Semantic Memory Flashcards

Question

Describe prototype theory

Answer 1

Categories are organized, in part, using unconscious mental statistics. One idea of how this is done is the prototype model. For this view, categories are determined by a mental representation that is an average of all category members. This averaged representation is a prototype, which may or may not correspond to an actual entity in the world.

Answer 2

An example of prototype extraction using dot patterns: When people first learned the categories, they were shown deviations from the prototypes. The prototypes were never shown during learning. However, when people were later asked to sort both old and new patterns the prototypes were identified and correctly sorted at a high rate of accuracy, suggesting that they were derived and used to make decisions If people can readily derive prototypes from things as meaningless as dot patterns, surely the same mental mechanisms are involved for deriving categories of our everyday experiences with everyday objects.

Answer 3

For example, if photographs of faces are used for making preference judgments along with morphed composites of faces, people rate the composite faces, which are closer to the prototype face, as more attractive. That is, people prefer faces that are averages of others. Because they are averages, they have fewer unusual and distinguishing characteristics and so are easier to mentally process. A pretty face is a boring face. This is also part of the reason why attractive faces are harder to remember

Answer 4

They provide a clear explanation for the central tendencies of categories (which is the prototype itself) and a graded category structure. The closer an instance is to the prototype, the better a member of the category it is.

Answer 5

For example, people are often aware of a category size —that is, about how many different members are in the category. Prototypes convey no information about the variability among category members. Also, a caricature (a category member with exaggerated features) is thought to better represent a category than a prototype when a category is considered in the context of other, related categories. That is because the caricature captures distinctive features and emphasizes them. This helps distinguish one category from other, similar categories.

Answer 6

Exemplar theory: people use all the category members to make decisions (e.g. thinking of all birds to decide whether something is a bird) Explanation-based views: people try to have reasons for why things should be grouped together, categories are theories or explanations (feathers and wings tend to go together because feathers are suited for flying).

Answer 7

Captures central tendency, graded membership, and family resemblance, as well as information about category size, variability, correlated attributes, and any new information about the category (Because categorization is always using all of the memory traces, new experiences can have an influence). They can explain the context sensitivity of categories. For example, the color gray is more similar to white in the context of hair color but is more similar to black in the context of clouds Previously activated semantic meanings can bias how new information is interpreted. For example, a phrase like “adolescent doctor” is easier to interpret if it follows the phrase “animal doctor” than if it follows “country doctor.”

Answer 8

it seems that both are used, though in different circumstances. Specifically, it is more adaptive in a natural environment to move early on from a more exemplar-based form of categorization to a more prototype based form of categorization. This allows people to deal with the family resemblance of many natural kinds and be less led astray by more peripheral and irrelevant features of individual category members

Answer 9

An inherent circularity. Specifically, categories are defined by experiences with members of that category. That is, the members of the category all contribute to defining it. However, the memory traces that are selected are those that conform to the criteria of the category. In short, how can memory traces be selected to define a category if the category is needed to select them in the first place?

Answer 10

In general, people place an emphasis on causal factors as compared to the effects. For example, knowing that an animal swims is a more fundamental characteristic than knowing that an animal has webbed feet (presumably they have webbed feet because they swim). Also, categories are defined, in part, by how people interact with things not just statistical regularities. For example, what makes something a chair has more to do with your sitting in it than with the materials used to make it.

Answer 11

People can make new categories on the fly. These are ad hoc categories. For example, coffee, perfume, leather, and skunks are all members of the category things with a distinctive smell. Ad hoc categories are interesting because people generate them off the cuff but they have many of the same properties as standard categories. They have a central tendency, graded structures, and family resemblance. Thus, some semantic memory structures are generated spontaneously. This raises questions about the stability of semantic memory in general.

Answer 12

Consistent with the idea that we create categories in the pursuit of explanations is psychological essentialism. This is the idea that members of a category share an underlying essence, of which people may or may not be aware.

Answer 13

It usually applies to natural kind categories, which can be defined by chemical structure or DNA, such as water or skunk. That said, some artifact categories, such as scientific instruments, are treated as though they have essential qualities, and some natural kind categories, such as humans, are not.

Answer 14

When we engage in stereotyping. Stereotypes are categories for groups of people. When you stereotype people, you are treating them as if they are essentially the same as other members of that group. These stereotypes are activated automatically and we use this semantic knowledge to make assumptions about people and are surprised when those assumptions are violated

Answer 15

Another characteristic of semantic memory is the influence of knowledge that is ordered along some dimension, such as size, intelligence, or age. These are linear order effects and reflect the organization of information as it is stored or processed in semantic memory.

Answer 16

Semantic distance effect: people make faster judgments about the order of two items as the distance between them increases (size of rabbit and elephant vs dog and rabbit, more likely to confuse adjacent days of the week rather than days that are separated from one another). Semantic congruity effect: people are faster to judge the relationship between two items if the valence of the comparison term matches the end of the dimension they are on (easier to judge that Jefferson was president before Monroe than it is to judge that Monroe was president after Jefferson because both are at the “early” end of the dimension) Serial position effect, people are faster to make judgments about two or more items at the extremes of a dimension than those in the middle (e.g cork is bigger than castleblayney vs sligo is bigger than Clare).

Answer 17

Dimensional information is stored in semantic memory along with the concepts. Jefferson and Monroe are both thought of as early U.S. presidents, so the attribute “early” is stored directly with them. When this information is needed, if the attributes match the judgment people respond faster than if there is a mismatch. With a mismatch, people need to do more thinking to get the information lined up properly.

Answer 18

The spatial-numerical association of response codes, or SNARC effect. When people made judgments about numbers, such as whether they were odd or even, judgments about smaller numbers were made faster with the left hand. The reverse was true for the larger numbers The SNARC effect is consistent with the idea that people have a mental number line in semantic memory going from left to right with small numbers on the left and large numbers appearing as one moves to the right, although this may be due to relative, rather than absolute, magnitudes

Answer 19

It is also observed in blind people, suggesting that it is not strictly visual. Moreover, the SNARC effect is not limited to numbers. Researchers also found a similar pattern with musical pitches. People respond faster to lower tones with the left hand and higher tones with the right

Answer 20

In life, there are many situations that are fairly regular in how they unfold and operate, and how we react to them. That is, common experiences shared some framework that unites them. We are able to capitalize on this to help us understand new situations, much as we use categories to understand new objects or creatures. A semantic memory that captures commonly encountered aspects of life is called a schema. A schema can be thought of as a blueprint for events that people can draw upon to understand a specific case.

Answer 21

selection: process to sort out which things are likely to be central and which are peripheral (which are important; e.g time elapsed in football v baseball) abstraction: converting the surface form of information (e.g., verbatim wording) into a more abstract representation that captures the underlying meaning interpretation: allows people to fill in the gaps for things that were missed integration: sometimes we come across event descriptions that are given out piecemeal, and we need this schema process to guide us in putting these pieces together into a coherent whole (e.g clues of a murder) reconstruction: Memories are not complete records of the past. Instead, they are fragmentary. Sometimes there are a lot of fragments, enough to recover almost the entire memory, whereas in other cases the fragments are few and far between. With reconstruction, people fill in the memory gaps.

Answer 22

Knowing which schema is relevant can greatly influence performance. E.g people read an ambiguous passage. If people are told ahead of time that the passage is entitled “Washing Clothes,” then they remember more of it later. The title allows them to activate the appropriate schema and they can then select what is relevant in the passage and interpret it. This occurs during encoding because this title benefit is only observed when it is given before reading

Answer 23

when people hear sentences and comprehend them, within a few minutes they are not be able to distinguish verbatim sentences from paraphrases. Similarly, if people see a picture, they are less likely to notice a change, such as adding or subtracting elements, rearranging entities in the scene, or changing the orientation of entities in the picture, if the rearranged picture fits their abstract, schematic memory of what they saw People see ambiguous line drawings with one of two labels (e.g curtains/ diamond in square). Each of these labels is placed next to the drawing. After a period of time, people were asked to draw what they remembered seeing. People tended to distort their drawings to conform to the label that was provided. People used their schemas to abstract away and lose the ambiguous information. Thus, what is remembered is more schema-consistent.

Answer 24

People may misremember having encountered things that they only inferred using a schema.

Answer 25

The first four are for the encoding of new information, and the fifth is important during retrieval.

Answer 26

In one set of experiments, he gave British Cambridge University students a Native American folktale to read, called “The War of the Ghosts.” Sometime after reading—often several days, weeks, or months later— people were asked to recall the tale. People not only forgot parts of the story but they added new elements. Often this new information was less consistent with the original story but more consistent with typical English folktales.

Answer 27

When knowledge refers to a sequence of events that occurs in a stereotyped fashion, these scripts are temporally ordered schemas that are structured according to the major components of the event with a preference for using script information in a forward order (e.g what to do at a restaurant.)

Answer 28

The use of scripts influences how information is retrieved and used. For example, when people read a text of a scripted event they take longer to read a sentence when the action is further along in the script from the prior sentence than if it is closer When the information is close in the script, less effort is required. However, when the information is far in the script, more effort is required because more of the script needs to be scanned to bring the person up to date. More knowledge must be inferred.

Answer 29

People are likely to make causal inferences because understanding causal relations is important for understanding how the world is structured and operates. However there is a bias to infer causes but not effects because people can easily infer how they may have gotten to a current point in time (what caused this) if they access the appropriate schema. However, knowing what will happen next (what effects to predict) is more difficult because in many cases any number of possible outcomes could exist. This is not to say that people never use schemas to make predictions about the future—only that they are much less likely to do so.

Answer 30

Students were given ambiguous texts, (e.g could be either about an escaped convict or a deer hunter.) Students were first asked to read this text from one perspective. After a brief delay, the students were asked to recall the story. Some did this from the same perspective as they read it. In contrast, others were led to believe that the experimenter had given them the wrong title initially. They were then given the “correct” other title. When there was a title switch, people recalled the same amount of information as when there was no switch. However, there were important differences. When there was a title switch, people made fewer intrusions, whereas those who did not have the switch made a large number of schema-consistent intrusions. Thus, people can disregard schemas and use more detailed, verbatim memories.

Answer 31

Semantic illusions- inappropriate retrieval of information when there is overlapping lexical information, such as a similar name (e.g moses illusion/ moses' ark) Naive physics- memory can incorporate physical principles of the world, such as gravity and friction. Some naïve physics knowledge is stored in semantic memory. However, when we consciously try to apply this knowledge, misunderstandings can be revealed

Answer 32

Semantic processing is very general unless people focus on the information of interest (we only do a cursory check of knowledge in semantic memory to see if the information is broadly consistent.) People engage in only a partial assessment of semantic information (people retrieve some of the information from semantic memory and, so long as it is a close fit, they are willing to go with it.) Similar language elements, such as a similar name, can inappropriately activate information in semantic memory, giving the illusion that it is known (if it sounds close, people are often willing to disregard some smaller inconsistencies)

Answer 33

Students were given diagrams (google docs). The task was to indicate (1) the trajectory of a ball shot out of the tube, (2) the trajectory of the ball when the string broke, and (3) the path of the bomb when the plane dropped it. Although people do give the correct responses in some cases, they also give incorrect responses using incorrect knowledge in semantic memory. What is interesting is that people are responding as if they are holding medieval impetus theories of motion. Such responses are more likely when people use static diagrams and are less likely when viewing moving displays although not always

Answer 34

With education, people can overcome their naïve beliefs about motion. That said, these beliefs never seem to go away completely. For example, if people are asked to verify statements that are inconsistent with early childhood beliefs, such as the idea that the earth goes around the sun, people respond more slowly compared to statements that are consistent with early beliefs, such as the idea that the moon goes around the earth. This is because there is some interference from older, incorrect knowledge. I

Answer 35

While education improves performance, it can also get in the way. Researchers reported what they called the Galileo bias, in which people mistakenly believed that if two balls of different weights dropped from 10 meters they will hit the ground simultaneously. However, this does not take into account air resistance. Even with extensive practice dropping these balls, students at Arizona State University continued to make errors based on semantic knowledge of what they had learned in elementary school about Galileo and gravity.

Answer 36

Not a good model because we cannot simply grab one section and we have all that knowledge in one place in the brain however it is a good model to an extent in that libraries tend to have a classification system (e.g hobbies, skiing, alpine skiing etc)

Answer 37

Networks of connections between concepts.

Answer 38

Assumption 1: If we know how a computer generates knowledge and uses it, we will know how human memory works Assumption 2: If we know how human memory works, we can make a super-powerful computer These have largely been refuted and is not how it works in practice

Answer 39

A network representation showing how concepts are tied to different concepts through many rules and relationships via coding type language e.g My dog Isa Dog Isa animal. agent, dog; relation, eats; Object; meat. Diagram in notes

Answer 40

By having many many rules for many many instances/ cases (e.g chess, fighting)

Answer 41

Herbert Simon studied chess grandmasters and low level students. He first placed pieces on the board to represent a game situation. He would show them, take it away and have them set it up again. The grandmasters were extremely good at this, students were a lot worse, showing an effect of expertise. Simon then set up the boards in ways which would never naturally occur in a game (e.g extra horses). The difference in performance disappeared between the students and masters.

Answer 42

Expert systems in which, given a situation, one consults their database of what can be done in certain situations they know and applying one which is suitable.

Answer 43

The Collins & Quillian Model is a hierarchal network model in which there are both categories and characteristics. Categories of animals could include birds, fish etc which could include canary, ostrich; shark, salmon as subcategories. These categories have associated characteristics such as bird- wings, fly, feathers and ostrich- long thin legs, feathers, can't fly. Connections between categories are called links and the categories they connect are called buttons.

Answer 44

The sentence 'the seahorse has eyes' takes longer to comprehend than the 'male seahorse carries the baby' because eyes are something that is not unique to seahorse but you have to travel back up to fish and animals to access that characteristic, however a male seahorse carrying the child is something uniquely associated with them and not fish or animals so it is accessed more quickly.

Answer 45

It was found through later studies that this theory did not hold up and makes several incorrect predictions. It has no explanation for "typicality' effects e.g a robin being one of the first birds when you think of a robin, thought of faster than a chicken or a penguin because it is more typical of a bird. It also does not take into account that frequency of association is more important than distance in the hierarchey- i.e a cat is an animal is quicker registered of than a cat is a mammal.

Answer 46

Collins & Loftus made a spreading activation model in which, for example, if you think of the a firetruck, then the concept of red could be activated and therefore red objects such as cherries, fire etc could be more easily reached (primed.)

Answer 47

Associative priming- e.g 'cats and dogs' may prime 'weather' from the expression; found together in the real world Semantic priming- when there is a semantic bond between the too e.g Labrador primes dog.

Answer 48

Robin- quicker RT if primed with the word bird or not: semantic prime. Box- quicker RT if primed with word bird: associative prime Arm: Slower RT if primed with word bird: slower response due to unrelated prime

Answer 49

SOA or Stimulus Onset Asynchrony: interval between prime and target Facilitation = RT non-primed trial -RT primed trial

Answer 50

By informing the participant that when there is a e.g a body part, there will be part of a building. Then there will be a faster RT for 'door' than 'heart'. There is then a slower RT as if it is an unrelated prime

Answer 51

- A good classification is reached after many examples - This learning is pretty difficult - The prototype is typically misrecognised as 'seen', even if it was not - This was taken to suggest that subjects induce a type of prototype

Answer 52

(Anterograde) amnesia patients can also learn this type of categories, suggesting that this learning is implicit

Answer 53

Examplar theory holds up the best compared to prototype but makes the really weird remember-all items in the category that is likely wrong

Answer 54

Neural network models

Answer 55

Subjects had to sit and wait in an office until a study began. When they got into the study, participants had to remember everything they could about the office. Since it was quite a typical office setting, it activated this schema and participants would add stuff that they would expect to find in an office but were not present such as books and did not report some items which are not typical of offices such as the notice board.

Answer 56

Benefits: - Creates expectations - Makes it much easier to make inferences - You remember more Disadvantages - Distorts memory (without realising it) - Allows manipulation by others (e.g magicians, politicians and advertisers)

Answer 57

They have a progressive loss of semantic knowledge. You can speak to thme completely fine and then mention grandchildren and they could say "what are grandchildren?" If you explore further you will find that whole categories of knowledge are missing from their index. If getting them to talk about a topic in which they have lost a lot of information on, they may use words like 'thingy' and try to talk around it (word finding problems and some comprehension difficulties). However there are no problems with new learning.

Answer 58

infero-lateral temporal cortex but (early in the disease) with sparing of the disease. It is considered a sub-type of fronto-temporal dementia

Answer 59

It may be a bunch of episodic knowledge that they can use as if it were semantic knowledge e.g e.g., ‘ducks’ are ‘those animals that I sometimes feed in the park’ withno generalization (e.g., lays eggs???). This is also context dependent (toothbrush in kitchen drawer may not be recognized; only in bathroom)

Answer 60

E.g you could show them a camel. Then take it away for a few minutes then ask them to draw a camel or the thing they just seen and often it will not look like one, instead integrating different aspects such as wheels instead of legs or a human head.

Answer 61

Picture naming responses to the same stimuli can also be recorded such as (over three years): Swan: swan, bird, animal Rooster: chicken, chicken, dog

Answer 62

First the very specific knowledge (subordinate) leaves and is replaced by the more general knowledge (superordinate) which survives longer. And then this can go and can be replaced by some kind of (high frequent) prototypical response (e.g dog for rooster or duck) or even something non-living for something which is living for example (e.g vehicle for peacock).

Answer 63

Quillian’s hierachical model could not predict reaction times correctly in many important cases. But it seems that knowledge erodes in a hierarchical manner with semantic dementia–From the subordinate to superordinate levels