A Theory About Why We Forget What We Once Knew

Loading...

Taking too long?

Reload document

| Open in new tab

CURRENT DIRE CTIONS IN PSYCHO LOGICAL SCIENCE

A Theory About Why We Forget
What We Once Knew

John T. Wixted

University of California, San Diego

ABSTRACT—Traditional theories of forgetting assume that
everyday forgetting is a cue-overload phenomenon, and
the primary laboratory method used for investigating that
phenomenon has long been the A-B, A-C paired-associates
procedure. A great deal of research in psychology, psy-
chopharmacology, and neuroscience suggests that this
approach to the study of forgetting may not be very rele-
vant to the kind of interference that induces most forget-
ting in everyday life. An alternative interference theory
holds that recently formed memories that have not yet had
a chance to consolidate are vulnerable to the interfering
force of mental activity and memory formation, even if the
interfering activity does not involve material similar to
what was previously learned. This account helps to ex-
plain why sleep, alcohol, and benzodiazepines all forestall
forgetting of a recently learned list, and it is consistent
with recent work on the variables that affect the induction
and maintenance of long-term potentiation in the hippo-
campus.

KEYWORDS— forgetting; interference; consolidation; ret-
rograde facilitation; long-term potentiation (LTP)

In common parlance, the verb ‘‘to forget’’ refers to the inability
to remember a prior event whether or not that event was ever
committed to memory. Thus, for example, I might claim to forget
where I placed my keys, but the truth may be that I neglected to
encode the relevant information in the first place. People with
amnesia resulting from damage to the brain’s medial temporal
lobes typically experience this kind of forgetting. That is, they
fail to encode new information, so they seem to forget everything
quickly.

In the field of psychology, theories of forgetting are typically
concerned with the loss of information that was once success-
fully encoded. Thus, although you once remembered the name

Address correspondence to John T. Wixted, Department of Psy-
chology, 0109, University of California, San Diego, La Jolla, CA
92093-0109; e-mail: jwixted@ucsd.edu.

of your first-grade teacher, thereby proving that the information
was encoded, you may now discover that the name has slipped
away (forever, perhaps). Why does that happen? Traditionally,
the answer has been that forgetting occurs either because
memories naturally decay or because they succumb to the forces
of interference. The story that most students of psychology learn
is that a classic sleep study conducted by Jenkins and Dal-
lenbach (1924) tilted the balance in favor of interference theory.
In that study, subjects remembered more of a previously learned
list when they slept before taking the memory test than when
they remained awake. Because interference is much less likely
to be encountered while sleeping, whereas a natural decay
process should unfold whether one is asleep or awake, the re-
sults pointed to a dominant role for interference. In a recent
review of a century of work on forgetting, I argued that the field
accurately interpreted Jenkins and Dallenbach’s study (i.e., the
results suggest a role for interference) but then modeled the
interference process in a way that is not likely to be very rele-
vant to ordinary forgetting (Wixted, 2004).

In the early part of the 20th century, two views of interference
struggled for supremacy. According to one view, interference
consists of new memories degrading previously established
memory traces that have not yet had a chance to consolidate.
For such interference to occur, the new memories do not need to
be especially similar to the ones they impair (Skaggs, 1925).
The formation of new memories, per se, is the interfering force.
According to the other view, interference consists of what later
came to be called a cue-overload effect
(Robinson, 1920;
Watkins & Watkins, 1975). Cue overload occurs when more than
one memory is associated with the same retrieval cue. Thus, for
example, if I memorize ‘‘Learned Hand’’ as the name of a famous
judge, the retrieval cue ‘‘famous judge’’ will become ever less
effective in retrieving that name as I memorize the names of
more famous judges. From this point of view, similarity of the
interfering material is critical because similar items tend to be
subserved by the same retrieval cue (whereas dissimilar items
tend not to be). Thus, instead of interference being viewed as the
degradation of a memory trace, it is viewed as a competition
phenomenon that occurs at the time of retrieval.

6

Copyright r 2005 American Psychological Society

Volume 14—Number 1

John T. Wixted

The study of cue-overload interference is exemplified by the
well-known A-B, A-C paired-associates design. In this proce-
dure, the experimental group of subjects learns an A-B list of
cue-target word pairs (e.g., hero-prison, water-valley, tiger-im-
age, etc.) followed by a second, A-C, list involving the same cue
words but different target words (e.g., hero-women, water-salad,
tiger-infant, etc.). A control group learns the A-B list followed
by a C-D list, which involves a completely different set of words.
Both groups are then tested by giving the A cues and asking for
recall of the B targets. The typical finding is that the experi-
mental group performs more poorly than the control group, and
this implies that the more items are attached to the A cue, the
less likely that cue will be to occasion the retrieval of any one of
its associates. Because paired-associates procedures generate
powerful effects in the laboratory, the cue-overload view of
forgetting came to dominate the trace-interference account
during the latter half of the 20th century.

Specific theories of cue-overload interference died an inglori-
ous death in the late 1960s, in part because concepts such as
unlearning (the idea that forgetting is caused by the unlearning
of previously learned associations) and spontaneous recovery
(the idea that unlearned associations spontaneously reestablish
themselves) failed to find much empirical support (Tulving &
Madigan, 1970). Still, the general notion that interference is
mainly due to cue overload at the time of retrieval lived on, despite
the fact that multiple efforts to demonstrate that cue-overload
effects apply to everyday forgetting (not just laboratory forgetting)
generally suggested just the opposite (Underwood & Ekstrand,
1967). In fact, the only reason to believe that cue-overload effects
play any role at all in everyday forgetting is that almost everyone
can point to a few examples from their own lives where it surely
does. But the mere fact that cue-overload effects sometimes play
a role does not mean that the role is a substantial one.

Summarizing the state of the art late in his career, Underwood
(1983) said: ‘‘A relatively few years ago it seemed that a fairly
comprehensive theoretical account of forgetting was close at hand,
but that has slipped away. Some investigators have lost confidence
in interference as a major cause of forgetting, but none of the
proposed replacements thus far has created a feeling that things
are on a productive new track. But that will surely come’’ (p. 262).
In search of a productive new track, I recently reviewed a
century of research on forgetting and concluded that an idea
long ago abandoned by the field—namely, that everyday for-
getting is largely the result of trace interference—was ahead of
its time (Wixted, 2004). The evidence bearing on that claim
derives from the fields of psychology, psychopharmacology, and
neuroscience, and I briefly review that evidence next.

PSYCHOLOGY: THE TEMPORAL GRADIENT OF
RETROACTIVE INTERFERENCE

An idea that the field of psychology has never fully embraced
(even though it has become the ‘‘standard story’’ in the field of

neuroscience) holds that memories consolidate for a period of
time after they are formed. During the consolidation period,
memories are especially vulnerable to disruption. Perhaps the
most compelling piece of evidence in favor of consolidation
theory is the temporal gradient of retrograde amnesia (Ribot,
1882). The temporal gradient is observed when a brain structure
known as the hippocampus is damaged. Bilateral hippocampal
damage induces anterograde amnesia (which is the inability
to form new memories), but it also impairs, to some degree,
previously formed memories, with memories formed just prior
to brain injury being more impaired than memories that were
formed longer ago (i.e., retrograde amnesia is temporally
graded). Older memories are assumed to be relatively spared
because they are more completely consolidated than newer
memories (and, therefore, are less dependent on the hippo-
campus).

If memories do consolidate over time, and if the formation of
new memories does interfere with previously formed memories,
then one should also expect to see a temporal gradient of ret-
roactive interference. That is, the encoding of new information
should interfere more with recently formed memories than with
older memories. This issue was first addressed by Mu¨ ller and
Pilzecker (1900), and they found clear evidence for the pre-
dicted temporal gradient (i.e., retroactive interference was more
pronounced if the interfering material occurred early rather
than late in the retention interval before the memory test).
However, most other studies did not. In fact, all reviews of the
relevant literature conducted after 1930 concluded that there is
little or no evidence for the predicted temporal gradient of
retroactive interference. To many psychologists, this meant that
the process of consolidation, even if it is real, is a physiological
process that has little to do with the psychological processes
that are responsible for forgetting (such as unlearning and
spontaneous recovery). Such findings may help to explain why
the field of psychology has never really embraced the notion of
memory consolidation.

I have argued that this interpretation of the literature is
inaccurate and that it results, in part, from a failure to dis-
tinguish between interference based on trace degradation (a
storage phenomenon) and interference based on cue overload
(a retrieval phenomenon; Wixted, 2004). No consolidation
theory predicts that the retrieval competition induced by cue
overload will be more pronounced if the retrieval cue is
overloaded early in the retention interval. Even so, manipu-
lating the timing of cue overload was the preferred method
of investigation. The few studies that varied the temporal point
of memory formation itself against a background of mental
quietude (e.g., Skaggs, 1925), which are the kinds of studies
that speak more directly to the issue, found evidence for the
temporal gradient that is predicted by consolidation theory.
In light of such evidence, the long-neglected notion that
memories consolidate may be an important part of the story of
why we forget.

Volume 14—Number 1

7

Why We Forget

PSYCHOPHARMACOLOGY: RETROGRADE
FACILITATION

NEUROSCIENCE: A POSSIBLE NEURAL MECHANISM

Alcohol and benzodiazepines (e.g., Valium, Xanax) both induce
temporary anterograde amnesia. Perhaps the best-known ex-
ample is the alcoholic ‘‘blackout,’’ which refers to the individ-
ual’s complete amnesia for events that occurred while he or she
was intoxicated. Also well established, but less well known in
the field of psychology, is that these drugs (like sleep) result in
retrograde facilitation. That is, material learned just prior to
taking one of these drugs is, at a later point, remembered better
than it otherwise would have been (i.e., less forgetting will have
taken place). An obvious explanation for this interesting phe-
nomenon is that the reduced rate of memory formation while one
is under the influence of the drug protects recently formed
memories during a period of time when they are especially
vulnerable. This explanation is sometimes rejected in the psy-
chopharmacology literature because retrograde facilitation is
observed in laboratory studies even when no explicit interfering
task is arranged for either the subjects who are administered the
drug or the control subjects who receive a placebo. According to
this way of thinking, if a subject memorizes a list of words and
then drinks alcohol, a reduction in interference (relative to a
control group) ought to be observed only if a similar interfering
task is subsequently arranged. But memory for the list is en-
hanced even when an interfering task is not presented, and
some researchers have taken this to mean that the drug directly
enhances the consolidation process.

The idea that an interference effect should be observed only if
an explicit (and similar) interfering task is arranged makes sense
from the point of view of traditional interference theory. In fact,
Underwood (1957) made this very point when he observed that
the sleeping subjects in Jenkins and Dallenbach’s (1924) classic
sleep study did not avoid learning material that was similar to the
study list, so avoidance of retroactive interference could not
explain why sleep facilitates memory. To Underwood, this rea-
soning implied that interference must have been proactive in
nature (i.e., arising from prior similar learning) and that sleep
somehow held the forces of proactive interference at bay. Were
he alive today, he might advance the same account of alcohol-
induced retrograde facilitation. However, in the absence of any
compelling evidence, this unnatural theory seems hard to accept.
An alternative possibility is that alcohol and benzodiazepines
directly enhance the consolidation process, but this notion is
hard to accept as well. Why would a drug that facilitates the
consolidation process induce anterograde amnesia? Facilitating
the consolidation process should result in anterograde and
retrograde facilitation, not anterograde amnesia and retrograde
facilitation. A more likely possibility is that these drugs mini-
mize memory formation in general and that, in so doing, they
protect recently formed (and, therefore, fragile) memories from
the deleterious forces of nonspecific retroactive interference
that arise from the process of memory formation itself.

What is the source of retrograde facilitation? One possibility is
that drugs like alcohol and benzodiazepines close the hippo-
campus to new input, thereby inducing anterograde amnesia
without compromising the ability of the hippocampus to con-
solidate previously formed memories. Because new input is
prevented, recently formed (and, therefore, incompletely con-
solidated) memories are protected from the retroactive inter-
ference that they would otherwise encounter. Thus, these drugs
may act in the same way that sleep does even though the in-
dividual remains conscious. By contrast, hippocampal lesions
both prevent new input (resulting in anterograde amnesia) and
terminate the hippocampal-dependent consolidation of recently
formed memories (resulting in retrograde amnesia as well).

The prevailing view of how memories are initially encoded is
that the process involves a change in the probability that
postsynaptic neurons in the hippocampus will fire in response to
neurotransmitters released from presynaptic neurons. That is,
neurons are in a chain of communication, and when a neuron
that is earlier in the chain (the presynaptic neuron) releases a
substance known as a neurotransmitter, it causes the next
neuron in the chain (the postsynaptic neuron) to fire. Memories
of an experience may be initially encoded in the brain when that
experience causes certain postsynaptic neurons to become more
reactive to neurotransmitters released by presynaptic neurons.
The laboratory analogue of this theoretical memory mechanism
is long-term potentiation (LTP), which is a durable increase in
synaptic transmission in response to high-frequency stimulation
of presynaptic neurons. This artificially induced increase in
synaptic efficacy typically lasts only a few days or weeks, so it
presumably does not represent the way in which memories are
permanently coded. Still, in their fragile, preconsolidation
state, memories may exist by virtue of an LTP-like process
taking place in the hippocampus.

Alcohol and benzodiazepines, which interfere with the en-
coding of new information, both block the induction of LTP in
the hippocampus (Givens & McMahon, 1995). However, alco-
hol does not impair the maintenance of hippocampal LTP in-
duced an hour prior to administration of the drug (Givens &
McMahon, 1995). Blocking the induction of LTP without im-
pairing the maintenance of previously established LTP parallels
the effects of alcohol on memory. That is, alcohol blocks memory
formation without impairing previously formed memories.

Several recent studies show that the induction of LTP in the
hippocampus (like the formation of a new memory) serves as an
interfering event. Thus, the induction of hippocampal LTP
causes partial forgetting of a previously learned task (Brun,
Ytterb(cid:1), Morris, Moser, & Moser, 2001). It also impairs the
maintenance of previously established LTP (Villarreal, Do,
Haddad, & Derrick, 2001), which may be the very reason why it
induces forgetting of prior learning. Findings like these suggest
that the hippocampus may not be an unlimited resource system

8

Volume 14—Number 1

John T. Wixted

when it comes to the induction and maintenance of LTP. If a
recently formed memory relies on an LTP-like process that re-
quires biophysical resources to be maintained, then it would be
impaired by the subsequent formation of memories that pref-
erentially draw on the same resources. Only after a memory is no
longer dependent on the maintenance of an LTP-like process in
the hippocampus (i.e., only after it has consolidated) will it be
impervious to the interfering effects of retroactive interference.

CONCLUSION

To summarize, my reading of the literature is that forgetting is
largely a function of nonspecific retroactive interference acting
on memory traces that have not yet had a chance to consolidate.
In the early stages of consolidation, memory traces are theo-
retically maintained by an LTP-like process in the hippocam-
pus. However, the maintenance of these traces is vulnerable to
the subsequent induction of LTP-like processes associated with
the formation of new memories.

Although it is now possible to tell a story of forgetting that
integrates findings from psychology, psychopharmacology, and
neuroscience, every part of that story needs to be solidified by
further research. In psychology, some very old evidence sup-
ports the idea that a temporal gradient of retroactive interfer-
ence is observed when memories are formed against a
background of mental quietude, but that work is far from de-
finitive. Similarly, in psychopharmacology, we know that alcohol
and benzodiazepines induce retrograde facilitation, but we do
not know if a temporal gradient exists. Thus, for example,
memory tested 24 hr after learning should be enhanced to a
greater extent if alcohol is consumed immediately after learning
than if it is consumed 8 hr after learning. And in neuroscience, a
new question concerns whether it is newly formed memories or
recently activated memories that undergo a process of consol-
idation. Further inquiry into these matters is sure to provide a
more complete answer to the question of why we forget what we
once knew.

Recommended Reading
Ekstrand, B.R. (1972). To sleep, perchance to dream (about why we
forget). In C.P. Duncan, L. Sechrest, & A.W. Melton (Eds.),

Human memory: Festschrift for Benton J. Underwood (pp. 59–82).
New York: Appelton-Century-Crofts.
Underwood, B.J. (1957). (See References)
Villarreal, D.M., Do, V., Haddad, E., & Derrick, B.E. (2002). (See

References)

Wixted, J.T. (2004). (See References)

REFERENCES

Brun, V.H., Ytterb(cid:1), K., Morris, R.G.M., Moser, M., & Moser, E.I.
(2001). Retrograde amnesia for spatial memory induced by
NMDA receptor-mediated long-term potentiation. The Journal of
Neuroscience, 21, 356–362.

Givens, B., & McMahon, K. (1995). Ethanol suppresses the induction

of long-term potentiation in vivo. Brain Research, 688, 27–33.

Jenkins, J.B., & Dallenbach, K.M. (1924). Oblivescence during sleep
and waking. American Journal of Psychology, 35, 605–612.
Mu¨ ller, G.E., & Pilzecker, A. (1900). Experimentelle Beitra¨ge zur
Lehre vom Geda¨chtnis [Experimental contributions to the sci-
ence of memory]. Zeitschrift fu¨r Psychologie Erga¨nzungsband, 1,
1–300.

Ribot, T. (1882). Diseases of memory: An essay in positive psychology.

London: Kegan Paul, Trench & Co.

Robinson, E.S. (1920). Studies from the psychological laboratory of the
University of Chicago: Some factors determining the degree of
retroactive inhibition. Psychological Monographs, 28(Whole No.
128), 1–57.

Skaggs, E.B. (1925). Further studies in retroactive inhibition. Psy-

chological Monographs, 34(Whole No. 161), 1–60.

Tulving, E., & Madigan, S.A. (1970). Memory and verbal learning.

Annual Review of Psychology, 21, 437–484.

Underwood, B.J. (1957). Interference and forgetting. Psychological

Underwood, B.J. (1983). Attributes of memory. Glenview, IL: Scott,

Review, 64, 49–60.

Foresman and Co.

Underwood, B.J., & Ekstrand, B.R. (1967). Studies of distributed
practice: XXIV. Differentiation and proactive inhibition. Journal
of Experimental Psychology, 74, 574–580.

Villarreal, D.M., Do, V., Haddad, E., & Derrick, B.E. (2002). NMDA
receptor antagonists sustain LTP and spatial memory: Active
processes mediate LTP decay. Nature Neuroscience, 5, 48–52.
Watkins, C., & Watkins, M.J. (1975). Buildup of proactive inhibition as
a cue-overload effect. Journal of Experimental Psychology: Hu-
man Learning and Memory, 1, 442–452.

Wixted, J.T. (2004). The psychology and neuroscience of forgetting.

Annual Review of Psychology, 55, 235–269.

Volume 14—Number 1

9