Transcript Stable aspects of a familiar situation

Lecture 8 – Visual Memory
1
How is new learning influenced by WWAK?
• Through associations between new material and
familiar material.
• But how do we develop associations?
• In particular, do we associate individual objects
in the world with other individual objects? Or
classes of objects with other classes?
Lecture 8 – Visual Memory
Example:
What do you think of when I say, “Cat?”
Probably, “Dog.”
But which dog?
A specific image? Or a general image which
captures some essential ‘dogness’?
2
Lecture 8 – Visual Memory
Dog vs. Dogs
In the next slide, notice that every dog is
different in some way. They vary in colour, in
posture, even in size a little bit.
Yet, there is some general ‘dogness’ on display,
as well.
That suggests two ways we can represent things.
3
Lecture 8 – Visual Memory
4
Lecture 8 – Visual Memory
5
Two ways to represent things
1. As a specific object, with great detail
• Useful for controlling specific responses (e.g.
picking up a given cup in a given location)
• Does not generalize well (e.g., handle on next
cup may be thinner than handle on this cup)
• Store long enough to execute response.
Lecture 8 – Visual Memory
Two ways to represent things
2. As a general object, with less detail
• Contains the information that may be
relevant in a similar but not identical situation
• E.g., are cups as delicate as soap bubbles?
How do dogs behave when you reach for the
milkbones?
• These we may store for years.
6
Lecture 8 – Visual Memory
Two ways to represent things
Each of these ways has its own purpose.
If you want to behave towards some object in the
world, you need a detailed representation of that
specific object – size, shape, location, texture…
If you want to learn a lesson for future reference,
you need a less detailed (visual) representation
which generalizes better.
7
Lecture 8 – Visual Memory
Three levels of detail
1. Detailed current representation.
2. Stable aspects of a familiar situation (e.g.,
your living room). Abstract representation, less
detail.
3. General aspects of visual form (e.g., what
does a sofa look like – not your sofa, but sofas
in general?). Least detailed
8
Lecture 8 – Visual Memory
Three levels of detail
In what follows, I will make these points:
• Representations at different levels of detail vary
in how much it is worth remembering them over
the long term.
• With more general representations, we use
imagery to improve memory either deliberately
(mnemonics) or automatically (dual-coding).
9
Lecture 8 – Visual Memory
Three levels of detail and their role in memory
1. Detailed current representation
• Melcher (2001)
• Simons & Levin (1998)
• Standing (1973)
2a. Stable aspects of familiar situation
• Method of loci
• Nickerson & Adams (1979)
2b. General aspects of visual form
• Paivio’s Dual Code theory
10
Lecture 8 – Visual Memory
Memory for detailed current stimulus
1. How much detail can we get into a stimulus
representation?
• For a short time – a lot
• Melcher (2001) – memory for scenes
accumulated over a series of brief views as
though scene had never been out of sight.
• Effect did not build up across days
11
Lecture 8 – Visual Memory
12
Lecture 8 – Visual Memory
13
Lecture 8 – Visual Memory
Memory for detailed current stimulus
Simons & Levin (1998)
• Change blindness – people don’t notice
changes in visual scene after brief interruption
(when the scene is of little importance)
• abstract expectations about a scene allow us
to focus on some small part that we use to
check for moment-to-moment consistency
14
Lecture 8 – Visual Memory
15
But what about Standing (1973)?
Standing (1973):
• Good performance on recognition of 10,000
faces seen in photographs over several days
• Recognition tested with a set of several hundred
of these photographs (plus foils)
• But did subjects store whole faces or just
salient bits?
Lecture 8 – Visual Memory
What do these studies tell us?
• We store detailed information about a visual
scene only on the scale of seconds/minutes
(long enough to execute a response).
• We may store over a long period (e.g., days)
just enough information to distinguish one
scene from another.
• Don’t use these repns. for mnemonics.
16
Lecture 8 – Visual Memory
Memory for stable aspects of a situation
Stable aspects of a situation worth remembering,
because likely to occur again
Because these aspects are well-learned, they can
be used to form associations to new material.
Method of loci
• associate to-be-learned material with familiar
locations, such as rooms in your house.
17
Lecture 8 – Visual Memory
Memory for stable aspects of a situation
However, stable aspects of a situation that never
influence a response are not worth remembering.
Nickerson & Adams (1979)
• US students were poor at discriminating linedrawing of real dime from fakes.
• How many maple leaves on a penny?
18
Lecture 8 – Visual Memory
19
What do these studies tell us?
Some stable aspects of a situation are meaningful,
and these are stored in LTM.
Other aspects are not meaningful and these are
not stored in LTM.
Memories that we keep around in the long term
can be used as the basis for associations in new
learning (e.g., through method of loci).
Lecture 8 – Visual Memory
20
General aspects of visual form
The most abstract forms are the most general –
they have the least detail in them, so apply most
widely to new situations.
Because they are general, such forms are highly
overlearned (frequently experienced).
Because they are overlearned, such forms can help
us remember experiences, even when we don’t
deliberately use mnemonics.
Lecture 8 – Visual Memory
General aspects of visual form
Paivio’s Dual Code theory:
• Least-detailed representation of any object,
situation, or event is a word that names it.
• For example, the word ‘dog’ captures that
essential dogness without capturing any of the
details that make a particular dog particular.
• Imageable words are more memorable than
words that name non-imageable things
21
Lecture 8 – Visual Memory
22
Paivio’s Dual Code theory
1. Distinguish between high-imageability words and
low-imageability words.
High imageability
Low imageability
Firetruck
Tree
Cake
Volcano
Duty
Idea
Truth
Love
Lecture 8 – Visual Memory
23
Paivio’s Dual Code theory
2. Compare recall for high- vs. low-imageability
words.
Paivio, Smythe, & Yuille (1968)
• Cued recall. Cue & target varied in imageability.
• Recall was best when both were high, worst
when both were low, intermediate for one of each.
Lecture 8 – Visual Memory
24
Paivio’s Dual Code theory
In method of loci, there is no necessary connection
between two associates (e.g., hamburger and
bathroom).
• connection must be made explicitly
In Paivio’s study, pre-existing associations (e.g.,
“tree” with image of tree) are used.
Lecture 8 – Visual Memory
Review – Memory for detailed representations
• contains significant detail
• supports responding in a dynamic world
• valid (accurate as a description of the world)
only on the time scale of seconds or minutes
• memory persists for a short time only
25
Lecture 8 – Visual Memory
Memory for stable aspects of familiar situations
• contains less detail, but still distinguishes
between individual objects/locations (e.g., my
sofa and yours)
• supports activities that don’t require dynamic,
constantly-updated information (e.g., planning
routes).
• supports method of loci
26
Lecture 8 – Visual Memory
Review – Memory for general visual forms
• general visual forms are highly-overlearned.
• as a result, they make good associates for tobe-learned material.
• Paivio: imageable words more likely to be
recalled, because automatically encoded using
both verbal and image representations.
27