Cognitive Maps: How the Brain Organizes Knowledge Ling 411 – 18
The Cognitive Map Hypothesis Hypothesis: Knowledge is organized in the cortex as maps Established (hence not hypothetical): The cognitive map of the body Primary motor and somatosensory areas The map of pitch frequency In primary auditory area Hypothesized: Conceptual Phonological
Properties of cognitive maps Established for somatic and frequency maps Local specificity Every cortical location has a specific function Adjacency Adjacent locations for adjacent functions Nearby locations for related functions Comes in degrees Hypothesis: these properties apply to all homotypical cortical areas all types of knowledge represented in the cortex
First step in exploring the hypothesis: Categories Understanding phonology Phonological structure is organized around phonological categories E.g., vowels and consonants, voiceless stops Understanding semantics Semantic structure is largely a matter of conceptual categories Understanding how categories work is the key to unlock the mysteries of semantics To understand how categories work we need to understand how the brain manages categorial information
What is a concept? Concepts vs. percepts Percept: one sensory modality Locations are known Auditory: temporal lobe Visual: occipital lobe Somatosensory: parietal lobe Concept: more than one sensory modality Higher level (more abstract) Locations, for nominal concepts: Angular gyrus (?)MTG
Types of Conceptual Categories Discrete Even integers Counties in Texas Radial Birds Vehicles Family resemblance Games Furniture Ill-defined Thought Mind
Phenomena associated with conceptual categories 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world 5. Categories and their memberships vary from one language/culture system to another 6. Categories influence thinking, in both appropriate and inappropriate ways 7. Subcategories, and sub-subcategories, in hierarchical chains
Phenomena associated with categories: 1 1. No small set of defining features (with rare exceptions) The feature-attribute model fails Works for some mathematical objects, but doesn’t apply to the way people’s cognitive systems apprehend most things Example: CUP
Phenomena associated with categories: 2 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries Example: VEHICLE Car, truck, bus Airplane? Boat? Toy car, model airplane? Raft? Roller skate? Snowboard?
Fuzzy Categories No fixed boundaries Membership comes in degrees Prototypical Less prototypical Peripheral Metaphorical The property of fuzziness relates closely to the phenomenon of prototypicality
Phenomena associated with categories: 3 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members Prototypical CAR, TRUCK, BUS Peripheral: AIRPLANE, TOY CAR, RAFT, ROLLER SKATE, etc. Varying degrees of peripherality
Prototypicality phenomena The category BIRD Some members are prototypical ROBIN, SPARROW Others are peripheral EMU, PENGUIN The category VEHICLE Prototypical : CAR, TRUCK, BUS Peripheral: ROLLER SKATE, HANG GLIDER
Phenomena associated with categories: 4 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world In the world, everything is unique lacks clear boundaries changes from day to day (even moment to moment) Whorf: “kaleidoscopic flux”
Phenomena associated with categories: 5 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world 5. Categories and their memberships vary from one language/culture system to another cloche(of a church) clochette(on a cow) sonnette(of a door) grelot(of a sleigh) timbre(on a desk) glas(to announce a death) English: French: bell
Phenomena associated with categories No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world 5. Categories and their memberships vary from one language/culture system to another 6. Categories influence thinking, in both appropriate and inappropriate ways B.L. Whorf Example: Racial profiling
Phenomena associated with categories No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world 5. Categories and their memberships vary from one language/culture system to another 6. Categories influence thinking, in both appropriate and inappropriate ways 7. Subcategories, and sub-subcategories, in hierarchical chains ANIMAL – MAMMAL – CARNIVORE – CANINE – DOG –TERRIER – JACK RUSSELL TERRIER – EDDIE Each subcategory has the properties of the category plus additional properties Smallest subcategory has the most properties
Beyond description to explanation How can we explain these phenomena? The answer this question depends on how our information about categories is represented in the brain The brain is where our linguistic and cultural knowledge is represented
Facts and a hypothesis that we can build on Fact: The brain is a network Composed, ultimately, of neurons Cortical neurons are clustered in columns Columns come in different sizes Each minicolumn acts as a unit Therefore a person’s linguistic and conceptual system is a network Hypothesis: Every word and every concept is represented as a sub-network Term: functional web (Pulvermüller 2002)
Properties of functional webs I: Functional Webs A concept is represented as a functional web II: Columnar Nodes Nodes are implemented as cortical columns III: Nodal Specificity Every node in a functional web has a specific function III(a): Adjacency Nodes of related function are in adjacent locations More closely related function, more closely adjacent
Property III(a): Adjacency Nodes of related function are in adjacent locations More closely related function, more closely adjacent Examples: Adjacent locations on cat’s paw represented by adjacent cortical locations Similar line orientations represented by adjacent cortical locations
Hypotheses concerning conceptual webs Hypothesis I: Extrapolation to Humans The findings about cortical structure and function from experiments on cats, monkeys, and rats can be extrapolated to humans Hypothesis I(a): The extrapolation can be extended to linguistic and conceptual structures and functions Hypothesis II: Hierarchy A functional web is hierarchically organized Hypothesis III: Cardinal nodes Every functional web has a cardinal node Hypothesis III(a): Each subweb likewise has a cardinal node
(Part of) the functional web for CAT V P A M C The cardinal node for the entire functional web T Cardinal nodes of the subwebs
Phenomena associated with categories 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Subcategories, and sub-subcategories, in hierarchical chains 5. Categories are in the mind, not in the real world 6. Categories and their memberships vary from one language/culture system to another 7. Categories influence thinking, in both appropriate and inappropriate ways REVIEW
How to explain? Description is fine, but its only a start Next step: Explanation How to explain? By answering the question of how categories are represented in the brain REVIEW
Phenomena associated with categories: No small set of defining features (with rare exceptions) Example: CUP More realistic alternative: radial categories 2. Fuzzy boundaries Example: VEHICLE 3. Prototypical members and peripheral members VEHICLE Prototypical: CAR, TRUCK, BUS Peripheral: AIRPLANE, TOY CAR, RAFT, ROLLER SKATE, etc. Varying degrees of peripherality These three phenomena are interdependent
How do radial categories work? Different connections have different strengths (weights) More important properties have greater strengths For CUP, Important (but not necessary!) properties: Short (as compared with a glass) Ceramic Having a handle Cups with these properties are more prototypical
The properties of a category have different weights T CUP MADE OF GLASS CERAMIC SHORT HAS HANDLE The properties are represented by nodes which are connected to lower-level nodes The cardinal node The threshold More important properties have greater weights, represented by greater thicknesses of lines
Activation of a category node The node will be activated by any of many different combinations of properties The key word is enough – it takes enough activation from enough properties to satisfy the threshold The node will be activated to different degrees by different combinations of properties When strongly activated, it transmits stronger activation to its downstream nodes.
Prototypical exemplars provide stronger and more rapid activation T CUP MADE OF GLASS CERAMIC SHORT HAS HANDLE Stronger connections carry more activation Activation threshold (can be satisfied to varying degrees ) Inhibitory connection
Explaining Prototypicality Cardinal category nodes get more activation from the prototypical exemplars More heavily weighted property nodes E.g., FLYING is strongly connected to BIRD Property nodes more strongly activated Peripheral items (e.g. EMU) provide only weak activation, weakly satisfying the threshold (emus can’t fly) Borderline items may or may not produce enough activation to satisfy threshold
Activation of different sets of properties produces greater or lesser satisfaction of the activation threshold of the cardinal node CUP MADE OF GLASS CERAMIC SHORT HAS HANDLE
Explaining prototypicality: Summary Variation in strength of connections Many connecting properties of varying strength Varying degrees of activation Prototypical members receive stronger activation from more associated properties BIRD is strongly connected to the property FLYING Emus and ostriches don’t fly But they have some properties connected with BIRD Sparrows and robins do fly And as commonly occurring birds they have been experienced often, leading to entrenchment – stronger connections
Phenomena associated with categories: 4 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world In the world, everything is unique lacks clear boundaries changes from day to day (even moment to moment) Whorf: “kaleidoscopic flux”
Phenomena associated with categories: 5 1. No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world 5. Categories and their memberships vary from one language/culture system to another cloche(of a church) clochette(on a cow) sonnette(of a door) grelot(of a sleigh) timbre(on a desk) glas(to announce a death) English: French: bell REVIEW
Phenomena associated with categories No small set of defining features (with rare exceptions) 2. Fuzzy boundaries 3. Prototypical members and peripheral members 4. Categories are in the mind, not in the real world 5. Categories and their memberships vary from one language/culture system to another 6. Categories influence thinking, in both appropriate and inappropriate ways B.L. Whorf Example: Racial profiling
These phenomena (4-6) are interrelated 4. Categories are in the mind, not in the real world 5. Categories and their memberships vary from one language/culture system to another 6. Categories influence thinking, in both appropriate and inappropriate ways B.L. Whorf Example: Racial profiling
Bidirectional processing and inference T CUP MADE OF GLASS CERAMIC SHORT HANDLE These connections are bidirectional Separate fibers for the two directions; shown as one line in the notation
Bidirectional processing and inference T CUP SHORT HANDLE Thought process: 1. The cardinal concept node is activated by a subset of its property nodes 2. Feed-backward processing activates other property nodes Consequence: We “apprehend” properties that are not actually perceived
Another hypothesis of Whorf Grammatical categories of a language influence the thinking of people who speak the language Can we explain this too in terms of brain structure?
Example: Grammatical gender Does talking about inanimate objects as if they were masculine or feminine actually lead people to think of inanimate objects as having a gender? Could the grammatical genders assigned to objects by a language influence people’s mental representation of objects? Boroditsky (2003)
Experiment: Gender and Associations (Boroditsky et al. 2002) Subjects: speakers of Spanish or German All were fluent also in English English used as language of experiment Task: Write down the 1 st 3 adjectives that come to mind to describe each object All the (24) objects have opposite gender in German and Spanish Raters of adjectives: Native English speakers
Examples: Key (masc in German, fem in Spanish) Adjectives used by German speakers: Hard, heavy, jagged, metal, serrated, useful Adjectives used by Spanish speakers: Golden, intricate, little, lovely, shiny, tiny Bridge (fem in German, masc in spanish) Adjectives used by German speakers: Beautiful, elegant, fragile, peaceful, pretty Adjectives used by Spanish speakers: Big, dangerous, long, strong, sturdy, towering
Results of the Experiment (Boroditsky et al. 2002) Raters of adjectives were native English speakers Result: Adjectives were rated as masculine or feminine in agreement with the gender in subject’s native language
Categories and the brain All of these phenomena associated with categories can be explained as inevitable consequences of the structure and function of the human brain
Phenomena associated with categories: 7 7. Subcategories, and sub-subcategories, in hierarchical chains ANIMAL – MAMMAL – CARNIVORE – CANINE – DOG –TERRIER – JACK RUSSELL TERRIER – EDDIE Each subcategory has the properties of the category plus additional properties Smallest subcategory has the most properties
How to explain? Perceptual Neuroscience Hypothesis I: Extrapolation The findings described by Mountcastle can be extrapolated to humans Hypothesis I(a): Extrapolation can be extended to linguistic and conceptual structures Why? Cortical structure, viewed locally, is Uniform across mammalian species Uniform across different cortical regions Cortical structure and function, locally, are essentially the same in humans as in cats and monkeys and rats Moreover, in humans, the regions that support language have the same structure locally as other cortical regions
Support for the extrapolation hypothesis Conceptual systems in humans evidently use the same structures as perceptual systems Therefore it is not too great a stretch to suppose that experimental findings on the structure of perceptual systems in monkeys can be applied to an understanding of the structure of conceptual systems of human beings In particular to the structures of conceptual categories REVIEW
Columns of different sizes Minicolumn Basic anatomically described unit neurons (avg 75-80) Diameter barely more than that of pyramidal cell body (30-50 μ) Maxicolumn (term used by Mountcastle) Diameter μ Bundle of about 100 continuous minicolumns Hypercolumn – up to 1 mm diameter Can be long and narrow rather than cylindrical Functional column Intermediate between minicolumn and maxicolumn A contiguous group of minicolumns
Functional Columns Intermediate in size between minicolumn and maxicolumn Hypothesized functional unit whose size is determined by experience/learning A maxicolumn consists of multiple functional columns A functional column consists of multiple minicolumns Functional column may be further subdivided with learning of finer distinctions
Columns of different sizes In order according to size Minicolumn The smallest unit neurons Functional column Variable size – depends on experience Intermediate between minicolumn and maxicolumn Maxicolumn (a.k.a. column) 100 to a few hundred minicolumns Hypercolumn Several contiguous maxicolumns
Hypercolums: Modules of maxicolumns A visual area in temporal lobe of a macaque monkey
Perceptual subcategories and columnar subdivisions of larger columns Nodal specificity applies for maxicolumns as well as for minicolumns The adjacency hypothesis likewise applies to larger categories and columns Adjacency applies for adjacent maxicolumns Subcategories of a category have similar function Therefore their cardinal nodes should be in adjacent locations
Functional columns The minicolumns within a maxicolumn respond to a common set of features Functional columns are intermediate in size between minicolumns and maxicolumns Different functional columns within a maxicolumn are distinct because of non-shared additional features Shared within the functional column Not shared with the rest of the maxicolumn Mountcastle: “The neurons of a [maxi]column have certain sets of static and dynamic properties in common, upon which others that may differ are superimposed.”
Similarly.. Neurons of a hypercolumn may have similar response features, upon which others that differ may be superimposed Result is maxicolumns in the hypercolumn sharing certain basic features while differing with respect to others Such maxicolumns may be further subdivided into functional columns on the basis of additional features That is, columnar structure directly maps categories and subcategories
Hypercolums: Modules of maxicolumns A visual area in the temporal lobe of a macaque monkey Category (hypercolumn) Subcategory (can be further subdivided)
Category representations in the cortex Hypercolumn Maxicolumn Functional column Sub-functional column Supercategory Category Subcategory Sub-subcategory
Hypothesis applied to conceptual categories A whole maxicolumn gets activated for a category Example: BEAR Different functional columns within the maxicolumn for subcategories BROWN BEAR, GRIZZLY, POLAR BEAR, etc. Adjacent maxicolumns for categories related to BEAR (sharing various features) I.e., other carnivores Similarly, CUP has a column surrounded by columns for other drinking vessels
Perceptual subcategories and columnar subdivisions of larger columns Nodal specificity applies for maxicolumns as well as for minicolumns The adjacency hypothesis likewise applies to larger categories and columns Adjacency applies for adjacent maxicolumns Subcategories of a category have similar function Therefore their cardinal nodes should be in adjacent locations
Support from patients with brain damage (from Rapp & Caramazza 1995) J.B.R. and S.B.Y. (905b-906a) Herpes simplex encephalitis Both temporal lobes affected Could not define animate objects ostrich, snail, wasp, duck, holly Much better at defining inanimate objects tent, briefcase, compass, wheelbarrow, submarine, umbrella Conclusion: cortical areas for conceptual categories
Additional support from cases of brain damage J.J. and P.S. (Hillis & Caramazza 1991) (906-7) J.J. – left temporal, basal ganglia (CVA) Selective preservation of animal concepts P.S. – mostly left temporal (injury) Selective impairment of animate category P.S J.J.
Two different patients with anomia Deficit in retrieval of animal names (Damage from stroke) Inability to retrieve words for unique entities (Left temporal lobectomy)
Two more patients with anomia Deficit in retrieval of words for man-made manipulable objects (Damage from stroke) Severe deficit in retrieval of words for concrete entities (Herpes simplex encephalitis)
What is it that determines location? Logical categories like ANIMALS vs. TOOLS/UTENSILS ? If so, why? Abstract categories based on cognitively salient properties?
Animals vs. Tools/Utensils? These two categories have been studied most extensively in the literature What is it that determines location? Observations: Most animals are known mostly in the visual modality Many tools and utensils are known largely in the somatosensory and motor modalities
Proximity principle and nominal concepts Supramarginal gyrus, angular gyrus, and middle temporal gyrus are all close to Wernicke’s area Angular gyrus occupies intermediate location between the major perceptual modalities Supramarginal gyrus especially close to somatosensory perception Middle temporal gyrus especially close to visual perception
Functional columns in phonological recognition A hypothesis Demisyllable (e.g. /de-/) activates a maxicolumn Different functional columns within the maxicolumn for syllables with this demisyllable /ded/, /deb/, /det/, /dek/, /den/, /del/
Demisyllables [di, de, da, du] F1 and F2 For [de] It is unlikely that [d] is represented as a unit in perception
Functional columns in phonological recognition A hypothesis [de-] A maxicolumn (ca. 100 minicolumns ) Divided into functional columns (Note that all respond to /de-/) deb ded den de- det del dek
Phonological hypercolumns (a hypothesis) Maybe we have Hypercolumn of contiguous maxicolumns for /e/ With maxicolumns for /de-/, /be-/, etc. Each such maxicolumn subdivided into functional columns for different finals /det/, /ded/, /den/, /deb/, /dem/. /dek/ N.B.: This is a hypothesis, not proven But there is indirect evidence Maybe someday soon we’ll be able to test with sensitive brain imaging
Adjacent maxicolumns in phonological cortex? ge- ke- be- pe- te- de- A module of six contiguous maxicolumns Each of these maxicolumns is divided into functional columns Note that the entire module responds to [-e-] Hypercolum
Adjacent maxicolumns in phonological cortex? ge- ke- be- pe- te- de- A module of six contiguous maxicolumns The entire module responds to [-e-] deb ded den de- det del dek The entire maxicolumn responds to [de-]
Learning phonological distinctions: A hypothesis ge- ke- be- pe- te- de- 1. In learning, this hypercolumn gets established first, responding to [-e-] 2. It gets subdivided into maxicolumns for demisyllables deb ded den de- det del dek 3. The maxicolumn gets divided into functional columns
Indirect evidence for the hypothesis Fits the structural organization demonstrated in monkey vision Cortical structure and function have a high degree of uniformity MEG is able to pick up different locations in Wernicke’s area for different vowels MEG can only detect activity of at least 10,000 contiguous apical dendrites (Papanicolaou) Requires perhaps at least 250 adjacent minicolumns The size of a maxicolumn or hypercolumn
Remaining question: The process of learning distinctions When a hypercolumn is first recruited, no lateral inhibition among its internal subdivisions Later, when finer distinctions are learned, they get reinforced by lateral inhibition Question: How does this work?
Inhibitory connections Based on Mountcastle (1998) Columnar specificity is maintained by pericolumnar inhibition (190) Activity in one column can suppress that in its immediate neighbors (191) Inhibitory cells can also inhibit other inhibitory cells (193) Inhibitory cells can connect to axons of other cells (“axoaxonal connections”) Large basket cells send myelinated projections as far as 1-2 mm horizontally (193)
Neural processes for learning Basic principle: when a connection is successfully used, it becomes stronger Successfully used if another connection to same node is simultaneously active Mechanisms of strengthening Biochemical changes at synapses Growth of dendritic spines Formation of new synapses Weakening: when neurons fire independently of each other their mutual connections (if any) weaken
Neural processes for learning A B C If connections AC and BC are active at the same time, and if their joint activation is strong enough to activate C, they both get strengthened (adapted from Hebb) Synapses here get strengthened
end