Week 8 - Semantic Memory Flashcards
Semantic Memory
semantic vs.. autobiographical (episodic memory)
Semantic memory
• Knowledge about the world
• What we know
Cognitive Neuropsychology – patients show
category specific problems
Hillis & Caramazza (1991)
• Task – name line drawings
• JJ - left temporal lobe and basal ganglia damage
– animals 91% correct
– other categories 20% correct
• PS – damage left temporal lobe and smaller
damaged areas in the right temporal lobe and
frontal lobes
– animals 39% correct vegies 25% correct
– other categories 95% correct
Category deficits = differences in
processing difficulty?
Greater number patients reported with living thing
category deficit
Possible explanation – living things less familiar and
usually more visually complex than non-living things
Processing Difficulty Hypothesis
Supporting evidence for this explanation:
Cognitive Neuropsychology patients
Normal participants – elderly and young participants –
living things harder even when items matched for
frequency, familiarity and prototypicality
Not all category deficits explained like this…
• JJ – greater problem with non-living than living things
• Number other studies demonstrated differences even
after controlling for processing difficulty between living
and non-living thing items
Some patients do show category specific deficits
Theories of Semantic Memory
Neural-structure principle Sensory/functional theory Domain specific hypothesis Correlated-structure principle Organised unitary content hypothesis Conceptual-structural account
Sensory/functional theory
Adheres to neural-structure principle
Information in brain segregated based on types of
information (perceptual vs. non-perceptual)
Information about an object represented in a
distributed fashion in brain depends on modality of
input
Allport (1985): each type of sensory information
represented in separate interconnected nodes
Nodes specific information: action-oriented
elements, kinaesthetic elements, visual elements,
tactile elements, and auditory elements
Sensory/functional theory
Warrington and Colleagues
Sensory/functional theory assumes
1) Organisation of semantic system based on modality
specific sub-systems
─ Visual/perceptual
─ Functional/associative
2) Naming living things – visual/ perceptual information
Naming non-living things – functional/ associative
information
Patient evidence in support of
Sensory/functional theory
Warrington & Shallice (1984)
• SBY – 75% correct non-living things – 0% correct living things • JBR – 94% correct non-living things – 4% correct living things – JBR – problem musical instruments, gem stones, metals, fabrics, food
Sensory/functional theory predictions
• 1) recognition of all living things depends
representations in the same semantic sub-system
(visual/ perceptual) therefore….
– Specific deficits within the living things category
should not be observed
• Patients show deficit for
– fruits/vegetables relative to animals
– Animals relative to fruits/vegetables
Warrington & Shallice - ok to have deficits outside the
living/ non-living things if these things have an
emphasis on the perceptual properties of an object
Fruits/ vegetables emphasis on colour
Sensory/functional theory predictions
2)
2) patients with category specific deficits (living vs..
non-living) will also have deficits for the modality or
type of information tapped into via the impaired
category
Problems providing information or knowledge about
visual/perceptual characteristics or
functional/associative characteristics
Initially patient data was consistent with the perdition
but more recent patient cases do not support this
prediction
Sensory/functional theory predictions
3)
3) patients with a deficit for visual/ perceptual or
functional/ associative knowledge should show a
categorical deficit that is most dependent on that type
of knowledge.
Living vs.. non-living things deficits.
Patients have shown greater deficit for visual/
perceptual knowledge than functional/ associative
knowledge but they do not show any performance
differences in naming living and non-living things.
Sensory/functional theory Argued that
Visual/ perceptual and functional/ associative
representations are interdependent so damage to
one would effect the other to some degree
Both living and non-living things have perceptual
and functional elements
Damage to one type of information will have some
type of impact on the other type of features
Evidence against the Sensory/functional
theory
Not well substantiated by patient data
Damage to perceptual or non-perceptual properties
damage categories most reliant on this type of
information
Not always the case
– Patient EW – problem with animals BUT ok fruits,
vegies, food, musical instruments
– Patients problems processing visual information
but do not have any category specific processing
deficit
Evidence against the Sensory/functional
theory
Lambon-Ralph et al (1998) – patient IW
– Select name from 5 choices given
– IW worse given perceptual than non-perceptual
information about an object
– IW – equivalent performance for living and nonliving
things (no category deficit)
– When asked questions about items – better at
providing non-perceptual information and provided
greater amounts of non-perceptual information
about an item when asked to provide a definition
– Number patient cases no longer show the living
vs. non-living thing deficit when familiarity is
controlled
`Evidence against the Sensory/functional
theory
Normals
Flores d’Arcais et al (1984; 1985) two studies
• Priming word pairs perceptually but not conceptually
related
• Paintbrush-carrot = priming
Fundamentally flawed
Pecher et al (1998) – using correct methodology – no
priming for perceptually but not conceptually related
items e.g., pizza-coin vs. pizza- hotdog
Problems with the Sensory/functional theory
Sensory/functional theory of semantics cannot
account for patients with sub-category specific
deficits
Early supporting studies (normals and patients)
methodological problems
Domain-specific hypothesis
Assumes evolutionary pressures resulted in
specialized and functionally distinct neural circuits
that process perceptually and conceptually distinct
categories of objects
Semantic knowledge is organised into categories
(domains) that reflect evolutionary salient distinctions
Specialised neural mechanisms for recognising
and understanding certain categories of
knowledge
Categories are those which require rapid and
efficient identification for survival and reproduction
Animals, fruits/vegetables, conspecifics and possibly
tools important for survival.
Evidence to support domain-specific
hypothesis
Accounts for patient data sparing categories of animals,
fruits/vegies or body parts or patients with just one
category impaired
Consistent with developmental studies:
– infants distinguish between living & non-living things
– 9 month olds – animals vs. non-animals
Domain-specific hypothesis predictions
1) Assuming distinct neural systems dedicated to
animals, fruits/vegetables, conspecifics and tools
then the other systems cannot compensate for the
damaged neural system.
Data from 16 year old patient with poorer
performance for living than non-living things had
this deficit since age 1.
More patient cases show deficits for living than
non-living things
Patients with specific-category deficits
Domain-specific hypothesis predictions
2
2) No association between a deficit for a type or
modality of knowledge and a conceptual deficit for a
specific category.
Patients with category-specific semantic deficits
present with equivalent impairments to visual/
perceptual and functional/ associative knowledge.
Domain-specific hypothesis predictions
3
3) Perceptual stages of object recognition models
might be functionally organised by domain specific
constraints and this predicts category specific visual
agnosia despite intact early visual processes
Supported by patients who have equivalent
impairments in their performance on tasks testing
visual/perceptual and functional/ associative
conceptual knowledge for living things BUT visual
agnosia for living things
Problems with domain-specific hypothesis
Lack specificity how knowledge is represented within categories Selective deficits categorical organisation but approach does not explain how things are represented within each category – Plants – food or poison – Animals – attack or food – Tools – function, developed later
Organised Unitary Content Hypothesis
OUCH
• OUCH unitary amodal semantic (conceptual)
representation system
• Members of semantic category share attributes
(humans breathe, composed certain type of
matter)
• Core semantic properties of an object tend to
be highly intercorrelated (breathing, being
composed of water, etc.)
• Category comprises highly correlated concepts
because of overlapping features
• OUCH unitary amodal semantic (conceptual)
representation system
• Members of semantic category share attributes
(humans breathe, composed certain type of
matter)
• Core semantic properties of an object tend to
be highly intercorrelated (breathing, being
composed of water, etc.)
• Category comprises highly correlated concepts
because of overlapping features
1) how concepts are represented within semantic
memory
Members of semantic category cluster close
together in feature space
Within a category representation occurs within
semantic space and is lumpy
OUCH 2
Assumptions
2) Category specific patient deficits due to damage to
lumpy region(s) within semantic space
Brain damage affects category members because it
affects objects/concepts with similar properties
and these are stored in adjacent neural areas OR
Affects the highly correlated features of category
members
OUch can account for?
Can account for very severe deficits
e.g., problem with animal category is a problem
with the lumpy semantic space representation for
animals and associated features
Organised Unitary Content Hypothesis
(OUCH) predictions
• 1) OUCH predicts modality specific semantic effects
• optic aphasia – cannot name an object but patient
can mime the use of the object
– Fits with dissociation between word access to
semantics and object access to semantics and the
strong link between visual attributes of an object
and its use
OUCH prediction 2
• 2) OUCH predicts relative sparing of categorisation
performance
– Patients with impaired object naming or the
inability to name the attributes (features) of an
object can still name the object’s superordinate
category
Organised Unitary Content Hypothesis
(OUCH) predictions
3)
3) OUCH predicts category specific deficits
– Category impairments would occur due to
disruption of feature representations of items
within the impaired category.
Problems with Organised Unitary Content
Hypothesis (OUCH) predictions
- Type of featural representation not defined
- Ways information clusters together not specified
- Functional rather than neural approach
Conceptual-Structure account
Systematic relationship among properties of members
of a category e.g., animals, furniture, vehicles
– Bird: large, beak, wings, eats fish, flies etc.
• Relationship between features of concepts which links
them together into categories
Conceptual-Structure account 2
assumes
Assumes
• 1) living things have more shared features than
non-living things. In other words non-living
things have more distinct/ informative features
than living things
• 2) living things – biological information is highly
correlated with shared perceptual properties
(can see – has eyes) AND for artefacts
function information is highly correlated with
distinctive perceptual properties (used for
spearing – has tines)
Assumes
• 3) features that are highly correlated with other
features will be more resistant to damage than
features that are not highly correlated OR
disrupting access to a given feature will disrupt
access to highly correlated features
Conceptual-Structure account prediction
1a)
when brain damage is mild deficit for
living things
when brain damage is severe deficit for
non-living things as all that is left is the
highly correlated perceptual and functional
features of living things
Conceptual-Structure account prediction 1a
Moss et al (1998): distribution of distinctive vs.
shared properties differs between living and nonliving
things
living things will show impairment at any level of
damage, except severe since individual features
of animals are more unique and therefore more
likely to be damaged
Artefact impairment only occurs with severe
damage as the distinction between form and
function is strongly correlated and therefore
robust to brain damage
Category specific deficits based severity of damage
Conceptual-Structure account prediction
1b)
when brain damage is severe deficit for
living things as whole sets of intercorrelated
features are wiped out
when brain damage is mild deficit for nonliving
things (artefacts) as these concepts
have more informative/distinct features that
are wiped out at this level of damage
Conceptual-Structure account prediction 1b evidence
Devlin et al (1998):
living things have a large number of intercorrelated
properties and the degree of correlation between these
properties is higher for living than non-living things
severe brain damage impairment for living things as
severe damage would negate shared features
mild brain damage impairment for non-living things
because mild damage impairs the distinctive features
Category specific deficits based severity of damage
Evidence to support Conceptual-Structure
acco
Category specific deficits
Studies Alzheimer’s patients
Evidence to refute Conceptual-Structure
account
Alzheimer’s patient data – used to develop theories
not replicated
Cannot explain
• Selective deficit narrowly defined categories
– i.e., defuse brain damage – selective category
deficits such as fruits/vegies, body parts
Problem for assumption that the number and degree
of intercorrelated properties relevant to all living things
Degree of severity argument
• JJ and PS have similar levels of task performance
(assume similar degree damage), yet show a double
dissociation
Problems with Conceptual-Structure
account
• Theories ok with broad category deficits but have
problems accounting for fine-grained deficits
• Alternative views about degree of brain damage and
impairments to living vs. non-living things, views are
in opposition
• Theory not well supported by empirical studies and
patient case
Neuroanatomical basis of semantic memory
Patients with deficits for living things damage to
• Left temporal lobe
• Left and right temporal lobe
• Some cases right temporal lobe only
• Damage to frontal and inferior parietal areas
• Widespread brain damage
Neuroanatomical basis of semantic memory
Deficit artefacts
Left temporal lobe and basal ganglia
• Left temporal lobe
• Left frontal and inferior parietal areas
Neuroanatomical basis of semantic memory PET
Living things/animals:
• Inferior temporal lobe (bilateral or left hemisphere)
• Bilateral occipital lobes
Non-living things:
• Posterior middle and inferior temporal gyri
• Fusiform gyri of temporal lobes and left
inferior frontal region
• Lingual, parahippocampal gyri, middle
occipital gyrus and dorsolateral frontal regions
