Memory Flashcards
Learning and memory involve a series of processing stages:
- Encoding of information into memory;
2.Storage of information within the memory system; - Retrieval of stored information from memory.
Atkinson & Shiffrin’s (1968) multi-store model,
consisting of modality-specific sensory stores, a short-term store of very limited capacity and a long-term store of essentially unlimited capacity capable of holding information over time.
Explain sensory store
Sensory store can be subdivided into the iconic store and the echonic store
The iconic store is the brief sensory store for visual information. George Sperling (1960) conducted a
number of experiments to understand the iconic store, for example involving brief presentation of 12 letters
in a grid array. Typically, participants could only report 4 or 5 of the letters correctly, but claimed to have seen
many more. In a “partial report” condition, following removal of the visual array, Sperling prompted
participants to report either the top, middle or bottom row of letters. Participants were able to report most of
the letters from the requested line, but only if the delay between removal of the array and presentation of the prompt was ~1 s or less. This suggests that information in iconic storage decays in less than a second.
An analogous sensory store for the auditory modality is the echoic store, which was investigated by Anne
Treisman (1964). She presented participants with an auditory message to one ear and asked them to repeat
the message back aloud while ignoring a second message being presented to the other ear. If the second, ignored, message was actually identical to the first but started at a different time, participants only noticed they were the same if they started within 2 seconds of each other. This suggests that the persistence of unattended information in the echoic store is ~2 s, otherwise information decays.
However, in apparent contradiction of Broadbent’s model, Treisman (1960) found that participants in shadowing
experiments tended to switch channels whenever the attended side was continued briefly on the
unattended side.
Explain short-term store
Atkinson & Shiffrin’s distinction between short-term and long-term stores was influenced by William James,
who distinguished between the “psychological present” and the “psychological past”. According to Atkinson
& Shiffrin, the short-term store contains information currently held “in mind”, and is of limited capacity.
George Miller (1956) investigated the capacity (or “span”) of short-term memory by asking subjects to recall digit strings. Typically, participants could recall strings correctly up to a length of 7 ± 2 digits. This result held for other stimuli like letters, and even for words, from which Miller suggested that short-term memory holds ~7 integrated units of information (termed “chunks”).
Information can be retained in short-term memory by rehearsing it (either out loud or sub-vocally).
According to Atkinson & Shiffrin, the longer an item is held in the short-term store, the greater likelihood
should be of long-term storage. To investigate this claim, Rundus (1971) presented participants with a list of 20 words and asked them to rehearse the list out loud. When participants were subsequently asked to recall the words, the more frequently a word had been rehearsed, the more likely it was to be recalled on the word recall test. The exception to this was the last few words in the list, which always had a high likelihood of recall irrespective of the amount of rehearsal. This is known as the recency effect (see below).
Interference
Peterson & Peterson (1959) studied the duration of short-term memory by asking participants to remember a
3-letter stimulus for a few seconds while counting backwards in threes. The ability to remember the stimulus
diminished rapidly, suggesting that information decays from short-term memory within a matter of seconds.
However, Waugh & Norman (1965) manipulated the speed with which digits that were to be remembered
were presented to participants and found that digit recall was (more or less) unaffected, suggesting that shortterm memory forgetting is due to interference from exposure to additional information, rather than the passage of time.
What is separation of Short-Term and Long-Term Stores (recency effect and primary effect)
As noted above, the recency effect refers to the observation that the last few items in a list are often much
better remembered than items from the middle of the list. Atkinson & Shiffrin attributed this effect to the
last few items still being present in the short-term store from the end of list presentation. Glanzer & Cunitz
(1966) showed that the recency effect could be eliminated if participants counted backwards prior to recall,
supporting the link with short-term memory. Earlier items in a list also tend to be better remembered than
those in the middle – the so-called primacy effect. Atkinson & Shiffrin proposed that these items are
recalled from long-term memory. Consistent with this proposal, Glanzer & Cunitz found that the primacy
effect was unaffected by counting backwards after list presentation.
Studies of patients with brain lesions have also been used as support for the separation of short-term and
long-term memory. For example, Scoville & Milner (1957) reported patient HM, who suffered medial
temporal damage and had impaired long-term memory but normal digit span. HM exhibited a preserved recency effect but a much reduced primacy effect. In contrast, patient KF, reported by Shallice & Warrington (1970), suffered parieto-occipital lobe damage and had intact long-term memory but very poor digit span. KF’s primacy effect was preserved, but there was no discernible recency effect.
Criticisms of the Multi-Store Model
There have been a number of criticisms of Atkinson & Shiffrin’s model in the years since its publication. One
criticism is that, according to the model, processing in the short-term store is required for encoding into longterm memory. However, patient KF, with defective short-term memory (digit span) but preserved long-term
learning and recall, provides evidence against this claim. As noted above, the multi-store model also predicts
that the longer an item is held in the short-term store, the greater the likelihood of long-term storage.
Although Rundus (1971) showed that rehearsing items in the short-term store does correlate with long-term
retention, other factors such as depth of processing (Craik & Tulving, 1975) are far more important for
determining whether information will subsequently be remembered after a delay of a few minutes or more.
Another assumption of the multi-store model is that the short-term and long-term stores are unitary,
operating in a single, uniform way. However, Warrington & Shallice (1972) reported that patient KF had
worse short-term memory for auditory letters and digits than for visual stimuli, suggesting that there may be distinct short-term memory stores for different kinds of material. Similarly, Baddeley & Hitch (1974), using dual-task methodology, found that auditory rehearsal of digits did not affect the number of errors made in a concurrent grammatical reasoning task. This was interpreted as suggesting that there may be a distinction between an auditory-verbal short-term store and a central information processing system involved in reasoning.
Explain the working memory model
In an attempt to address the limitations in the multi-store model, Baddeley & Hitch (1974) proposed the
concept of working memory. This highly influential model, which has been revised in recent years
(Baddeley, 2000), comprises 4 primary components: an auditory-verbal phonological loop for short-term
storage of speech-based information; a visuo-spatial sketchpad for short-term storage of spatial and visual
information; a multimodal episodic buffer which holds and integrates information from the phonological
loop, visuo-spatial sketchpad and long-term memory; and a modality-free central executive, responsible for
selecting and initiating cognitive processing routines.
WORKING MEMORY
- Phonological loop – a slave system for the temporary retention of spoken verbal material such as a
phone number - Visuospatial sketchpad – a slave system for the temporary storage and manipulation of spatial and visual
information, such as the location of a phone number on the page of a phone book - Episodic buffer – a limited capacity system for integrating phonological and visuospatial representations
with information from long-term memory - Central executive – a modality-free processing system involved in coordinating the operation of the other
systems for performing demanding cognitive tasks
The functional separability of the core Working Memory components was determined primarily on the basis
of numerous dual-task experiments. For example, Robbins et al. (1996) examined the involvement of working
memory components in chess move selection by testing the effect of several concurrent tasks: rapid word
repetition, sequential key pressing and random number generation. The key pressing task (which involved the
visuo-spatial sketchpad) and the random number generation task (which involved the central executive)
affected the quality of chess moves selected, whereas the word repetition task (which involved the
phonological loop) had no effect. Thus, Robbins et al. concluded that chess move selection involves the
central executive and visuospatial sketchpad, but not the phonological loop.
What is a phonological loop?
The results of the experiments described in the previous lecture support the existence of a speech-based temporary store, but two key empirical findings provide insights as to its structure. One of these is the
phonological similarity effect. Baddeley (1966) found that serial recall of a list of phonologically similar
words (such as FEE, HE, KNEE, etc.) was significantly worse than from a list of phonologically dissimilar
words (such as BAY, HOE, IT, etc.), whereas visual or semantic similarity had little effect on recall. This
suggests that speech-based representations are used in storing the words, and that recall requires
discrimination between memory traces, which is more difficult for similar phonological representations.
The second finding is the word-length effect (Baddeley et al., 1975), in which recall of a list of long words
(such as OPPORTUNITY, ALUMINIUM, etc.) is typically worse than recall of a list of short words (such as
WIT, SUM, etc.). The word length effect is illustrated in the fact that mean digit span differs across cultures:
for example, digit span in speakers of Chinese (in which digits are relatively quick to say) is typically
significantly greater than in speakers of Welsh (in which digits take longer to say). Baddeley et al. (1975)
confirmed that the word-length effect depends on the phonological loop by asking participants to silently
mouth digits (articulatory suppression) during presentation and recall of words. This manipulation eliminated
the word-length effect, suggesting that phonological storage capacity is determined by rate of rehearsal.
Baddeley (1990) drew a distinction between a phonological store and an articulatory control process. The phonological store is concerned with speech perception, whereas the articulatory control process is linked to speech production that gives access to the phonological store. By this account, the phonological similarity effect can be attributed to confusions between similar representations in the phonological store, and the word-length effect can be attributed to the time taken to rehearse longer words via the articulatory control process.
Visuospatial Sketchpad
The visuospatial sketchpad is used for the temporary storage and manipulation of spatial and visual
information, including visual imagery. Because of the history of word list learning experiments in the last
century, less research has been done on visuospatial working memory. What has been done has primarily
explored the effect on different tasks of concurrent visuospatial processing. For example, Baddeley et al.
(1975) asked participants to encode material using either rote verbal learning or an imagery-based strategy.
When this task was combined with pursuit rotor tracking (tracking a moving light), performance using the
imagery-based strategy was disrupted, whereas performance using the verbal strategy was unaffected.
Pursuit rotor tracking involves visual perception as well as spatial localization. To assess whether both of
these factors are important, Baddeley & Lieberman (1980) repeated the previous experiment, contrasting
specifically visual (making brightness judgements) and specifically spatial (pointing at a moving pendulum
while blindfolded, guided by an auditory tone) concurrent tasks. Learning using the imagery-based strategy
was most clearly disrupted by the spatial concurrent task. On the basis of data such as this, Logie (1995)
argued that visuospatial working memory could be divided into two components (similar to the distinction
within the phonological loop). The visual cache passively stores information about visual form and colour
and is subject to decay and interference by new visual information. The inner scribe processes spatial
information and allows active rehearsal of information in the visual cache. This distinction is supported by
neuropsychological data. For example, Beschin et al. (1997) reported patient NL, who had preserved
perceptual skills but could not describe details of a scene from memory. Similarly, Farah et al. (1988) reported
patient LH, who performed better on spatial processing tasks than on visual imagery tasks.
Episodic Buffer
The phonological loop and visuospatial sketchpad permit temporary storage of modality-specific kinds of
information, but various findings are difficult to explain with such a model. For example, Baddeley et al.
(1984) noted that articulatory suppression does reduce memory span for visually-presented material (e.g., from
~7 to ~5 items), but does not eliminate it as would be predicted by the phonological loop model.
Furthermore, patients with severely impaired short-term phonological memory, with an auditory span of only
1 digit, can typically recall up to 4 digits with visual presentation. In addition, Chincotta et al. (1999) studied
memory span for Arabic numerals (1, 2, 3, etc) and digit words (one, two, three, etc), finding that participants
used both verbal and visual representations in performing the task. These findings suggest that verbal and visual information must be combined and stored somewhere in working memory. A further issue is that
memory span for meaningful sentences can be as much as 15-16 words, vastly exceeding normal phonological loop capacity (Baddeley et al., 1987). The traditional explanation for this is that information from long-term memory is used to integrate words into meaningful ‘chunks’. However, Baddeley & Wilson (2002) found that densely amnesic patients, with grossly impaired long-term memory, can exhibit normal immediate sentence span. Similarly, Baddeley (2000) described a densely amnesic patient who was able to continue playing bridge, keeping track of all the cards that had been played.
On the basis of such data, Baddeley realised that the working memory model needed a new temporary storage system that could allow verbal and visual codes to be combined and linked into multi-modal representations.
Evidence for a distinct, multi-modal short-term store also comes from neuroimaging research. Prabhakaran et
al. (2000) asked participants to perform a working memory task that required the temporary retention of
integrated verbal and spatial information. Activation in right frontal cortex was greater for retention of
integrated rather than modality-specific information, consistent with an episodic buffer. By contrast, posterior regions exhibited material-specific working memory effects, consistent with the phonological loop and visuospatial sketchpad. Overall, the episodic buffer is a useful addition to the working memory model, but a detailed account of how it integrates information from the other components and from long-term memory is
currently lacking.
Central Executive
The central executive is conceived as an attentional system, the most important and versatile component of the working memory system. It has been likened to the central processing unit in a computer, utilised to
optimise performance in situations requiring the operation and controlled coordination of a number of
cognitive processes. However, it is also the component of working memory about which least is understood,
probably because its complexity resulted in a somewhat vague specification (“little more than a homunculus”; Baddeley, 1996). As a result, most of the research in working memory tackled more tractable problems relating to memory span, etc. However, Smith & Jonides (1999) identified the following major functions of the central executive: switching attention between tasks, planning sub-tasks to achieve a specified goal, selective attention to certain stimuli while ignoring others, and updating and checking the contents of other working memory stores.
One task Baddeley used to study the central executive is random generation. Although this task sounds easy,
it quickly becomes difficult to avoid stereotyped, non-random patterns (repetition, letters in alphabetical
order, acronyms like CIA, BBC, etc.). Baddeley (1996) asked participants to hold 1-8 digits in mind while
trying to generate a random sequence of key-presses on a key-pad. Randomness of the sequence decreased as the digit memory load increased, suggesting greater demands on a general-purpose, limited capacity central executive. Concurrently reciting the alphabet or counting did not affect performance, but alternating between letters and numbers (A-1-B-2-C-3, etc.) decreased randomness, suggesting that the rapid switching of attention between tasks may be one of the functions of the central executive. Impairment to the functions of the central executive is termed dysexecutive syndrome (Baddeley, 1996), and is traditionally associated with damage to the frontal lobes of the brain. Consistent with this view, D’Esposito et al. (1995) reported that dorsolateral regions of the frontal lobe showed greater activation under dual-task than single-task conditions.
Similarly, Duncan et al. (2000) identified lateral frontal cortex as the neural basis of “general intelligence”,
characterised as a specific system involved in control of diverse forms of behaviour.
LONG-TERM MEMORY
William James (1890) was one of the first to make the distinction between primary and secondary memory
stores. Primary memory was defined as information remaining in consciousness after perception: “the
psychological present”. In contrast, secondary memory referred to information about events that have left
consciousness: “the psychological past”. This distinction influenced Atkinson & Shiffrin’s (1968) multi-store model, which distinguished between short-term and long-term memory stores. In their model, the short-term store contains information currently held “in mind”, and has limited capacity. The long-term store retains information over time, and has essentially unlimited capacity.
The capacity of long-term memory may be theoretically unlimited, but in reality there must be a limit as there are a finite number of neurons and synapses in the brain (roughly 1015). Standing (1973) tried to locate the limit by testing memory for many thousands of pictures, but found remarkably good memory even for 10,000 or more.
Lindauer (1986) has suggested that capacity may be constrained by the rate of acquisition. He noted that humans can study a maximum of something like 100 stimuli per minute, which over a 70 year lifetime, he calculated would equate to a theoretical capacity limitation of ~3 billion stimuli.
The first rigorous experimental investigations of human long-term memory were conducted by Ebbinghaus (1885), who taught himself lists of nonsense syllables and repeated them until he could recite them correctly twice. He measured the time to learn the lists, and then the time saved when relearning them after various delays. He found that retention decreased with longer retention intervals, but that the rate of forgetting slowed down after the first hour or so.
Factors Affecting Long-Term Retention
There are a number of factors that can improve retention in long-term memory. One is practice which, as we considered previously, increases the likelihood of information being transferred from short-term to longterm memory. Pirolli & Anderson (1985) asked participants to practice sentences like “The sailor is in the park” and found that long-term memory retrieval improved as a direct function of the number of days of
practice. Another factor is the level of processing a stimulus receives, which Craik & Lockhart (1972)
showed has a considerable effect on its memorability. Thinking about a word’s meaning was associated with
considerably better memory than simply repeating it. Craik & Tulving (1975) went on to show that the depth
of processing the word receives is important, finding that semantic (meaning-based) processing led to better retention than phonological (sound-based) processing, which was better than perceptual (visual or auditory) processing. Anderson & Bower (1972) suggested that level of processing may be less important than the extent to which the to-be-remembered information can be related to associated information and previous knowledge. When they asked participants to elaborate on sentences (generating a continuation to the sentence, e.g. “The doctor hated the lawyer because of the malpractice suit”), memory for the sentence was
significantly better. Other factors demonstrated to be important include the degree to which to-beremembered information is organised in a systematic way, and the time over which the encoding of information occurs. Bahrick (1979) was one of the first to show that spacing learning, with increasing
intervals of time between study sessions, led to far better long-term retention. This suggests that “cramming” in the days before an exam may be less effective than going over the material repeatedly at different points during term.
Amnesia
Atypical forgetting can be a symptom of amnesia, often
associated with damage to the hippocampus and surrounding
structures in the medial temporal lobes. Patient HM (later
identified as Henry Molaison) was the most famous amnesic
patient who, in 1953 at the age of 27, had experimental
surgery that involved bilateral removal of his medial
temporal lobes to relieve symptoms of severe epilepsy.
Following surgery, HM’s epilepsy symptoms improved, but
he experienced the unfortunate side effect of “very grave,
recent memory loss”. On formal testing in 1955, HM
reported the date as 1953, and claimed that he was 27 years
old (Scoville & Milner, 1957). He was subsequently involved
in decades of experimental research, which found that he
was generally unable to remember people, names, words, or
objects, regardless of the sensory modality they were
presented in. In contrast, there was little change in HM’s
intellectual functioning, in his language comprehension,
perceptual function or reasoning abilities. His store of
semantic knowledge about the world and ability to learn new
motor skills appeared unaffected and, as we have seen
previously, his short-term memory was also intact. Evidence
from other patients suggests amnesia to be associated with
deficits both in remembering past events and imagining future experiences (Hassabis et al., 2007).
Atkinson & Shiffrin’s (1968) multi-store model assumed that there was a single, unitary long-term store. Since
1968, however, things have become rather more complicated. Recent instantiations (e.g., Squire, 1992)
consider that long-term memory comprises several separate memory systems.
Explicit and Implicit Memory
According to Schacter (1987), explicit memory typically requires conscious recollection of a previous
experience, whereas implicit memory is revealed when performance on a task is facilitated in the absence of
conscious recollection. These different memory systems tend to be studied using different memory tasks.
For example, if a participant is asked to study a list of words, explicit memory tasks might include recognition,
cued recall, or free recall. Implicit memory tasks might include fragment completion, word stem completion,
or degraded word naming.
Evidence has shown that manipulating various factors in a memory test will affect either explicit or implicit
memory more than the other. For example, Jacoby & Dallas (1981) examined the effect of different levels of
processing at encoding. Deeper levels of processing improved explicit memory (as previously shown by Craik
& Tulving, 1975), but had little effect on implicit memory. Jacoby & Dallas also examined the effect of
modality differences between study and test, finding that changing modality between study and test had no
effect on explicit memory, but significantly reduced implicit memory.
Evidence for the separation of explicit and implicit memory also comes from patients with amnesia as a result
of damage to medial temporal lobe. For example, on pursuit rotor or Gollin figures tasks, performance of
amnesics improves over trials, despite the patients not being able to recall having done the test before
(Warrington & Weiskrantz, 1970). Graf et al. (1984) found that patients with amnesia were impaired on all
the explicit memory tests, but showed no deficit on implicit memory.
Patients exhibiting a dissociation in the other direction are much rarer, but Gabrieli et al. (1995) reported one
such patient, MS, whose lesion affected occipital cortex. Consistent with the results Graf et al.’s study,
patients with medial temporal lobe amnesia were impaired at explicit but not implicit memory. Patient MS
performed as well as controls at explicit memory, but was severely impaired on the implicit memory task.
Taken together, these results suggest that dissociable brain regions support explicit and implicit memory and,
thus, that they are functionally separate memory systems.
Functional imaging studies support the view that that different brain areas are likely involved in explicit and
implicit memory. As we saw in the last lecture, Schacter & Wagner (1999) reviewed evidence reporting
activity in the medial temporal lobe during explicit encoding and/or retrieval. Conversely, Simons et al. (2003)
investigated brain areas associated with implicit memory (so-called “priming”), observing activation associated
with perceptual priming in a posterior region of the brain known as fusiform cortex.
Episodic and Semantic Memory
The distinction between episodic and semantic memory was first made by Tulving (1972). Episodic
memory is considered to be memory of specific events or episodes occurring in a particular time and place
(e.g., remembering what you had for breakfast this morning). Semantic memory, by contrast, is general
knowledge about objects, people, facts, concepts and the meanings of words, without awareness of where or
when the information was learned (e.g., knowing that breakfast is a meal that you eat in the morning).
Episodic memory is specific to each person, whereas semantic memory can be culturally shared. Episodic
memory often involves “mental time-travel”, thinking back to a specific moment in one’s personal past and
consciously “re-living” a prior episode as it was previously experienced.
As with the explicit/implicit distinction, manipulating various factors in a memory experiment affects either episodic or semantic memory more than the other. For example, Jacoby & Dallas (1981) found that different levels of processing at encoding improved episodic memory, but had no effect on semantic memory.
Providing stronger evidence, Jacoby (1983) reported a double dissociation, whereby reading a word out of context at encoding improved participants’ semantic memory, whereas generating the word themselves
improved their subsequent episodic memory.
Evidence from neuropsychological patients has also been used as support for the separation of episodic and
semantic memory, although this has proved to be a controversial issue. For example, patient HM, who had
extensive medial temporal lobe damage, showed a severe impairment on episodic memory tasks, but was able
to name pictures of everyday objects, define concepts, etc. Similarly, Vargha-Khadem et al. (1997) reported
children with selective hippocampal damage acquired early in life. All the children showed very impaired
episodic memory, but performed well on tests assessing semantic memory (e.g., vocabulary, etc.).
The Organisation of Memory Systems
On the basis of such data, Tulving (1998) concluded that episodic and semantic memory are separate, and
proposed a process-specific relationship between memory systems: serial encoding (S), parallel storage (P),
independent retrieval (I). The SPI model makes a number of predictions: preserved semantic memory should
be possible in the absence of a functioning episodic memory system (as demonstrated in the selective
hippocampal patients who are amnesic but can learn new semantic knowledge). However, one should not be
able to find patients who exhibit preserved episodic memory if their semantic knowledge is dysfunctional. In
other words, a double dissociation between episodic and semantic memory should not be possible.
However, patients with early progression of different forms of dementia can show evidence of a double
dissociation between episodic and semantic memory. Patients with early Alzheimer’s Disease have medial
temporal lobe atrophy, whereas patients with early Semantic Dementia have atrophy that is most apparent in
lateral temporal lobe areas. With progression of the disease, both kinds of patients eventually have atrophy in
both temporal lobe areas, and exhibit impairment in both forms of memory (any many other cognitive
functions), but in their very early stages, selective impairments can be seen. Patients with early Alzheimer’s
Disease resemble amnesics, with impaired episodic memory and preserved semantic knowledge. Patients with
Semantic Dementia suffer profound degradation of semantic knowledge (Hodges et al., 1992), but less
apparent episodic memory impairment. For example, Simons et al. (2002) asked patients to perform a
semantic task that involved identifying drawings that are associated with each other (e.g., a pyramid is
associated with a palm-tree rather than a fir-tree). The patients’ episodic memory was later tested for the same drawings. A double dissociation was observed between impaired episodic and preserved semantic memory in Alzheimer’s Disease, and impaired semantic and preserved episodic memory in Semantic Dementia.
If patients with Semantic Dementia have dysfunctional semantic knowledge, what information are they using
to perform the episodic memory task? Graham et al. (2000) proposed that perceptual information might be
able to feed directly into episodic memory. To test this Multiple Input hypothesis, they compared episodic
memory in patients with Semantic Dementia for perceptually identical and perceptually different pictures.
Episodic memory was normal when the perceptual representation was the same at study and test (e.g., the same picture of a phone as was encountered in the study phase), but just as impaired as in Alzheimer’s
Disease when the perceptual representation could not be relied on (e.g., identifying a different picture of a
phone as was encountered in the study phase). Graham et al. argued that episodic memory typically relies upon multiple inputs from perceptual and semantic systems, and, in the absence of meaningful semantic input, perceptual information alone can be sufficient to support successful episodic memory.
SPI model and Multiple input model
However, patients with early progression of different forms of dementia can show evidence of a double
dissociation between episodic and semantic memory. Patients with early Alzheimer’s Disease have medial
temporal lobe atrophy, whereas patients with early Semantic Dementia have atrophy that is most apparent in
lateral temporal lobe areas. With progression of the disease, both kinds of patients eventually have atrophy in
both temporal lobe areas, and exhibit impairment in both forms of memory (any many other cognitive
functions), but in their very early stages, selective impairments can be seen. Patients with early Alzheimer’s
Disease resemble amnesics, with impaired episodic memory and preserved semantic knowledge. Patients with
Semantic Dementia suffer profound degradation of semantic knowledge (Hodges et al., 1992), but less
apparent episodic memory impairment. For example, Simons et al. (2002) asked patients to perform a
semantic task that involved identifying drawings that are associated with each other (e.g., a pyramid is
associated with a palm-tree rather than a fir-tree). The patients’ episodic memory was later tested for the same
drawings. A double dissociation was observed between impaired episodic and preserved semantic memory in
Alzheimer’s Disease, and impaired semantic and preserved episodic memory in Semantic Dementia.
