Objective To compare outcomes between adjustable spectacles and conventional methods for refraction in young people.Design Cross sectional study.Setting Rural southern China.Participants 648 young people aged 12-18 (mean 14.9 (SD 0.98)), with uncorrected visual acuity ≤6/12 in either eye.Interventions All participants underwent self refraction without cycloplegia (paralysis of near focusing ability with topical eye drops), automated refraction without cycloplegia, and subjective refraction by an ophthalmologist with cycloplegia.Main outcome measures Uncorrected and corrected vision, improvement of vision (lines on a chart), and refractive error.Results Among the participants, 59% (384) were girls, 44% (288) wore spectacles, and 61% (393/648) had 2.00 dioptres or more of myopia in the right eye. All completed self refraction. The proportion with visual acuity ≥6/7.5 in the better eye was 5.2% (95% confidence interval 3.6% to 6.9%) for uncorrected vision, 30.2% (25.7% to 34.8%) for currently worn spectacles, 96.9% (95.5% to 98.3%) for self refraction, 98.4% (97.4% to 99.5%) for automated refraction, and 99.1% (98.3% to 99.9%) for subjective refraction (P=0.033 for self refraction v automated refraction, P=0.001 for self refraction v subjective refraction). Improvements over uncorrected vision in the better eye with self refraction and subjective refraction were within one line on the eye chart in 98% of participants. In logistic regression models, failure to achieve maximum recorded visual acuity of 6/7.5 in right eyes with self refraction was associated with greater absolute value of myopia/hyperopia (P<0.001), greater astigmatism (P=0.001), and not having previously worn spectacles (P=0.002), but not age or sex. Significant inaccuracies in power (≥1.00 dioptre) were less common in right eyes with self refraction than with automated refraction (5% v 11%, P<0.001).Conclusions Though visual acuity was slightly worse with self refraction than automated or subjective refraction, acuity was excellent in nearly all these young people with inadequately corrected refractive error at baseline. Inaccurate power was less common with self refraction than automated refraction. Self refraction could decrease the requirement for scarce trained personnel, expensive devices, and cycloplegia in children’s vision programmes in rural China.