MCAT content library / Psychology & sociology / Classical conditioning: learn the predictive relationship
Pairings tell only part of the story. Compare outcomes with and without a cue, then use the testing history to explain the response.
About 17 minutes, including the worked examples and checks.
From measurement to learning · Lesson 3 of 5 · Continue in this path
On this page
- Ask what the cue adds to the prediction
- Give each stimulus and response a role
- Worked comparison: equal pairings, different information
- Extinction changes what the cue predicts
- Similarity and learning history answer different questions
- Follow the relationship through the changed case
- Check your reasoning
- Continue learning
- Sources
Ask what the cue adds to the prediction
A tone occurs before food many times. That history gives a learner an opportunity to associate the events, but the number of pairings leaves out a crucial comparison: how often does food arrive without the tone? A cue that distinguishes food from no food carries different information from a cue that accompanies an outcome already common in the background. This lesson develops the classical-conditioning topics in AAMC’s learning outline.1
Classical conditioning concerns learning a relationship between events. It can change a response elicited by a cue even when obtaining the outcome does not depend on making that response. In an operant task, by contrast, the learner’s action changes the consequence it receives. Both relationships can occur in the same setting; identify the actual contingency in the experiment.3
Give each stimulus and response a role
Imagine a laboratory animal that salivates when food enters its mouth. Before training, a particular tone does not evoke that salivary response. Food is the unconditioned stimulus (US), and salivation to food is the unconditioned response (UR). “Unconditioned” means the response does not require the specific tone–food learning under study. The tone begins as a neutral stimulus with respect to that response.2
If the tone subsequently elicits salivation before food arrives, the tone functions as a conditioned stimulus (CS), and salivation to it is a conditioned response (CR). Classify a response by what elicits it: saliva after food is not automatically evidence for a CR. The period in which the cue acquires this influence is acquisition.2 Measuring the response before the food arrives helps separate anticipation from a reaction to the food itself.
Worked comparison: equal pairings, different information
The following schedules and response records are original teaching examples. Two groups encounter twenty tone windows and twenty equally long background windows without the tone. In each window, food either arrives once or does not arrive. Within a group, the sequence varies so that window order provides no additional reliable cue.
| Group | Food in tone windows | Food in background windows | Probability difference |
|---|---|---|---|
| Predictive cue | 16/20 = 0.80 | 2/20 = 0.10 | 0.70 |
| Uninformative cue | 16/20 = 0.80 | 16/20 = 0.80 | 0.00 |
Both groups receive sixteen tone–food pairings. In the first, hearing the tone changes the estimated chance of food from 0.10 to 0.80. In the second, the tone adds no information about whether food occurs in a window. The comparison can be summarized as ΔP = P(US | CS) − P(US | no CS). The vertical bar means “given”: compute each probability within its own condition.
Rescorla’s experiments showed why this background comparison matters: changing outcome probability in the absence of a cue changed conditioning even when cue–outcome pairings were available. A positive contingency can support a cue’s excitatory predictive role; pairing counts alone are insufficient.4 For the table, the first cue is the stronger predictor of increased food probability. This is a prediction about the supplied schedule, not a measurement of how much either group learned.
To evaluate learning, add a behavioral measure under comparable testing conditions. For example, compare salivation during a cue-only probe while accounting for hunger and response capacity. The schedules deliver different total amounts of food, so an unqualified difference in salivation could also reflect different motivational states. Separating programmed information from observed behavior prevents a correct probability calculation from becoming an unsupported causal conclusion.
Timing, previous learning, and the kind of cue and outcome also matter. ΔP is a useful description of this comparison, not a complete learning theory or a guarantee that every individual will produce a particular response. A learner need not consciously calculate probabilities for behavior to become sensitive to predictive relationships.3
Extinction changes what the cue predicts
After acquisition, presenting the CS repeatedly without the US can reduce the CR: this is extinction.2 In the illustrative record below, salivation is scored during the tone, before any scheduled food. Acquisition blocks contain tone–food trials; extinction blocks contain tone-only trials. The final block follows a pause, with no intervening tone–food trials and no change of setting.
| Block | Condition | Trials with a CR (%) |
|---|---|---|
| 1 | Acquisition | 20 |
| 2 | Acquisition | 40 |
| 3 | Acquisition | 70 |
| 4 | Acquisition | 80 |
| 5 | Extinction | 60 |
| 6 | Extinction | 30 |
| 7 | Extinction | 10 |
| 8 | Extinction | 0 |
| 9 | Tone-only test after a pause | 20 |
The response increases during acquisition and decreases when the tone stops predicting food. Its partial return after a delay is spontaneous recovery. That return is evidence against treating the end of extinction as proof that the original learning has been erased. Extinction can produce new learning whose expression depends on time and context.3
Read the test history before choosing a label. If tone–food training resumed before the response returned, renewed acquisition would be a competing explanation. If only the setting changed, context would need attention. In this record, the controlled detail is a pause without new pairings in the same setting. A low endpoint alone cannot tell these histories apart.
Similarity and learning history answer different questions
Generalization occurs when learning involving one cue influences responding to another. Discrimination is differential responding to cues associated with different outcomes.2 To test transfer to a similar tone, present that tone before giving it its own food pairings. Otherwise, a response might reflect new training with that tone. If later training pairs one tone with food and another with no food, a growing response difference supports discrimination.
Not every cue–outcome combination is equally easy to learn. Garcia and Koelling found that rats’ learned avoidance depended on the relation between the cue type and the subsequent aversive event: taste and audiovisual cues did not become interchangeable merely because both had been present. This supports biological constraints on associative learning.5 In a supplied experiment, use the stated species, cues, and outcome. Do not convert such a result into a rule that one kind of association is impossible for every organism.
Follow the relationship through the changed case
For a new passage, write down what the cue predicts before and after the manipulation. Then identify the response’s eliciting event, whether that response was needed to obtain the outcome, and what happened between training and the test. Those details distinguish acquisition, extinction, recovery, and transfer more reliably than the mere presence of a tone, food, or repeated trials.
ACTIVE PRACTICE · CHANGE ONE VARIABLE
Manipulate a Cue’s Predictive Value, Not Just Its Pairings
The original conditioning design uses 20 tone windows and 20 equal-duration windows without the tone. Food is delivered in 16 tone windows in both groups. Its background occurrence is either 2/20 (predictive) or 16/20 (uninformative). Programmed probabilities are not measured learning.
P(food | tone) =; P(food | no tone) =.
Predictive difference ΔP = .
Positive predictive relationship under the programmed schedule.
Changing the sliders creates hypothetical schedules with 20 windows of each kind; it does not simulate acquired responding or equate exposure count with conditioning strength.
Check your reasoning
Choose an answer before opening its feedback. Each option explains the reasoning behind that choice.
Check 1
Two groups receive a light cue in twenty equal-duration windows. Food arrives in fifteen cue windows for each group. Group R also receives food in five of twenty background windows; Group S receives it in fifteen of twenty background windows. Assume window order offers no extra prediction. Which comparison correctly describes the light’s information about food?
A. It raises food probability by 0.50 in R and by 0.00 in S.
Correct. For R, ΔP = 15/20 − 5/20 = 0.75 − 0.25 = 0.50. For S, ΔP = 15/20 − 15/20 = 0.00. Equal pairings therefore coexist with different predictive information.
B. It raises food probability by 0.50 in R and by 0.75 in S.
Review this choice. The R calculation is correct, but 0.75 for S is the probability during the cue, not the increase relative to its background. S’s background probability is also 0.75.
C. It raises food probability by 0.75 in R and by 0.00 in S.
Review this choice. The S comparison is correct. In R, however, 0.75 omits the 0.25 background probability. The cue-associated increase is 0.50.
D. It raises food probability by 0.75 in R and by 0.75 in S.
Review this choice. These are the two probabilities during the cue. A predictive increase requires subtracting each group’s probability without the cue.
The principle to carry forward
The same number of pairings does not imply the same contingency. This calculation identifies information in the schedule; a behavioral test is still needed to measure the learned response.
Check 2
A tone previously paired with a mild air puff evokes blinking on 80% of trials. Repeated tone-only trials reduce this to 5%. After a two-day interval with neither tone nor air-puff presentations, an initial tone-only test block in the same setting evokes blinking on 30% of trials. Which conclusion is best supported?
A. The delay generated new tone–air-puff pairings that rebuilt the response.
Review this choice. The stem excludes tone and air-puff presentations during the interval. New pairings cannot explain recovery in the stated history.
B. The air-puff-elicited response became stronger during the tone-only test.
Review this choice. No air puff occurs in the test. The measured blink is elicited by the previously trained tone, not by the unconditioned air-puff stimulus.
C. The reduction after tone-only training did not demonstrate complete loss of the original learning.
Correct. The trained response partly returns without new pairings, a spontaneous-recovery pattern. The low response at the end of extinction therefore did not establish that the original learning had disappeared.
D. The response returned because the test moved to a different setting.
Review this choice. The test occurs in the same setting. A changed-context explanation is unavailable in this design, whereas time since extinction changes.
The principle to carry forward
Use the intervening history to explain recovery. A response returning after time alone differs from a response returning after new pairings or a change of context.
Check 3
Animals are trained with a 1000 Hz tone followed by food. Researchers want to know whether this learning transfers to a 1200 Hz tone. Which result most directly supports transfer from the original learning rather than a new association with the 1200 Hz tone?
A. A response to 1000 Hz becomes larger after additional 1000 Hz–food trials.
Review this choice. This shows a change in responding to the trained tone after more training. It does not test transfer to 1200 Hz.
B. A response to 1200 Hz appears on its first presentation before any food follows it.
Correct. The new tone elicits the response before it has its own food-pairing history. Given the earlier 1000 Hz training, this is the direct generalization test among the options.
C. A response to 1200 Hz grows over repeated 1200 Hz–food trials.
Review this choice. Repeated pairings can establish a new association with 1200 Hz. The result does not isolate transfer from the earlier 1000 Hz learning.
D. A response to both tones disappears after many tone-only presentations.
Review this choice. This describes the outcome of nonreinforced exposure after some unspecified learning history. It does not establish how responding to 1200 Hz initially arose.
The principle to carry forward
To isolate generalization, test the new cue before teaching a new cue–outcome relationship. Later acquisition can otherwise mask the distinction.
Continue learning
From measurement to learning
Lesson 3 of 5.
- Previous in From measurement to learning: Sensory thresholds, adaptation, and habituation
- Next in From measurement to learning: Operant learning: consequences and response patterns
- Return to this path’s overview
See every step in From measurement to learning
From measurement to learning
Use behavioral evidence to distinguish adaptation, predictive associations, consequences, and learning from a model.
- Psychological Measurement: What Does the Score Mean?
- Sensory thresholds, adaptation, and habituation
- Classical conditioning: learn the predictive relationship · You are here
- Operant learning: consequences and response patterns
- Observational learning and reciprocal social influences
Related lessons
- Sensory thresholds, adaptation, and habituation
- Operant learning: consequences and response patterns
- Observational learning and reciprocal social influences
- Experimental Controls: What a Comparison Can Show
Browse psychology & sociology · MCAT content library
MCAT preparation and roadmap · Aim High, Score Higher
Sources
1. AAMC: What’s on the MCAT Exam? — 7C Learning
2. OpenStax Psychology 2e: Classical Conditioning
3. Mark E. Bouton, Noba: Conditioning and Learning
4. Rescorla (1968): Probability of shock in the presence and absence of CS in fear conditioning
5. Garcia and Koelling (1966): Relation of cue to consequence in avoidance learning