Behaviour and Environment 3
Mental representations
”Any mental content or operation that stands for something else in the world” (Payne & Cameron,
2013)
Examples of mental representations: Categories, exemplars, symbols, mental images,
memories, truth, values, probabilities, schemas etc.
Function of mental representation
”For the most part we do not first see and then define, but define first and then see” Lippmann,
(1922) Mental representations help us to understand the world around us. We use mental
representations and then observe the world around us.
Classification E.g. Classifying humans or objects
Additional attributes E.g. What can we do with this glass of water? It’s healthy to drink
water
Steering attention and interpretation E.g. When you are thirsty, and drinking is on your
mind, you’ll be able to detect glasses of water faster
Communication E.g. Communicating about specific objects or people
Thinking
Availability & Accessibility
Availability: The idea that all kinds of knowledge is available to us. It’s available within
everybody’s head. At the same time, only a part of the available information is accessible.
Accessibility: “Accessibility can be defined as the activation potential of available
knowledge.” Higgins (1996)” Some of the available knowledge is also accessible.
Example: A research done by Macrae et al. 1995.
They studied whether activating mental representations of a Chinese woman is either being
Chinese or is being a female.
They showed participants either a video of a Chinese women putting on make-up (as an
example of something that’s relevant for females) or a video of a Chinese women eating with
chopsticks (stereotypical for Chinese people)
After watching the videos, the participants performed a
decision task. The task is designed to measure the
activation of activated concepts.
Hypothesis: After watching the video of a Chinese
woman putting on make-up, the ‘make-up’
representation will become more accessible.
Results: (see figure). After watching a make-up video,
participants were faster in categorizing words that were
related to women. At the same time, they were slower
in categorizing women-related words in the chopstick-
related condition/
How is knowledge stored? Different models for mental representatinons
(Payne, 2013): Theories of mental representations are always about metaphors, e.g. Bins or nodes.
Associative network models
Schema models (Predictive coding)
Connectionist models
Multiple format models
Embodied cognition
Situated cognition
,Associated network models
This type of models is based on the idea that the mind works like a computer. For each association,
there are nodes which can become active.
The nodes have varying levels of activation based on prior experience and current conditions,
and the links vary in their strength.
Influential for understanding accessibility and priming effects.
Parallel search models: activation spreads automatically among connected nodes.
People are unaware of the nodes themselves or the spreading among them, but they
become aware of the concept that results from nodes that have reached the threshold of
consciousness. Difference between subthreshold activation and suprathreshold activation.
Serial search models: rather than automatically spreading activation, there is an intentional search
following linked nodes until a concept is retrieved. The nodes and pathways are unconscious, but the
search process (e.g. Remembering someone by remembering how they look like) is conscious.
Representations are only momentarily implicit or explicit. The direction of attention determines what
is implicit or explicit.
Dual process model: uses both the parallel search (implicitly) and serial search (explicitly). The
implicit representations can become explicit by spreading activation that crosses a threshold
(quantitative) and intentionally performing serial search (qualitative distinction; different ways of
searching).
When I see coffee, all kinds of nodes can become active (e.g. Awake, caffeine). Activation is
facilitated for connections in the network that are more powerful (e.g. Tasty).
Associated network models are a-modal. Embodied cognition is modal (via sensory information).
Schema models
Perceivers “go beyond the information given”, Bruner (1957) Information is stored in an
abstract form.
Once a schema is operated, it operates as a lens through which you perceive the world
around you.
Directs attention, memory and judgement
Top-down approach: schemas are broad representations that structure and make sense of
psychological experiences. E.g. Activation of the black-hostility schema made people act more hostile
in line with the schema, and then interpret retaliatory hostility as due to the inherent hostility of the
interaction partner.
Schema theories suggest that knowledge structures are used implicitly in the processing of new
information, but not that the content of the schemas themselves are unconscious. Researchers often
,measure the contents of schemas using self-reports. The content is implicit or explicit only
temporarily, as attention is direct toward or away from any given aspect of the knowledge base.
Predictive coding
Predictive coding (also known as predictive processing) is a theory of brain function in which the
brain is constantly generating and updating a mental model of the environment. The model is used to
generate predictions of sensory input that are compared to actual sensory input. This comparison
results in prediction errors that are then used to update and revise the mental model.
Bayesian processes:
Priors affect perception We use prior knowledge in perceiving the world around us. Priors
are e.g. Coffee is tasty, being careful with hot coffee etc.
Posterior (‘comparison between perception and prior’) The prior about coffee at a new
specific bar is updated (when e.g. The coffee is not so tasty over there).
Bruner’s schemas can be interpreted as prior within predictive coding.
Connectionist models
Connectionism: parallel distribution processing
Nodes, that may vary in their level of activation
Facilitative and inhibitive links One node activates another
node. Just like in the associative network models. There are
also inhibiting links.
Concepts exist by means of dynamic interplay of distributed
elements A node has no semantic meaning. A whole
‘system’ of nodes and their including weights form a concept,
once activated. The representations are not static, but strongly activated by the
environment.
Input, hidden, and output elements
Connectionist models don’t assume that nodes have semantic meaning. Rather, representations are
distributed as emergent patterns across the entire set of connected nodes. When given a set of
inputs, the network eventually settles into a pattern that satisfies the parallel constraints of the
activated nodes and the weighted connections between them. Importantly, distributed
representations are not discrete because there are not distinct representations “stored” anywhere in
the connectionist network.
Different theories:
Local connection models “grandmother cells”—single neurons that represent specifi c single neurons that represent specifi c
things, places, or persons, such as your grandmother. These models are more similar to
associative network models than to connectionist models because these models abandon
the assumption of distributed pattern completion. They are distinguished by the operating
Principle of parallel constraint satisfaction.
Distributed connection models The distributed representations of connectionist networks
might therefore be described as efficient and unintentional.
See table!!
, Multiple format models
Multiple types of mental representations, e.g. The memory system model (each system within a
different part of the brain, mainly used within implicit bias):
Semantic memory system learns through more conceptual means that are implemented
by neocortical regions.
Affective memory system learns through mechanisms, including fear conditioning, that
are implemented by subcortical pathways in which the amygdala is central
Procedural memory system m learns through a network connecting the striatum and basal
ganglia to prefrontal cortex and motor regions
So: learning and behaviour are linked based on neurological pathways.
For all three systems, processing is presumed to be unconscious, though the outcome (e.g., a
semantic idea, an aff ective reaction, or a habitual response) may become temporarily conscious.
Carlston’s associated systems theory (AST; Carlston, 1994) posits four distinct processing systems:
verbal, visual, behavioral, and affective.
Different kinds of representations might be implicit to differing degrees. AST posits that all
lower level representations are permanently inaccessible to conscious awareness. Higher
level representations can all be made temporarily conscious, but the ease with which this
happens varies from system to system
