How good a bet is good form?

'Good form' theories of perception leave some latitude concerning how casually or cautiously order is imposed on the stimulus. Exploring this issue, the present experiment introduced a series of figures admitting up to three alternative 'good' readings. Eight college students estimated two angles in each of fifty-six pictured spatial forms. Theory predicted that geometrical regularities of rectangularity and symmetry would dominate their estimates. But a regularity would rarely appear when inconsistent with projective geometry and the given figure. The accuracy of subjects' estimates was also assessed. The results confirmed the hypotheses at high significance levels, arguing that such figures are interpreted through an order-imposing process restrained by projective geometry, and that subjects could make roughly accurate estimates based on the imposed order. Parallels with computer scene analysis are discussed. It is concluded that perceptual presumption of certain 'good forms' runs little risk of misinterpreting the stimulus.