Many aspects of maintenance are affected by uncertainty. Observations obtained through inspection and monitoring are often not more than probabilistic indicators of the underlying complex deterioration processes. Therefore, analysis in support of maintenance must usually cope with incomplete information. Probabilistic methods are suggested for various aspects of maintenance analysis, such as inspection timing, flaw detection, failure rate prediction, reliability assessment, benefit cost analysis, and evaluation of decision alternatives. Examples demonstrate the relationship between life span, failure rates and reliability, and the probabilistic evaluation of decision criteria and decision alternatives. The probabilistic approach allows the inclusion of subjective experience and judgment in the quantitative analysis. This is shown to be especially important when analyzing risky decisions.