The purpose of this study was to compare refractions measured with an autorefractor and by retinoscopy with and without cycloplegia. The objective refractions were performed in 199 right eyes from 199 healthy young adults with a mean age of 21.6 +/- 2.66 years. The measurements were performed first without cycloplegia and repeated 30 min later with cycloplegia. Data were analysed using Fourier decomposition of the power profile. More negative values of component M and J(0) were given by non-cycloplegic autorefraction compared with cycloplegic autorefraction (p < 0.0001). However more positive values for the J(45) vector were given by non-cycloplegic autorefraction, although this difference was not statistically significant (p = 0.233). By retinoscopy, more negative values of component M were obtained with non-cycloplegic retinoscopy (p < 0.0001); for the cylindrical vectors J(0) and J(45) the retinoscopy without cycloplegia yields more negative values (p = 0.234; p = 0.112, respectively). Accepting that differences between cycloplegic and non-cycloplegic retinoscopy are only due to the accommodative response, the present results confirm that when performed by an experienced clinician, retinoscopy is a more reliable method to obtain the objective starting point for refraction under non-cycloplegic conditions.