week 1 chapter 4

Question 1

Agnosia

Answer 1

difficulty recognising objects due to brain issue. Cannot recognise the stimulus (may be visual, auditory, tactile agnosia etc).

Question 2

Apperceptive Agnosia

Answer 2

inability to assemble the various components of an object, into an integrated perceived whole.
*

Question 3

Associative agnosia

Answer 3

can see but cannot make sense of it or unable to appreciate its function, although may have encountered it before.

Question 4

Feature Net

Answer 4

Model explaining how detection of features might activate detectors (eg when see a letter or part of a letter). When detector receives input, activation level is increased. sufficient activation will trigger response threshold, detector fires, and leads to increased activation of next detector in chain.

Question 5

Lateral Occipital Complex

Answer 5

brain region involved in object recognition

*

Question 6

Bottom-up Processing

Answer 6

Also called Data driven processing. recognition/perception driven by details of features

Question 7

Top-down Processing

Answer 7

also called Concept driven processing.Less about the features, and more about the context to make sense of the features.

Question 8

Visual Search Tasks

Answer 8

subject asked to examine a display and judge whether particular item present or not. as combination of features etc requested gets more complex, discovery time increases.

Question 9

Integrative Agnosia

Answer 9

arises from damage to parietal lobe.
integrative agnosia has symptoms of both apperceptive agnosia, and associative agnosia. can usually achieve drawing an object, but it is very labour intensive and effortful.

Question 10

WHAT IS THE EVIDENCE that features play a special role in object recognition?

Answer 10

we are very efficient in simple visual search tasks (eg where a unique item amongst non unique), but become slower when need to search for a combination of features hidden amongst other combinations of features etc.

Question 11

Tachistoscopic Presentations

Answer 11

Special device (outdated now), which showed visual stimulus for precise length of (short) time. Now use computers.

Question 12

Repetitive Priming

Answer 12

causes increased recognition because have viewed recently.

Question 13

Word Superiority Effect

Answer 13

a letter is actually easier to recognise when it is within a word. demonstrated by being flushed a letter, then a mask (or blind) then asked whether was option a) or b). Compared with when flashed a word then asked a) or b) . better at picking the whole word. Only applies when words are actual, not gobbledygook.

Question 14

Well-Formedness

Answer 14

this applies to a few letters as a combination. well-formed=the combination is very common as a word string (part) in the studied language. as well-formedness increases, so does the Word Superiority Effect.ie even an incomplete word, which is a very common string, will be easily recognised.

Question 15

Summary of how easier to recognise (making mistakes)

Answer 15

easier to recognise words than letters, and more so well known words or even well-formed strings. uncommon miss-spellings forming non words that are close to common words, are frequently mis-read as the
word.

Question 16

Desirable Difficulty

Answer 16

sometimes, if text is just a little harder to read (due to font, layout etc), one is forced to pay more attention, and sometimes this is advantageous to comprehension

Question 17

Font Effects Summary

Answer 17

a) .one study showed text in harder to read font, made reader think author less intelligent (reader struggling to read blamed author for not being clear)
b) .All capitals harder to read as less distinct features.

Question 18

Feature Nets & Activation level

Answer 18

A Feature Net is a Model for how letters/words/features might be recognised. Maybe each feature detector is a neural network. proposed that info is bottom up. Activation level of a detector likely to be influenced by frequency or recency.
*

Question 19

Response Threshold

Answer 19

thought in feature net model that with sufficient activation, detector reaches response threshold and fires (similar theory to a neuron).

Question 20

Recency and frequency

Answer 20

in feature net model, thought frequency or recency of having fired, leaves activation level sitting higher (so will reach threshold with minimal input)

Question 21

Bigram Detector

Answer 21

explanation of a mid-level detector in a feature network, which will detect pairs of letters.
*

Question 22

Visual processing Pathway Theory

Answer 22

proposed to consist of;
feature detectors,
letter detectors,
bigram detectors,
word detectors.
*

Question 23

Recognition Errors

Answer 23

*

Question 24

Distributed Representation(as opposed to locally represented)

Answer 24

The Feature Net seems capable of “deciding” or “inferring” corrections as necessary. Yet it is proposed that the Net does not actually “know” anything about language rules (not locally recognised), but rather is purely automatic with vast networks of interconnections and “decision” as to what is right is based on the relative activations of the collective (distributed knowledge).

Question 25

McClelland & Rumelhart Model

Answer 25

Further proposed that the Feature net Model may also have inhibitory functions by some firings upon other detectors. Also proposes that sequencing of firing as well as going from lower levels to higher, can also move in the vice versa direction. *

Question 26

Recognition by Components Model

Answer 26

Hummel and Biederman further proposed refinement of the Feature Net model by proposing that one could recognise objects by their components (or geons). (previous models were more explaining print word recognition etc, whereas this model worked for words and objects). *

Question 27

Geon

Answer 27

geometric ion. simple shapes which are the components of objects. Proposed only need 30 to have the whole "alphabet" of objects.

Question 28

Viewpoint Independent

Answer 28

Object can be readily recognised irrespective of viewpoint. Thought that geons can be recognised from any angle.

Question 29

Recognition via Multiple Viewpoints Model

Answer 29

data does suggest that objects have some degree of being viewpoint dependent-some aspects take longer to recognise. proposes Lower order detectors for lines and angles and higher order detectors for entire object aspects etc. both models agree on some form of hierarchical firing.

Question 30

Viewpoint Dependent

Answer 30

proposed that for some objects there is a limit on stored viewpoints and if not presented in align with one of them, takes longer to process (and mentally rotate) and recognise.

Question 31

What Pathway

Answer 31

Part of inferotemporal cortex with many neurons here found to fire preferentially to certain objects being viewed. Many of these neurons appear to have a view prefence also.

Question 32

Perceiver's Task

Answer 32

Arguments as to whether object recognition is viewpoint dependent or not. Possibly yes and at other times no. The perceiver's task may determine partly how objects are processed (eg is this a cup?vs is this the cup I showed you before?) etc.

Question 33

Occipital Cortices differences when viewing objects

Answer 33

133

Question 34

Prosopagnosia

Answer 34

inability to recognise faces. Can often tell if male or female though, and can recognise objects. Cannot even recognise own face. Either due to brain damage or may be congenital.

Question 35

Facial Recognition summary

Answer 35

Seems to require specially dedicated processing. Is strongly viewpoint dependent. (harder to recognise upside down faces than objects). has a powerful inversion effect.

Question 36

Inversion Effect

Answer 36

If object viewed from aberrant aspect, recognition is reduced. Severe(or powerful) inversion effect = recognition vastly decreased in aberrant view. eg upside down face is very difficult to recognise. * *

Question 37

Fusiform Face Area (FFA)

Answer 37

There seems to be in humans, a special system for recognising faces, housed in the Fusiform Face Area of the brain. Damage here may result in prosopagnosia. Arguments as to whether only used for faces, or if may also be recruited in specialised tasks such as recognition of individual tigers etc. It is recognised that people find it easier to recognise faces within own ethnicity.

Question 38

Holistic Perception

Answer 38

In contrast to feature networks proposed for object recognition etc, facial recognition seems to rely on holistic recognition, whereby recognised due to overall configuration, not as summation of components. Components contribute more in terms of their relationship with other components (ie eye spacing from nose etc).

Question 39

Brain Area involved in facial recognition

Answer 39

``` several areas seem to be involved in facial recognition, inclusing Fusiform Face Area Occipital Face Area Superior Temporal Sulcus. * ```

Question 40

Composite Effect (in facial recognition)

Answer 40

Has been demonstrated that easier to recognise top half of composite face if misaligned halves, but when halves aligned, whole face is considered, thus harder to work out who is top half. *

Question 41

Understanding of the Larger Context

Answer 41

Has also been demonstrated that some instructions/priming/expectations can influence what seen. Thus larger context means that there is definitely times when Top-Down Priming is occurring and this relies on prior memory and prior knowledge. (ie different instructions prior to a task and result in different perception)

Question 42

Speed Reading Summary

Answer 42

Speed reading actually relies on skipping more information, but if done appropriately, will still accurately "read". In time, can train eyes to follow pointer (as opposed to vs versa) and so eyes go faster and rely on more inference. However, for hard to understand material, relying on more guesswork (speed reading) will lead to inaccurate guesswork and problems with comprehension.