An informal model of latent inhibition (LI), the trace hypothesis, is described. This hypothesis holds that preexposure (PE) to the to-be-conditioned stimulus alters the salience (associability) of the CS such that only the onset and initial segments, but not the later segments, are capable of supporting conditioning when paired with a US. Thus, LI is effectively a trace conditioning phenomenon. Four experiments used rats and a one-trial fear-conditioning task to test predictions which stem from this hypothesis. Experiment 1 showed conditioning performance decreased as trace intervals increased between 0 and 12 s, thus demonstrating the sensitivity of this task to trace intervals proposed by the trace hypothesis to produce LI. Consistent with the trace hypothesis, LI was found to be an increasing function of the CS duration (Experiment 2), the number of PE trials (Experiment 3), and the US onset during a 15-s CS (Experiment 4). These data are somewhat problematic for several explanations of LI, but, in general, the trace hypothesis extends extant explanations by focusing on changes in salience within a given presentation of the CS.