Mental models of the earth: A study of conceptual change in childhood

60 children, 20 each in grades 1, 3 and 5

Elementary school children at one school in Urbana, Illinois, USA; individual 30-45 minute interviews about the shape of the earth.

Cross-sectional interview study, no control group and no intervention - grade D, and its influence in this field is out of all proportion to its design. Read in full text. Each child answered a fixed set of factual, drawing and GENERATIVE questions ("If you walked for many days in a straight line where would you end up?", "Can you fall off the end or edge?"), with follow-up probing "when needed". Two features must travel with any citation of it. FIRST, the coding scheme was built partly POST HOC: the authors state that during scoring "the data suggested the need for a new model - the flattened sphere", and they created categories of "acceptable" and "unacceptable" deviations, allowing each child exactly one acceptable deviation before being excluded from a model. SECOND, the authors themselves ran the obvious falsification test and it passed - they randomly reassigned responses across children and re-ran the classification, and only 23% of the shuffled subjects could be assigned to any non-mixed category, versus 82% of the real data, chi-square(1, N = 60) = 114.13, p < .001. That test is a genuine defence against "the models are an artefact of loose criteria", but it is not a defence against "the models are an artefact of ambiguous questions", which is the charge Panagiotaki, Nobes & Potton later made and partly sustained.

Of 60 children, only 23 held the scientific spherical-earth model. The rest were assigned hollow sphere (12), dual earth (8), flattened sphere (4), disc earth (1), rectangular earth (1), or mixed/unclassifiable (11). The developmental gradient is steep: 3 of 20 first-graders held the sphere model versus 12 of 20 fifth-graders. Item-level responses show the apparent incoherence directly - 54 of 60 drew a circle for the earth, yet 16 said the earth has an end or edge, 14 said you can fall off it, and 23 said you look "up" to see the earth. Inter-judge agreement was 94% at item level and 80% (16 of 20) at model level before discussion. The theoretical claim, which became the dominant framework in conceptual-change research, is that children do not simply absorb the taught fact; they build SYNTHETIC MODELS (hollow sphere, flattened sphere) that reconcile the culturally transmitted "the earth is round" with entrenched presuppositions derived from everyday experience (the ground is flat, unsupported things fall). For instruction the implication is uncomfortable and is why the paper belongs in this archive: telling children a true fact reliably produced a NEW WRONG MODEL rather than either the correct model or the original naive one - the misconception assimilated the instruction.

Not the operative confound. The operative threat is measurement: the models are inferred from open-ended drawings and verbal answers to questions the child may parse differently from the experimenter.

Contested. Panagiotaki, Nobes & Potton 2009 gave 127 six- and seven-year-olds either the original drawing task or a version with clarified instructions and questions and found substantially more evidence of scientific understanding and less of naive mental models under the revised version; Nobes & Panagiotaki 2007 gave the original instructions to ADULTS and got flat, hollow and dual earths from them too. The synthetic-model construct therefore survives as a description of what children say under one interview protocol, not as an established fact about what they believe.