Spike trains of individual neurons can be recorded in awake behaving subjects using extracellular electrodes that are chronically implanted into the brain. Analyzing the information contained in these spike trains involves multiple steps, the first one being spike detection. In one widely used method, all spikes that exceed a researcher-defined threshold are treated as action potential candidates and suprathreshold waveforms are classified as spike or noise. In this study, a new method is developed to compute the spike detection threshold, for the first time completely independent of the researcher, based on hypothesis testing using truncated probability distributions. The use of the thresholds, which are called "truncation thresholds", is illustrated in extracellular recordings from cortical area M1 in awake behaving rats.