Concept Discovery Through Examples and Non-examples¶

Strategy · All strategies
Evidence · 5 claims (4 for, 1 mixed) · 15 studies (6 causal, 4 quant-synthesis, 4 review, 1 theoretical), q1–q4 · 3 of 15 report an effect size

Description¶

Learners are presented with a carefully sequenced set of examples and non-examples of a target concept and asked to identify shared characteristics, classify new instances, and articulate the concept's defining attributes. The instructor or system supports exploration with questions and context-sensitive feedback that corrects misconceptions as they emerge. The strategy treats concept acquisition as pattern discrimination built from varied instances rather than transmission of a definition.

Design Implications¶

Concept learning depends on exposure to varied, well-chosen instances: comparing multiple cases helps learners abstract the deep structure that defines a concept rather than surface features of any single instance Multiple varied cases support flexible transfer of concepts. [+M]. Non-examples are as important as examples — they establish the boundaries of the concept and prevent overgeneralization Erroneous examples build conceptual knowledge by making boundaries explicit. [+M]. However, unguided discovery is unreliable: learners left to explore without prompts, sequencing, or feedback frequently fail to induce the intended concept, and guided forms of discovery consistently outperform pure discovery Guided discovery outperforms pure discovery across outcome measures. [+S].

Context¶

Requirements¶

• A rich, representative set of examples and non-examples, varying surface features while holding defining attributes constant
• Sequencing that starts with clear, matched pairs (example/non-example differing on one attribute) before moving to harder discriminations
• Interactive classification tasks with immediate, context-sensitive feedback (Practice, Feedback)
• Prompts that direct attention to relevant attributes rather than leaving search entirely open (Advance Organizers)

Constraints¶

• Unguided exploration with minimal prior knowledge produces weak or erroneous inductions; novices need more scaffolding than experts Minimal guidance is less effective for novices than explicit guidance. [-S]
• Effectiveness reverses with expertise: highly knowledgeable learners benefit less from heavily scaffolded example sets and may profit from more open exploration Guidance benefits decline as learner expertise increases. [~M]
• Poorly chosen non-examples (differing on irrelevant features) teach spurious discriminations
• Time-intensive: building a sufficient instance set and feedback logic costs more design effort than presenting a definition

Implementation Variability¶

• Rational set design: systematically vary variables across examples (easy/difficult, divergent/range of divergence) per Tennyson & Park's concept-teaching model
• Matched pairs: present an example and non-example side by side differing on exactly one defining attribute
• Learner-generated instances: after classification practice, learners produce their own examples and non-examples, which deepens encoding
• Erroneous examples: present flawed instances for diagnosis, leveraging error analysis rather than avoidance Erroneous examples build conceptual knowledge. [+M]

Target Learners¶

• Learners with some relevant prior knowledge who can notice attribute patterns with prompting Activating prior knowledge improves learning of new related concepts. [+M]
• Intermediate learners refining fuzzy or misconceived concepts (e.g., distinguishing mass from weight)
• Novices benefit most when discovery is strongly guided; pure exploration suits them poorly Minimal guidance is less effective for novices. [-S]

Target Learning Goals¶

• Concept acquisition: learning defining attributes and category boundaries
• Discrimination and classification: accurately sorting novel instances
• Conceptual change: replacing misconceptions through confrontation with counterexamples
• Inductive reasoning: practicing abstraction from instances to rule

Instructions¶

1. Activate relevant prior knowledge and surface likely misconceptions before presenting instances (Activation)
2. Present a matched example/non-example pair and ask learners to identify the distinguishing attribute (Comparing Cases)
3. Present a varied sequence of additional examples and non-examples, prompting learners to test and refine their hypothesis about the concept's definition
4. Require classification of novel instances with immediate feedback on errors (Practice, Feedback)
5. Ask learners to state the rule and generate their own example and non-example (Articulation)
6. Fade support: move from prompted comparison to independent classification as accuracy grows (Fading)

Related Strategies¶

• Use Worked Examples — the procedural analogue: studying solved instances rather than inducing a concept
• Comparing Cases — the core comparison mechanism this strategy sequences into a concept lesson
• Guided Inquiry — broader inquiry framing in which example sets serve as the data

Related Elements¶

• Non-Examples — the boundary-defining counterpart to examples; essential for preventing overgeneralization
• Comparing Cases — the comparison activity that drives attribute abstraction
• Feedback — context-sensitive correction of misclassifications
• Advance Organizers — frames what learners should look for before exploration

Patterns That Use This Strategy¶

• Concept Attainment — the canonical pattern: hypothesis testing against positive and negative instances
• Cognitive Flexibility Theory — multiple varied cases across contexts for complex concepts
• Direct Instruction — uses example/non-example sequences but with teacher-led rule statement, contrasting on the discovery dimension

Examples¶

• Concept Attainment model (Bruner, Goodnow & Austin) — learners see labeled YES/NO instances (e.g., of "democracy") and hypothesize the defining attributes, then classify new cases.
• Impressionism lesson — learners view a set of Impressionist paintings alongside Academic Salon works, identify shared characteristics (visible brushwork, light effects), then classify unfamiliar paintings.
• Wolfram Demonstrations — interactive parameter manipulation lets learners discover what makes a function continuous by generating examples and non-examples on demand.
• Merrill's Pebble-in-the-Pond concept lessons — a portrayal (specific instance) followed by a sequence of similar portrayals for classification practice.

Key Sources¶

• Tennyson, R. D., & Park, O.-C. (1980). The teaching of concepts: A review of instructional design research literature. Review of Educational Research, 50(1), 55–70. doi:10.3102/00346543050001055
• Alfieri, L., Brooks, P. J., Aldrich, N. J., & Tenenbaum, H. R. (2011). Does discovery-based instruction enhance learning? Journal of Educational Psychology, 103(1), 1–18. doi:10.1037/a0021017
• Kirschner, P. A., Sweller, J., & Clark, R. E. (2006). Why minimal guidance during instruction does not work: An analysis of the failure of constructivist, discovery, problem-based, experiential, and inquiry-based teaching. Educational Psychologist, 41(2), 75–86. doi:10.1207/s15326985ep4102_1
• Merrill, M. D. (2002). First principles of instruction. Educational Technology Research and Development, 50(3), 43–59. doi:10.1007/bf02505024
• Marton, F., & Booth, S. (1997). Learning and awareness. Lawrence Erlbaum Associates.