Cognitive Deck 1 Flashcards
Briefly describe how cognitive psychology developed over time?
1) Armchair psychology - based on logic and reasoning, not evidence. John Locke.
2) Wilhelm Wundt - one of first psychology labs, studying mind and behaviour as a science. Introspection, people analysing their own thoughts. But people not always reliable, biases.
2) Behaviourism - first half of 1900s, regards everything has a behaviour, learnt rather than innate. eg Pavlov and Skinner.
3) Nativism. Chomsky.
How can the mind be studies scientifically?
- Experimental studies. “Mental chronometry” – measuring the time that a mental process takes to be carried out. Pioneered by Donders in the 19th Century.
- Neuroimaging. MRI, PET, CAT.
- Case studies.
- Single and double dissociations. Single = documenting a patient who is impaired in domain X, but (relatively) well-functioning in all (most) other domains. Double = can we find another patient with the opposite impairment?
Define the following:
Agnosia, Prosopagnosia, Aphasia, Amusia.
Agnosia – loss of ability to recognise objects (despite vision being intact).
Prosopagnosia – inability to recognise faces (but objects can be recognised).
Aphasia – specific language impairment (in the absence of intellectual impairment).
Amusia – deficit in musical pitch processing, music memory and recognition (but auditory system is intact).
Sensation vs Perception
Sensation - detection of simple properties, e.g.: brightness, colour, loudness, sweetness etc.
Perception - interpretation of sensory signals, e.g., object recognition, identification of properties such as location, size, movement, etc.
Biologically, how does perception work?
Brain relies on signals received from sense organs (eyes, ears, skin, etc.) via afferent (incoming) nerves.
Neurons transmit brief electrical pulses with fixed amplitude and duration. Make excitatory or inhibitory connections with each other to create networks with one-to-many and many-to-one linkages.
85 billion neurons roughly in the brain.
Give some information about vision.
* How much of brain devoted to vision
* Receptors in eye
* How we see colour
1/3 of our brain space devoted to vision and corresponding action.
Optic nerve is formed from axons of approximately one million ganglion cells.
Receptors (photoreceptors) in the retina of the eye are rods and cones (about 120 million rods and 7 million cones) – modified neurons containing photosensitive pigment (rhodopsin).
Rods function at low light levels, cones in bright light.
Cones are colour-tuned – peak sensitivity to either red, green or blue (different rhodopsins). Output depends on wavelength of light.
White light is a mix of all visible wavelengths, all at the same intensity. “White” is therefore coded as equal output of red, green, blue receptors.
What is the structure of the retina?
Rods and cones at back layer of retina.
Network of connections between receptors and optic nerve performs local computations (basic analysis).
Response of photoreceptors and bipolar cells to illumination is graded. (response to light changes smoothly).
Amacrine cells and horizontal cells combine and contrast signals from adjacent photoreceptors. (refine the visual signal).
Ganglion cells generate action potentials and form the optic nerve.
Retinal processing involves ‘cleaning up’ of image and beginnings of feature extraction.
What is single-cell recording?
- One way we gained knowledge on visual system.
- By Hubel and Wiesel, won a Nobel prize.
- Rods and cones at back layer of retina.
- You take an animal, stick a tiny electrode in parts of brain and measure electric potentials going down pathways of the brain. Response of a single neuron.
What happens to visual info after the retina?
Signal via the LGN at the base of brain is delivered to back of brain (generally in occipital lobe).
Input ultimately feeds into the visual cortex.
Separate structures exist in visual cortex for extracting information about shape, colour, position, motion, etc.
Feature processing structures in visual cortex arranged in layers.
Collections of ‘simple cells’, each responding to a line or edge in illumination reaching a particular region in the retina.
What is segmentation?
Part of perception.
- First stage of object recognition – visual features that belong to the same object are grouped together.
- Figure-ground perception - a prerequisite for object recognition (applies to all objects in visual field).
- Assignment of figure and ground can be ambiguous but distinction must always be made.
What are the Gestalt principles of grouping?
Grouping of elements to make a “figure” is determined by a set of basic principles that are automatic, and innate.
By similarity, connectedness, proximity, continuity, closure, common fate.
What is high level perception?
In everyday cognition, we recognise perceptual information effortlessly, e.g.,
identifying objects (or faces)
reading printed words
understanding speech
Processing is typically extremely rapid and efficient.
Template theories and their issues.
When we recognise something, we match it up with the closest instance (‘template’) of things stored in our memory. Pattern recognition is based on global similarity match between sensory input and templates stored in memory; best match is output of recognition process.
E.g., letter recognition. Theory is intuitive and computationally simple – works for some machine recognition systems, e.g., bar code readers or Automated number plate recognition (ANPR).
“Real” input is often quite variable (eg letters can look different).
Needs to be able to deal with changes in viewpoint, quantification of “similarity”.
Feature matching theories and their issues.
Pattern recognition is based on identification of features in the visual array. “Features” are fragments or elementary components of a larger pattern. For recognition purposes, objects can be defined in terms of their component features (eg ‘A’ is made of two diagonal lines and a horizontal line). Advantage: a limited number of features can be used to represent a
very large number of objects.
Visual variability for different fonts, eg hard to come up with one feature set for letter B. Same for more complex objects eg a cat has different visual features.
What is a potential solution to issues for both template and feature matching theories?
- Represent objects in object-centred fashion:
- object is perceived in a coordinate system centered on the object, not the viewer.
- brain aligns a reference frame, using object’s axes of elongation and symmetry.
- uses that reference frame to measure relative positions of object components.
Recognition by components with evidence. Limitations.
Biederman.
- Objects can be described in terms of small set of geometrical parts named geons - about 24.
- geons are simple 3D shapes: cylinders, cones, wedges, etc., each in ~15 sizes and builds.
- Brief (100 msec) presentation of images with 65% deletion of contour:(A) deletion from middles of segments, (less diagnostic of geon structure).(B) deletion at vertices (critical for identifying geon structure).
-
Correctly identified objects:
- condition A: 70%
- condition B: 45%
- Vertices (which provide info about geon structure) seem indeed critical for object recognition.
- difficult to distinguish between objects with identical (or very similar) geon structure, e.g., horse and cow?
- Recognition of specific individuals (e.g., faces) - if there is a generic geon construct of face, how does the visual system distinguish between different faces?
- Works well for artefacts, but less so for natural objects (mountains, trees, etc.) – what is the geon representation of a puddle?
How is face recognition different to object recognition?
Inverted objects are easier to identify than inverted faces (eg upside down faces).
This could imply that face recognition is more ‘holistic’ than object recognition – relies on whole face (configuration of features), not just on individual features.
- We process faces using local features (e.g., nose, eyes, etc.) and most importantly, their configurations.
- When faces are inverted, only local features are available – which are not very dissimilar between normal and inverted view.
- But when the faces are right-side-up, we can process the configural information as well, and so the inverted mouth and eyes look odd.
Dichotic listening task.
- Two different messages are transmitted to the two ears through headphones.
- Participant is asked to “shadow” one of the messages while ignoring the other.
- After a particular episode/trial, s/he can be asked questions about
the unattended information.
Cherry (1953)
- Participants almost entirely unaware of information presented to the
unattended ear:
- knew whether or not a voice was presented
- could report physical attributes of the voice (e.g., gender of the speaker, etc.)
- knew very little about the content of the message
- unable to report the language
- Implication: information in the unattended message is processed only to a very “shallow” degree.
Broadbent’s filter model
- “Chunks” of sensorial information are represented as balls.
- Attentional selection symbolised as Y- shaped tube through which information must pass.
- Information enters through sensory channels and is filtered as it proceeds.
-
Tube accepts only one ball at a time, with hinged flap acting as a filter.
Claim that switching from one channel to another takes effort comes from this model.
What counters Broadbent’s filter model?
Moray (1959) …
- reported a surprising finding: participants able to report if their name is presented on the unattended channel.
- “Cocktail party phenomenon” – highly pertinent stimuli (such as one’s name) can suddenly capture one’s attention in a noisy environment.
- Such effects of “high priority” appear problematic for static filter model such as Broadbent’s.
Treisman’s attenuation model,
- Unattended information is not entirely blocked (as e.g., in Broadbent’s model).
- Instead, attention acts as a selective filter.
- Physical properties are analysed, then knowledge about words is assessed (mental lexicon), entries in the mental lexicon are stored in terms of frequency of occurrence and relevance.
- Recognition thresholds can be temporarily lowered by current priorities.
- If signal passes both filters, meaning is analysed.
This counters the cocktail party effect as ones own name has high relevance and frequency.
- Expectations can temporarily lower the thresholds of words in mental dictionary.
- Hence, fragments of unattended information will sometimes access meaning if they fit into the context of the attended message.
Late selection theories, with evidence.
- Provocative claim: selection of information regarding conscious awareness occurs only after analysis of meaning.
- Claim: contrary to theories outlined so far, all sensorial information is always processed nonselectively and in parallel, up to the level of meaning (!)
- Output of sensorial processing is placed in short-term memory.
- Information in STM is quickly lost, and this loss acts as “attentional bottleneck”.
- Display with two dimensions is presented.
- Participants are instructed to attend to one dimension, and ignore the other.
- For instance (Allport et al.,1985): Name red picture, ignore green picture. Time it takes to name the target is measured with voice-activated trigger. In critical condition, object you were ignoring in previous trial then becomes object of focus. On critical trials, average naming times to the target (i.e., red) object are slower if ignored on the previous. Suggests that ignored information is not simply discarded, but is actively suppressed.
Capacity theory.
Kahneman’s theory of attention.
- Inability to perform two tasks at once is not the result of built-in attentional bottleneck (early, late, etc.).
- Rather, people have limited-capacity pool of attention to carry out mental activities.
- If an activity is easy, very little attentional capacity is used up
- if an activity is difficult, it uses up all or most of resources
Depends on
- task demands (some activities are more demanding than others).
- arousal (determined by alertness, circadian rhythm, medication, etc.).
- individual differences/dispositions (some individuals are better than others in paying attention).
- momentary intentions (how important it is to you).
Evidence for capacity theory.
Dual-task experiment: individuals carry out two activities at once, and we measure the impact of one task on performance on the other.
Johnston & Heinz (1978)
- Primary task – a light flashes repeatedly and at random intervals, and participants press a button as quickly as possible when they detect it.
-
Secondary task – at the same time, individuals “shadow” words simultaneously presented to both ears, either
- by simple repetition (i.e., identical message was delivered to both ears) - EASY
- according to physical category (“shadow what the female voice said”) - HARDER
- according to semantic category (“shadow the word that is a city”) – HARDEST
- fastest when participants simply shadowed the message (“one list” condition)
- slower when participants shadowed according to physical criterion
- slowest when participants shadowed according to semantic criterion.
Automatic vs Conscious attention processing.
Automatic
→ The process occurs without intention, without a conscious decision
→ Not open to conscious awareness or introspection
→ Consumes few if any conscious resources; i.e., it consumes little if any conscious attention
→ Operates very rapidly, usually within 1 sec
Conscious
→ The process occurs only with intention, with a deliberate decision
→ Is open to awareness and introspection
→ Uses conscious resources; i.e., drains the pool of conscious attentional capacity
→ Is slow, taking more than a second or two for completion
Shift from attention-demanding to automatic mode of processing occurs through practice.
Can link to troop task (reading word is practiced so automatic, naming of colours less so, requires more effort).
STM vs LTM
Short-term memory (STM
- Holds information only very briefly, and has severely limited capacity
Long-term memory (LTM)
- Holds information over long periods, and has potentially unlimited capacity
What did Miller find about STM?
- Individuals can hold ~7 items simultaneously in STM (Miller, 1956)
- Capacity largely independent of nature of items:
- digits: 9 2 4 7 3 8 1
- letters: T M F J R L B
- words: HAT CUP BOOT LOG TEA ROOM WINE
- But capacity is less than seven for items that don’t mean anything:
- ‘nonwords’: PLOCK TWEB DRUNG VUKE DRAND BLAY
- If possible, short-term memory makes contact with representations which are permanently stored in our minds.
What is the digit span task?
- Common component of many IQ tests.
- Screening test for clinical impairments (e.g., dementia).
- Read out three digits (3, 1, 4) in a monotonous voice, with one digit read per second.
- If correctly recalled, increase to four digits** (7, 3, 5, 1).
- If correctly recalled, increase to five, etc. (in a clinical setting, task is often stopped after five digits).
“Superspan” lists present lists of items longer than max. capacity of STM (~7). Free recall task: recall the list of items, in any order.
HM.
- suffered from intractable epilepsy.
- part of his temporal lobes surgically removed, including hippocampus.
- showed defective LTM learning: inability to acquire new information post surgery (“anterograde amnesia”).
- but had normal STM!
KF.
- brain damage following motorbike accident.
-
LTM performance in normal range (learning of word pairs, lists of words,
etc.) - STM dramatically impaired: usual digit span of 7±2 reduced to max. 2 (!)
Double dissociation with HM, therefore likely STM and LTM are separate.
Modal model of memory, and criticism.
Atkinson and Shiffrin, 1968.
- For information to enter LTM, it must pass through STM
- STM has limited capacity of about seven or so items
- Information can be ‘refreshed’ in STM by means of rehearsal
- The longer an item held in STM, the more likely to be transferred to LTM
Then why do patients with STM deficits function well in everyday life if everything must mass through STM?
Working memory model.
- STM consists of more than one component.
- Central executive: integrates information. “Amodal” (not tied to a particular type of code; language, vision, etc.).
- Supplemented by two “peripheral” systems which are ‘modality specific’: visuo-spatial sketchpad and phonological loop.
Word length effect in memory.
- STM performance is better for shorter than for longer words. E.g., wit sum pill bag top.
- easier to recall (i.e., more words recalled on average) than university opportunity aluminium constitution auditorium.
Alternative view of Miller for STM capacity.
Revised, time-based, view (Baddeley et al., 1975):
- No inherent (built-in) limitation of the PL capacity.
- Rather, STM capacity is constrained by the time it takes to report the items.
- Measured capacity is determined by number of items that can be reported before they are lost from the phonological loop.
- Amount of items correctly recalled (i.e., 90% or higher accuracy) is not 7, but whatever can be uttered in about 2 seconds.
Intuitive view of concepts.
Most things should be definable by their properties. Eg cat = catches mice, furry, whiskers, sleeps a lot.
Classical view of concepts.
Plato, Locke etc.
Concepts can be defined by an exhaustive list of necessary and sufficient semantic features.
Issues with the classical view of concepts.
- Hard to come up with an exhaustive list of features.
- Some birds fly and some don’t, but still all birds.
Rosch prototypes for concepts.
Eg all 9 family members have similar features but no two members have exactly the same.
Prototype has all the features of it’s category.
We organise mental representations of a conceptual category around the average of various instances we are familiar with.
Therefore a graded membership - some closer to the prototype than others (penguins kinda far).
Fuzzy boundaries between conceptual categories.
Exemplar theory for concepts.
We don’t compare a new stimulus to a prototype (prototype theory), instead we compare to all stored instances of members of that category.
Classifying a new instance is related to how easily it brings instances of other category members to mind.
Which is better, exemplar theory or prototype theory?
Evidence from neuroimaging and acquired brain damage suggests both are used.
Visual cortex involves prototypical mental representations, prefrontal cortex and basal ganglia involved in learning exemplars.
What are the categories for category-specific deficits in concepts?
Animate objects, inanimate biological objects, and artefacts.
What is homoeconomics?
A discipline which explores how people interact when they exchange goods and services. A person has infinite ability to make rational decisions.
What is behavioural economics and heuristics?
Studies effect of psychological, cognitive, emotional, cultural, and social factors on the decisions of individuals and institutions.
People often apply heuristics (mental shortcuts or rules of thumb when making a fast decision). Often wrong and lead to biases, violating rationality.
What is availability bias?
Probability estimate of occurrence is judged by its availability in memory. Eg after 9/11 people opted to travel by car over plane.
For low-probability but high risk events, humans tend to use mental shortcuts rather than computing the objective risk.
What is conjunction fallacy?
People often believe that with more information it is more likely to be true, even though the probability of the simple answer is more likely.
