Disclosed are systems and methods for managing testing unit latency and measurement uncertainty in computer-based stimulus-response tests. An estimated latency LE and an associated measurement uncertainty are determined as characteristics of a particular testing unit. LE is used as offset for all measurements taken on the testing unit, and results treated subject to the characteristic measurement uncertainty when determining test-taker performance. Estimated actual response times RTE are processed subject to a confidence value determined from the uncertainty. Uncertainty propagation determines test metrics involving a plurality of estimated actual response times RTE, where cumulative uncertainty is reported as a confidence rating in the metric. Overall test results (e.g., pass vs. fail) based on one or more metrics are also reported according to a confidence rating associated with the cumulative uncertainty propagated through the relevant metrics. Various calibration techniques are disclosed for determining the latency estimate LE and associated uncertainty values.