Transcript Components of memory - University of Leicester

Long-term memory (LTM)
Long-term memory: episodic and semantic
memory
PS2016: Cognitive Psychology
John Beech
1
LTM: episodic and semantic memory – a preview
1. An examination of earlier research on memory.
– (e.g. Atkinson and Shiffrin, 1968)
2. An overview of the components of LTM (Squire, 1992)
3. Procedural memory
– (e.g. about amnesia)
4. Encoding, storage and retrieval.
– (e.g. levels of processing)
5. Relatedness vs. distinctiveness
6. Semantic vs episodic memory
– Comparisons and contrasts
7. Retrieval in episodic memory
8. The cognitive neurospsychology of episodic memory
9. Conclusions
2
Atkinson & Shiffrin (1968) and 3 memories
3
Long-term memory (LTM)
Atkinson & Shiffrin proposed 3 types of memory:
Sensory memory: We process information very briefly in its
raw form. Large capacity. Rapid decay.
Short-term memory (STM): This has limited capacity for
period of 15-30 sec. Rehearsal maintains it. Forgetting
slow.
Long-term memory (LTM): This has large (unlimited?)
capacity that holds information for a long period. Perhaps
all that can be remembered is held for virtually a lifetime,
or at least while the brain is capable of retrieval.
It is believed that LTM is organised into components which
contain different types of information.
4
Working Memory
The concept of Working Memory came later, which as its name
suggests, is a kind of STM involved in everyday mental
activity.
Examples of Working Memory:
• Listening to or reading text
• Remembering a phone number.
Known as working memory and can be used visually, spatially,
in mental arithmetic, in reasoning and problem solving as
well.
It can be important for transferring information into long-term
memory. It works as a temporary holding store during
processing.
5
LTM
Ryle (1949) proposed an important difference between
remembering information based on “knowing that” and
when we remember in terms of “knowing how”.
We know that:
• our bike is currently parked in a bike shed
• we had to take the cat to the vet last Wednesday
• the capital of England is London
We know how:
• to ride a bike
• to swim
6
Squire (1987, 1992) and her LTM distinctions
LTM
Declarative (fact)
Episodic
Semantic
Procedural (skill)
Skills
Priming
Conditioning
7
Declarative vs.
procedural memory
Declarative (facts) memory – involves factual information, such
as ‘Henry VIII had 6 wives’. It means that information is stored
that is explicitly available for later—which is often referred to as
explicit memory. It can be tested by free recall, cued recall and
recognition memory. we are conscious of these memories.
Procedural (skills) memory – involves the knowledge we have for
performing a skill such as using a computer mouse. Although we
learn skills, they are implicit memories in that we don’t have
access to how they work. In fact if you try to think about what
you’re doing, your performance can get worse. (E.g. if you’ve
learned to type, but then try to be more self-conscious about
what you’re typing, you can make typing errors.) We are not
conscious of information in implicit memory.
8
Procedural memory
• Amnesics are able to continue with their normal skills and
even develop new ones. Just to explain further—patient
HM had his medial temporal lobe removed as part of an
operation for epilepsy when 27 in 1953. He was the worst
case in a series of 10 patients reported by Brenda Milner.
• Removal of HM’s medial temporal lobe, including the
hippocampal formation, resulted in dramatic memory
impairments, because it was bilateral.
• He remained of normal intelligence (IQ 112) and had no
psychological illness. However, the surgery resulted in
intense anterograde amnesia (Events taking place after
surgery are never remembered for more than 60 sec).
9
Procedural memory
• But to return to the main point—if HM does a jigsaw,
he will get faster and faster each time he does it. In
other words his implicit (or procedural) memory is
intact. However, HM, like other amnesiacs would not
remember doing that particular jigsaw before. In
other words, explicit memory is severely affected.
• There a number of procedural memory skills that
amnesiacs can do without difficulty: mirror
drawing/writing; finger mazes; Tower of Hanoi—but
what is the commonality across these tasks?
• This leads to the problem of circularity—if an
amnesiac can do a task, then this is labelled as
‘procedural’.
10
Procedural memory
• Baddeley (1997) thinks that it is important to study the
sub-components of procedural memory. Squire (1992)
thought that these were: (1) skills and habits; (2) priming;
(3) simple classical conditioning and (4) non-associative
learning (e.g. habituation). As shown in the previous fig.
• However as mentioned at the beginning, the focus of this
topic is on episodic memory (and also how it compares to
semantic memory), so examining procedural memory just
serves as a contrast for our purposes.
11
Encoding, storage and retrieval
Another perspective is that memory has 3 stages:
•
Encoding—inputting information
•
Storage—keeping in memory
•
Retrieval—pulling out of memory
Beginning with encoding—this is how we represent information
(e.g. events). But this encoding also depends on how it is
processed. This might be just perceptually, but information
can be processed much further than this…
12
Encoding: Levels of processing models
•
A model to challenge the Atkinson and Shiffrin
was one proposed by Fergus Craik and Robert
Lockhart (1972).
•
They proposed the levels of processing model in
which items that were processed “deeper” were
consolidated into LTM.
Written words were presented in 3 conditions:
1. Ps had to identify if they were in upper or lower
case letters – this was a superficial level.
2. Ps judged if words rhymed with each other – an
intermediate level as processing for meaning still
not involved.
3. Ps had to make a judgement about a word (e.g.
Lamp post -can it rotate?) – this was considered to
be deep processing.
13
Encoding: Levels of processing models
Written words were presented in 3 conditions:
1.
2.
3.
•
•
Ps had to identify if they were in upper or lower case letters
– this was a superficial level.
Ps judged if words rhymed with each other – an intermediate
level as processing for meaning still not involved.
Ps had to make a judgement about a word (e.g. Lamppost can it rotate?) – this was considered to be deep processing.
Found that memory better if deep processing
involving semantic processing was used.
This is compared to information coded
visually or phonologically.
Thus there’s an incompatibility with the STMLTM formulation of Atkinson & Shiffrin
because it wasn’t about just holding
information in STM long enough, it was
instead about the type (superficial vs. deep)
of processing taking place.
14
Problems with levels of processing
• But the levels of processing model isn’t ideal. It’s
major problem is one of circularity. What constitutes
shallow versus deep processing? It comes down to the
definition of the experimenter. This means that it is a
theoretical formulation that can’t be falsified—one can
always shift the definition to suit one’s findings. This
is not actually too much of a problem as there are other
formulations in psychology that are also very difficult
to falsify.
15
Problems with levels of processing
• Another problem is that it cannot explain all memory
phenomena (see Rutherford, 2005); but again, this is
perhaps not surprising. A relatively simple model is
not going to explain phenomena outside its domain.
• At least the levels of processing model meant that
researchers shifted from viewing memory as a
structural phenomenon to one in which memory was
regarded as a process, rather than residing in different
storage areas. This is not to say that the concept of
memory storage isn’t also viable.
16
A paradox: relatedness vs.
distinctiveness
Relational and item-specific processing
Research evidence in memory has a paradox: on the
one hand if items are distinctive in some way we
tend to remember them. (E.g. we find it difficult
to remember a set of faces if they are all very
similar.) On the other hand, we know that if we
organise items together they are easier to
remember. (E.g. if you had to remember: “table,
cat, chair, mouse, bed, cow, etc”. This is easier
than this more disconnected list: “coat, pen,
organ, broom, gun, car, etc.”)
17
Relational and item-specific processing
• Hunt and McDaniel (1993) refer to relation
processing for similarity in processing and itemspecific processing for when there is
distinctiveness.
• To illustrate relational processing vs. item-specific
processing division consider a study by Hunt &
Einstein (1981) that looked at this in the context of
recall and recognition (which are both tests of
explicit memory).
18
Hunt & Einstein (1981) study
They used a 2 x 2 design looking at (1) categorised lists
(6 words from 6 categories e.g. (a) “table, chair,
bed…” (b) “cat, mouse, cow,… etc”) vs. a list of 36
unrelated words. And (2) relational processing vs
item-specific processing—to be explained next.
Type of
processing
Relational
Specific
Grouped
Ungrouped
1
2
3
4
19
Hunt & Einstein (1981) stages of the expt
1. For relational processing two groups sorted their list
into categories (pretty difficult when dealing with the
unrelated lists!). The other two groups had to undergo
item-specific processing: they rated each word for
pleasantness. Thus the relational group were
processing for similarity – looking at how the items
could be associated together, while the other group
(the item-specific group) was processing
distinctiveness by assessing its particular pleasantness.
2. All Ps read a short story for one minute to stop
rehearsal.
3. (3) They ‘free recalled’ the 36 words.
4. (4) They were given a recognition test – the 36 words
were again presented along with 36 new words and
they had to say whether each item was ‘old’ (i.e.
previously presented) or ‘new’.
20
Hunt & Einstein (1981) study-findings
100
90
80
70
60
Relational
50
Item-specific 40
30
20
10
0
50
45
40
35
30
25
20
15
10
5
0
grouped
ungrouped
Relational
Item-specific
grouped
ungrouped
LH figure shows the results of free recall (FR) and the RH figure for
recognition. An interpretation for FR is that when items are related
together, performance is better. Thus the one condition (the
ungrouped – item-specific group) in which the items were unrelated
and Ps were concentrated on individual properties, performance was
the worst. In the other red condition, the items were grouped
together, even though this wasn’t explicit, it helped recall.
In recognition there is an advantage for item-specific processing (red)
21
as performance is best in both the pleasantness rating conditions.
Hunt & Einstein (1981) study
Conclusions
(1) In free recall it helps to relate the items with each
other. Trying to process each items by itself is not so
helpful. Perhaps because when retrieving one item it
helps you to retrieve others?
(2) In recognition, it helps to concentrate on the
individual properties of the items to produce the best
performance. This helps to differentiate it from
others?
(3) Note that actual performance in free recall is much
worse than in recognition. It is much more difficult
trying to retrieve information “out of the blue” than
judging if you have seen something before.
22
Semantic vs. episodic memory (Tulving,
1972, 1983, Schacter et al. (2001)
Semantic memory is about facts,
concepts and ideas;
whereas episodic memory is about
personal experiences and is
temporally dated. As the label implies,
episodic memory about the episodes
(i.e. autobiographical experiences)
that have happened to us throughout
life.
Evidence: Lefrancois (2000) KC, an
amnesiac, plays chess and knows that
he can, but cannot remember playing a
game.
23
The distinction has plausibility:
• Episodic, semantic and procedural
memories are interactive with one
another: if we are learning a new
language our first time we encounter a
certain word it is within our episodic
memory.
• Gradually its meaning assimilates into
semantic memory and we may forget
the first encounter. This makes for an
efficient system—the less useful
episodic information gets cleaned out
gradually. What is left resides in
semantic memory.
• Learning a new skill requires episodic
memory while we consciously master,
e.g. the gear positions in a car. But
again, this information eventually gets
cleared away. What is left resides
within procedural memory.
• This is a system that has a built-in
avoidance of redundancy, so that we
disconnect the context in which we
learned something. This is efficient in
terms of storage.
Tulving’s distinction
between episodic and
semantic memory
Semantic
Episodic
Procedural
24
Comparing episodic and semantic memory
Episodic
Semantic
1.
2.
3.
1.
4.
5.
It is organised in time.
It makes a reference to me.
I consciously retrieve events
in my life.
My memories are subject to
forgetting.
The context of what I
remember is important-to be
shown later.
2.
3.
4.
5.
Time is irrelevant—I can’t
remember when I learned
about Henry VIII.
It refers to my general
knowledge—it is not about
me.
These are ‘events’ that
haven’t happened to me
personally—I just know them.
This is permanent
knowledge—unless my brain
deteriorates or gets damaged.
The context of learning is not
important.
25
Episodic vs. semantic memory
However, Rutherford (2005) argues against making a case for
independent episodic and semantic memories, contrary to
the proposals of Tulving and Schacter. Although it is a
useful heuristic, he thinks that a better way would be to
think of semantic memory as ‘an abstraction of episodic
experience’. That is, we are continuously have episodic
experiences in everyday life and these continue to input
into some form of distillation of these experiences, that we
can refer to as semantic memory.
26
Episodic vs. semantic memory
There are contrasting ways to test semantic and episodic
memory:
Sentence verification tests semantic memory e.g. ‘a spaniel
is a dog’ true or false?
Sentence recognition tests episodic memory—where you test
if individual sentences had occurred earlier in the
experimental session.
27
Episodic vs. semantic memory
A study Rutherford cites to show this contrast between
episodic and semantic memory comes from Anderson and
Ross (1980):
Anderson & Ross found that the time to verify sentences (i.e.
using semantic memory) was affected by episodic
information that they gave them in previous sentences.
In the episodic part of the experiment people were given
sentences to learn (e.g. “A spaniel retrieves a ball”) and
later verified sentences (e.g. “A spaniel is a dog”).
Retrieval time in this later semantic part was affected by the
nature of the episodic information given beforehand.
If episodic memory was affecting retrieval time from semantic
memory, this suggests a relationship between them.
28
Retrieval in episodic memory
Retrieval – let’s examine further aspects of episodic
memory
How good is our ability to retrieve from LTM (or
how permanent is our storage)?
• It depends on the type of retrieval. Tulving &
Pearlstone (1966) showed that after presenting lists
of words from categories (e.g. furniture, cooking,
colours, etc) that if they used cued recall—by giving
the category names performance was 75% compared
with just free recall which produced 40%.
29
Retrieval in episodic memory
How good is our ability to retrieve from LTM (or
how permanent is our storage)?
• Willem Wagenaar (1986) kept a diary for 6 years
and put down 1 or 2 events a day => 2400 entries
altogether—without ever looking back. Each
given cue: who, what, where and when. Every
day he did 5 retrievals using 1, 2 or 3 of these
cues. After 5 years recall was down to about 30%
and when was least effective, probably because
least visual.
30
Retrieval in episodic memory
Example (invented)
Who: My Aunty Vera
What: I went with her to an art gallery during a visit
Where: The local art gallery
When: Monday, January 15, 1983
Question: Who did we encounter just outside the art gallery?
Answer: A nurse who’d spotted that she’d dropped a glove.
31
The permanence of semantic memory
• By contrast how well is information stored in
semantic memory – if we can make a distinction
with episodic memory?
• According to Bahrick (1984) semantic memory is
pretty stable. He found that 40% of Spanish
vocabulary was retained 50 years later.
• For this reason it has been called a ‘permastore”.
32
Retrieval in episodic memory
• Retrieval takes place when information held in memory
becomes available. There appears to be an element of
activation as part of this retrieval process.
• The encoding specificity hypothesis was introduced by
Tulving and Osler (1968). Ps were given a list of 24 cue-target
pairs (e.g. thimble-HAND).
• Then they had immediate recall in one of 3 conditions; (1)
thimble-HAND—same cue as at encoding; (Thus they were
given the cue word “thimble” and they had to say “hand” to be
correct.) (2) thumb-HAND—another weakly associated cue (3)
just free recall—ie no cue. Condition (1) was best relative to
(3), but (2) was ineffective.
• Thus these retrieval cues need to be stored at the time of initial
learning to be effective.
33
Retrieval in episodic memory
Transfer appropriate processing (TAP):
Morris, Bransford and Franks (1977) gave Ps words to learn under one of
two different processing tasks followed by one of two different tests:
• Two processing tasks:
rhyme: Does the word rhyme with…
semantic: e.g. Can you feed it?
• Two types of test
standard yes/no recognition test
rhyming test: e.g. Were you given a word that rhymes with cable?
Test
Encoding task
Standard
Rhyme
Semantic
83%
33%
Rhyme
62%
49%
34
Retrieval in episodic memory
Transfer appropriate processing (TAP):
Morris, Bransford and Franks (1977)
Test
Encoding task
Standard
Rhyme
Semantic
83%
33%
Rhyme
62%
49%
• The result was that the Ps did better in the tasks that matched (ie.
semantic encoding better in standard test [83 v 62] & rhyme better in
the rhyme test [49 v 33]) .
• This supports the TAP hypothesis. In other words, performance
appears to be best if the processes one undertakes while learning are
the same type of processes that one uses during retrieval. An
implication would be if you are going to be given an exam needing
recall, you would be better to revise by testing yourself using recall,
35
rather than by recognition or cued recall.
Retrieval in episodic memory
Transfer appropriate processing (TAP):
Morris Bransford and Franks (1977)
Test
Encoding task
Standard
Rhyme
Semantic
83%
33%
Rhyme
62%
49%
• A slight problem with the Morris et al. study is that
performance levels on the standard vs. the rhyming tests
are not matched—but matching is not likely to change the
interaction.
36
Retrieval in episodic memory
Encoding specificity
Another aspect is the importance of context during encoding. Grant et al. (1998) had Ps
read an article in noise or silence followed a test of reading comprehension under noise
or silence and found the following results…
The top figure shows the design for Grant et al.’s (1998) “studying” experiment. The
bottom shows the results of the experiment. Asterisks indicate situations in which study
and test conditions matched. This suggests an advantage if one studies in quiet 37
conditions.
Retrieval in episodic memory
Encoding specificity – state dependent memory
Eich & Metcalfe (1989) looked at the effects of mood in this
context.
Musical mood induction- they tested on mood several times
each session:
happy mood Ps – rated happier and more aroused
sad mood Ps- rated sadder and less aroused.
Later they studied a word list either in a happy state or in a sad
state.
E &M found that the best performances were either (1) when
they encoded while happy and were tested while happy, or
(2) when they encoded when sad and tested when sad.
Here’s the data…
38
Retrieval in episodic memory
The top figure is the design of Eich and Metcalfe’s (1989) “mood”
experiment and the bottom shows the results. Asterisks show when
encoding and retrieval sessions match.
39
Retrieval in episodic memory
Conclusion of encoding specificity – state dependent
memory
The concepts of encoding specificity and Transfer
Appropriate Processing (TAP) give importance to the
relationship between the processes involved in encoding
and retrieval. It seems that performance is improved
considerably if there is a similarity in information and
indeed a similarity in the mental state of the individual.
40
The cognitive neuropsychology of episodic memory
• Earlier we look at the case of HM who had amnesia because of bilateral
surgery to the hippocampus. This demonstrates how new episodic
memories need the hippocampus. Thus when this structure is damaged on
both sides of the brain, new procedural memories can be learned, but new
events (episodes) in one’s life cannot be retained.
• There is evidence that the prefrontal cortex (PFC) is also needed for the
establishment of new episodic memories. Patients have such damage due
to aneurisms or head injury (esp. car accident).
• When the PFC is damaged new information can be learned but it tends to
be disordered.
• Incidentally, there is also evidence (Nyberg et al., 2000) from studying the
PFC that women are better than men in retrieval from episodic memory.
41
The prefrontal cortex and episodic memory
• Janowsky et al. (1989) showed how patients with PFC damage could
recognise an object seen before, but could not say where or when it had
been seen.
• Nyberg et al.’s (1996) HERA model (hemispheric encoding/retrieval
asymmetry) is based on a meta-analysis of many neuroimaging studies.
The left PFC is more involved in episodic memory encoding than the right
PFC. On the other hand, the right PFC is more important for retrieval
within episodic memory.
42
The hippocampus and
episodic memory
• Although there is general
agreement that the
hippocampus is involved
in episodic memory, due to
the amnesic evidence,
there is controversy about
how long such memories
are stored for normal
people.
• Views range from the
hippocampus storing
information only for a
brief period before being
processed elsewhere to
episodic memories being
stored permanently in the
hippocampus.
43
Conclusions on the relationship between
episodic and semantic memory
• As a reminder, episodic memory is a ‘one-off’ learning situation. An
episode happens (e.g. I post a letter) and then it moves away in time,
perhaps repeated later, perhaps not. But only one occurrence is all that
is required.
• Tulving (2002): “An event happens, a person experiences it, memory
traces are laid down representing the event, the past vanishes and is
replaced by the present.”
• With semantic memory there could be several exposures, but not
necessarily. E.g. we can have a flash-bulb memory of “9/11”.
• When there is a repetition, there is an update and perhaps a modification
of semantic memory, if we learn something to add to this.
• Some think of episodic memories as transmuting into semantic memory
over the passage of time. This is done by making some generalisations
from these similar events, while the specific context in which these
encounters occurred is lost.
44
Conclusions on the relationship between
episodic and semantic memory
• Others think that episodic memories remain in that form, so what is
retrieved is still an episodic memory. Nevertheless they also help
semantic knowledge. Thus the destiny of episodic memories is NOT
inevitably to become semantic memories.
• There is a plausibility in the notion of an episodic-semantic distinction,
although many are skeptical (e.g. Rutherford, 2005). It feels like one
is undergoing different types of task when comparing what one ate in
the evening 3 nights ago (episodic) with retrieving the name of the
prime minister (semantic).
• But there is not strong evidence that there are two completely different
memory systems.
• At the moment, it looks like the controversy will continue…
45