AIMS To assess the agreement between the hand-held autorefractor Retinomax(R) and three different on-table autorefractors when measuring cycloplegic refraction in subjects with small and high ametropia. To assess the agreement between the cycloplegic refraction using the Retinomax(R) and by retinoscopy in children with small and high ametropia. METHODS Part A.276 subjects were refracted under cycloplegia using both the Retinomax(R) and an on-table infrared automated refractor (Topcon RM-A 6000, Nidek AR 800 or Nikon NR 5000). They were separated into subjects withsmall ametropia (mean sphere </= 3.5 D hyperopia, </= 3 D myopia) and high ametropia (mean sphere > 3.5 D hyperopia, > 3 D myopia). The agreement between both types of refractors regarding the different refractive components was assessed for the whole group and for the two subgroups of small and high ametropia. Part B. 48 infants were refracted under cycloplegia by retinoscopy and by the Retinomax(R). The agreement between both methods of refraction was analyzed in the same manner as in part A. RESULTS Part A. No significant bias was found between the two types of refractors with regard to the spherical equivalent. The 95% limits of agreement were +/- 1 D. Although no clinically significant bias was found with regard to the cylinder power in the 276 subjects, it was found that the 95% limits of agreement were much better (+/- 0.75 D) in small ametropia subjects than in high ametropia subjects (-2.1 to +1.3 D). No significant bias was found with regard to the axis determination. Part B. No significant bias was found between the Retinomax(R) and retinoscopic measurements with regard to the spherical equivalent. The 95% limits of agreement were -1.36 to +1.76 D. However, the mean difference for spheres and cylinders showed a positive bias and a negative bias, respectively, suggesting more positive spheres and larger cylinders when measured by the Retinomax(R) compared to retinoscopy. This was particularly obvious in cases of high ametropia. CONCLUSION Compared to retinoscopy and on-table autorefraction, the hand-held refractor Retinomax(R) is accurate in any ametropia with respect to the spherical equivalent. In small ametropia, there is a good accuracy when measuring the three refractive components (sphere, cylinder and axis). The accuracy decreases in high ametropia, especially with regard to the cylinder power.