COG PSYCH 4

• Patients with apperceptive agnosia can perceive an object’s features but not the object in its entirety

  • They cannot link this input to visual knowledge but can draw well from memory

What is apperceptive agnosia?

• There are variations in “stimulus input”

• Uses contextual influences

• This word is ‘THE’, even if the H looks like an A, we know that it spells ‘THE’.

How is the process of object recognition complex?

• Top-Down – processes shaped by knowledge

  • Concept-driven

  • Prior knowledge and experience

  • EX) Context effects

• In everyday life, we tend to do top-down processing. 

  • You have your answers already, you know how things will go.

    • EX) “How are you?” “I’m doing good, you?”

What is Top-Down Processing?

Process shaped by the stimulus

• Data-driven, raw sensory details

• EX) “How are you?” “I’m running late to a class.”

What is Bottom-Up Processing?

• Recognition begins with identifying visual features in the input pattern.

  • Vertical lines, curves, diagonals, etc.

  • Evidence for feature detectors in the visual system

• Larger units can then be detected by assembling the smaller pieces

How are features important?

• Tasks in which participants examine a display and judge whether a particular target is present

  • Efficient when target is defined by a simple feature

  • Slow when target is defined by a combination of features

• These findings suggest that feature analysis is a separate step from the step in which detected features are combined

What are Visual Search Tasks?

• A device used to present stimuli for precise amounts of time

  • Tachitoscopic presentations with modern computers

What is a Tachitoscope?

feature detection

What does object recognition start with?

• It’s easier to perceive and recognize letters-in-context than if they appear in isolation

• The word-superiority effect can be demonstrated using a “two-alternative, forced-choice” procedure

  • Participants are shown a word (i.e. ‘dark’) or a single letter (i.e. ‘E’)

  • Did the display contain an E or a K?

• More accuracy when the original stimulus was a word rather than a single letter.

What is the Word-Superiority Effect?

• Well-formedness – how closely a letter sequence conforms to the typical patterns of spelling in the language

• The more well-forced a letter sequence..

  • The easier it is to recognize the sentence

  • The greater the context effects produced by the sequence on recognition

• EX) HZYQ vs. FIKE vs. HIKE

What is the Degree of Well-Formedness?

• Well-formedness also influences errors

  • Likely: DPUM misread as DRUM

  • Unlikely: DRUM misread as DPUM

  • This goes to show that there are automatic processes

• People can perceive stimuli as being more regular than they actually are

  • TPUM misread as TRUMPET

How does Well-Formedness influence errors?

A hierarchical network of mental "detectors" used to explain how the human brain recognizes patterns, letters, and words.Starts from bottom-up

What is a Feature Net?

• Each detector in the network has an activation level.

  • With input, this activation level increases

  • Some detectors will be easier to activate than others

• Detectors fire when their response threshold is reached

  • Similar to a neuron’s threshold for firing an AP

• Individual detectors are probably complex assemblies of neural tissue, not individual neurons or groups of neurons

How is the Feature Net designed?

• Recency – detectors that have fired recently will have higher activation levels

  • A warm-up effect

• Frequency – detects that have fired frequently will have higher activation levels

  • An exercise effect

In a Feature Net, what do the starting activation levels depend on?

Yes, but only when it is a well-primed detector.

The priming will depend on frequency and recency

• The network is biased to respond to inputs similarly to how it has responded previously

Will a weak signal be enough to trigger something?

• The bias to recognize frequent or primed words can result in errors, but it also results in correct recognition more often than incorrect recognition

  • It helps more than it hurts

How can we make recognition errors?

No. Knowledge is not locally represented. Rather, feature nets contain distributed representation.

• Knowledge in a network is reflected by relationships across detectors.

Is knowledge locally represented?

• The same mechanisms that enable the network to resolve ambiguous inputs (i.e. sloppy handwriting) and recover from errors can also result in recognition errors

• Perfect accuracy is sacrificed for efficiency

What is ‘Efficiency vs. Accuracy’?

• McClelland and Rumelhart Model

  • Emphasizes the role of inhibitory connections among detectors

  • Information flows bottom-up, top-down, and within the same level

  • Includes excitatory connections and inhibitory connections

• Recognition by components (RBC) model

  • Applies the feature net model to recognition of three-dimensional objects

McClelland and Rumelhart Model on this card. RBC model on the other card

What are the descendants of the Feature Net?

• Researchers continue to develop recognition models

  • Models designed to explain human capacities

  • Models designed to build computer systems that can recognize objects

• How do the models work?

  • Detection and cataloguing of visual input

  • Analysis of incoming patterns

  • Inhibitory and activating connections

What are the known approaches and new frontiers to Object Recognition?

• Prosopagnosia – an inability to recognize individual faces (including their own) despite otherwise normal vision

  • Unable to assign identity.

• Some people are super-recognizers, with extremely accurate face recognition

  • Opposite of prosopagnosia

  • Advantages for politicians and salespeople

How do we know that recognizing faces specifically seems to involve specialized neural structures?

Faces.

• Faces show a powerful inversion effect

  • The effect is much larger than for other types of stimuli

• The difference between words and faces is that faces are much more similar, whereas words have more variability

Are faces or words more recognizable?

Yes

• Some researchers suggest that other types of recognition are special in the same way a facial recognition

  • EX) A bird-watcher who developed prosopagnosia lost the ability to distinguish faces and types of warblers

• The fusiform face area (FFA) is particularly responsive to faces, but high levels of activation can also be produced by tasks requiring subtle distinctions

Are other types of recognition special in the same way as facial recognition?

• Holistic Perception – perception of the overall configuration rather than an assemblage of parts

  • Evidence from the composite effect

• Same-race vs. Cross-race facial recognition

  • People tend to be more accurate in recognizing faces from their own racial background

  • Implies potentially different recognition mechanisms

What is the Holistic Recognition style?

• Limits of Feature Nets

  • Some target objects depend on configurations, not individual features

  • Knowledge that is external to object recognition nevertheless influences recognition

What are the limitations of Feature Nets?

• Some top-down effects can be explained by feature nets.

  • EX) The Word-Superiority Effect

• Other Top-Down effects, however, need more explanation

  • Larger priming effects are evident for words viewed in a sentence rather than in isolation

  • Context and expectations influence perception

What are the Benefits of Larger Contexts?