Thema 4: Semantic Memory / Thema 5: Memory and The
Changing Brain
Baddeley, A., Eysenck, M. W., & Anderson, M. A. (2015). Chapter 7: Semantic memory and stored knowledge (pp.
165-193).
Semantic memory: an individual's store of knowledge about the world - conceptual/generalised
Episodic memory: memory for events occurring at a specific time in a specific place
Semantic versus episodic memory
Episodic memory is often associated by a sense of conscious recollection of the past - whereas semantic memory does not. They
share a lot of features but also have a lot of differences.
• Findings: separate systems
Research on brain damage suggests that the two types of memory are distinctly different from each other. Retrograde
amnesia for episodic memories often spans several years but for semantic memories only shortly before the onset of the
amnesia.
People with semantic dementia often suffer from severe losses of concept knowledge even though episodic memory and
most cognitive functions are relatively intact. It is associated with damage to the anterior frontal temporal lobes.
• Findings: interdependent systems
In functioning, the semantic and episodic memory often combine in an interdependent fashion - the same neural
network, including frontal, temporal and parietal areas, was activated during retrieval of semantic, episodic and
autobiographical memory. Other research also show performance on certain task to depend on both episodic and
semantic knowledge. Sensory and perceptual information can be used to recall recent autobiographical memories but not
remote ones. Recall of remote memories requires semantic knowledge to provide a framework or scaffolding facilitating
the retrieval of episodic information.
• Conclusions
In general, retrieval from episodic memory is more likely to involve conscious recollection than retrieval from semantic
memory. The strongest evidence comes from the study of brain damaged people - amnesic patients often have more
severe problems with long-term episodic memory than do patients with semantic dementia, but the opposite is the case
for long-term semantic memory. Yet, many long-term memories consist of a mixture of episodic and semantic memory.
Organisation of concepts: traditional views
• Hierarchical network model
The first systematic model of semantic memory was put forward by Collins and Quillian - with the assumption that
semantic memory is organised into a series of hierarchical networks. Major concepts are represented as nodes, and
properties or features are associated with each concept. The longer the concept and properties are separated, the longer
it takes to respond. The model is on the right lines in its claim that we often use semantic memory successfully, by
inferring the right answer.
o Limitations:
• Not only can the properties differ in hierarchical distance, but also in familiarity. It seems that when you
control for familiarity, the effect disappears.
, • They differ also on typicality - the typicality effect: if something is not typical, it takes you longer to respond.
This tells us that most categories are loosely defined.
• Spreading activation model
The spreading activation model of Collins and Loftus was designed to resolve the problems of the hierarchical model -
they argued that it was too inflexible. Instead, they argue that semantic memory is organised on the basis of semantic
relatedness or semantic distance. The length of the links indicates the degree of semantic relatedness (the closer, the
more related). When something gets activated, the activation spreads the strongest to the concepts closest to the
activated node. In this way, the model predicts the typicality effect, and the facilitation effect (semantic priming) - when
words are closely related.
So, activation spreads amongst semantically related concepts. It is more successful than the hierarchical model because it
is more flexible - which however does mean that the model does not make very precise predictions, making it difficult to
assess its overall accuracy. Other limitations: (1) the representation of a semantic concept in a single node is too simple,
and (2) the model implies that each concept has a single, fixed representation.
• Naming objects
There is also research suggesting that concepts are organised into hierarchies. 3 levels: (1) superordinate categories at
the top (like item, or furniture), (2) basic-level categories at intermediate level (like chair), and (3) subordinate categories
at the bottom (like easy chair). Generally there is a preference for using basic-level categories, because it gives the best
balance, it is the most general level for motor movements for interacting and it is usually first acquired in small children.
When you have expertise you often use subordinate categories, and when talking about faces. When you're familiar with
it, it is also often most quickly categorised at the subordinate level. It can also be quicker in superordinate levels than
basic level, since basic level requires more information
In general, categorisation is fasted at a superordinate level since it requires the least cognitive processing. Categorisation
at basic level is often preferred since it combines informativeness and distinctiveness, and categorisation at the
subordinate level is often preferred to basic level by those with expertise.
Using concepts
Concept representations have the following characteristics:
• They are abstract in nature and detached from sensory input and output motor processes
• They are stable; the same representations for different occasions
• Different people have generally similar representations
• Situated simulation theory
Barsalou argued that all the above assumptions are incorrect - he argues that the representation of any given concept
varies across situations depending on the individual's current goals and major features of the situation.
Concept processing would be influenced by the context or setting, and have perceptual or emotional properties
So, according to the situated simulation theory, how we use conceptual knowledge in everyday life often involves the
perceptual and motor systems, explaining why concept processing varies across situations depending on the individual's
goal. The way we process a concept depends on the situation and the perceptual and motor processes engaged by the
current task.
Limitations:
o Barsalou exaggerates the extent to which concept processing varies across situations (concepts have a stable core and
their structures are context-dependent).
, o According to Barsalou the perceptual and motor processes are central to understanding the meanings of concepts, but
it is also argued that these processes occur after the interpretation of the meaning
Concepts and the brain
• Concepts within the brain
Semantic memories are actually not just stored in a single place - they are stored in way more complex ways. Different
kinds of information about a certain objects are stored in different locations (visual/auditory). This is a feature-based
approach and in line with Barsalou's emphasis in the role of perceptual and motor features in concept use.
• Hub-and-spoke model
The spokes in the model consists of several modality-specific brain areas in which sensory and motor processing occur.
The six spokes shown in the figure relate to visual features, verbal descriptors, olfaction (smell), sounds, praxis (motor
information) and somatosensory information. In addition, each concept has a 'hub' - a modality-independent unified
conceptual representation that provides an efficient way of integrating knowledge of any given concept. It is assumed
these hubs are located in the anterior temporal lobes. Evidence from this comes from research on patients with semantic
dementia. From research it seems that semantic dementia causes blurring of the boundary separating members of a
category from non-members (they would find it difficult to judge atypical category members or non-members that look
like they would belong).
Category-specific deficits show that people can be impaired in specific categories - colour, taste, smell, visual motion, and
function.
The hub-and-spoke model provides a comprehensive account of semantic memory - there is considerable support for the
notion that concepts are represented in semantic memory by a combination of abstract core (hub) and modality-specific
information (spoke). There has been good progress in identifying the brain areas associated with the hubs and various
types of spokes.
Limitations:
o More remains to be discovered within the concept hubs
o How is modality specific 'spoke information' integrated with hub information
o No consensus concerning the number and nature of concept spokes
Schemas
Much of the knowledge we have stored in semantic memory consists of larger structures of information - schemas: an
integrated chunk of knowledge about the world, events, people or actions. They include scripts and frames:
• Scripts: deal with knowledge about events and the sequences of events
• Frames: knowledge structures referring to some aspect of the world, containing fixed structural information and slots
for variable information
• Schemas versus concepts
So now we have 2 major types of information in semantic memory; (1) abstract concepts, corresponding to individual
words, and (2) broader and more flexible organisational structures based on schemas and scripts.
Patients with semantic dementia often have more problems with accessing concept-based information, whereas
patients with damage to the prefrontal cortex have more problems with accessing script related information rather than
concept information. This is because scripts often have goal-directed qualities.
Research on brain-damaged patients provides the support for the distinction between knowledge of concepts and script
knowledge.
• Usefulness of schematic knowledge
Schematic knowledge is useful because it allows us to form realistic expectations. They allow us to make predictions.
They also play a role in reading and listening because they allow us to fill in the gaps and draw inferences. Next, they help
us prevent cognitive overload (with stereotypes for example).
• Errors and distortions
