Purpose: An orientation-specific visual evoked potential (osVEP) protocol was developed to probe meridional anisotropies in children with refractive amblyopia. The aim was to characterise the osVEP response in children with bilateral refractive amblyopia, evaluate the intra-session repeatability of the main osVEP components (C1, C2 and C3), coefficient of repeatability (CoR) of the response to gratings in different meridians and determine if refractive amblyopes have poorer repeatability as compared with non-amblyopic controls. Methods: Children aged 4–7 years with newly diagnosed and untreated bilateral refractive amblyopia and non-amblyopic controls were recruited. Orientation-specific pattern-onset VEPs were recorded in response to an achromatic sinewave grating stimulus of 4 cycles per degree under monocular and binocular stimulation. The grating lines used for monocular stimulation were parallel with the subjects’ most positive and negative astigmatic meridians when considered in sphero-minus cylinder form (Meridians 1 and 2, respectively). In subjects without astigmatism, meridians 1 and 2 were designated horizontal and vertical gratings, respectively. Binocular stimuli were presented with grating lines parallel to meridians 45, 90, 135 and 180°. The repeatability of latencies of the main osVEP components (C1, C2 and C3) were investigated using two successive osVEPs recordings for each stimulus meridian and the CoR for each component’s latencies were assessed. Results: Seven amblyopic children (Visual acuity (VA) ranging from 0.08 to 0.40 LogMAR in the less amblyopic eye and 0.26–0.52 LogMAR in the more amblyopic eye) and 7 non-amblyopic controls (VA ranging from 0.00 to 0.02 LogMAR in either eye), with a median age of 4.6 and 7.0 years, respectively, completed the study. C1 had the highest CoR for most conditions assessed. Ratio of CoRs C1:C2 was > 2 for all binocular meridians in controls and the 90 and 180 meridians in the amblyopes; C1:C3 was > 2 for the binocularly assessed 45, 90 and 135 meridians in the controls and the 90 and 180 meridians in the amblyopes; C2:C3 were all < 2 for all meridians assessed in both groups. Conclusions: The osVEP waveforms are reliable and useful for future investigations into the meridional anisotropies in children with refractive amblyopia, particularly the C3 component. Component C1 had the poorest repeatability, which consequentially affected C2 amplitude estimation. Only C3 amplitude and latency could be consistently estimated as C2 and C3 latencies were similarly repeatable. Coefficients of repeatability of osVEP latencies did not appear to systematically differ between non-amblyopic and amblyopic children.