Stereoscopic system weights correlation- and match-based disparity representations to support perception of binocular depth. Fine depth perception mainly uses the match-based representation, while coarse depth perception combines both (Doi et al., 2011, 2013). We previously reported that neural responses in macaque area V4 is consistent with match-based representation in that the disparity-tuning amplitude of V4 neural ensemble gradually decreased as the proportion of binocularly contrast-matched elements (match level) reduced. However, the precise relationship between these responses and depth perception remains unclear. To address this question, we applied a standard decision-making model to our recorded data and derived psychometric functions. We constructed two neuronal pools with opposite preferences by resampling from our dataset. We assumed a weak noise correlation (i.e. covariance in the activity of pairs of neurons) within and between neuronal pools according to previously reported values in V4. The model estimated the sign of stimulus disparity (i.e., near vs. far) through an optimal linear readout of population (pooled) responses: a weighted sum of responses of near-preferring neurons were compared with that of far-preferring ones and the pool with larger response dictated the decision. We simulated psychometric functions by feeding into the model the neuronal responses at different disparity magnitudes (i.e. fine and coarse) and different match levels. We found that the simulated psychometric function had qualitatively similar shape to that observed for human subjects in a fine depth discrimination task, although the simulated sensitivity was poorer than behavioral sensitivity. In coarse task, the neurometric and psychometric functions had qualitative differences. These results suggest that the population responses of area V4 is more consistent with fine stereopsis. A further refinement of V4 responses in downstream areas might fill the gap between neurometric and psychometric functions.