ch. 5 Knowledge Representation

Where is "knowledge" in the brain?

- Prefrontal regions: associated with extracting meaning from pictures and sentence - LPFC dominant: verbal material - RPFC dominant: visual materials

The Analogue/picture metaphor theory

- a visual code is used that closely resembles the original state of affairs - visual images are "pictures in the head" - we know that they are not completly identical

propositional networks

- can remember components, then combine them into a whole

The propositional position

- visual imagery uses a symbolic code, may not be images at all. - visual images are descriptions constructed of abstract propositions

Example of pegword method

1. Create an image of memory items with another set of ordered cues 2. Create an easily recalled list of nouns (the ordered cues) 3. Then picture each memory item interacting with one of the nouns

the 2 important principles that underlie the method of loci technique.

1. Make info more organized 2. Ensures the info is processed more meaningfully and elaborately with visual imagery.

3 Mnemonic techniques

1. Method of Loci - Place items in familiar locations 2. Pegword Method - Interact items with learned cues (easy rhymes for example, one-bun) 3. Interacting Images - Have two images interact (learning language)

Implications of mnemonics

A keyword method for learning foreign-language vocabulary - Can convert foreign words to sound-alike native language terms - Can create meaningful connection between sound- alike and meaning Probably not the best way.

what is a propositional representation

A representation of meaning as a set of propositions. It exists as concepts, think of an image, we think of the concepts. Epiphenomenon, visualize what it looks like

part hierarchy

Additional to generalized hierarchy, schemas have part hierarchy that describes parts of that category

Perceptual Symbol System

All info is represented in terms that are modality specific and basically perceptual. Similar to Pavio's Dual Code Theory; info represented as combined verbal and visual codes

Conceptual knowledge is put together in ____.

Categories (one that we already know). - system to ease representation for communication and prediction. -they influence how we experience, encode, remember information.

fluid category boundaries

Category boundaries are fluid and depend on use.

Child studies on learning categories

Children assume that if something is part of a member of a biological category, it is an inherent part of all members of that category. This pattern is opposite with respect to actions. Children come to believe that all things in a biological category have the same parts (like pectin in apples) and that all things in an artifact category have the same function (like "holding" for cups).

Schema; slots & default values

Cluster of information - Abstractions from specific instances that can be used to make inferences about instances of the concepts they represent - People will infer that an object has the default values for its category, unless they explicitly notice otherwise. - slots: specify values of various attributes that members of a category possess. - default values: A typical value for a slot in a schema representation

Shepard (1967) Memory for Visual Info experiment

Compared memory for verbal and picture information. Found memory for verbal information is good, memory for visual information is nearly perfect.

Embodied Cognition fMRI study

Contribution of motor action and its connection to the environment. They recorded brain activation while people listened to verbs that involved the face, arm, or leg. Results: there was greater activation in the part of the motor cortex that would produce such an action.

Bower, Karlin, and Dueck (1975) "droodles" study

Dhowed participants "droodles" which they then had to redraw from memory - Some subjects were given a verbal explanation, others were not. - Subjects given verbal labels were better able to redraw the pictures - People's good memory for pictures is tied to their interpretation of these pictures (changed depending on the label given to the image).

Amodal Symbol System

Elements within the system are inherently non-perceptual.

Event concepts: Scripts

Event schemas that involve stereotypic sequences of actions. People use them to reason about prototypical events.

Pavio's Dual Coding

Experiment with nouns Noun pairs: - CC: both words concrete objects (e.g., book/table) - CA: one concrete, the second abstract (e.g., chair/justice) - AC: reverse of previous (e.g., freedom/dress) - AA: both abstract (e.g., beauty/truth) Results: CC = 71.3%; CA = 62.5%; AC = 46.0%; AA = 37.8%. Best predictor of memory is imageability of a word (i.e. abstract < concrete)

T or F: people tend to always agree on typical members of a category and people judgements about a-typical categories to not vary greatly.

False. There is a lot of variation. Some disagree and some even change their minds from instance to instance (people are also unsure within themselves).

Artificial animals PET study

Found that people sometimes use rule-based abstractions -prefrontal cortex- and sometimes use memory-based-occipital and cerebellum- instances to represent categories. Hipocampus when exemplars were using memory Conclusion: People sometimes use abstraction and other times use instances.

NIckerson and Adams (1979) Failure of visual memory

Given a bunch of pennies, only one is the real penny. Thought that since people handle pennies all the time the task should be easy. People struggled. We encode certain features, but not minor details Meaning will override style, coins are all about the meaning.

Example of method of loci

Given a list to remember. Make an association between each word and different locations in a house (or a familiar place).

Rosch typical member of categories experiment 2

Had participants generate sentences for a category (ie. bird). Then the category was replaced by either a typical member (ie. robin), less typical member (ie. eagle), or a peripheral member (ie. chicken). Results: participants rated sensibleness of typical member more than less typical, and peripheral very low. Results indicate that participants had the typical member in mind when generating the sentences.

Bower scripts experiment

Had participants study stories that included some typical events from a script. Participants then had to recall the stories or recognize them. Participants tended to recall statements that were part of the script but had been left out of the story. Similarly seen with recognition.

Although an ostrich cannot fly, we still recognize it as a bird.

It can accomodate exceptions to the general rules.

What is the downside of using categories?

It can lead to stereotyping. Stereotypes lead to prejudice (action).

Fluid Scripts

Just like categories are fluid, so are scripts. Research findings indicate that new events are encoded with respect to general schemas and subsequent recall is influenced by the schema.

Reality of schema "office" study

Left participants in an office, then asked them to describe it. Results: Participants made default assumptions of what was in the room (Eg. desk, chair, books on shelves, etc.). People will infer that an object has the default values for its category, unless they explicitly notice otherwise.

verbal meaning vs. style results

Memory for meaning is equally good whether people are warned or not. Style did not really mean anything, but does not mean we can never remember style because warning (cueing) did have an effect on memory for the stylistic change.

Mirror Neurons

Neurons that fire both when an animal is performing an action and when it observes another animal performing this action. -implications for Autism

are biological and artifact categories represented in the same brain regions?

No. Damage to temporal lobes showed deficits in biological categories. Damage to frontoparietal regions are impaired in artifact categories

Bransford & Johnson (1972) Importance of meaning to memory laundry passage experiment

Participants had to read a passage and remember it, but it had no meaning (no context). People could not remember anything very well. Then told the context of the passage was "doing laundry". People were much better at remembering the story.

Dunning et al. stereotyping experiment

Participants read sentences like: "Elizabeth was not surprise when receiving her math SAT score." and " Bob was not surprise when receiving his math SAT score." Participants were more likely to falsly believe they had hear "low score" in the first case and "high score" in the second sentence after controlling for sexist attitudes showing implicit gender stereotyping.

Wanner study on remembering exact wording

Participants warned/not warned about exact wording memory test. - Measured ability to remember meaning versus style of the sentence and effect of warning participants - Memory is better for changes in wording that result in changes of meaning than for changes in wording that result just in changes of style., whether they were warned or not. - People normally extract the meaning (deeper brain structres) from a linguistic message and do not remember its exact wording. - People are actually bad at listening to style, better at recalling based on meaning. - Warning did have an effect on memory of the stylistic change, those who were warned remembered styled changes better than those who were not warned.

Rosch typical member of categories experiment 1

Participants were asked to identify category of a presented picture. They were faster to judge typical rather than a-typical pictures. Eg. Faster to judge apples as fruits than to judge watermelon as fruits. Typical members of a category seem to have an advantage in perceptual recognition too.

What do people do for concrete nouns?

People spontaneously make images for concrete nouns. Imagery varies with concreteness. Concrete nouns are dual coded while abstract nouns are only coded verbally. The first noun in the dual coding task acts as a conceptual peg on which the second noun is hooked (imageability of the first noun is critical- explains difference in CA and AC)

Labov cups vs. bowls experiment

Presetented participants with various ambiguous cup/ bows that gradually increased in width and shape. There was a neutral condition, and percentage of "cup" responses decreased with increasing width. In the food condition, where participants were told to imaging eating mashed potatoes out of the object, more responses were "bowls? Results: classification is continuous with properties of an object and is dependent of presented context.

Exemplar Theories

Proposed that we actually store specific instances, with the more general inferences emerging from those instances. Find things that have features in common. Look for degree of similarity between a new item and stored representations. Eg. Remembers every single "bird" and then we pull general inferences from this.

Degree of category membership

Schema allow for variation in the object that might fit a particular schema. - based on functional utility. - some category membership is more typical. - Different instances are judged to be members of a category to different degrees, with the more typical members of a category having an advantage in processing.

Oates & Reder ambiguous figures task

Showed abstract images and words. Abstract photos were no better remembered than words

Natural vs. Artificial Categories

Supported by different brain regions.

Mnemonics

Technique to help remember things - easier to remember less meaningful material when you apply meaning to them

What does our cognitive system gain by categorizing?

The ability to make predictions. This gives us economy in representation and communication.

Do people remember two items separate but together? Or two items interacting in a weird way?

The latter. Eg. piano and cigar are better remembered when we imagine the piano smoking a cigar.

Amodal hypothesis mapping of representations

There is an intermediate abstract system ("meaning") - we have systems for converting back and forth between the perceptual and motor representations and this abstract information

What can semantic networks not do?

They just store properties with concepts, they cannot capture the nature of general knowledge (Eg. typical size, shape, etc.)

Semantic Networks

Used to encode conceptual knowledge as well as propositional knowledge. -(Quinllian) Proposed that people store information about various categories in a network structure. - semantic networks account for behavioural data at different levels of the network

Abstraction Theory

What we do is store general information. We create a prototype. We actually abstract general properties from the instances we have studied. - people might store a single prototype and judge specific instances in terms of their similarity to that prototype. - people might store representation that also encode some idea of the allowable variation around the prototype.

What happened when people were given a coin to study and then were asked about it a week later?

When asked to remember detail, they achieved much higher accuracy than in the penny study. When people are told to pay attention to detail, they remember it better.

Perceptual Interpretation study (Barsalou et al.)

When people understand a sentence, they actually come up with a perceptual representation of that sentence. - Horizontal vs. Vertical Nails. Results: people were faster a recognitizing verticle nails when reading about nailing into a floor, but faster at recognizing the verticle nails if they read about nailing into a wall. - Concepts are represented in brain areas similar to those that process perceptions. - comprehending a sentence requires a perceptual interpretation of that sentence.

What happens when typical scripts are rearranged?

When recalling stories, participants tended to put the actions back into their normal order. Schemas for events strongly influence memory and encoding new events relates to present schemas.

Dual Coding Hypothesis

Why do visual mnemonics improve memory? A Return to Pavio's Dual Code hypothesis Memory contains two distinct coding systems... 1. Verbal (abstract, linguistic meaning) 2. Imagery (mental pictures) Having a dual code improves memory over having only a single code.

Can abstraction systems involve unconscious recognition? What brain region is involved?

Yes. Basal ganglia. Damage to this region results in deficits in learning categories (implicit category learning)

Implications of Semantic Networks

a. If a fact about a concept is encountered frequently, it will be stored with that concept, even if it could be inferred from a higher-order concept. b. The more frequently a fact about a concept is encountered, the more strongly that fact will be associated with the concept. c. The more strongly facts are associated with concepts, the more rapidly they are verified. d. Inferring facts that are not directly stored with a concept takes a relatively long time.

Schema theory is an _______theory

abstraction.

What kind of knowledge are respresented in posterior regions?

categorical information

Both abstraction and exemplar theories predict better processing of ____ of a category

central members. - Abstraction: central members are more similar to the abstract representation. - Exemplar: central members will have more similarities on average.

isa links can create the structure _____

generalized hierarchy.

Schemas help us make _____ about instances and concepts

inferences. This aspect also deals with exceptions and understand constraints. Eg. A house built underground would not have windows, or an ostrich cannot fly but is still a bird.

ISA LINK (slot)

link in a semantic network that supports of categories. EG. "____ is a ____"

Mandler & Ritchey (1977) scenes changes and memory

participants studied pictures of scenes. Then had to decide if they had already studied it or if it was a novel scenes. Made one token/style change, no change to the picture's meaning. The other was a type change (changed the meaning of the picture). Results: subjects were sensitive to type rather than token change. Meaning of a picture is more memorable than the picture style.

Snowflake task (Goldstein et al.)

showed people studied snowflakes or faces. Found that people could not remember snowflakes, but could remeber faces even though snowflakes were more different than faces.

Schemas represent categorical knowledge according to a ___ structure.

slot: attributes that members of categories process filled with one or more values or particular instances.

there seems to be greater ____ involvement in the representation of animals and a greater ______ involvement in the representation of artifacts.

visual, motor.

Multimodal hypothesis mapping of representations

we have various representations tied to different perceptual and motor systems, we have means to directly converting one representation to another. Systems can understand eachother.

what happens following a long delay?

we lose the information about the style/structure after a long delay

Argument (propositional analysis)

‣ An element of a propositional representation that corresponds to a time, place, person, or object - changing the proposition changes the meaning

Relation (propositional analysis)

‣ An element that organizes the arguments of a propositional representation (i.e., verbs, adjectives)

Propositional Analysis

‣ Involves memory for complex sentences in terms of memory for simple, abstract propositional units ‣ People tend to remember the propositions they encounter but are insensitive to the actual combination of propositions.

what is a proposition?

‣ Proposition ‣ The smallest unit of knowledge that can stand as a separate assertion ‣ A proposition specifies relationships between concepts. ‣ It is neither visual or verbal.