Generalization in AP Psychology

The claim: a conditioned response spreads by similarity

Generalization only happens after acquisition has already worked. A neutral stimulus has been paired with an unconditioned stimulus (UCS) until it became a conditioned stimulus (CS) that triggers a conditioned response (CR). Generalization describes what happens next. The learner also responds to things that look, sound, or feel like the CS.

The pattern is graded. The original CS gets the strongest CR, close look-alikes get a slightly weaker one, and distant look-alikes get little or none. Psychologists call this falloff a generalization gradient.

This sits in Unit 3: Development and Learning, under Topic 3.7: Classical Conditioning, where the course asks how associating one stimulus with another shapes both behavior and mental processes. Generalization and stimulus discrimination are the pair of findings the course names as coming out of classical conditioning research. Generalization explains how far a learned association reaches beyond the exact stimulus that was trained.

Evidence from Pavlov's tones and Little Albert

Pavlov's dogs were conditioned to salivate to a specific tone. When tested with tones that were somewhat higher or lower, they still salivated, just less than to the trained tone. That's the gradient in action.

A historical example of research on conditioned emotions is Watson and Rayner's Little Albert study (1920). Albert was conditioned to fear a white rat by pairing it with a loud noise. Afterward, his fear showed up to other furry things he had never been conditioned with, including a rabbit, a fur coat, and a Santa Claus mask. The fear spread because those stimuli resembled the rat.

That case also shows why generalization matters for the Topic 3.7 point about treating mental disorders. A phobia often covers a whole category (all dogs, all water) after one bad experience with a single example. Counterconditioning has to address that whole category along with the original trigger.

Reading a generalization gradient in a data set

Set-based multiple-choice questions can hand you data and ask what it shows. Here's a typical setup. Participants learn to blink to a 1000 Hz tone paired with an air puff. Then they're tested on several tones.

• 1000 Hz: 92% blinked
• 900 Hz: 78%
• 800 Hz: 61%
• 700 Hz: 45%
• 600 Hz: 28%

The supported conclusion is that blinking weakens as the test tone becomes less similar to the original CS. Two details are worth reading carefully. First, some participants blinked even at the 600 Hz tone, but people blink on their own too, and this data set has no baseline (blink rates to these tones before conditioning). Without that comparison, you can't say how far generalization actually reached. Second, the drop is gradual, so this isn't an all-or-nothing switch.

A wrong answer choice might claim the participants responded equally to every tone. The data rule out equal blink rates across tones, but without a baseline they cannot establish that conditioning occurred. So a choice claiming the data prove learning happened goes further than the numbers allow.

In a scenario question without numbers, the move is the same. Name generalization, then say the CR occurred to a stimulus similar to the original CS. Naming the term alone usually isn't enough explanation.

Where generalization stops: discrimination and higher-order conditioning

Generalization has limits, and the neighboring concepts help show where they are.

Stimulus discrimination is the opposite outcome. The learner responds to the CS but not to similar stimuli. Getting nervous at every bell is generalization, and getting nervous only at the school bell is discrimination. Discrimination usually develops when similar stimuli keep showing up without the UCS, so the learner stops responding to them.

Higher-order conditioning is the other common mix-up. There, a new stimulus becomes a trigger because it was paired with an existing CS, which then acts like a UCS. The new trigger doesn't have to resemble the old one at all. Say a statistics textbook first became a CS for anxiety. If a previously neutral calculator were then paired with that textbook, and the calculator began to cause anxiety on its own, that would illustrate higher-order conditioning. Ask whether the new trigger looks like the CS (generalization) or was paired with it (higher-order).

One more limit: similarity has to be the explanation. If a person's fear of one river spreads to the ocean after watching vivid news stories about shark attacks, the news exposure is a competing explanation. Generalization alone doesn't account for that pattern cleanly.

Three other "generalizations" that show up in AP Psych

The word appears in other places in the course with different meanings, so read carefully.

In Topic 3.8: Operant Conditioning, generalization shows up too. Here, a response reinforced in one situation may occur in similar situations, rather than a conditioned response occurring to stimuli similar to a CS. The similarity that matters is between the situations where the behavior happens.

In Topic 3.5: Communication and Language Development, overgeneralization is a language error, where a child applies a grammar rule too broadly (saying "goed" for "went"). That's about grammar rules, and it has nothing to do with conditioned responses.

In the Article Analysis free-response question, you judge a study's generalizability, meaning whether its findings apply beyond the sample. That's a research-methods idea. For example, one in-depth case study of a person who developed an aversion to all citrus after getting sick from spoiled orange juice shows stimulus generalization in that individual. It can't support the claim that this generalization happens universally, because a single case doesn't represent a population. Both meanings can show up in the same question, so keep track of which one you're being asked about.