Note: Some of the memory demonstrations that we will do in class
will be messed up if you have advance knowledge of them. I
encourage
you to preview the notes before class, but try to skip the parts about
the memory demonstrations.
I. Goals.
A. Where we are/themes.
B. Memory codes, encoding specificity, and
transfer
appropriate processing.
C. Distinctiveness.
D. Sum up memory so far. II. Where we are/themes.
A. Here are some situations:
1. What are some specific strategies to increase the amount I
remember after I study?
2. I study for hours and hours, but don’t seem to remember much
(especially long-term). Why isn’t studying working for me?
We’ll see if we can answer these questions.
B. Where we are. We’ve worked out a pretty complicated
model of the processing system. The original model had boxes for
sensory register, short-term memory, and long-term memory.
In our discussion of memory (both short-term memory and long-term
memory),
we added a few extra components to this. Now, we’re scrapping the
box approach and turning to processing. It’s not where you put
information
that counts, it’s what you do with the information.
Levels of processing: Material can be processed at a shallow
level (attention to physical features) up to a deep level (attention to
meaning). Generally, the deeper the processing the better your
memory
will be.
C. Themes.
1. Do we really need to separate long-term memory from short-term
memory?
2. What is deeper processing? What does it do to memory?
Top III. Memory codes, encoding specificity, and transfer
appropriate
processing.
A. Memory codes. Deeper processing is really just different
types of memory codes. Attending to meaning is different from
attending
to the sounds of a word. Generating a memory using each code
requires
different amounts of effort. Let’s have a demonstration of that.
Demonstration: We'll do a word list with tasks.
We'll repeat it, but I'll read the list.
This demonstration is based on Craik and Tulving (1975). They
had similar tasks and found a memory order that went from structural to
sound to meaning. So, we found that deeper is really more work
(different
from more time). In particular, when you changed the format of
the
word from the presentation modality (verbal, visual) it helped.
Meaning
processing is even better. Can we show that more work with
meaning
will lead to different amounts of memory? Yes. Here’s a
demonstration
with three kinds of semantic tasks.
Demonstration: I’ll present a list of words. There
will be three tasks. We'll look at the effect of different
semantic tasks.
Now, we can answer the first question. What strategies can I
use to remember more? Make your studying more effortful. To
the extent that you can access more prior information (like your self
concept)
the better you’ll do. To the extent that you mix in different
modalities,
the better you’ll do.
This version of deeper processing also suggests one reason for better
memory for deep processing. Deeper processing generates more cues
to use in recall. We saw earlier that more cues improves the
search
part of retrieval (the effortful part).
B. Encoding specificity. Another aspect of a processing
account is encoding specificity. The most effective retrieval
cues
will be the ones that you used during the original learning.
Here’s
the rule: The best retrieval cues will be ones that match the
cues
used to learn. For example, context can matter. Baddeley
had
divers learn some words on land and recall under water. He also
had
the reverse situation. When learning and remembering took place
in
different contexts, memory was worse than for the same context.
But,
for less extreme situations, these effects can be weak. One kind
of context is the other stuff you’re studying. If stuff that’s
studied
together is recalled together, that should be good (the encoding
context
matches the retrieval context). If they don’t match, that’s bad.
This was investigated by Thomson and Tulving (1970). They made
a list of word pairs that were strongly related (black-white) or weakly
related (black-train). They presented the words together for
learning.
For recall, if you learned with weak associates, memory was better if
you
were tested with the weak associates (present associate, recall related
word). If the strong associates were presented as cues when
people
learned with the weak associates, memory was worse. So, the
context
can screw you up if the retrieval context is different from the study
context.
This is true even if the retrieval context is really better (more
related).
CogLab: We'll look at the
results of our encoding specificity demonstration.
This also says a lot about what "forgetting" is. We've considered
interference and decay so far. We need to add retrieval
failure.
The information isn't gone at all, you just can't access it. Why
not? No cues. If you could construct the right retrieval
cue,
you would be surprised to find that the information is in there just as
it was when you could get at it. In other words, there's no such
thing as forgetting, it's a function of not being able to construct a
cue.
A good analogy is to think of the library. If a book is in its
right
place it's both available and accessible. If a book is
mis-shelved,
it's available in the sense that its contents are still in the library,
but it's not accessible because the usual retrieval cue (call number)
won't
be useful in finding it. If someone could come up with a proper
cue,
the contents of the book would be accessible again. (Maybe that's
why when you hear a song you haven't heard in a long time, a flood of
memories
comes back. I know I have summer and winter songs based on when I
first heard them.) Note that this is probably restricted to long
term memory. What we said about interference in short term memory
is still true, and interference probably plays a role in retrieval
failure
as well. Let's have a demonstration of retrieval failure (I found
this to be
pretty
stunning when I tried it on myself).
Demonstration: I’ll present a list of sentences.
You'll have three seconds per sentence, and you will be asked to recall
these later. Present sentences, recall. Now, here's a list
of retrieval cues. Write as many sentences as you can think of
based
on these cues. What you should find is that even though you
couldn't
recall a lot of the sentences (suggesting that they had been
forgotten),
once you have a cue they're right there.
Implication for studying: Think of retrieval cues when you're
studying. Make rich representations with lots of potential cues,
relate information in the notes in as many ways as possible so that one
topic could cue a different topic. Think about it like
this:
If you treated each sentence in the demonstration as an isolated
element,
then the cue for a sentence will only buy you that sentence.
However,
if you also made cross-associations (for example a sheet is a sail and
a boat is a shelter, so imagine some conglomeration of those things),
then
a cue to one sentence will buy you two or more sentences. In your
notes, enriching the encoding cues opens up new retrieval avenues and
will
help prevent retrieval failure. Note how this is a kind of deeper
processing.
Another issue to consider is cue overload. If you try to relate
too much information to a single cue, it won't be much use at
retrieving
any of the information. By expanding the connections and opening
up more cues, you can prevent that from happening.
C. Transfer-appropriate processing. Another aspect of a
processing account is transfer-appropriate processing. The type
of
deep task you use should be based on the type of retrieval task you’ll
be using. Two retrieval tasks you know about are recognition
(multiple
choice) and recall (essay). The best studying strategy for each
of
these is different. For essay, you want to organize and attend to
the overall structure of the material. For multiple choice, you
want
to pay more attention to details, and less to overall structure.
Morris, Bransford, and Franks (1977) tested this general idea.
They had two kinds of tasks and two kinds of tests. The tasks
were
meaning and rhyming. The tests were a standard recognition test
and
a rhyming test. Here’s a layout:
Encoding task
Standard recognition test
Rhyming test
Does _____ rhyme with legal (eagle)
63%
49%
Does _____ have feathers (eagle)
84%
33%
So, for a regular memory test, deeper processing was better (attending
to meaning beats rhyming). But, for a rhyming test (which
emphasizes
sound) memory is better if you attended to sound when learning (rhyme
beats
meaning).
One implication of this is the effect maintenance rehearsal will
have.
Theoretically, maintenance rehearsal should increase familiarity.
The more times you repeat something, the stronger its representation in
memory should become. However, this won't be much use in a recall
task because you need to think of a cue to access the item before you
can
decide if it's familiar. So, maintenance rehearsal will do
virtually
nothing to recall (we saw that in the last set of notes), but that's
just
part of the story. If you're doing a recognition task the test
itself
provides the cue, familiarity would be the only thing you'd need to
decide
if you'd seen it before. Since it's automatically computed,
maintenance
rehearsal should make recognition better. Glenberg, Smith, and
Green
(1977) investigated this. People did various amounts of
maintenance
rehearsal using a task similar to the one I used to get you to do
various
amounts of maintenance rehearsal. Then, they either recalled the
list or recognized it. More maintenance rehearsal did nothing to
improve recall, but more maintenance rehearsal did increase
recognition.
So, the process used worked OK for one kind of retrieval task, but not
for another.
The take-home message: Match the encoding (learning) processes
with the retrieval (test) processes to maximize recall. In spite
of Glenberg, et al. (1977), I would still advise that deeper processing
is better. More maintenance rehearsal will improve memory on a
recognition
(multiple choice) type task, but richer cues will still give you
more.
The real issue is do you want to learn it for the test, or do you want
to learn it forever?
D. Kind of elaboration. You want to make precise
elaborations
when you do deeper processing. So, try to emphasize the relevant
features in your studying. Example: “the fat man read the
sign
about thin ice” will help you remember “fat” better than “the fat man
read
the sign that was about two feet tall.” That’s why there are
differences
between meaning tasks. Some make more precise elaborations.
Anderson and Ortony (1975) presented some evidence that remembering
something is a constructive process that includes the context
surrounding
the information. For example, if people read Pianos can be
pleasing
to listen to, music would be a better recall cue for piano
than
if people read Pianos can be difficult to move.
Alternatively,
heavy is a better cue for the second sentence. The context
itself provides elaboration of the concept and determines what the
effective
cues will be.
Implication: Different nuances of the words will be cued by the
context, and it will take something like that context to retrieve those
nuances. You can control the nuances that get into memory and the
kinds of cues that are required by controlling elaboration during
studying.
E. Put it all together: Why is deeper processing better?
1. It leads to different memory codes. This might give
more retrieval cues.
2. The better the learning context and the retrieval context
match, the better. (The more effective those cues will be.)
3. The more the code generated at learning matches the code
needed
for retrieval, the better.
4. The more precise the elaboration, the better. (To ensure
that the appropriate cues are available, make the right elaborations.
Top IV. Distinctiveness. A last way that deeper
processing
helps is by increasing distinctiveness. Schmidt (1991)
described four kinds of distinctiveness.
A. Primary: If an item on a list is different from other
items on the same list, it is distinctive. (A red word on a list
of words in black ink.)
B. Secondary: If an item is different from other items
in long-term memory, it is distinctive. (Oddly shaped words like
“khaki” or “lymph”.)
C. Emotional: If a strong emotional context accompanies
an item, it is distinctive. (Space shuttle explosion.)
D. Processing: If a word receives a different kind of
processing,
it is distinctive. (Categorize vs. rhyme if you only rhyme one
word.)
Does deeper processing make words more distinctive?
Theoretically,
yes. There are only 26 letters, so there will be a lot of overlap
in a visual representation, which means lots of proactive
interference.
There are more sounds, so less overlap. There are even more
meaning
features, so that’s the best. When you construct a retrieval cue,
the less stuff it matches, the more useful it will be. If
encoding
makes items distinctive, then you’ll have fewer matches to each
retrieval
cue, and you’ll get better memory.
Take-home message: Make the items you study distinctive so you
reduce buildup of proactive interference.
Top V. Sum up memory so far. We have a pretty
complicated
set of memory facts on our plate. Let’s have one more
demonstration
to make it all come together.
Demonstration: I will present one more list of
words.
Memorize the whole list. Present at 2 second rate. After
last
word, recall. Now, let’s use the score sheet.