cognition midterm 2

Children and Gender Essentialism

- "I'm going to tell you about a baby girl named Zillah. After she was born she went to live with her uncle on an island. On this island there were only boys and men." OR "I'm going to tell you about a baby boy named Arza. After he was born he went to live with her aunt on an island. On this island there were only girls and women." - Ss then made forcedchoice judgments about stereotypical properties. When Zillah/Arza is a big kid, what does he/she like to do? Collect dolls/tools Sew/build things Put on make-up/go fishing Play with tea set/toy truck etc. Five year olds choose nature over nurture: They predict that the grown child will prefer gender stereotyped behaviors over those common in their environment.

Sensory Selection -- Attenuation Theory

- An attenuator modulates channel strength • Attended channel: full strength • Unattended channels: partial strength • If they are relevant to current WM content. - A dictionary then prioritizes items • Depending on their activation threshold in semantic LTM. - Attended information enters WM. Unattended information might enter as well, depending on its strength.

1. Transformation Experiments

- Children are told about an accident or operation that changes an animal's appearance to that of another animal. - There were asked whether the animal's category membership had changed. -e.g Biological Kind, minerals, artifacts Category membership is sometimes a matter of an object's underlying essence or nature, not just its observable features: - Children and adults were less essentialist when it came to artifacts: You are what you can do! - Children behaved as similarity-based classifiers: You are what you look like. - Older children and adults behaved as essentialists: You are what you have on the "inside." Differences - Natural kinds • Existence does not depend on humans. • There are sciences that study them. - Artifacts • Existence depends on humans. Subsequent research showed that even young children have essentialist beliefs for some categories

Limits on Automaticity

- In the consistent mapping condition, Ss were able to learn to detect particular letters against a field of particular distractors. - But in the varied mapping condition, they couldn't learn to detect just any letter against a field of random distractors.

Testing the Hierarchical model -- sentence verification

- It predicts that response times should increase with the network distance needs to be traversed. Examples - Fast: can A canary sing? No traversal of the network is required (sing is stored directly with canary) - Slower: can A canary fly? - Slow: Does A canary have skin? One must go from canary to bird to animal (to verifty the presence of has skin.)

Outcomes of attention -- Precueing

- S must maintain fixation (not move their eyes) during the entire trial. - Fixation → Cue (80% Accurate) → Fixation → Target (50ms) → Left or Right? (RT recorded) - Valid cue, neutral (no cue), or invalid cue - Faster RTs if cue is valid. - Slower RTs if cue is invalid. - Visual attention can shift without eye movements!

Processing capacity and perceptual load -- Stroop Effect

- Some stimuli have a load that is so low that they are almost impossible to ignore - Word reading interferes with color naming. (Word reading slows color naming on conflict trials; Word reading speeds color naming on congruent trials.) - Number naming interferes with counting.

Processing capacity and perceptual load -- Testing Load Theory

- Ss asked to detect identity of 1 of 2 target characters in a display. For example, Please indicate whether the display has an X or N. - Subject were given either a low-load (easy) or high-load (difficult) primary task. (i.e. Easy, hard, easy + distractor, Hard + distractor) - On easy trials there is excess capacity, so some attention gets allocated to the distractor. RTs slow down. - On hard trials there is no excess capacity, and so little attention gets allocated to the distractor. RTs slow down less

Testing For Defining Features

- Ss listed features of eight categories (Fruit, fish, sport vegetable, bird, furniture, vehicle kitchen utensil). Only features produced by a substantial number of Ss were kept. -Ss were presented with 30 test words for each category (e.g. Test words for category fish: trout, shrimp, shark, lobster, tadpole, etc.). They judged if the word was a member of the category (on a 1-7 scale). - Other Ss listed the features of the test words. - Each category was analyzed to identify necessary and sufficient features. No evidence for defining features in many categories (against the definitional approach). However, the fact that people aren't conscious of defining features doesn't mean that they don't exist. There are stronger tests.

Evidence — Concrete vs. Abstract Materials

- Ss study pairs of nouns - Nouns were concrete (Ex fork-tape potato-tent etc.) or abstract (Ex justice-sarcasm substance-deficit etc) - "Link the pair of nouns with a mental image." Ss had to draw image they generated OR "Link the pair of nouns with word, phrase, or sentence." Ss had to write it down. - Cued Recall Test: fork-??? OR justice-??? - Imagery enhances memory of concrete material. - But it hurts memory of abstract material. For abstract material, semantic elaboration is better.

Evidence for Causal Relations and Classification

- Ss told about three features of a novel category (e.g., Kehoe ants). - Ss told about causal relations linking the features into a causal chain OR Ss not provided with any causal information. - Classification Test → Ss then rated the category membership of items with all features but one. Missing-X item (Had Y and Z but not X) and Missing-Z item (Had X and Y but not Z). - A missing cause feature (X) has a greater impact on categorization than a missing effect (Z). This is known as the causal status effect. - Also, any inconsistency (or contradiction) in which a cause is present and effect absent or vice versa lowers classification ratings drastically. This is known as the coherence effect.

Picture superiority effect

- Subjects studied items presented for 5 sec. each. Maximum of 2000 items (3 hours) per day - words, Simple images of everyday objects (e.g., a dog, a hat, etc.), Vivid images - different number of items: 20, 40, 200, 400, 1000, 4000, or 10000. - 2 days later: Recognition Test —> Two items presented, one studied and one new. Ss choose studied item. - Memory for images is very good (especially when vivid) and substantially better than for words. - 95% correct on 1,000 vivid pictures! - 83% correct on 10,000 normal pictures!

Example of Activation Thresholds

- The activation thresholds of words vary with their strength of encoding in LTM. - Strongly encoded ➝ low threshold ➝ easily activated (e.g., one's name). - Weakly encoded ➝ high threshold ➝ difficult to activate (e.g., rutabaga).

Sensory Selection -- Late Selection Model

- the filter operates after semantic processing. - Therefore, all stimuli get semantically processed, but not all enter WM and consciousness. - The filter selects which information should be passed on for further processing after it has been semantically processed. Irrelevant information is discarded.

Propositional Networks

A proposition Is a language-like structure that represents a simple fact. - The cat is on the mat. - A dog is a type of animal. - A chair has a seat. - Sunrises cause roosters to crow. A propositional network is a graphical representation of a set of related propositions - taxonomic hierarchy - part-whole relations - causal relations

Essentialism

According to Plato, all concepts have an essential reality behind them (an idea or form) that makes the concept what it is. - Worldly instantiations of a concept are imperfect manifestations of the underlying form. • Also known as Platonic idealism. Essentialism has influenced thinking in numerous intellectual domains. - Metaphysics, mathematics, etc.

2. Classification and Diagnostic Reasoning

According to essentialism, an essence CAUSES its observable features. Classification is often an act of diagnostic reasoning in which one reasons from observable features to an essence - Diagnostic reasoning occurs when one reasons from effects to causes on the basis of causal knowledge - ex. doctors diagnose diseases on the basis of observed symptoms. Diagnostic evidence can be undermined when a symptom has an alternative explanation.

Processing capacity and perceptual load -- Summary of Selective Attention

According to load theory, whether selective attention appears to operate early or late depends on: - Whether the system has left over capacity [more excess capacity → more processing of distractors] - The capacity the distractor requires to be processed [low load distractors → hard to ignore].

2. Ambiguous category membership.

According to the definitional approach, Category membership should be unambiguous. e.g. - Either a shape is a triangle, or its not. - Subjects judged category membership of everyday categories. • E.g., which are fruits?: strawberry, acorn, pear, etc. - Subjects returned one month later and answered the same questions. - There's uncertainty between people • Many borderline cases: Some objects are half in and half out of a category. - However, people might differ from one another • e.g. A pumpkin might definitely be a fruit for half the people, and definitely not for other half. - The question of where there is uncertainty within individuals can be answered by looking at the results one month later • Subjects changed their mind 22% of the time for the borderline cases!

3. Graded category membership.

According to the definitional approach, Category membership should be ungraded. - All category members are equal. E.g - An uncle is as good an uncle as any other. Subjects judged (on a 1-7 scale) "how good an example" category members were of their category. • E.g., which are good birds?: robin, swallow, eagle, etc. Category members vary in how good an example they were of the category -- typicality effects. Are typical item those with many high family resemblance features (features that appear in many category members)? - Ss rated the typicality of instances of many categories. • 120 members of 6 superordinate categories. Furniture, vehicle, vegetable, weapon, fruit, clothing. • 90 members of 6 basic level categories. Car, truck, airplane, chair, table, lamp. - Other Ss listed features of the same items. • A family resemblance score was computed for each feature: The number of category members it appeared in. A family resemblance score was computed for each category member: The sum of the family resemblance of its features. A category member's family resemblance score determined its typicality: Correlations over categories between .84-.94. Typical category members are those that share features with many other category members.

1. Existence of defining features.

According to the definitional approach, People should be able to name the defining features of categories. But, for many categories identifying necessary features is not so easy. - Many features that come to mind first are not really necessary. - And those that necessary are rarely jointly sufficient. e.g. - When asked what defines dogs, people will often respond with features like... Has four legs Plays fetch, But it is easy to see that in fact these are not defining - Some features of dogs ARE necessary. All dogs have lungs, need food, will die in the cold, and breath air. But they are not jointly sufficient, because other animals have them.

Automaticity in Visual Search

After training, subjects were presented with simpler test trials. - The number of frames was 1 rather than 20. - During test, the key dependent variable is response time (RT): How fast could Ss correctly judgment the presence or absence of a target letter? - Consisting Mapping Condition: After extensive training, RTs was about the same for different memory set and frame sizes. The detection of the target letters became insensitive to load i.e. automatic, turned into a very low load task. - Varied Mapping Condition: RTs continued to increase with both memory set size and frame size.

Essentialism and Social Kinds

Although biological essentialism is technically incorrect, it serves as a useful first order approximation. However, when it comes to social kinds, essentialism can be seriously wrong. Many social kinds are partially or even entirely socially constructed. - They are associated with many properties, norms, and expectations that are based on historical cultural conventions. In contrast, such kinds become essentialized when they are treated as reflecting only biological distinctions. - Gender essentialism: the belief that the distinctive qualities of men and women are due to biological factors and so are fixed, intrinsic, and innate. - Racial essentialism: the belief that racial groups form discrete genetic categories that are fundamentally different.

Attention and binding -- Feature Integration Theory

An important step in visual processing is object perception. - The act of recognizing that a set of features all belong to the same object. According to feature integration theory... - Primitive features (color, shape, texture, etc.) are first perceived independently, each associated with a spatial location but not with each other or any object. - They then undergo feature binding to represent which go with which objects Feature integration theory stipulates that feature binding requires cognitive resources (i.e., attention). This prediction has been tested by use of: - Feature conjunction tasks - Visual search tasks

Outcomes of attention -- Covert Attention

Attention is a mental phenomenon in which some stimuli are processed. - E.g., listening to the left or right ear. In vision, we can shift attention overtly by moving our eyes. - But we can also shift visual attention without eye movements. - This is referred to as covert attention.

Imagery and memory

Bower and Winzenz presented Ss with paired associates (ex. Boat-tree, plan-leaf) and compared two encoding conditions: - imagery condition — Ss were asked to form a mental image. - repetition condition — Ss were asked to repeat the words. Their memory for the word pairs was then tested (cued recall). Creating visual images is a dramatic way to increase memory.

1. Variability

Categories differ in how variable they are. - E.g., dogs are more variable than cats. Prototypes lose information about category variability, exemplars don't. - In fact, people are sensitivity to variability information

2. Correlations

Categories often exhibit correlations between their features. - E.g., small birds sing, large ones don't. Prototypes lose information about correlations, exemplars don't. - In fact, people are sensitivity to correlations

Function of Categories and Concepts

Classification - Allows us to treat different things as the same. Communication - We communicate with each other by using words that refer to concepts. Prediction - Predict invisible properties.

Summary of Causal Model Theory

Classification is not just based on feature-based similarity. - As in prototype or exemplar theory In fact, some categories share no features at all... Goal-derived (or ad hoc) categories: - Have members that satisfy a goal. - Often formed spontaneously. e.g. "Things to take out of the house in case of fire." → Money, children, photos, passport, pets, etc.

Summary of Filters

Clearly, people can choose which incoming sensory information to prioritize. - Yet, filter theories are too simplistic. • Unattended channel is never fully filtered out. • Supposedly unattended information can even influence attended information unconsciously.

Real-World Example of Causal Status

Clinical psychologists diagnose mental illnesses given patients' symptoms Based on the DSM*. Kim and Ahn assessed clinicians' theories of mental disorders. - How symptoms cause one another. - How those relations affect diagnosis. Individual clinicians have their own theories of mental disorders. - Not part of the DSM itself. For better or worse, those theories affect how Clinicians diagnose patients. • More causal symptoms have more influence on their diagnosis of patients

Feature Integration Theory

Conjunction: connection of two or more things. Integration: the act of making a whole. Feature integration theory: the mind first represents features (color, shape, texture, etc.), then must bind/conjoin/integrate those features together into singular representations of objects.

Summary of Theories of categorical structure

Defining (theory) → Defining features (Mental representation) → Rule application (Method of classification) → Rule discovery (Method of learning) → Classes defined by people (Most applicable domain) Prototype (theory) → Prototype (Mental representation) → Similarity to prototype (Method of classification) → Abstracting a prototype (Method of learning) → Social groups (Most applicable domain) Exemplar (theory) → Exemplars (Mental representation) → Similarity to Exemplars (Method of classification) → Memorizing Exemplars (Method of learning) → Classes of objects (Most applicable domain)

Advantages of Exemplar Theory

Exemplar theory also accounts for the key empirical results: - No defining features → All feature dimensions contribute to the computation of similarity. - Borderline cases → Unclear cases are equidistant from the members of two (or more) alternative categories. - Typicality effects → Atypical items are more distant from other category members. Our brains appear to track the particular category members we observe. As a result, it keeps track of subtle details that prototypes alone don't. - Category Variability - Feature correlations - Exceptions - context, etc.

Exemplar theory and memory

Exemplar theory does not claim that you are consciously recalling each past category member. - If you were, classification would take forever. Instead, your memory is accessing all past category members in parallel. - And telling you to which category the new object has the greatest total similarity. e.g. A is a dog because its total similarity to dogs is greater than its total similarity to any other category.

Feature Conjunction Task

Fixation → Display with two numbers and four figures (200 ms) → Random dot visual mask → Cue to-be-reported object - Load condition: Stimulus → Cue → Report number s (e.g., 1 and 8) → S reports features of object at cued location. No. of conjunction errors* measured. - No-load condition → Cue → Stimulus → S reports features of object at cued location. No. of conjunction errors* measured. - On 12% of trials, subjects reported seeing objects made up of features from two different objects. (load trials had more conjunctive errors than no-load conditions) - When there's not enough processing capacity, you get illusory conjunctions. It implies that attention is needed to bind features together to form mental representations of objects.

Sensory Selection -- Evidence for early selection model (1)

In 1956, Colin Cherry assessed the filter's properties using dichotic listening. - Subjects hear two messages, one in each ear, and are asked to attend to one of them. - They shadow by repeating what they hear in that ear. Afterwards, Cherry asked what they heard in the unattended ear. Cherry found little processing of the unattended channel. Subjects didn't notice when... - ...speech was reversed. - ...the language changed. - ...a word repeated 35 times. In particular, virtually nothing about the meaning of the sentence was noticed.

Outcomes of attention

In general, attended stimuli are processed more quickly and accurately than irrelevant stimuli

Spatial Knowledge

Information in semantic memory also includes spatial knowledge. Spatial knowledge appears to have a hierarchical structure. - Images have parts, which have subparts, etc. Evidence for hierarchical organization comes in the form of distortions

Multidimensional Spaces

It is often useful to represent memory as a multidimensional space in which memory items vary along many dimensions. This convention allows us to represent similarity as distance: Nearby objects are similar, distant objects are dissimilar.

Imagery and the history of psychology

Many philosophers speculated that thought consists of nothing but a sequence of visual images. - Aristotle - The British empiricist John Locke And visual imagery seems to have played a special role in the history of science - Friedrich Kekule had dream with snake biting its tail, helped him discover arrangement of benzene atoms. - Albert Einstein's special theory of relativity emerged when he imagined himself traveling next to a light wave.

2. Classification by Similarity

New objects are classified according to their similarity to the prototypes. - Similarity is the distance between two objects in psychological space. Similarity to the prototype determines typicality e.g. New face is an Abbot because it is closer (more similar) to the Abbot prototype than the Baker prototype.

Attention and Mental Disorders

Our attentional systems are vulnerable. Basic ADHD facts - ~6% of children ages 3-17 have been diagnosed with ADHD at least once. - Diagnosed approximately 2-3 times more often in males than in females. - Inherited: A child with an ADHD sibling is 300-500% more likely to have ADHD. Diagnostic tests for ADHD use some of the tasks presented in class Larger Stroop effect in those with ADHD. In those with ADHD, brain images show less activity in the frontal lobes and more activity in sensory regions. - The PFC's ability to suppress irrelevant tasks and stimuli is compromised. Ritalin is a stimulant that increases frontal lobe activity. - Person feels calmer, more "focused."

Properties of Biological Essentialism

Properties - Immutability → Biological kinds do not change (evolve) over time. - Strong inductive potential→ Features displayed by one instance of a kind are often displayed by other instances. - Sharp category boundaries→ Something can be a member of only one biological kind. - Innate potential→ Members of biological kinds have predetermined developmental trajectories. Problems: - Theory of evolution shows that species are NOT static, unchanging entities. - And, biology has shown that boundaries between species are not so sharp. E.g., species cannot be identified as sets of interbreeding individuals. • Xs can breed successfully with Ys, Ys with Zs, but not Xs with Zs. • X males can breed successfully with Y females but not vice versa.

Advantages of Prototype Theory

Prototype theory accounts for the key empirical results: - No defining features → All feature dimensions contribute to the computation of similarity. - Borderline cases → Unclear cases are equidistant from two or more prototypes. - Typicality effects → Atypical items are more distant from the prototype. - Statements about prototypical objects are verified faster. RT (An apple is a fruit) > RT (A kiwi is a fruit) - Typicality effects → When listing examples of birds, people list robin before ostrich. - Typical objects are affected more by priming → See pp. 270-272 in book

Prototype Theory and Memory

Prototype theory does not claim that you are consciously recalling every prototype. - If you were, classification would take forever. Instead, your memory is accessing all prototypes in parallel. - And telling you to which prototype the new object has the greatest similarity.

Summary of Prototype Theory

Prototype theory is very successful theory of categorization. - It accounts for many important empirical results. • No defining features, typicality effect, borderline cases, statement verification times, etc. Nevertheless, we shall see that the theory also faces some important challenges

Summary of Essentialism

Psychological essentialism is an innate tendency to ascribe essences to kinds, even if we don't know what they are. - Technically wrong for biological kinds but at least approximately correct. - Often very wrong for social kinds. Even young children treat some categories as essentialized (e.g., gender).

Prototype Theory

Rather than definitions, many everyday categories have a probabilistic structure. - Category membership is a matter of degree. Categories have a family resemblance structure, e.g., robins share some features with sparrows, but there may be no set of necessary & sufficient features for my bird concept According to prototype theory, a category is represented by a prototype (mental representations of the average member of a category). - The theory has two components. 1. Learning by prototype abstraction 2. Classification by similarity to prototype

Imagery vs perception

Researchers asked if visual images have some of the properties of real vision. - If they do it is more likely they are based on spatial representation, not propositions. Evidence 1. Transformations 2. Zooming 3. Scanning

Visual Search

Single feature -- The time it takes to detect some visual targets is independent of the number of distracters! The target stands out (pop-out effects). When the target is a conjunction of two features, detection time increases with the number of distracters (correct trials only). It's as if we have to allocate attention to each item one-by-one in order to detect the conjunction. For single-feature search, because brain processes low-level features in parallel, increasing # of distractors doesn't increase detection time. -But for conjunction search, no such parallel mechanism exists.

3. Exceptions

Some categories have exceptions, members that are more similar to another category's prototype. - E.g., whales and dolphins are more similar to fish than other mammals. Exemplar models represent exceptions (Category membership is simply memorized), prototype models don't (exceptions are misclassified).

Hybrid Theories

Some theorists argue that Prototypes might be learned first and then augmented with exemplars later. - Categories with few members may end up being represented primarily by exemplars. And, different kinds of categories may be represented differently. - Object categories → exemplars - Social categories → prototypes - Categories defined by people → definitions. • Mathematical objects, kinship terms, legal concepts, sports, etc.

Imagery vs perception — mental scanning (1)

Ss are asked to memorize a number of images. Each image given a name Ss are given name of one image and asked to form visual image whole condition: asked to focus on the whole image focus condition: asked to focus on one end of the image (top, bottom, left, or right) name of property given. Ss indicate whether that property is in the image. RT recorded. RTs increased as the "distance" in the visual image between focus point and property increased

Evidence for Classification by Diagnostic Reasoning

Ss are presented with an object with 3 features that are typical of a target category (e.g., four legs, tail, very smelly are typical features of skunk). - Undermined condition → A 4th feature provided an alternative explanation for one of the first three (e.g., had been wading in the sewer provides an alternative cause of very smelly). - Control condition → A 4th feature is unrelated to the first three (e.g., has just eaten). - Baseline condition → No other information is provided. Ss rate the probability that the object is a member of the target category (e.g., skunk). - In the control and baseline conditions, very smelly provided evidence for a skunk, because one reasons diagnostically from very smelly to skunk. - But in the nondiagnostic condition when an alternative explanation for very smelly is provided, it no longer implies the presence of a skunk. The feature had been wading in the sewer undermines the evidence that very smelly provides for skunk category membership.

Imagery vs perception — zooming

Ss asked to image a pair of animals - imagine a rabbit next to a fly or a rabbit next to an elephant. does the rabbit have a pink nose? Responses are slower when one must "zoom in" on a visual image

Testing Divided Attention

Ss were trained to detect a target letter from a memory set in a frame of letters. - Memory set size was either 1, 2, or 4; Frame size was either 1, 2, or 4; means there's variation in the number of comparisons - They were trained on this task for 30 days, a few hours a day! - 20 frames rapidly presented, a target letter may or may not be present in one frame. S responds "yes" upon seeing target. - This task requires divided attention: One must attend to both the memory set and the current frame. - During training, the key dependent variable is accuracy: Did they correctly judge whether a target letter was present? On each trial, the memory set and frame were randomly chosen in two ways: - Consistent Mapping Condition: there's Potential Memory Set Letters and Potential Frame Letters - Varied Mapping Condition: Memory set and frame were drawn from the same pool. Consistent Mapping Training Results: - Ss found the task initially difficult but gradually improved. They were able to divide their attention between the memory set and the contents of each frame. - After about 600 trials, Ss reported that they were no longer explicitly comparing the memory set to each element in the frame. They were able to automaticity detect the target character in each frame! (pop out effects)

Processing capacity and perceptual load -- Load Theory and Distractors

The chance that a distractor is processed (attended) increases... - As the capacity it requires decreases. • E.g., detecting one's own name, "Fire!", words with a low activation threshold, etc. (Similar to Attenuation Theory's activation thresholds.) - As the cognitive system's spare capacity increases.

Sensory Selection -- Broadbent's Early Selection Model

The theory posits a filter that selects stimuli on the basis of low-level properties. - It is an "early" filter because it is early in the stream in cognitive processing, in particular before semantic processing - Incoming sensory stimuli are analyzed into their low-level, physical properties (color, pitch, loudness, etc.) and channels (left ear, right visual field, etc.) - The filter selects which information should be passed on for further processing. Irrelevant information is discarded. - A process that Broadbent called the detector determines higher level semantic properties and deposits the results into WM

Reasons for Picture Superiority Effect

There are two reasons why memory is better for pictures than words. 1. Distinctiveness 2. Elaborations that produce multiple codes.

Summary of Causal Models

There is little doubt that classification is sometimes based on just similarity. - E.g., to prototypes or exemplars. But classification can be about more than just what things look like. - Coherence: whether features cohere with a category's causal knowledge. - Diagnostic reasoning: whether essences can be inferred from observed features.

Specialization in Face Recognition

There is some evidence that there are mental processes dedicated to the perception of and memory for faces. -Robert Yin asked if Ss would recognize four types of familiar images, including faces, presented either upright or inverted. - All objects are harder to recognize when inverted, but the deficit is especially dramatic for faces! Many objects are recognized by virtue of the features or characteristics they exhibit. - Example: round + red = apple • Thus, features can be recognized even when the object is inverted. But faces are recognized holistically. - Relations are critical:, e.g., the nose is above the mouth. • Thus, inverted faces are more difficult. More evidence that faces are special: - People suffering from prosopagnosia recognize objects but not faces! - Newborns' preference to view faces. - Certain areas of the brain (the fusiform gyrus) are active during face perception

Summary of Imagery vs. Perception

There is strong evidence that visual Imagery involves vision-like spatial representations. In particular, propositional networks do not suffice as an explanation of visual Imagery experiments

Imagery vs perception — 1. Transforming Visual Images

Transforming Visual Images take as long as the corresponding change in actual visual objects. - Mental rotations, foldings, etc

Model vs. Amodal Representations

Types of cognitive representations: - Modal representations -- in the same modality as one of the senses. • Visual-spatial sketchpad: Spatial representations • Phonological loop: Acoustic representations - Amodal representations -- not tied to any sense modality. • Propositional networks

Outcomes of attention -- Visual Attention as a "Spotlight"

Visual attention goes to one spatial location The spotlight metaphor is a useful but incomplete notion of attention. - This is because one can attend to not only spatial locations but also to objects (object-based attention)! - Fixation (1000ms) → Cue (75% Accurate) (100ms) → Fixation (200ms) → S responds when rectangle fills (RT recorded) - Valid cue; Invalid cue, same object; Invalid cue, different object - The spatial distance between the cue and the rectangle are the same. Thus, if attention is a spotlight, RTs should be the same in both invalid conditions. - But if attention is object-based, then invalid cue, same object trials should be faster (because attention is already allocated to the correct object). - Invalid cue RTs are faster when the target appears in the same object as the cue as compared to a different object.

memory improvement techniques

Visual imagery is a very effective way to help remember concrete material. Famous memory improvement techniques make use of imagery: 1. Keyword technique 2. Method of loci

Summary of Visual Imagery

Visual imagery is indeed based on visionlike spatial representations. Visual images are much more likely to be remembered. - But not when material is abstract. - But inverted faces are hard to identify. - But spatial knowledge is organized hierarchically, leading to distortions.

Pros and Cons of Visual Memory

Visual memory is generally very good. - Many memorization techniques rely on forming mental images. • The more bizarre or unusual, the better. But imagery may not help, or even hurt, memory for abstract material for which images are not easily formed. - (And don't forget: mental imagery can contribute to false memories!)

Sensory Selection

We can attend to many things. Which do we choose? Which do we ignore? -- senses based Humans can select which incoming sensory information they pay attend to. Early attention researchers assumed that selection was carried out by a filter. - The filter passes through stimuli that warranted additional processing and discards those that should be ignored.

Summary of Attention

We have powerful mechanisms to select what to process and what to ignore. - This reduces energy consumption (The brain consumes about 20% of the body's energy but weighs only 2%) - It improves performance on attended stimuli. They are so powerful we can fail to see stimuli right in front of us (Change blindness)! Yet our ability to filter out unattended information is not absolute. - Whether we like it or not. low-load tasks are processed automatically. • E.g., the Stroop effect. But this too is often a good thing. - There must be a balance between concentrating on the current task but noticing important unexpected events. Within limits, attention can be divided. Automaticity can be acquired with (extensive) practice. Even some some basic perceptual processes (e.g., feature binding) require attention.

Summary of Definitional Approach

When applied to everyday categories, the definitional approach has multiple problems. - No defining features - Borderline cases - Typicality effects Nevertheless, remember that definitions seem to apply to many man-made categories. - E.g., mathematics (e.g., triangles), kinship terms (e.g., uncle), sports (e.g., a strike).

inattention and change blindness

Whereas highly practiced stimuli demand attention, some novel stimuli receive no attention whatsoever Our attentional systems are very powerful. They can exclude stimuli that are right in front of us. - Nevertheless, remember that some stimuli (our names, written words, etc.) are automatically processed. Only 10% of Ss notice the changes. - Scarf disappears and reappears. - Hand on chin or not. - Color of plates change. - etc 7 of 15 Ss didn't notice that a new person suddenly appeared! - But people generally do notice changes in gender, race, and age. 46% fail to notice a character that walks into the scene. - A woman with umbrella or person in gorilla suit.

Processing capacity and perceptual load -- Automaticity

With (a lot) of practice tasks can become automatic. - They consume few cognitive resources. Some tasks are SO automatic they've become obligatory, they have to be performed! - And so interfere with the current task. • E.g., the Stroop task (Reading influences color naming, but not vice versa. Word reading is automatized, color naming is not.)

Summary of Divided Attention

Within limits, attention can be divided among multiple tasks. - E.g., attending to both a memory set and a frame. Dual task performance will succeed to the extent that one task is automatic. - E.g., with sufficient practice a detection task may eventually require no cognitive resources . The good news is that automaticity can be acquired. - The bad news is that takes many hours of practice. Also, not all tasks are susceptible to becoming automatic. - After 30 days, target letters were detected automatically in the consistent but not the varied mapping condition.

The Hierarchical Model

a propositional network that represents inter-concept semantic relations... - Between concepts • A canary IS-A (type of) bird, a bird is IS-A (type of) animal, a tree IS-A (type of) plant. - Between concepts and features • Canaries ARE yellow, oaks HAVE leaves, robins CAN sing, etc. To achieve cognitive economy, most features are associated with higher level concepts. - Flies is related to birds rather than individual birds (canaries, robins, etc.). - Exceptional cases are stored with the individual concepts. • Ostriches CANNOT fly Testing the model: - sentence verification - lexical priming

1. Learning by Prototype Abstraction

averaging all observed category members on each dimension - The prototype for the Abbot family is formed by averaging the individual family members over the dimensions of length of nose, height of eyes, etc.

Categores and Their Features

categories are usually thought of as having features, including: parts, attributes, properties, etc. - e.g. propositional networks

Computational Models of Cognition

computer programs that serve as a theory of some cognitive phenomena. - We "run" the program to see what predictions it makes for a cognitive task. • And then we conduct an experiment to test if people behave that way! There are many models of how people learn, use, and represent concepts

The Definitional Approach

concepts are just like definitions. - They have defining features. Definitions support two inferences. - Necessity: If something is a category member, it has the defining features. - Joint sufficiency: If something has the defining features, it is a category member. Inclusion is binary: either X satisfies the definition, or it doesn't. Example: Even Number -- Evenly divisible by two

Imagery vs perception — mental scanning (2)

could we get some of these effects from propositions? - For example, perhaps time to scan is a function of the number of propositions in a propositional network.. Ss memorize map Ss asked to mentally scan from one location to the another. S presses key when they "arrive" at the destination. RT recorded predictions - visual image: RT (beach to rocks) > RT (beach to well) - propositional network: RT (beach to rocks) < RT (beach to well) RTs were longer as the "distance" in the Image increased rather than the number of Intervening items

Causal Model Theory

emphasizes the causal knowledge that people have about categories → people have knowledge of not just the features of category members look but also how they work - In 1998, Sloman et al. asked subjects to list features of everyday categories (e.g. Apples, chairs, robins, and guitars). - They were then asked to draw how the features "depended on" one another. - More than 50% of interfeature relations in everyday concepts are causal A causal model is like a prototype augmented with causal knowledge between its features.

Prototype vs. Exemplar Theory

exemplar theory accounts for results that prototype models can't. 1. category variability 2. feature correlations 3. exceptions

Learning Definition-Based Categories

hypothesis testing or rule discovery: - A search for the defining features. researchers tested whether it applies for many everyday categories: 1. Existence of defining features. 2. Ambiguous category membership. 3. Graded category membership.

Sensory Selection -- What are the Problems with Early Selection Theories?

new experiments revealed substantial processing of the unattended channel. - subjects noticed Gender Speech ➝ tone, Fast vs. slow speech, Taboo words, "Fire!" ,Subject's own name! Moreover, it was shown that subjects can "shadow content" - Ss could switch ears to follow content (e.g. ear 1 --dog six fleas, ear 2 -- eight scratch two, heard -- dog scratch fleas) - According to early selection theories, this switching on the basis of semantics should be impossible! Meaning on a (supposedly) unattended channel can unconsciously influence attended information: - An ambiguous sentence was presented on the attended channel. • They threw stones at the bank yesterday. - River prime vs. Money prime - Subjects are then asked which of the following was most like the sentence they heard: They threw stones toward the side of the river yesterday OR They threw stones toward the savings and loan yesterday. - Ss were more likely to choose river interpretation when the biasing information on the unattended channel was river as compared to money - Moreover, Ss reported no awareness of the word they heard on the unattended channel. This is an example of subliminal semantic priming.

Sensory Selection -- What are some issues with Filters model?

none of these theories provide a complete account of the empirical data. - Early selection Doesn't account for awareness of some semantic information on unattended channel - Late selection Doesn't account for the fact that attended information is more likely to be semantically processed than unattended information - Attenuation Doesn't account for that the degree of processing of unattended information depends on the difficulty of the attended task. - Researchers asked subjects to attend to just one channel (e.g., ear) but did they? - Shadowing was intended to verify they did. • But perhaps they briefly switched attention to other ear even while shadowing. - This possibility of rapid attention switching greatly complicated the interpretation of experimental results.

1. Keyword Technique

one associates foreign vocabulary to familiar English words and phrases. - Spanish word daronico means leopard. • Sounds like door. • So, imagine a leopard walking through a door.

What is imagery?

our ability to imagine mental representation that aren't currently present. - Takes place in each sense modality. • Vision, audition, touch, smell, pain, etc. Visual imagery in particular has received a great deal of attention from both philosophers and psychologists.

Evidence — Memory for Faces

our memory for faces is extraordinary - Bahrick et al. obtained high school yearbooks from 392 subjects, ages 17 to 74. - Subjects were asked to pick the former classmate from group of five pictures. People's memory for faces was intact even after 34 years!

Psychological Essentialism

people assume that objects have an underlying nature or essence. - Even if they have no concrete idea of what the essence is. has been shown to affect numerous category-based judgments: 1. Transformation experiments 2. Classification as diagnostic reasoning

Divided attention and automaticity

people can divide their attention among multiple stimuli or tasks. - Success at doing dual tasks depends critically on their "load." (Combining Low-Load Tasks and a High-Load and a Very Low-Load (or Automatic) Tasks → Likely success)

Processing capacity and perceptual load -- load theory of attention

selection is determined by: - Processing capacity — the amount of information the system can handle. - Perceptual load — the capacity required to fully process a stimulus. • Low-load tasks: Less capacity required • High-load tasks: More capacity required

Picture Superiority Effect -- Distinctiveness

similar memory items interfere with one another during retrieval. Because they have large amounts of detail, images are very distinct. - Hence, there is little competition between them, minimizing interference

Sensory Selection -- Evidence for early selection model (2)

the cocktail party effect - the fact we can follow a voice in a crowded room. - The filter can track the low-level physical properties of the person's voice. Yet, it also explains how our attention can be captured by intense stimuli (e.g., a loud fire alarm).

Exemplar Theory

the mental representation of a category consists all the category exemplars ever observed. - Learning • Memorize category members - Classification • Classify a new object into the category with which it has the greatest total similarity (the sum of the similarity with all stored category members). e.g. - Total similarity to the Abbots is the sum of the similarity to each Abbot family member - Total similarity to the Bakers is the sum of the similarity to each Baker family member. - Test face is an Abbot because its total similarity to Abbots is greater than its total similarity to Bakers.

Thoughts Without Images

there are thoughts that don't involve visual images. - Classes of objects • Mammal, vehicle, even number, etc. - Abstract concepts • Justice, democracy, sarcasm, substance, etc. - Verbal rehearsal Accordingly, many have proposed that cognition doesn't require images. - Francis Galton: Visual images may be epiphenomenal, that is, not involved in how we think. - Behaviorist John Watson: Mental images as "unproven" and "mythological." Moreover, there are ways to represent thinking besides images

2. Method of Loci

used in ancient Greece and Rome. - Attributed to the Greek poet Simonedes To-be-memorized items are associated with locations on a familiar path. - Memorizer constructs an image which includes both new item and familiar location. • The more bizarre or unusual, the better! The method of loci can yield 70-100% improvement in recall. - Method of loci subjects recalled 38 of 40 study words on immediate recall test. - Even a day later they recalled 34 of 40! • Control subjects remembered very little.

Picture Superiority Effect -- Elaborations that produce multiple codes

visual stimuli can be elaborated with verbal information. - elaborations allow retrieval via multiple cues. This is referred to as the dual-coding hypothesis. - Pictures can be stored in two ways. • An image and a verbal/propositional code. -e.g. pic of a car crash + "A very tight parking spot!"