9 - conceptual knowledge Flashcards
what is conceptual knowledge
knowledge that allows us to recognize objects and events and make inferences about their properties
what is a concept
- mental representation of a category
- category of objects
what is a category
the set of all possible examples of a concept
- concepts provide rules for creating categories
-
what is categorization
the process by which things are placed in categories
why are categories called ‘pointers to knowledge’
because once something is placed into a category, we can begin to look at its unique traits rather than its normal ones
how does categorization influence our ability to take action
must be able to recognize something as a member of a class in order to properly engage with it
explain the definitional approach to categories
determining whether or not an object meets a category by looking at its list of properties and matching them up
- Wittgenstein did away with this
what is the family resemblance approach to categories
objects within a class share in some family resemblance, but may be very different in other was
what is the prototype approach to categories
category membership is determine by comparing the object to a prototype representative of the class
what is a prototype
- what is it based on
a typical or ideal member of a class
- Rorsch; the average of the members of the category (not an actual member but more of an abstractum
what is Rorsch’s notion of typicality
the degree to which an object is similar to the prototype
- high typicality means very close to the average
- Rorsch ran an exp. with birds and found that there were clear intutitions for this concept
is there a relationship between family resemblance and prototypically
yes, objects that share high family resemblance are also like, good candidates for being prototypical??
what is the sentence verification technique
- outcomes in prototypical terms
used by Smith et al to determine how fast ppl can answer questions about an objects category
- presented with a category statement ; an x is a y
- time to respond yes or n
- more prototypical cases have faster response times - prototypically effect
how is prototupicality related to item listing
more prototypical items are typically listed first in listing tasks
relationship between priming and prototypical items
Explain Rorsch’s findings
they are more affected by priming
- Rorsch; hear a prime (green) then two colours, had to indicate if they were the same or different
- three conditions
1. colours are the same, good prototypes
2. colours were the same but were poor indications
3. colours were different - priming results in faster ‘same’ judgements for the prototypical colours
- when ppl hear ‘green’ they imagine a ‘good’ green, so when they see one they can just send it
what is the exemplar approach to categorization
prototype with two differences
- there are more than one - there are many examples
- they are real members of the set - objects the person has seen before
compare the benefits of examplars and prototypes
- exemplars do better with atypical cases - there is a ‘penguin’ exemplar for birds, for example
- doesn’t discard useful information - exemplars fair better with variable categories like games
do we use prototypes or examplars irl
both, maybe?
- seem to av.g examplars into protos as we learn a category, then go back to exemplars
explain hierarchical organization
the fact that categories are organized hierarchically in tree like structures
what are the three levels of categories
- super ordinate / globe; (animal)
- basic (cat)
- subordinate / specific *(tuxedo)
what is unique about basic level categories
Rorsch; going above results in a major loss of information, but going down results in very little gain
- seen in listing tasks; basic level terms have more properties associated than global by a lot, but less than specific by only a little (3 vs 9 vs 10 items each)
are basic level categories universal
no, dependent on experience with the category
- more learning results in basic categories shifting down to the specific categories
what is Collins’ and Quillian’s semantic network approach
- computer model of human memory
- hierarchical tree-like structure
- store information at highest possible category to preserve the cognitive economy (no need for each bird to be able to ‘fly’, just say birds ‘fly’ and hold exceptions at the ‘penguin’ and ‘turkey’ nodes)
- not meant to mirror physiology
what are the predictions made by the semantic network approach (2)
- the time it takes for someone to retrieve conceptual information is dependent on how long they have to travel along the tree - (canarys are birds should be faster than canarys are animals)
- this prediction was right - spreading activation; activity should spread along the links from activated nodes, priming them
- found to be accurate by the lexical decision task
what are the main criticisms of the semantic network approach (3)
- can’t explain the typicality effect (canary is a bird and penguin is a bird would be predicted to be just as fast to recognize bc the distance between the two specific categories and the basic is the same, but not the case)
- questioned the concept of cognitive economy - evidence people store specific conceptual properties at the conceptual node
- Rips et al., - ranking animals as animals is faster than ranking then as mammals even tho mammals are technically more basic
what is connectionism
- what are connectionist models also called?
an approach to creating computer models for representing cognitive processes
- also called parallel distributed processing models bcc the propose concepts are represented by activity distributed across a network
wat are the components of a connectionist network
- input units- units activated by environmental stimuli
- hidden units, which connect input units to output units
- output units
- connection weights - high weights are very excitatory, vice versa
what does activation in a connectionist network depend on
- signal that originates in the input units
2. connection weights throughout the network
what is the basic principle of connectionism
a stimulus presented to the input units is represented by a pattern of activity distributed across the network
main differences between connectionist and semantic network approaches in representation
- connectionist - not hierarchical perse, concepts represented in the input, hidden, and property units in a network that are not nested in tree
- connection weights are very relevant
how do we train a connectionist network
- have to weight connections appropriately in order to have accurate outputs
- done through error signals, which happen when an output signal is wrong
- this back propogates throughout the network back to the input unit
- representation units (between input and hidden) provide information about how connection weights should be adjusted to fix the error
what are the (tfour) best pieces of evidence for connectionism
- resemble brain networks of representation
- we’ve developed networks that can simulate normal cognitive functioning like language processing, memory and cognitive dev.
- graceful degredation - the operation fo these networks is not entirely disrupted bye damage - similar to actual cases of brain damage
- can explain learning generalizations - similar concepts have similar patterns in the network, so training a system to recognize the properties of one concept can also help train other concepts
what are the four presented proposals about how concepts are represented in the brain
- sensory-functional hypothesis
- multiple factor approach
- semantic category approach
- embodied approach
explain the sensory-functional hypothesis to conceptual representation in the brain
a) what led to it
b) what this led them to think
c) explain the sensory functional hypth. as an outcome of a and b
a) ctegory-specific memory impairment due to encepnhalitis - lost the ability to identify one type object but retains the ability to recognize others (artefacts vs. animals)
b) artefacts are distinguished by their function while animals are distinguished by their sensory features
c) our ability to differentiate living things and artifacts depends on a memory system that distinguishes sensory attributes and anther system that distinguishes functions
what are three empirical limitations to the sensory functional hypothesis
- patients with sensory deficits who performed better identifying animals than artefacts
- patients that are poor at detecting mechanical artifacts but not other types
- patients w poor comprehension of small artifacts like tools, better for larger artists like vehicles
- general; artifacts are not just one group so this is overly simplistic
what is the central feature of the multiple factor approach
distributed representation - factors beyond sensory/funcitonal that determine how concepts are divided in categories
explain Hoffman and Ralph’s ‘experimental philosophical’ studies on the multiple factor approach and mechanical artefacts
- asked ppl about categories and how associated they are with colours, forms, motions ect.
animals ; sensory!
artefacts - functional. so far so good for the SF hypothesis - but mechanical divices and musical instruments overlapped with both categories
- ## middle ground
explain crowding and how it stands as evidenced for the multiple factor approach
- a factor that has been proposed to differentiate animals from artifacts
- animals tend to share a lot of properties while artefacts do not
- proposal; those w category specific impairments do not actually have category specific impairments, rather have trouble recognizing living things bc they can’t distinguish items with similar features
explain the semantic category approach
there exists specific neural circuits in the brain for some specific categories
- mahon and Caramazza; limited number of categories innately determined bc of their importance for survival
what is the evidence for the semantic category approach
- brain areas respond to specific types of stimuli (faces, places, bodies)
- Huth et al study on the category map in the brain
explain Huth’s category map study on abstract concepts
used a similar procedure to the first study, ut had them listen to stories rather than a film
- maps were very similar for the seven participants involved
does thee semantic category approach consider distributed representation
yes, concepts are mapped by a specific brain area (sometims) but also require information from a variety of networks
- think faces, need emotions, etc.
explain the embodied approach
our knowledge of concepts is based on the reactivation of sensory and motor processes that occur when we interact with the object
how are mirror neurons relevant to the embodied approach
it is these mirror neurons that are bnelieved to activate during conceptual thinking
explain Hauk et al.,s study on brain activity relevant to the embodied approach
- measured brain activity with fmri in 2 conditions
1. as participants moved a part of their body
2. as they read action words using these same body pats - activation is more extensive in actual movement, but modelled very similarly in space by the thinking task
what is the semantic somatotopy
the correspondence between the brain activity words related to specific body areas and the controlled movement of those parts
explain the arguments against the embodied approach
- Garcea et al - studied a patient who had a stroke that affected his ability to produce actions associated with obejcts
- should result in problematic identification of these objects
- but he did fine - bad at explaining abstract concepts
- responses to both of these, though
what is the overlap between the four theories of concepts in the brain
distributed representation
why is semantic dementia relevant to these theories
- results in a general loss of knowledge for all concepts due to anterior temporal lobe damage
- these other theories primarily consider graceful degredation to be the primary form
what as semantic dementia led to in the conceptual literature
the hub and spoke model of semantic knowledge
- areas of the brain associated with specific functions are connection to the anterior temp. lobe (damaged in semantic dementia) which serves as the integration hub for these other modalities
- functions include speech, valence, manipulation, etc.
what is the evidence for the hub and spoke model
- damage to one of thee specialized brain areas (spokes) causes specific deficits, while dmg to the ATL results n general deficits
what is transcranial magnetic stimulation
application of a pulsating magnetic field to specific brain regions that disrupt functioning in this specific area
- if we find specific functions damaged by this, the stimulated region is likely relevant for that function
what did Pobric et al find about the hub and spoke model using TMS/
jpreseented pictures off living things and artifacts to participants and measured response time for naming each one
- repeated procedure in two conditions
1. TMS applied to the ATL
2. TMS applied to an area of the parietal lobe typically activated during object manipulation - condition 1 slowed responses to bth, 2 only to objects
- condition 1 slowed responses for all object types. while 2 only slowed reaction for highly manipulatable artifacts but not for non-manipulatable ones
.is the ATL the only possible hub
no, there may be other ones
- ppl also argue that thee hubs are not that important, the pattern formed with the spokes is
