Deck 8: Associative Theories of Long-Term Memory

Question 1

In a memory network a connection tying two nodes together is called a(n):

A) associative link
B) nodal highway
C) axon
D) memory tie

Answer 1

A

Question 2

Participants are asked their mother's first name and then their father's first name. Their answers to the second question are MOST likely to be:

A) faster because activation spreads between two well-connected nodes
B) faster because the first answer will be locally represented
C) slower because of semantic interference
D) slower because of fan effects

Answer 2

A

Question 3

In a sentence-verification task, a participant sees sentences such as "A penguin is a bird," or "A dog is an animal." Generally, these sentences are of the form "An X is a Y." The speed of a participant's response to these sentences will be slower if:

A) several associative links must be traversed to trace a path between the node for X and the node for Y
B) the participants have recently seen the items named (e.g., penguin or dog)
C) the object named by X is a typical member of the category named by Y
D) there is a direct connection between the node for X and the node for Y

Answer 3

A

Question 4

Once a node is activated, it will serve as a source of further activation, and this activation will:

A) spread through all the associations radiating from the node
B) "choose" only one association to activate
C) inhibit the activities of all other nodes
D) direct the person's attention away from the node's contents

Question 5

Sentence-verification tasks are typically used to:

A) assess the extent of participants' knowledge about a topic
B) probe for a participant's false beliefs
C) determine how information is organized and connected in our mental encyclopedia
D) quantify the contribution of schemata to memory errors

Question 6

According to the text, hints help us remember because:

A) nodes representing an effective hint generally have a high degree of fan
B) nodes for the hint are likely to be connected to a memory central executive
C) the target node receives activation from both the hint's nodes and from other nodes
D) the hint's node is likely to be well primed by the context of the memory test

Question 7

A lexical-decision task is generally used to assess:

A) the degree of fan for the nodes representing the target word
B) the holding capacity of a participant's working memory
C) how rapidly participants can "look up" a word in their "mental dictionary"
D) the content of a participant's memory nodes

Question 8

Node L within a memory network has a low degree of fan. Node H within the network has a high degree of fan. All other things equal, a comparison of the nodes will show that Node H:

A) will, when activated, send activation to a smaller number of neighbors
B) will, when activated, send less activation to each of its associated nodes
C) can be reached via a smaller number of retrieval paths
D) has a higher response threshold than Node L

Question 9

In a sentence-verification task, participants respond to these two sentences: "A canary has feathers," and "A canary is yellow." We should predict faster response times to the second sentence because:

A) participants are slowed by semantic ambiguity
B) participants' responses are faster to perceptual properties than to conceptual ones
C) participants are faster in judging "isa" associations than they are in judging "hasa" associations
D) the property of having feathers is associated with the bird node, not the canary node

Question 10

Which of the following does NOT help a node to fire?

A) recent activation of the node so that it has been "warmed up"
B) the summation of two subthreshold inputs into the node
C) increasing the activation level of the node
D) increasing the response threshold of the node

Question 11

In an experiment, participants are shown a pair of letter strings and are asked to respond "yes" if both the letter strings are words and "no" otherwise. Which of the following pairs of letter strings are MOST likely to produce the fastest response?

A) bread-bed
B) bread-gert
C) bread-butter
D) bread-moth

Question 12

A node's activation level is defined in terms of:

A) how effective the node will be, when it fires, in triggering neighboring nodes
B) the current status of the node, determined (in part) by activation arriving from neighboring nodes
C) where the node is located within the processing network
D) the node's degree of fan

Question 13

Node A is connected to Node B; Node C is connected to Node D. Node A has five associative links radiating out from it, Node C has 20 associative links radiating out from it. Assume Node A and Node C receive the same amount of activation, had the same activation levels at the start, and have the same response threshold. In this case, we expect:

A) Node B to receive more activation than Node D
B) Node D to receive more activation than Node B
C) both Nodes B and D to receive the same amount of activation
D) the activation from Nodes A and C to cancel each other out so that neither Node B or D will be activated

Question 14

The priming effect observed in lexical-decision procedures reflects all of the following EXCEPT:

A) nodes are activated individually, so the activation of each node is largely independent of the activation level of neighboring nodes
B) once a node is activated, activation is likely to spread to other nodes, representing semantically related concepts
C) a node can be more quickly activated if it has already received some subthreshold activation from neighboring nodes
D) activation is likely to spread between closely associated nodes and is unlikely to spread between more distantly related nodes

Question 15

In Trial 18 of a lexical-decision task, a participant sees the word "chair." Of the following, which will produce the STRONGEST priming for this trial?

A) In Trial 17, the stimulus presented was "care."
B) In Trial 17, the stimulus presented was "table."
C) In Trial 17, the stimulus presented was "fair."
D) In Trial 17, the stimulus presented was "tree."

Question 16

In a lexical-decision task, participants are shown:

A) simple sentences and must decide whether each is true or false
B) letter strings and must decide whether each is a word
C) sequences of numbers and must decide whether each conforms to a specific pattern
D) word pairs and must decide whether the words in the pair are related

Question 17

Participants are better able to remember material learned earlier if they are in the same state at the time of recall that they were in at the time of learning. In network terms, this reflects the fact that nodes for the target materials:

A) are receiving activation from both the nodes representing the learned material and the nodes representing the participants' state
B) are associated indirectly only with the material that had been learned earlier
C) have higher response thresholds in some states than in others
D) probably have a low degree of fan

Question 18

In network models, subthreshold activation of a node:

A) causes that node to fire
B) can add together with other subthreshold activation
C) has no effect on the node
D) influences a node only if the node has not been primed

Question 19

The concept of "degree of fan" refers to:

A) the relationship between thoughts and concepts
B) the speed with which a node will respond to a given input
C) how many associative links radiate out from a node
D) the rate at which activation of a node returns to baseline levels

Question 20

A node's "response threshold" refers to:

A) the beginning moments of the node's response
B) how strongly the node has to respond in order to influence other nodes
C) the level of activation that will cause the node to fire
D) the maximum rate at which the node will respond

Question 21

Connectionist models are sometimes referred to as neural nets. This makes sense because:

A) the associative links in a network tie neurons together
B) the long-term goal of connectionist models is to provide replacements for damaged neurons in the brain
C) advocates of the connectionist approach argue that their models make good biological sense and function much as the brain does
D) research has found that most people do not understand the term "connectionist"

Question 22

Which of the following is MOST important in allowing network models to represent specific knowledge, in contrast to more general knowledge?

A) type nodes, specifying the context in which an event took place
B) agent nodes, identifying the main participants in an event
C) input nodes, tying propositions to perceptual qualities
D) token nodes, representing individual instances of a category

Question 23

Marni knows many facts about cowboy hats. However, while each of these facts is associated with cowboy hats, the various facts are not associated with one another. As a result, Marni may be less able to recall a particular fact about cowboy hats in comparison to someone who knows just that fact and nothing more. This is because:

A) activation spreads to only one node at a time, so the process of spread will take too long and Marni will give up before she finds the sought-after information
B) there are simply too many nodes to access
C) all of Marni's facts will become activated at once, causing anterograde amnesia
D) the large degree of fan from Marni's cowboy hat node causes the activation to be spread too thinly over the numerous connected nodes for it to activate the node she needs

Question 24

Which of the following is a disadvantage of a winner-takes-all system?

A) It doesn't allow a person to concentrate his or her thoughts on one topic.
B) If a distracting node is strongly activated, it can inhibit the target node.
C) It tends to prevent thoughts from being followed through to their completion.
D) It wastes energy on excessive activation and inhibition.

Question 25

Connectionist or parallel distributed processing (PDP) models differ from classic associative networks in that connectionism:

A) employs distributed processes
B) relies on a central executive to coordinate processing
C) uses local representations rather than distributed representations
D) draws mostly on serial processing

Question 26

In a winner-takes-all system:

A) a stronger node inhibits a weaker node, so that the weaker node does not activate its associated nodes
B) a weaker node can win over a stronger node by inhibiting the stronger node
C) all activation flows toward a stronger node
D) the stronger node can "shut down" if overstimulated

Question 27

A search through memory can be launched in several different ways. Which of the following is NOT one of these ways?

A) A stimulus is detected by the appropriate input nodes, and the nodes then trigger other nodes within the memory network.
B) A participant is contemplating an idea, and the nodes representing the idea then send activation to related nodes.
C) A participant is remembering an event he or she once witnessed, and activation then spreads outward from the memory nodes representing the event.
D) The memory executive identifies the focus of the memory question and directs activation toward the address of the target material.

Question 28

In connectionist networks, beliefs and memories are contained within:

A) the individual nodes
B) templatelike structures
C) the pattern of connections, independent of current activation states
D) the current state of activation of the relevant nodes

Question 29

The term "connectionism" refers to:

A) the study of communicative connections among individual neurons
B) a theoretical perspective useful in stimulating human memory
C) the phenomena observed when cognitive schemata interact with visual inputs
D) the claim that complex cognitive phenomena are best understood in terms of the parallel operation of many local representations

Question 30

The tip-of-the-tongue (TOT) state can be accounted for if the incorrectly activated node inhibits the correct node. This is commonly known as:

A) false activation
B) associative decline
C) retrieval blockage
D) memory inactivation

Question 31

In a system using distributed representations, each idea or mental content is represented by:

A) a single node
B) a group of nodes, so that the nodes are active only when the particular content is being contemplated
C) a widespread pattern of activation involving many different nodes
D) many different nodes, so that the content is brought to mind whenever any of the nodes are activated

Question 32

In Anderson's network account of propositional knowledge:

A) propositions are encoded by means of a small number of types of links, such as "isa" links or "hasa" links
B) propositions are linked to an image depicting the proposition's content
C) associations are identified in terms of their syntactic role within the proposition
D) propositions are limited to simple sentences such as "Dogs eat bones" or "My son fed the pigeons"

Question 33

In order for a connectionist or parallel distributed processing (PDP) network to function, each of the nodes must function:

A) in parallel with all of the other nodes
B) in a fashion coordinated by the central executive
C) in a manner that makes sophisticated use of the information available to that node
D) only when the specific content assigned to that node is relevant to the ongoing operation

Question 34

In connectionist theorizing, the strength of the association between two nodes is referred to as the association's:

A) activation strength
B) associativity
C) activation potential
D) connection weight

Question 35

At this moment, Enriqueta is thinking about her garden. In network terms, this "thinking about garden" is directly reflected in:

A) the connection weights among the nodes representing "garden"
B) the activation levels of the nodes representing "garden"
C) the response thresholds for the nodes representing "garden"
D) a temporary change in the connection weights among the nodes representing "garden"

Question 36

Which of the following is an advantage of a winner-takes-all system?

A) It allows the largest possible number of nodes to be activated.
B) It prevents stronger nodes from excessively inhibiting the input of weaker nodes.
C) It provides a selective mechanism so that distracting thoughts do not trigger other distracting thoughts.
D) It promotes activation of side-thoughts, which in turn promotes the discovery of novel associations.

Question 37

Often mental processes (such as the hunt through memory or the search for a problem solution) must obey several requirements at the same time. According to connectionists, this requirement is achieved through:

A) multiple distributed processing
B) simultaneous multiple constraint satisfaction
C) the connection weight principle
D) converging associations

Question 38

In a connectionist model, "knowledge" is BEST defined as:

A) a state of activation
B) a potential for activation to flow in a certain way
C) a particular group of nodes
D) the potential for creating new connections among nodes

Question 39

If the associative links between all nodes were the same type, with no markers for how the nodes are related:

A) all of the nodes would have the same response threshold
B) we would not be able to tell the difference between the statements "Sam has a dog" and "Sam is a dog"
C) the activation of one node would cause the activation of all other nodes, regardless of whether they were linked
D) the activation between two unrelated nodes would cancel each other out

Question 40

Which of the following is NOT a proposition?

A) My son eats too much candy.
B) All children eat too much candy.
C) All children despise candy.
D) Too much candy.

Question 41

Input nodes differ from other nodes in that they:

A) receive activation from other nodes
B) receive activation detectors associated with sensory organs
C) have associative links that connect to other nodes
D) are not affected by the strength of connections between nodes

Question 42

In learning algorithms, error signals do NOT:

A) come from an external source
B) cause a decrease in the node's connections to the other nodes that led to the error
C) allow an entire network to make use of feedback
D) affect only the node that directly led to the error

Question 43

Learning algorithms have been developed to try to accomplish learning in a connectionist network model. These algorithms:

A) require a few trials
B) provide feedback about a response
C) suggest that nodes that are activated at the same time will inhibit each other, and thus decrease learning
D) suggest that learning is accomplished only through forward connections (Node A activates Node B), and backwards signals (Node B sends a signal back to Node A) are not allowed

Question 44

Researchers are divided in their assessments of connectionist models. Which of the following is NOT a concern that has been raised about these models?

A) There has been some debate about the biological realism of the models.
B) Some have argued that connectionist models can accomplish complex tasks only when the input, or teaching examples, are adjusted in precisely the right way.
C) So far, connectionist models have been able to accomplish only narrowly defined tasks and have not achieved the broad intelligence displayed by humans.
D) Connectionist models seem limited in the quantity of information they can store.

Question 45

The notion of associative networks is based on the general idea that memory connections are:

A) hard to construct
B) our memories
C) sparse
D) not really important

Question 46

The strength of connections between nodes are important in associative networks because they explain:

A) why nodes are linked
B) the relationship between two connected nodes
C) how efficiently activation will spread between nodes
D) how activation can spread forward and backwards

Question 47

Using the terminology and theories of a network model, the tip-of-the-tongue (TOT) phenomenon would BEST be described as:

A) being in the right vicinity, but activation has not spread to the correct node
B) activation has spread to the correct node, but the level of activation is insufficient to cause the node to fire
C) being in the wrong vicinity
D) activation has spread to too many unrelated nodes to cause the correct node to fire

Question 48

Martha knows that London is the capital of England, but at this moment she is thinking about neither London nor England. In network terms, this bit of knowledge about London is BEST reflected in the:

A) current activation levels within the nodes representing London
B) response thresholds for the nodes for London and England
C) pattern of how activation will flow once the nodes representing London are activated
D) degree of fan for the nodes for England

Question 49

Which of the following results has NOT been found in studies on the tip-of-the-tongue (TOT) phenomenon?

A) It is a state where you are certain you knew the answer, but are unable to come up with it at that time.
B) People in the TOT state can often remember what the word begins with.
C) People in the TOT state can often give information about the word they are searching for, like how many syllables it has.
D) Later on, when you are no longer thinking about the word, it will pop into your head.

Question 50

There are several proposals for how nodes are connected in networks. Which of the following is NOT a proposal mentioned in this chapter?

A) Connections take the form of "isa" and "hasa."
B) Connections are propositions.
C) Connections include information about time and location.
D) The strength of a connection depends on how recently it has been used.

Question 51

Which of the following is NOT an example of the tip-of-the-tongue (TOT) phenomenon?

A) You recognize an actor in a movie, and can't remember where you have seen him or her before.
B) You're given a multiple-choice question on a test, but can't remember the answer.
C) You're filling out a crossword and can't remember a word, even though you know you know it.
D) You're writing a paper and can't think of a good word to use that means "to lessen the severity of . . . ."

Question 52

Connectionist theorists have proposed that connection weights are adjusted:

A) genetically
B) by learning algorithms driven entirely by local forces
C) by a central processor
D) randomly

Question 53

One way to test the network idea would be to build a working model based on the theoretical claims of the model, and see how well the model replicates the human data. What evidence could be derived from this testing method?

A) If the model does not do as well as humans, something is missing from the model.
B) If the model does as well as humans, the claims of the model are sufficient to accomplish human tasks.
C) If the model does as well as humans, we know humans use the same techniques as the model claims.
D) If the model does as well as humans, it would provide strong evidence for the model.

Question 54

There are still questions surrounding exactly how associative networks function in the larger context of long-term storage. If our long-term storage system can be thought of as a "warehouse," it may be the case that associative networks function as:

A) a library in the warehouse
B) the loading dock outside the warehouse
C) a conveyor belt moving information from outside the warehouse into the warehouse
D) the manager of the warehouse

Question 55

The concept of simultaneous multiple constraints satisfaction states that multiple demands are addressed at the same time. This concept is crucial for network models in all of the following ways EXCEPT:

A) in that many concepts are so broad (e.g., "water") that they could activate millions of nodes
B) in that addressing multiple constraints at once makes it more likely you will arrive at the correct node quickly
C) because addressing multiple constraints means you will reach a balance point efficiently
D) in that it explains the need for a central executive to oversee the activation of all nodes in the model

Question 56

According to connectionist models, "learning" occurs through an adjustment of:

A) a single connection weight
B) the number of connections
C) the degree of fan between two related nodes
D) a great many connection weights