Question 1

A bigram detector fires in response to the

Accepted Answer

A bigram detector specifically detects a pair of adjacent letters, known as a bigram. Therefore, the appropriate letter pair (option D) is the best choice. The other options do not apply to the function of a bigram detector.

Question 2

We often encounter ambiguous letters when reading handwritten words but can still interpret them. For example, the same shape can be interpreted as an A in CAT but an H in THE. At what level of analysis does the feature net resolve this issue?

Accepted Answer

The answer of We often encounter ambiguous letters when reading...

Question 3

English nonwords (e.g., "HICE") are easier to perceive than strings of letters not resembling English words (e.g., "RSFK") because

Accepted Answer

Nonwords composed of common letter combinations (such as "HICE" containing "HI" and "CE") activate the brain's bigram detectors, which recognize pairs of letters that occur frequently in English. This makes them easier to perceive than nonword strings of random letters, like "RSFK", which do not activate these detectors.

Question 4

Which of the following statements about a feature-net model is FALSE?

Accepted Answer

The answer of Which of the following statements about a...

Question 5

In a feature-net model, knowledge of spelling patterns

Accepted Answer

The answer of In a feature-net model, knowledge of spelling...

Question 6

Compared to detectors that have not fired recently, a detector that has fired recently is likely to

Accepted Answer

A detector that has fired recently is more likely to have a higher activation level, as firing increases the activation level of the detector. The position within the network and response threshold are not necessarily related to recent firing, and priming requirements would depend on the specific detector and stimuli.

Question 7

The bigram layer of a feature net is

Accepted Answer

The bigram layer of a feature net is developed with experience. It is based on the frequency and co-occurrence of pairs of letters in a given language, so it varies between languages and can be further optimized with data and experience. It is not rigid or something with which we are born, and it differs between languages, so answers A, B, and C are incorrect.

Question 8

Mistakes in word reading occur under a feature-net model of recognition. This results because the feature net encourages ________ over ________.

Accepted Answer

The feature net model of recognition encourages efficiency over accuracy, which means that the model prioritizes quickly recognizing familiar patterns rather than accurately decoding unfamiliar ones. This can lead to mistakes in word reading, as the model may rely too heavily on previously learned patterns and fail to accurately identify new or irregular words.

Question 9

A response threshold is the

Accepted Answer

A response threshold refers to the activation level at which a response occurs. It is the point at which a stimulus is strong enough to elicit a response from an individual.

Question 10

Participants in a tachistoscopic procedure are shown the sequence CQRN. Participants misperceive this string as CORN. In a feature-net account, which of the following statements probably does NOT contribute to this effect?

Accepted Answer

The fact that feature nets are generally unable to identify nonwords is not relevant to the misperception of CQRN as CORN. The other statements are all relevant to the priming of specific features and bigrams, which can influence how the sequence is perceived.

Question 11

The efficiency of a feature-net model occasionally leads to errors. We could avoid making word-recognition mistakes if we scrutinized each letter. Why is this a suboptimal strategy (and not the one we use)?

Accepted Answer

The answer of The efficiency of a feature-net model occasionally...

Question 12

A feature net is a

Accepted Answer

A feature net is a network of cognitive "detectors" organized in hierarchical layers. It is a computational model of how the visual system in the brain works, where each layer of detectors extracts specific features of an object, such as edges, corners, and textures. The extracted features are then combined to form a complete representation of the object's form.

Question 13

The bigram CO is more common in English than CQ, which makes identifying CORN more likely than identifying CQRN. According to feature nets, how is knowledge of bigram frequency stored?

Accepted Answer

Feature nets do not explicitly store knowledge of bigram frequency, as this type of information is considered general knowledge and not tied to a specific bigram-detection process.

Question 14

A vertical line (like the "l" in "line") could activate all of the following nodes in a feature net EXCEPT

Accepted Answer

A vertical line would activate nodes that contain vertical lines (like the K and the CK in CLICK) but it would not activate nodes that do not contain any vertical lines (like the O in CLICK).

Question 15

One type of error that can result from feature nets is overregularization. Is overregularization a significant problem?

Accepted Answer

Overregularization errors are infrequent and usually not problematic, meaning they are not a significant problem.

Question 16

The middle layer of a basic feature net contains ________ detectors.

Accepted Answer

The answer of The middle layer of a basic feature...

Question 17

Which of the following would be considered a benefit of a feature net?

Accepted Answer

A feature net's flexibility to deal with unclear inputs is a benefit because it allows it to adapt to new or ambiguous information without making errors or becoming rigid in its organization. Slow processing, errors being impossible, and new information not affecting past organization are not necessarily benefits of a feature net.

Question 18

Participants' recognition thresholds are

Accepted Answer

Participants' recognition thresholds are lower for frequently seen words due to the process of priming. The more exposure someone has to a word, the easier it is for them to recognize it again.

Question 19

On one trial of an experiment, a participant is shown the sequence GWXT. On a different trial, the participant is shown the sequence PAFE. On the basis of prior research, we should expect that

Accepted Answer

Prior research suggests that detectors for common letter combinations or bigrams, such as PA and FE, are likely to be well-primed in the brain. Therefore, the letter sequence PAFE is expected to be easier to perceive than GWXT, which does not contain any common bigram patterns.

Question 20

Recognition errors (like mistaking CQRN for CORN) lead us to what conclusion about feature nets?

Accepted Answer

Recognition errors suggest that feature nets are interactive and can lead to errors, but do not necessarily negate their ability to accurately represent our cognitive processing. Other evidence supports the idea that feature nets can explain both accurate performance and errors.

Quiz 4: Recognizing Objects