A cognitive analysis of subjective probability is applied to the evaluation of techniques used by decision analysts for eliciting probabilities from experts. The construction of a subjective probability requires both the formation of a belief and the assessment of a probability that qualifies the belief. The former process involves judgment and reasoning; the latter is purely judgmental. Subjective probabilities have traditionally been portrayed and studied as arising from judgment. Consequently, belief assessment procedures have been particularly underdeveloped. Procedures currently used by analysts to aid belief assessment are classified and evaluated. Although such procedures facilitate the communication of beliefs and offer important guidance for constructing probabilities, additional prescriptive development is possible. It is argued that significant improvements in assessment practice can be realized by providing better support for the reasoning employed by experts in belief assessment. Opportunities for descriptive and prescriptive research in belief assessment are identified.