Question 1

One of the key properties of the _____ approach is that a specific concept is represented by activity that is distributed over many units in the network.

Accepted Answer

The given property is a characteristic of the connectionist approach, also known as the neural network approach, which describes concepts as distributed patterns of activity over a network of interconnected units. The semantic network approach, on the other hand, represents concepts as nodes connected by links that indicate the relations between them. The hierarchical approach organizes concepts into levels of abstraction or specificity, while the spreading activation approach proposes that concepts are activated through a process of association and priming.

Question 2

The connectionist network has learned the correct pattern for a concept when

Accepted Answer

When the back propagated error signal is zero, it means that the network has successfully learned and can accurately reproduce the desired output pattern for a given input pattern. The other choices do not necessarily indicate successful learning, as connection weights that add up to exactly +1.00 may not be relevant to the learning process, output pattern matching the initial input pattern may not always be desirable, and output unit response being greater than zero does not necessarily indicate accurate learning.

Question 3

Learning takes place in a connectionist network through a process of _____ in which an error signal is transmitted from output units towards the input units.

Accepted Answer

The process described in the question is specifically describing back propagation, which is the transmission of error signals from output units back towards input units in a connectionist network. Therefore, the correct choice is D.

Question 4

Collins and Loftus modified the original semantic network theory of Collins and Quillian to satisfy some of the criticisms of the original model.However, their revised model was not immune to criticism.One criticism of Collins and Loftus' semantic network theory is that it

Accepted Answer

The criticism is that the model is too flexible and can be adjusted to explain any result, making it of little explanatory value.

Question 5

Collins and Quillian's semantic network model predicts that the reaction time to verify "a canary is a bird" is _____ the reaction time to verify "an ostrich is a bird."

Accepted Answer

According to Collins and Quillian's model, both "canary" and "ostrich" are associate with the concept "bird" at the same level in the hierarchy, so the reaction time to verify them as birds would be the same.

Question 6

Collins and Loftus modified the original semantic network theory of Collins and Quillian to satisfy some of the criticisms of the original model.In their modification, Collins and Loftus account for the typicality effect by

Accepted Answer

Collins and Loftus modified the semantic network model by using shorter links to represent more closely related concepts, which accounts for the typicality effect where more typical items are recognized faster.

Question 7

The activity that represents a particular object is established in a connectionist network through a process of learning that involves

Accepted Answer

In a connectionist network, the learning process involves adjusting the weights of inhibitory and excitatory connections between units. This is done through a process called backpropagation, where errors in the network's output are propagated backward through the connections and used to adjust the weights to improve performance.

Question 8

Which of the following is most closely modeled on the way the nervous system operates?

Accepted Answer

Parallel distributed processing theory, also known as connectionism, is a computational approach to modeling the brain that involves networks of simple processing units that work together to process information in a distributed, parallel way. This is similar to the way neurons in the brain work together to process information. The other options (semantic network theory, prototype approach, and enhancement due to priming) do not closely model the way the nervous system operates.

Question 9

One beneficial property of connectionist networks is graceful degradation, which refers to the property that

Accepted Answer

Graceful degradation in connectionist networks refers to the property that even if some of the nodes or connections are damaged or lost, the entire network does not completely stop functioning. Instead, the performance may degrade gradually as more and more of the network is damaged, but it can still continue to produce output. This is similar to the way that the brain is able to compensate for damage or injuries and still function to some degree.

Question 10

In a lexical decision task, participants have to decide whether

Accepted Answer

In a lexical decision task, participants are presented with a letter string and must decide whether it is a word or not. Therefore, the best choice is B - a letter string is a word.

Question 11

Connectionist networks are modeled after neural networks in the nervous system and incorporate all of the following features of the nervous system EXCEPT

Accepted Answer

Connectionist networks do incorporate excitatory and inhibitory connections, strength of firing patterns determined by inputs, and distributed coding. However, they do not necessarily represent specific concepts with individual nodes in the same way the nervous system does.

Question 12

The process of back propagation is most closely associated with

Accepted Answer

Back propagation is a machine learning technique that is commonly used in connectionist networks, which are also known as neural networks. The back propagation algorithm involves adjusting the weights in a network based on the difference between the predicted output and the actual output, and it is typically used in supervised learning tasks such as classification or regression.

Question 13

Connectionist theory states that a particular object (like a canary)is identified by activity in the specific "canary" output unit of the network.

Accepted Answer

Connectionist theory, or parallel distributed processing, suggests that objects are identified by patterns of activation across a network of units, rather than a single unit representing a specific object.

Question 14

Collins and Loftus modified the original semantic network theory of Collins and Quillian to satisfy some of the criticisms of the original model.People consider this to be a strong theory because it is powerful enough to explain just about any result.

Accepted Answer

While the modified semantic network theory addresses some criticisms of the original model, it is not considered a universally strong theory that can explain any result.

Question 15

According to Collins and Quillian's semantic network model, it should take longest to verify which statement below?

Accepted Answer

According to the semantic network model, verifying statements that have a longer path between the concept and its properties takes longer. In this case, verifying that a pig is an animal requires traversing two levels (pig → animal) in the network, which takes longer than verifying that a pig is a pig, which only requires one level of traversal (pig → pig). The other two statements (pigs are pigs and a pig is a mammal) also only require one level of traversal in the semantic network, so they should be easier and quicker to verify.

Question 16

Good psychological theories must have all of the following properties EXCEPT being

Accepted Answer

A theory that is too powerful to be refuted by empirical evidence is unfalsifiable, which goes against the scientific method. A good psychological theory should be able to be tested and potentially proven wrong by experimental results.

Question 17

Spreading activation

Accepted Answer

Spreading activation is a theory that suggests when a concept is activated, it also triggers the activation of associated concepts or closely related concepts in memory. This means that activation spreads along the links between the concepts, priming them and making them more easily accessible. It does not inhibit unrelated concepts or weaken the link between unrelated concepts. However, it may create new links between associated concepts as the activation spreads across them.

Question 18

Learning in the connectionist network is represented by adjustments to network

Accepted Answer

Learning in the connectionist network involves adjusting the weights between nodes based on the input data and expected output. The nodes, hidden units, and output units themselves do not change, but the strength of the connections between them do.

Question 19

Which of the following is NOT a property of the connectionist approach?

Accepted Answer

This statement is incorrect. Before learning, the weights in the network are usually initialized to random values, not zero.

Question 20

Collins and Quillian explained the results of priming experiments by introducing the concept of _____ into their network model.

Accepted Answer

Collins and Quillian's network model included the idea of spreading activation, which referred to the spread of activation from one node to another in the network as a result of priming. Therefore, to explain the results of priming experiments, they used the concept of spreading activation.

Quiz 9: Knowledge