chapter 2 Flashcards
Why should we study similarity?
Similarity underlies many cognitive processes such as object recognition, memory retrieval, problem-solving, and reasoning. It helps categorize information, make predictions, and link new knowledge to existing concepts. Similarity is also a diagnostic tool used in fields like medicine, science, and engineering.
How does similarity influence thinking processes like memory, object recognition, problem-solving, and predictions?
Object Recognition: We identify objects by comparing them to stored mental representations.
Memory Retrieval: Similar concepts activate related memories, making recall easier.
Problem-Solving: Recognizing similarities between problems allows us to apply known solutions.
Predictions & Inductive Reasoning: We make generalizations based on perceived similarity between past and present experiences.
What are the four main theories of similarity?
Geometric Model – Similarity is measured as psychological distance in a multidimensional space.
Contrast Model – Similarity depends on shared vs. unique features, accounting for asymmetric judgments.
Alignment Model – Similarity is based on how well features align in context, emphasizing relational comparisons.
Transformational Model – Objects are similar if one can be transformed into the other through a series of steps.
What is the difference between city block and Euclidean metrics in similarity comparisons?
City Block Metric: Measures similarity by summing differences along independent dimensions (e.g., counting feature differences). Best for separable dimensions.
Euclidean Metric: Measures similarity using a straight-line distance in a multidimensional space, integrating dimensions holistically (e.g., color perception).
What are the central assumptions of the geometric approach to similarity?
Minimality – An object is always most similar to itself (but violations occur in psychological judgments).
Symmetry – The similarity of A to B should be the same as B to A (but often, people judge similarity asymmetrically).
Triangle Inequality – If A is similar to B and B is similar to C, then A should be similar to C (but context influences judgments, violating this assumption).
What is an example of multidimensional scaling (MDS) in similarity research?
Nosofsky’s rock categorization study used MDS to analyze how people classified rocks based on perceptual dimensions like texture and color, rather than geological categories. This showed how similarity judgments rely on perceptual rather than conceptual features.
How does the contrast model determine the similarity between two items?
The contrast model (Tversky, 1977) calculates similarity by weighing shared and distinctive features:
S(A, B) = θ * f(A ∩ B) - α * f(A - B) - β * f(B - A)
Shared features increase similarity.
Unique features reduce similarity.
Asymmetry occurs when one object has more distinct features than another.
How does the alignment model assume that similarity is determined?
The alignment model suggests that similarity is based on relational comparisons rather than absolute features. Instead of comparing independent features, it aligns corresponding features in context (e.g., matching the smallest object in one set to the smallest in another, even if their sizes differ).
How does the transformation model explain similarity?
The transformation model suggests that objects are similar if one can be turned into the other through a sequence of steps. The fewer transformations required, the more similar they appear. For example, water and steam are similar because converting one to the other requires just one step.
What was the experiment with the quarter and the pizza, and what did it show?
In Lance Rips’ (1989) study, participants were asked to judge the similarity and categorization of a 3-inch round object.
Similarity judgment: The object was rated as more similar to a quarter due to shared shape and size.
Categorization: It was categorized as a pizza because of typicality and context.
This showed that similarity and categorization are distinct processes—one based on feature matching and the other on conceptual classification.
