Observational learning: knowing versus performing

Observational learning: knowing versus performing

Distinguish acquisition of an observed action from its immediate performance and test alternative explanations.

Question and prediction

Does failure to imitate demonstrate failure to learn?

Predict what happens if people who saw different consequences for a model later receive the same demonstration request.

Mechanism and evidence

Observational learning can occur when someone watches a model and acquires information about actions or outcomes. Attention, retention, ability to reproduce an action and motivation all matter. Seeing consequences for another person can affect whether an observer performs an action. An observer need not personally receive the model’s consequence to learn from the event.

Acquisition and performance are different constructs. A learner may know the steps but choose not to display them, or may remember the goal but lack the motor skill. Conversely, immediate copying does not establish durable retention or transfer. A study should measure the proposed mechanism rather than treating one visible response as a complete readout of learning.

The fictional task uses a harmless paper-folding sequence. After watching the same sequence, observers see the model either praised or neutrally acknowledged. More in the praise group copy immediately. When both groups later receive the same request and opportunity to demonstrate, scores become similar. This pattern is compatible with different initial willingness to perform, but it is not proof that motivation was the only difference.

Random assignment, equal exposure and a common later test improve the comparison. The follow-up test can itself change motivation or provide retrieval cues. To investigate retention, add a delayed test without replaying the sequence; to investigate transfer, change the paper orientation or ask for the underlying rule. These are proposed designs, not instructions to manipulate classmates or conduct unapproved research.

Inspect the relationships

Test	Praise group	Neutral group	Inference limit
Immediate copying	14/20	6/20	May reflect willingness as well as knowledge
Common demonstration	16/20	15/20	Not yet a delayed-retention test

Worked example

Immediate imitation is 70% versus 30%, a 40-percentage-point difference. On the common test it is 80% versus 75%, a 5-point difference. This narrowing weakens the inference that the neutral group learned nothing. Sample uncertainty and test effects still matter.

Your investigation

Create an acquisition/performance evidence table. Write a 200-word explanation comparing motivation, attention and retention as rival accounts. Propose one delayed or transfer test and specify a result that would weaken your preferred explanation.

Unfamiliar transfer

A learner describes all steps but cannot execute a fine motor movement. Which distinction matters? Explain your reasoning before checking feedback.

Reading

OpenStax Psychology 2e: observational learning. The records below are original fictional teaching cases, not research results from this source.

Fictional immediate imitation Two randomly assigned groups of 20 view identical folding steps. Model praised: 14 imitate immediately. Model neutrally acknowledged: 6 imitate.

Fictional common test Later, both groups receive identical materials and a common request to show the sequence. Praise group: 16 correct; neutral group: 15 correct. No delayed test is available.

Learning can precede its visible demonstration

Watching another person is one way to acquire information. Exploring an environment without a model is different. Latent learning describes learning whose effects are not apparent in the measured behavior until conditions for performance change. It does not mean unconscious learning in every case, and it does not follow merely because a score improves later.

A cognitive map is a proposed representation of relationships among places, not necessarily a picture consciously seen in the mind. A fixed sequence of turns and knowledge of where destinations lie can produce the same successful journey along a familiar route. A changed start or blocked route can make their predictions differ.

The notebook includes a short paraphrase of Edward Tolman’s 1948 paper. He used delayed-reward maze findings and spatial-transfer experiments to argue for learning beyond an immediately displayed response sequence. His paper is an advocate’s theoretical synthesis, not a neutral verdict on every competing theory. It reports earlier work by Blodgett and Honzik as well as other researchers. Our numerical and map examples below are invented and are not measurements from those studies. No animal experiment or deprivation activity is required.

Predict two routes before inspecting an outcome

Use this fictional floor plan. Every line is a two-way, equal-length corridor. The room labels mark locations, and north is up. There are no diagonal corridors. The original practiced route is S → A → G, starting at S while facing north. Thus its action sequence is “forward, then right.”

Text equivalent: the north row is A–G–C; the south row is S–B–D. Vertical corridors connect A–S, G–B and C–D. All seven links are available initially. The drawing shows connectivity; its screen distances do not encode travel costs. Every corridor costs one step under the teaching rule.

Make a prediction ledger for two explicitly simplified accounts:

1. Repeated-action account: at a new start, repeat the learned “forward, then right” sequence without using room relationships.
2. Relational-route account: use learned corridor relationships to choose an available path to G, allowing more than one equally short path.

At test, start at D facing north. Trace both predictions. Then block corridor C–G and predict again. Treat a wall as blocked movement, not a new corridor invented to make a rule work.

Review route predictions and the blocked corridor

Repeating forward takes D to C. Turning right now points east, where no corridor exists: the proposed second move is blocked. It does not reach G.

With the full map, D → C → G and D → B → G both reach G in two steps. There is no unique shortest-route prediction. Blocking C–G removes the first path but leaves D → B → G. A representation of available connections can support a new route without repeating the old action sequence.

A successful detour alone does not prove that a learner used a cognitive map. A visible goal, a sign, a familiar landmark association or immediate exploration could also guide the choice. To distinguish accounts, control those cues, define the first-choice measure before repeated test learning occurs, and compare predictions across several changed starts. Failure also need not prove no representation: instruction, attention or ability to act can interfere with performance.

Separate previous exploration from learning during the test

Consider an original hypothetical study with three randomly assigned groups of 20. X previously explores the same layout with no destination reward; Y spends the same time exploring a different layout; Z practices the destination route with reward. Everyone receives the same instructions, materials, incentive and changed-start test. The destination is not visible and no signs identify it. Scores below are counts choosing a valid two-step route on their first test attempt, before corrective feedback.

Group	Prior experience	Correct first transfer choice
X	Same-layout exploration, no destination reward	14 of 20
Y	Different-layout exploration, no destination reward	6 of 20
Z	Rewarded destination practice	16 of 20

Calculate X–Y and Z–X differences in percentage points. State which comparison tests the value of same-layout exposure without a destination reward, and which comparison cannot isolate reward alone. Identify one alternative to a full relational map that could still explain X’s advantage.

Review the exposure comparison

X scores70%, Y30% and Z80%. X–Y is40 percentage points; Z–X is10. The X–Y contrast is consistent with useful information acquired during same-layout exploration before the common test incentive. It does not establish that all learning requires no rewards, that exploration had no other consequences, or that the information was a complete map. Familiarity with particular cues may help.

Z differs from X in both reward and directed route practice; their difference cannot identify a pure reward effect. The groups are small, so a complete empirical analysis also needs uncertainty, methods and replication. These invented counts are a reasoning exercise, not evidence about actual learners or a prediction of these exact percentages.

If the reported scores instead came after five feedback-rich test attempts, performance could reflect new learning during the test. Preserve first-choice records to assess prior learning and analyze later improvement separately.

Try a new route and propose a discriminating test

Return to the map with all corridors open. Start at C facing west, with G still the destination. Predict the literal “forward, then right” sequence and the relational route. Explain why arriving at the goal after the first step can make this test less informative than the D-start test. Do not assume every change of starting point distinguishes the accounts.

Review the new-start test

One forward step from C reaches G. A learner who stops upon recognizing G succeeds without needing the second turn. Relational knowledge also supports the direct C → G path. Success at this start alone therefore does not distinguish these accounts when goal recognition permits stopping. Specify stopping rules and goal cues before interpreting observed behavior.

For a stronger test, choose a start or blockage where the accounts make different predictions, keep other cues comparable, record the first choice and state which result would weaken your preferred account. Do not design the scoring rule after seeing the result.

Save your route predictions, rate calculations and test-design revision using the latent-learning source in the notebook. The map is a worked teaching model, not a simulated mind. These reviews support guided reasoning; independent assessment of a fresh design remains separate.