Lecture 15 - Working and Long Term Memory Flashcards
short term memory and working memory are ____ of the same kinds of things
different models
they are trying to explain the same kind of phenomena but they are different approaches
Short-term memory is the…
temporary storing and maintenance of
immediately useful information.
It is sometimes called primary memory
The Atkinson-Shiffrin Memory model (1968)
took material from sensory memory, brought it to a short term storage module so you could start using control process from attention and start doing some things (rehearsing, coding, trying to encode memories so they’re easier to retrieve later on from long term memory)
as you were working on things (remembering things in the short term) - this was the contents of your consciousness: all the things you’ve got in mind right now: where the responses come out of
In our last lecture, we began talking about how information is coded in short-term memory. One experiment found that consonants (e.g. MSTLJX) could be erroneously recalled (e.g. as MSVLKX) if participants spoke during rehearsal. This kind of interference was said to suggest:
STM may use phonological (verbal) encoding of these different consonants
sounds of the consonants were important
used later when you try to recall
some phonological encoding of the material initially (first presentation, that’s how you encoded it.)
Coding (also called encoding) establishes
the form of the representations with STM.
phonological similarity effect.
these characteristic errors you would make when repeating individual consonants or whole words (since you encoded them with that verbal internal voice you make errors with things that sound alike)
The errors made suggested acoustic confusions:
MSVLKX, where V sounds like T and K sounds like K.
Do we remember information in terms of how it looks, how it sounds, or some underlying meaning? This can be tested using….
interference effects.
Conrad (1964)
interference effects
presented subjects with a series of consonants (e.g. MSTLJX).
• Subjects read aloud some random digits for a time
and were then told to recall the consonants
• The errors made suggested acoustic confusions: MSVLKX, where V sounds like T and K sounds like K. This is the phonological similarity effect.
• Sounds of letters interfered with each other, suggesting phonological (verbal) encoding.
It was later found that STM could use visual and semantic codes.
• Wickens (1972)
- used a paradigm called release from proactive interference to show SEMANTIC CODING.
- Subjects are given three words from a category (e.g. fruits: apple, banana, pear). They also had a distractor task (e.g. count backwards by threes - so they can’t rehearse). Next, they are asked to recall the words.
- They are next given a second list of three words from the same category, with the same procedure.
- A third list from the same category is presented with the same procedure.
- Finally, they are given a fourth list that can either come from the same category or a new category. The countdown and recall procedure is the same.
proactive interference
something you’ve already learned in the past (recent past) will interfere proactively with something you’re trying to learn now
past info interfering proactively with new info
Results Wickens (1972)
The results suggests that STM could use semantic codes.
• Recall on the first three lists got progressively worse.
- as long as the list came from the same category they got worse and worse
- The older lists seem to proactively interfere with the related incoming information.
- When the list is from a new category, the percent correct goes back up.
- The relatedness seemed to moderate the improvement.
because you’re activating in memory “fruits” , it became more and more difficult to know what you had just seen but when given something completely different (wasn’t getting proactive interference from same semantic category)
when you’re encoding something ____ is important, because you’re not just encoding individual items but you’re looking for the relatedness of those items and that’s extending in time
meaning, semantics
It seems that representations can also take the form of CHUNKS of semantic information.
• George Miller (1956)
• did a review of people’s abilities to process information from a variety of domains (tones, words, letters, smells, etc.). He found that people could consistently recall about 7 items, +/- 2.
• He suggested that the limits on STM are not physical cues,
but chunks of meaningful information.
Chunks?
are based on the application of top-down, prior
knowledge and will vary by individual expertise.
people have different ways of chunking semantic info cause it’s based on what you know, how you make sense of the environment
e.g.
Fourscoreandsevenyearsago = 25 letters, nearly impossible.
Four score and seven years ago = 6 words, difficult if scrambled.
“Four score and seven years ago” = One famous phrase, easy
–> as long as you have the right background
We have limited capacity in STM
When studying, task switching may use up some of this capacity (you are remembering to go
back to the other task).
So, no distractions!
Keeping things available in STM can depend on…
…semantic links (e.g. chunking). Other information can act as distractors and push the relevant information out.
If you memorize individual facts,
you are increasing the
processing demands. If you combine items semantically as chunks, you make storage (and retrieval) easier
_____ can take up short-term memory and cause a
decrease in performance (Beilock, 2010).
stress and anxiety
An interesting finding from Miller’s 1956 work was that…
…the mode (most frequently found number) for number of items recalled seemed to be 5.
most often people could remember 5 chunks
Nelson Cowan (2001) points out that
• the number of items in
STM seems to be closer to four semantically distinct items .
• While Miller was originally concerned with how many items could be temporarily stored, Cowan and other are concerned with how many items can be used and manipulated.
- problem solving, creativity - short term memory is a work space
• This (and other shortcomings) led to a reconceptualization of
STM by Baddeley & Hitch (1974) as working memory (st. that is a workspace in memory).
Working memory
is an alternative to short-term memory.
a workspace in memory
working vs. short terms
Whereas short-term memory was primary conceived as a short-term
storage and encoding space, the working memory model describes how you manipulate current and long-term memories.
how do you change things?
how do you bring stuff from long term memory and mix it in with stuff you’re experiencing right now?
Baddeley & Hitch (1974) originally envisioned three components:
1) central executive - monitoring, updating, rerouting
2) phonological loop - verbal
3) visuospatial - object and location
Baddeley & Hitch (1974)
visuospatial sketchpad
visual world, things you’re seeing
Baddeley & Hitch (1974)
The phonological loop
is the subsystem of working memory
that maintains and processes auditory information.
It has two components:
1) Phonological store
2) Articulatory control process
Phonological store:
This can hold speech-based information for
about 2 seconds.
the immediate auditory memory - the stuff that you’re bringing over from sensory memory and that’s what you’re working on: you’re internal voice
- you can also read stuff and encode it has auditory info
Articulatory control process
This is the inner, subvocal inner voice that allows you to rehearse and refresh the items in the phonological store.
Phonological similarity effect
Letters or words that sound similar are often confused. (e.g. V sounds like T, mop sounds like cop).
word length effect
Since the phonological store is only 2 second long (on average), it can be difficult to remember longer
words.
− Wash, tap, bit, stop, pot, zip (remembered better)
vs.
− Carwash, tapping, bitten, stoplight, potholder, zipper.
(take longer to rehearse in internal voice)
In general, we have worse recall for longer words because…?
they can’t be rehearsed as much in the 2 second window.
The phonological loop can be studied through the use of…
articulatory suppression.
• Task 1: Read the following list, then turn away and recall as many
words as you can.
- e.g.Applesauce, mouthwash, carwash, theatre, treehouse, saxophone
• Task 2: Read the following list while repeating “the, the, the…” out loud. Then turn away and recall as many as you can. (producing some phonological info)
- suppressing the loop with your own production
-e.g. Football, milkshake, tumbleweed, bluebird, residence, hospital
• With articulatory suppression, the second list is more difficult because
speaking our loud interferes with the phonological loop.
what is the phonological loop for?
• May be important for early language acquisition
–> articulatory work, rehearsal
phonological loop in language comprehension?
as you’re talking to someone or listening to someone, the words are on a loop and you can monitor those words and start pulling out the meaning?
NOOOOO
you don’t need the loop to monitor speech in real-time
articulatory suppression also seems to eliminate
the word length effect (e.g. shorter words have no advantage because they can’t be rehearsed).
The visuospatial sketchpad
part of working memory
involved with visual imagery –
the creation of visual images once the physical stimulus is no longer available.
could be something that you look out and see something and you have a mental image of it (representing the physical stimuli currently there)
OR
you’re asked “what did your fifth grade teacher look like?” (re-creating an image of physical stimuli that are no longer there)
Shepard & Metzler (1971) experiment with mental rotation tasks
- measured reaction times to decide whether pairs of objects were the same or different. This is called a mental rotation task. Objects could have a difference between 0 and 180 degrees.
- Press A if same, B if different.
Shepard & Metzler (1971) results
found that reaction times increased as the differences in angular rotation got larger.
the greater the angular separation, the more time it took in the visuospatial sketchpad to take that image and move it
shows that it takes effort and takes capacity that can be measured in time differences
The central executive
Baddeley model?
fulfills the function already discussed in the attention section. However, in the Baddeley model, it’s
primary importance is in the manipulation of information.
• Information taken from visuospatial sketchpad and phonological loop
is reviewed and ordered.
• **Relevant background knowledge from long-term memory is retrieved.
- “what is this like? what have i seen before? what’s relevant that i’ve experienced in the past with the things i’m experiencing now?”
• But how does this happen? Where is this combined information stored?
Baddeley (2000) updated his model to include an
episodic buffer
• This is something with a larger capacity system that takes information from longterm memory and combines it with incoming data.
- glueing all the things (LTM, current stuff, visuospatial sketchpad, phonological loop)together
- (maybe where binding takes place)
• It can reconstruct an ‘episode’ that extends in space and time (e.g. some current event).
• It is multi-dimensional,
combining information from
different senses.
fluid systems
- visuospatial sketchpad
- episodic buffer
- phonological loop
you’re manipulating them, they’re changing, they’re in flux,
crystalized systems
- visual semantics
- episodic LTM
- language
these things don’t change as often
e.g. think about your fifth grade teacher: her name or look is less likely to change
what is an episodic buffer?
key difference?
reconstructing an episode
taking all the stuff “right now”
in multidimensional space
**key difference: also from long-term memory
memory in the moment.
• Short-term or working memory models take limited information from a larger ‘sensory memory’.
• Computational complexity is reduced by using attention to
primarily focus, enhance, and bind the most relevant stimuli.
• Within STM/WM, we maintain and control/manipulate real-time
information.
semantic links
important for encoding info at first and retrieving info later
chunking!
making meaningful links between what’s happening now and what’s happened in the past
Computational complexity is reduced by
using attention to
primarily focus, enhance, and bind the most relevant stimuli
Within STM/WM
we maintain and control/manipulate real-time information.
Short-term or working memory models take limited information
from a larger ‘sensory memory’.
there’s more out there in sensory memory then we can use
these take this info and takes into a bottleneck
reducing computational complexity!!
What about long-term memory?
• These are memories that are maintained for days, months, and
years. They are your ‘top down’ information.
- There are many distinctions/types of LTM.
- How do we encode, store, and retrieve in this different space?
Baddeley & Hitch (1974) proposed their working memory model…
… as an alternative to Atkinson’s short-term memory.
primary difference between WM and STM?
The primary difference is that WM focuses on the manipulation and
use of information, while STM was mainly storage.
The phonological loop is made up of the ____
and _____.
phonological store
articulatory control process
The phonological loop can hold about _____
worth of verbal information. This coding and capacity limitation
leaves it vulnerable to the ____ and
_____.
2 seconds
phonological similarity effect
the word length effect
The visuospatial sketchpad allows you to maintain
visual imagery
The visuospatial sketchpad is measured with the _____
mental rotation task.
To make use of the central executive as a manipulator of information, an ___ was added.
episodic buffer
episodic buffer
This is a multidimensional
space that combines sensory data with long-term
memories to aid in organization and planning.
