Exam 2 Flashcards
(30 cards)
What did Sternberg find in his first experiment?
Both slopes different from zero. First, significant difference between present and absent slopes. This difference was due to set size one. When set size one was excluded, no significant difference between present and absent slopes. This suggests serial non-self terminating search.
Give details of Sternberg’s first experiment.
Question: how do you retrieve info from STM?
Task: same as our experiment but with digits
Logic: if selection of response requires retrieval from STM, then response time should tell you about retrieval process
IV: set size (1-6), target present vs. absent
What is a confound in Sternberg’s first experiment? What’s special about set size 1?
-Likelihood of a given item being the probe is dependent on set size
At set size one, the entropy of the probe is lower, so it’s easier to guess
What’s entropy?
Entropy of a probe item is the amount of info in that item and the opposite of it’s predictability (so, unpredictability)
As an event becomes more probable, it carries less info and becomes more predictable
How did Sternberg set up Experiment 2? What were the results?
Goal: replicate exp. 1, controlling probe entropy
Method: use a fixed memory set over a whole block
Conclusions of Sternberg’s memory search study.
1) Memory search is serial and non self terminating
2) The intercept of the search function = non-search processes (e.g., perceptual encoding and response selection)
3) Rate of search is faster than subvocal rehearsal –> search is not the same thing as rehearsal
What’s the horse race model?
Let’s say comparison process of probe to items in memory takes variable amounts of time and is done in parallel and you can’t respond until all searches have been completed
Then, as set size increases, the probability that you’ll have an item that takes a long time to compare increases –> looks like serial search
Why is the horse race model not a thing in Sternberg’s study?
The horse race model would give a logarithmic relationship between set size and response time, but Sternberg found a more linear relationship
What’s the difference between a category and a concept?
- category: set of items in the world
- concept: mental representation of that set
i. e., categories are extensions of concepts
What’s the classical view? What are some issues with it?
-Concept is a definition of a category: specification of necessary and sufficient conditions for category membership
Problem: some concepts have no definition
What does family resemblance mean for concepts? Define probabilistic and family resemblance.
-There is no single feature that all category members share, but there are some that are more common
SO, categories have a probabilistic family resemblance structure
-Probabilistic: any given feature may appear in some but not necessarily all category members
-Family resemblance: category members resemble one another like family members resemble one another
What is a prototype of a category?
-Prototype: most typical, best, most central category member
Average (central tendency) of all exemplars
May or may not actually exist (e.g., most prototypical dog)
What are prototype effects?
-prototypes are cognitively privileged
-more prototypical exemplars are…
categorized faster, listed first, share more features with other exemplars, learned earlier in childhood
-exposure to exemplars causes learning of a prototype
Give details of study that shows exposure to exemplars causes learning of prototype.
Method: generate exemplars by distorting a prototype
Training: present exemplars but not prototype
Test: view exemplars and rate confidence that seen previously. Exemplars included: previously seen exemplars, new exemplars, prototype
Result: confidence that exemplar had been studied was related to similarity to prototype, not related to whether exemplar had actually been studied
Conclusion: exposure to exemplars causes learning of prototype
What’s prototype theory?
- Prototype IS the mental representation of the category – i.e., prototype=concept
- Through exposure to exemplars, you compute their mean (prototype) and store this as the concept
- Categorize new exemplars by comparing them to prototypes in memory
Strengths and limitations of prototype theory.
Strength: provides a natural account of prototype effects
Limitations:
-fails to specify variance (how much deviation from prototype is okay?)
-fails to specify relation among exemplar’s features
-predicts incorrectly that only linearly separable categories are learnable
assume categorization based on feature-based similarity
What is exemplar theory?
- Store all category exemplars: exemplars are category representation
- Classify new exemplars by matching to most similar exemplar in memory
Strengths and limitations of exemplar theory.
-Strengths:
can account for prototype effects (if you store all the data, you have access to the mean when needed)
captures variance (and max, min, etc.)
captures correlations among features
can learn non-linearly separable categories
-Limitations:
assumes categorization based on feature-based similarity (storing/matching features, deciding membership all based on similarity)
What are some problems with similarity?
1) Similarity is intuitive but poorly defined
-similarity = shared features? any two objects share an infinite number of features
2) Similarity is context-sensitive
3) Similarity is bad at characterizing some kinds of categories
-e.g., superordinate, ad hoc
So, similarity may be an effect rather than a cause of categorization
What is schema theory?
Concepts as schemas or theories describing categories of things
-schema is a relational structure: gives relations as explicit entities
Specifies relations between features (and other concepts) instead of just listing features
Provides explanatory framework for understanding properties of category members
What’s psychological essentialism?
People assume objects (especially natural kinds) have an essence that makes them the way they are
- visible features are merely a reflection of this essence
- features do not define the concept so much as point to it
What does ANOVA allow you to do?
Test multiple means (more than one independent variable) and find interactions between the variables
Explain the logic of ANOVA.
We have two measures of variance:
-MSE based on variance within samples, which is a good estimate regardless if null is true
-MSM based on variance between samples, which is a good estimate only if null is true
If they agree (MSE = MSM), null is supported because it’s likely that these variances came from the same underlying distribution
If they disagree (MSE < MSM), null is rejected
How do we deal with unequal sample sizes in ANOVA?
Use sums of squared error because they are additive even when sample sizes are unequal
