PSYCH 240 LEC 12 (SMITH)

Coding experience by category

• Categories structured into a Taxonomy:

  • Hierarchy with successive lvls

    • Top = more specific objects

  • Store and understand experience by sorting it into categories instead of treating every object as brand new

  • Superordinate vs basic vs subordinate

    • basic is easy to recall

Categories allow inferences

• Either deductive

  • it is impossible for new belief to be false if old ones are true

• or inductive

  • is is improbable for new belief to be false if old are true

• Basic + Subordinate categories support more inferences vs Subordinate categories

  • will classify object as apple or McIntosh Apple vs fruit

• Natural-kind categories support more Inductive Inference

Greater similarity among items within category than between categories

• Maximize within-category similarity vs Minimize between-category similarity

• Best seen in basic level (apples similar to one another but different oranges

Measurement of similarity: Geometric approach

• Objects/Items represented as points in multidimensional space

• Distance between 2 points equals their dissimilarity

• Closer distance means more similar

Measurement of similarity: Metric axioms

• three assumptions psychological distance works like physical distance:

  • minimality

  • symmetry

  • triangle inequality.

Measurement of similarity: Violations of metric axioms

• Minimality: Familiar items have more features, so they share more with themselves. Apple scores a(5); pomegranate scores a(2).

• Symmetry: If b > c, then Sim(pomegranate, apple) > Sim(apple, pomegranate).

• Triangle inequality: If the weight on common features (a) is large, lemon–orange and orange–apricot (3 shared features each) are similar, but lemon–apricot (1 shared) is not.

Measurement of similarity: Featural approach to measuring similarity

• Items are sets of discrete features (red, round, hard).

• Similarity increases with shared features and decreases with distinctive ones.

Tversky’s contrast model

• Sim(I,J) = Common features

  • = a(shared) - b(I-J) - c(J-I)

• I-J = features only the first item

• J-I = features only the second item

• a, b, c = weights

Contrast model’s account of metric violations

• Minimality: Familiar items have more features, so they share more with themselves. Apple scores a(5); pomegranate scores a(2).

• Symmetry: If b > c, then Sim(pomegranate, apple) > Sim(apple, pomegranate).

• Triangle inequality: If the weight on common features (a) is large, lemon–orange and orange–apricot (3 shared features each) are similar, but lemon–apricot (1 shared) is not.

Similarity and categorization: Typicality effects

• People reliably rate how typical instances are; a robin is a typical bird vs penguin isn’t

• Typical items are categorized faster + fewer errors, in both word and picture tasks.

  • Children learn typical names first

  • Ratings are fairly unrelated to how frequent or familiar the items are.

Similarity and categorization: Typicality as similarity

• An instance’s typicality reflects its similarity to its category

• In Table 2.3, the contrast model (a=1, b=½, c=¼) gives robin, bluebird, and swallow a score of 6, while vulture, flamingo, and penguin score low

  • sorts the birds into high, medium, and low typicality

  • more typical items reach the similarity threshold faster, so they’re categorized faster

Similarity and categorization: Reasoning-based categorization vs. similarity-based categorization

• Similarity-based categorization

  • fast, automatic feature comparison.

• Reasoning-based categorization

  • slower and more deliberate, using background knowledge and arguments.

    • Rips (1989) study 1: A 3-inch object was judged more likely a pizza than a quarter, even though it was closer to a quarter’s size. Quarters have a fixed size and pizzas vary, so people weighed variability, not just similarity.

    • Rips (1989) study 2: A bird that accidentally became insect-like, but still had normal bird offspring, was judged more likely a bird even though it was more similar to an insect.