This study examined a proposed procedure for constructing theoretical descriptions of mechanics problems, in particular, to determine if the procedure led to explicit and correct descriptions of the motion and interaction of systems and if the resulting theoretical descriptions facilitated generation of correct constraint equations and, hence, correct solutions to the problems. Twenty-four paid volunteers (undergraduate physics students) were randomly assigned to one of three groups: (1) model group, induced to solve problems in accordance with the full version of the proposed procedure; (2) modified-model group, induced to work with a less complicated version; and (3) a comparison group which solved problems without any external guidance. Criteria used as measures of good problem-solving performance and major classes of errors were established and adequacy of solution was assessed with respect to these performance measures. Very explicit rules for constructing initial problem descriptions were found to lead reliably to explicit and correct descriptions of motion and interaction. In addition, these descriptions were found to facilitate achievement of correct solutions. Specific examples of typical difficulties subjects encountered during problem-solving are also discussed. (Author/JN)