulsions
when given in sufficient dosage, carbon disulfide gas, anesthetics in some
situations, spreading depression, microwave radiation, polarizing currents,
anoxia, and some types of subconvulsive electrical brain stimulation. Almost
anything that significantly alters the neural activity of the brain appears
to be a potential disruptive agent. However, it may not be necessary to have
gross changes in brain chemistry or electrical activity for memory disruption.
Jacobs and Sorenson (1969) reported having been able to produce retrograde
amnesia in mice by dunking them for 10 seconds in hot (48° C) or cold (1°
C) water.
A typical consolidation
experiment consists of training a group of animals in some learning task and
then, immediately after learning, giving the disruptive agent to half of the
animals (keeping the other half as a control group). Depending on the nature
of the disruptive agent, the control animals are given some other treatment,
such as shock to the tail, instead of the disruptive agent. If on later retention
tests the control animals perform better than the experimental animals, then
perhaps the disruptive agent did in fact impair consolidation. What makes
the consolidation literature difficult is that there are many things other
than im pairing consolidation that the disruptive agent might do that could
result in poor performance on the retention test. Lewis (1969) lists a few
of the alternative effects of the disruptive agent: (1) it might affect some
of the processes of the storage mechanisms, such as the way information is
catalogued for later retrieval; (2) the memory itself may be unimpaired, but
the animal cannot associate the memory with the appropriate cues; (3) the
subject might lose the motivation to express the memory in its behavior; (4)
the disruptive agent may produce various forms of suppression, competition,
and inhibition. All of this highlights an earlier point: that many variables
affect performance, of which learning is only one.
The type of learning
task used is very important in consolidation research. Most of the experimental
tasks are ones in which the animal can learn the task in one trial, because
learning tasks requiring more than one trial have the following complications.
If the animal is given several trials before the disruptive agent, then considerable
consolidation may be going on between and/or during the trials. Thus the effects
of the disruptive agent would be confounded with the number of trials, the
inter-trial interval, the disrupting aspects of a trial on prior consolidation,
and the varying times of consolidation. On the other hand, if the disruptive
agent were given after each trial, the possible effects on consolidation would
be confounded with artifacts of multiple treatments of the disruptive agent.
For example, repeated ECS treatments in rats may result in a decrease in general
activity, decreased heart rate, and weight loss. Although some of these effects
may occur with a single ECS (Reuttenberg & Kay, 1965), the effects are
less than with multiple ECS’s.
The most common type
of one-trial learning task with rats and mice is a variation of passive avoidance, a situation where the animal can avoid a shock
by not making some specified response. The type of passive avoidance
task used in consolidation research often involves the animal’s receiving
a shock when it first makes a fairly common response, such as jumping from
a platform onto a grid floor or going from a large bright compartment into
a smaller dark compartment. Usually one such shocked trial is enough for the
animal to refrain from making the response again.
Theoretically, by
applying the disruptive agent at various times after learning we should be
able to determine how long consolidation takes. For at the point at which
the disruptive agent no longer impairs retention, consolidation may be over.
Using such an approach many investigators suggest that consolidation takes
about an hour, but this is far from being commonly accepted. Other estimates
vary between 10 seconds and a week. Probably consolidation time varies according
to factors such as the difficulty of the learning task, the nature and duration
of the disruptive agent, and the species of the animal.
When the animal in
a passive avoidance task receives a shock, an emotional response is elicited
and may become conditioned to stimuli of the test apparatus, so that later
the test apparatus tends to elicit this conditioned
emotional response (CER).
The CER may cause a decrease in general locomotion by the animal, which would
facilitate later performance in the passive avoidance task since it decreases
the probability that the animal will make the punished response. If the disruptive
agent, such as ECS, impairs the CER, the animal will make more incorrect responses, giving the appearance that the ECS disrupted memory
of the punished response. Chorover and Schiller (1966) argued that when CER’s
are minimized, ECS has little effect when given more than 10 seconds after
training. Apparent disruptive effects after 10 seconds are due to the ECS’s
interfering with the CER. Spevack and Suboski (1969) have made a similar argument,
saying that the CER incubates (increases in strength) following the shock
and that ECS after a minute does not disrupt consolidation but only the incubation
of the CER. In a critique of the Spevack and Suboski theory Dawson (1971)
included the following arguments: (1) there is no evidence that the retrograde
amnesia gradient is short when no CER is produced; (2) there is no good evidence
for ECS halting the incubation of a CER; and (3) Spevack and Suboski have
not well specified how consolidation and CER incubation actually work and
can be measured independently. Thus ECS may in fact disrupt consolidation,
but considerably more research is necessary to separate out other possible
effects of ECS, such as its effects on CER’s.
Retrieval Explanations
As mentioned
earlier, following traumatic amnesia, humans generally regain most of their
lost memories. Similarly some animal experiments (Nielson, 1968; Zinkin &
Miller, 1967) have reported recovery of memories that appeared to be disrupted
by ECS. But if ECS disrupts the consolidation of memories, then the memories
should never return. Unfortunately the experimental data on memory recovery
following ECS is quite complex and contradictory (see McGaugh & Herz, 1972,
p. 14). Possible recovery is confounded by factors such as the species of
experimental animal, the experimental task and procedure, the strength of
original learning, the amount of ECS-produced amnesia, dissipation or counter-conditioning
of ECS artifacts, opportunity for new learning, and cues prior to the retrieval
test that remind the animal of his learning experience. McGaugh and Herz (1972,
p. 20), however, point out that there are numerous studies showing amnesia to
be stable over long periods of time and that, at least in some situations, the
memory loss is permanent. They argue that this is “all that is required by the
most general form of the consolidation hypothesis.”
The recovery of
memories has suggested to some theorists that the disruptive agent affects
retrieval, not consolidation (see Nielson, 1968; Thompson & Neely, 1970;
Weiskrantz, 1966). Most retrieval theories of the effects of ECS appeal to
state-dependent learning effects. That is, the ECS produces a state different
from the one under which the animal learned the response. This change in state
then impairs retrieval of the information. With time the brain returns to
“normal,” and recall improves.
To test this hypothesis,
Thompson and Neely (1970) gave rats ECS at varying times relative to learning
and retention of a one-trial passive avoidance task. They found that rats
given ECS 25 minutes before training showed no later retention under “normal”
conditions. Similar results occurred with rats that had not been given ECS
before training but had been given ECS before retrieval. However, there was
no disruption if ECS was given before both training and retention.
The best results were when the ECS was given the same amount of time before
training as it was before retention. Thus the more similar the brain states
are at training and retention, the better the retention is. In a second experiment,
Thompson and Neely gave rats ECS 5 seconds after learning and found the best
retention in those rats that were also given ECS before the retention trial.
As in all the consolidation literature, there are experiments that appear
to be contradictory to the data and theories just discussed (see McGaugh &
Herz, 1972, p. 19).
Miller and Springer
(1973) have pointed out problems in such state- dependent theories. First,
such theories would generally predict recovery over time from experimental
amnesia, which often does not occur. Secondly, memories which are recovered
by giving the subject an ECS at the time of retrieval do not disappear again
when the brain returns to “normal.” Miller and Springer suggest a retrieval
model in which the disruptive agent often impairs “the establishment or future
functioning of the cataloging system necessary for retrieving the information
from long-term storage at some later time.” That is, ECS might not affect
the memory itself, but rather it affects an associated system that catalogs
information for future retrieval.
Facilitation
of Consolidation
If consolidation can
be disrupted, then perhaps it can also be facilitated or speeded up. There is
some evidence that drugs, particularly central nervous system stimulants such
as amphetamines, can, when given in proper dosages, facilitate performance, and
perhaps consolidation (McGaugh & Herz, 1972, p. 48; McGaugh &
Petrinovich, 1965). To show that the drug affects consolidation rather than
original learning or motivation, it is necessary to give the subject the drug after
the learning trial rather than before. Again, there are many reports
contradictory to the studies supporting drug effects on consolidation. The
effect seems to depend on variables such as the environment in which the
subject was raised, the strain and age of the animal, the complexity of the
learning task, the amount of time between learning and drug administration, the
nature and dosage of the drug, and the animal’s post-training environment.
Finally we need to know a lot more about the exact physiological effects of the
different drugs.
The assumption of
consolidation seems reasonable, although experiments trying to pin it down
have produced a complex and ambiguous set of data and a variety of different
explanations. Future research will have to factor consolidation out of all
the other effects. Also it will be necessary to tie together the research
on consolidation with research on the physiological bases of memory.
ELECTROSHOCK THERAPY
From the observation
that epileptics appeared to be less inclined to become schizophrenic than were
“normals” came the following questionable inference: Perhaps many forms of
mental illness can be improved by artificially inducing the equivalent of an
epileptic seizure. This was the genesis of a therapeutic treatment called electroshock therapy or electroconvulsive therapy (ECT). ECT is essentially the application of ECS to humans. Surface
electrodes are attached to the head (sometimes the current is applied directly
to the brain) and a current is applied in a strength great enough to produce
highly abnormal activity throughout the brain. The ECT first causes the person
to go unconscious. Then there is a general convulsion throughout the body,
often accompanied by overt motor seizures. Many practitioners minimize the
motor effects of ECT by using muscle relaxants and anesthetics. In those cases where
such drugs are not used—and many practitioners believe that they shouldn’t be
used — the patient could thrash around in a fashion
similar to an epileptic suffering from a grand mal seizure. When the patient
later regains consciousness, he is very confused and has general memory
disturbances.
Although ECT has
been given for almost all forms of mental illness, it now is primarily used
for various forms of depression, for patients with very agitated mood changes,
and for very acute catatonics. The number of treatments a patient receives
varies according to the case and the institution, and might be as few as two
or three or as many as 75, spread over time.
The exact physiological
and psychological effects of ECT are basically unknown. As would be expected
from our discussion of consolidation, ECT does produce forgetfulness of events
immediately preceding the shock treatment and often of the shock itself. There
is some evidence that multiple
treatments of ECT may produce intracranial hemorrhaging. When such patients
have their heads opened up their brains are found to be engorged with blood.
ECT also often produces a general flattening of the EEG (Kimble, 1965, p.
250).
ECT has been reported
to be successful in many cases and the question is why. Although no one
really knows, there are scores of explanations, some of which border on
the ridiculous. One theory is that ECT is a punishment by which the patient
atones for his sins. Another is that the coma produced by ECT is a symbolic
form of death. A third theory is that ECT helps suicidal patients by making
them fight for their life. The most commonly accepted theory is that in
the confused state following ECT many (although not all) patients are more
suggestible and amenable to forms of therapy that they resisted before.
Therefore ECT is best when immediately coupled with psychotherapy.
It may well be
that most of the effects of ECT can be explained in terms of thinking about
ECT as an extreme form of punishment. To vastly oversimplify: if every time
you act depressed you are shocked unconscious, pretty soon you will stop
acting depressed in the presence of other people. The evidence that ECT
acts like a punishment is quite strong. Although patients receiving ECT
do not at first report pain or fear of the treatment (probably because of
the ECT-produced amnesia), they generally do develop an aversion to the
ECT applications, and often dread having another treatment (Gallinek, 1956).
With continued treatment this conditioned aversion generalizes to situations
which are similar to situations where the patient received the ECT. The
author worked with one patient who, following his release from a hospital
where he received ECT, became very anxious whenever he saw someone in a
white uniform because they reminded him of the attendants who forcibly took
him down to the ECT room. To try to minimize such conditioned aversion many
practitioners anesthetize the patient to the point of unconsciousness before
taking him for ECT.
An example of the
use of ECT as a punishment is provided by a report of an American psychiatrist
working in a Vietnamese mental hospital (Cotter, 1967). The main object
was to get the patients to work more. Patients who would not work were punished
with three ECT treatments a week. This program significantly increased the
number of people volunteering for work. (Some patients who wouldn’t work
in order to avoid ECT were made to work in order to eat.) Although Cotter
says that part of the effect of the ECT may have been a reduction of mental
illness, “with others it was simply a result of their dislike or fear of
ECT.” Cotter attempted to justify his program by saying that “The use of
effective reinforcements should not be neglected due to a misguided idea
of what constitutes kindness.”
If ECT produces
its major effect through punishment, then there are a number of problems.
First, ECT is probably a much stronger form of punishment than should be
necessary to cure mental illness. Second, the ECT-produced convulsions may
disrupt the consolidation of some of the learned effects of the punishment,
thus partially defeating the objective. Third, and most important, as will
be seen later, punishment is generally an inefficient and often undesirable
technique of behavioral change. There are many more powerful techniques
that do not produce conditioned aversion or possible brain damage.
There is some biochemical
evidence suggesting possible effects of ECT. For example, drugs such as
reserpine, which produce depressive reactions in humans, also produce abnormal
biogenic-amine content in the brains of animals. ECS has been reported to
increase the turnover of these biogenic amines. This suggests that some
forms of human depression might have a biochemical origin which can be offset
by ECT. However, there are still many gaps in such research. For example,
how similar biochemically are drug-induced depressions and other forms of
depression? Of all the different behaviors lumped under the term “depression,”
which may be due to a biochemical imbalance? Is ECT the most effective way
to alter the brain chemistry? Which comes first?—these changes in brain
chemistry or changes in behavior?
The use of ECT
as a treatment procedure has been on a gradual decline for a variety of
reasons: the advent of tranquilizing drugs, more effective procedures for
dealing with problems such as depression, questions about the effectiveness
and morality of ECT, and the fact that the effects of ECT are often short
lived. In many hospitals now it is extremely rare for anyone to receive
ECT. Unfortunately, in some hospitals it is still used indiscriminately,
perhaps many times a day, primarily by practitioners who do not know alternative
treatment procedures and/or are enchanted by the speed with which ECT can
be applied in comparison to the slowness of other forms of therapy. In this
latter case it is clear that the disadvantages of ECT greatly offset any
possible advantages. The important question is whether we will eventually
be able to isolate a small specific class of people (perhaps with a biochemical
imbalance) for which ECT is the most effective and desirable treatment procedure.
Meanwhile studies of ECT should provide information about the nature of
ECS, consolidation, and holding mechani sms.
SOME
THEORETICAL EXTENSIONS OF HOLDING MECHANISMS
A number of theorists
have developed psychological models, based on holding mechanisms, that make
predictions that extend beyond the original conceptualization of the holding
mechanism. One such theory of action decrement was
proposed by Walker (1958). Action decrement refers to the lowered capacity
for rearousal of the same event. That is, right after an animal makes a
particular response in a stimulus situation, this situation has less of
a chance of re-exciting the neural network necessary to produce the same
response. This decreased probability, or action decrement, persists for
some time and then dissipates. Action decrement is assumed to be a direct
result of the process which produces consolidation. Thus, information is
less accessible for retrieval during consolidation. Action decrement may
have some biological utility in that it protects the consolidation process
from interference.
Consider a rat
which on its first trial in a T-maze chooses to go into the right arm. According
to the action decrement theory, some of the neurons related to this choice
become less easily fired. If during this decremental phase the rat is given
a second trial in the T-maze, there is a tendency for the rat to go left
because the neural networks involved in turning right are less easily excited.
Rats do, in fact, often show this type of alternation of responses, called
spontaneous alternation, during early stages of learning.
The more the consolidation
that takes place, the greater the long term memory (LTM). At the same time,
the more the consolidation, the more the action decrement. Thus the theory
predicts that the amount of action decrement is highly correlated with the
amount of learning. It is also assumed that the more arousal or excitement
in the animal’s nervous system during learning, the more consolidation will
occur. Also, rewards are considered a powerful source of arousal. All of
this produces the following logical sequence: The more rewarded the animal
is for a behavior, the more aroused he is; the more the arousal, the more
the consolidation, and thus the more the LTM and action decrement. This
leads to the prediction that if we rewarded our rat for turning right on
the first trial in the T-maze, it would be even more probable that it would
turn left on the next trial. Walker (1956) showed that thirsty rats rewarded
with water have a greater tendency to alternate than non-rewarded rats.
This simple experiment gave great support to the action decrement theory,
since many other theories of learning would probably have predicted that
rewarding the animal would make alternation less probable.
Later Walker (1967,
p. 226) expanded on the relationship between consolidation and action decrement
by suggesting that the end of the perseverative consolidation process might
be signalled not only by the end of ECS effectiveness in disturbing permanent
memory but also by the duration of action decrement and the time between
trials required to minimize intertrial interference.
Another prediction
from the action decrement theory is that the more arousal is associated
with some information, the poorer it can be recalled at first, because of
action decrement, but the better it will be recalled later, as the result
of greater consolidation and better LTM. This then is a situation where
motivation (arousal) is assumed to affect learning and not just performance.
This prediction
was examined by Kleinsmith and Kaplan (1963). They had college students
learn paired associates in which eight words were each paired with a different
single digit. The words (e.g., kiss, vomit, swim) were chosen to differ
in the amount of arousal they elicited. (Remember that arousal refers just
to general neural excitement regardless of whether the subjective experience
is pleasant or unpleasant.) The amount of arousal elicited by each word
for each subject was determined by the Galvanic
skin response (GSR) — changes
in the electrical conductivity of the skin. For each subject, then, it was
possible to select the three words of highest arousal and those of lowest
arousal. After learning, different groups of subjects were tested for recall
of the paired associates at a recall interval of 2 minutes, 20 minutes,
45 minutes, 1 day, or 1 week. At 2 minutes the low arousal material was
remembered five times better than the high arousal material. At 20 minutes
they were recalled about the same. By 45 minutes the high arousal material
was recalled significantly better.
The explanation
suggested by Kleinsmith and Kaplan is that at 2 minutes there is more reverberatory
activity associated with the high arousal material and so the information
is not as available for STM (short term memory). However, this increased
reverberation produces better LTM, which is reflected by the fact that the
high arousal material is better remembered at 45 minutes. (The reader can
verify the effects of arousal on LTM by recalling memories from his distant
past and observing that most of them were associated with states of arousal,
i.e., the event was particularly pleasurable or painful. This is not proof,
of course, of the above theory.) Later Kaplan altered his theory somewhat,
suggesting that the relative unavailability of high arousal material at
first is the result of reverberation-generated neural fatigue, rather than
of the reverberation per se (Kaplan, 1972, personal communication; Pomerantz
et al., 1969). The later improved recall with high arousal material thus
is a joint effect of the dissipation of fatigue and the time required for
the biochemical changes underlying LTM.
Butter (1970) replicated the results of Kleinsmith and Kaplan with some procedural changes (e.g., controls for serial position and GSR habituation effects). In addition, Butter also used nouns that differed in noun-imagery, the ability of the word to elicit images (cf. Paivio’s research, discussed in Chapter 3). Butter found that nouns low in this imagery elicited high arousal; nouns high in imagery elicited low arousal; and recall associated with the material agreed with the expected crossover of low and high arousal material.
A number of experiments
report approximately the same types of effects of arousal on learning or
recall. However, interpretations become too complex and often confused when
one tries to interrelate these results with the subtleties of verbal learning
and other responses (such as the orienting response), which are also tied
in with arousal states (Maltzman et al., 1966; discussion in Walker, 1967).
This does not devaluate the phenomenon, but suggests the need for considerably
more research.
The concepts discussed
in this chapter, such as reverberatory circuits, short term memory, consolidation,
and action decrement, seem as if they could all be reduced to a few basic
constructs that could account for a wealth of disparate phenomena. Although
such a synthesis can be fairly easily made at a general theoretical level,
any proposed model readily falls apart when we investigate the specific
parameters of the various phenomena. For example, the time courses of STM,
consolidation, and action decrement appear to be substantially different
depending on a wide range of experimental variables. However, hopefully
such a synthesis of constructs will eventually be possible.
Many events happen
so quickly that the related information would pass us by unless we had some
storage procedures — information holding mechanisms — to
temporarily hold the information until it can be adequately processed. Such
holding mechanisms also permit us to associate events which do not occur
together in time. The existence, number, and nature of holding mechanisms
are debatable. One possible sequence is as follows: (1) perceived information
goes into the temporary holding mechanism of sensory storage; (2) some of
the information then goes from sensory storage into the temporary holding
mechanism of short term memory; and (3) some of the information from short
term memory goes into the permanent storage of long term memory.
Sensory storage
holds all information picked up by the sense
receptors. This information is held only a fraction of a second, after which
time part of the information is processed further and the rest is dissipated
away. Short term memory is usually conceived of as a storage with a
limited capacity; in short term memory, information dissipates out with
time and/or is readily displaced by newer information entering short term
memory. Long term memory, on the other hand, is generally conceived of
as a permanent storage with practically unlimited capacity. Information
stored in long term memory may never be lost, even though there may be problems
in retrieving the information from storage, as in forgetting. Many differences
have been proposed between short and long term memory, although some theorists
suggest that they are not really different storages at all but simply different
points on a continuum of storage. Apparent differences between short and
long term memory, according to these theorists, are due to the amount of
time the subject has to perform his learning task.
Memories may be
stored in long term memory according to their attributes — distinguishing
characteristics that help separate memories during initial storage and later
retrieval. One way of studying the processes of long term memory is by investigating
memory devices and systems, known as mnemonics.
One candidate for a physiological network that may function as a holding mechanism is the reverberatory circuit.
Reverberatory circuits are closed loops of neurons
that circle back on themselves and thus restimulate the circuit. In this
way information might be held for the duration of activity within the reverberatory
circuit. The exact functions of such circuits are still highly speculative.
Information cannot
be instantly stored in long term memory. Rather the information is held
until a process of consolidation results in the physiological changes necessary
for long term memory. Evidence for a consolidation process comes from many
sources, the primary one being the apparent disruption of consolidation
resulting in a failure to remember events that occurred just prior to the
disruption. This disruption occurs in accidents, as when a person is hit
on the head, and in experiments where electroconvulsive shock is applied
to the brains of test animals. Other evidence for consolidation comes from
studies that apparently facilitate consolidation through the use of drugs.
Noted in the literature
are many confounding effects of consolidation that allow
for various interpretations. For example, in addition to producing brain
seizures, electroconvulsive shock also acts as a punishment, which suppresses
some responses and alters the brain in ways that may impair retrieval. Some
theorists argue that consolidation of memories is seldom disrupted, and
that only retrieval is impaired.
Related to consolidation
is the theory of action
decrement, which postulates
that after an animal makes a response the underlying neural network has
a lowered capacity to be rearoused, probably because it is involved in consolidation.
This unavailability of the neural network makes it less probable that the
animal will repeat the same response to the same stimulus situation. It
is also assumed that the more aroused the animal, the greater the action
decrement, the greater the consolidation, and hence the better the long
term memory. This leads to experiments suggesting that high arousal material,
as opposed to low arousal material, will not be remembered as well immediately
after learning, but will be remembered better later when action decrement
and consolidation are more completed.
Hebb, D.O. Organization
of Behavior. New York: Wiley, 1949.
Kintsch, W. Learning,
Memory, and Conceptual Processes. New York: Wiley, 1970. Chapter 4.
McGaugh, J. L., & Herz, M. J. Memory Consolidation. San Francisco: Albion, 1972. Norman, D. A. Memory and Attention. New York: Wiley, 1969.