We sought to demonstrate cardinal-like (absolute), rather than ordinal (relative), numerical recognition in the primate species Macaca mulatta (rhesus monkeys), which has been insufficiently explored in the literature, to contribute to determining the features of the cognitive representations underlying such an ability. Two adult male rhesus monkeys-Augustus and Spike-were trained to identify numerosities from 1-9 in a delayed numerical matching-to-sample task that required subjects to select a previously shown numerosity among two distractor numerosity arrays, after a delay of one second. Stimuli were boxed numerical arrays composed of various geometric shapes. Attention to numerosity alone as the basis for identification of an array was ensured by controlling for non-numerical stimulus cues such as surface area, cumulative perimeter, density, spatial layout, shape and forecolor of the individual components of the arrays. Reaction time and accuracy data revealed end effects, with performance for items towards the ends of the stimulus continuum better than for middle items. Accuracy data for both subjects also revealed a distance effect for mean target-to-distractor distance, although reaction time data did not. Results indicate parallel processing for all numerosities trained, a continuous representation of small and large values alike, and no scalar variability in numerical representation, opposing the predictions made by Weber’s Law and the analog magnitude estimation model of representing numerosity.