This paper describes a proposal to classify the conceptual variation of terms represented by regular and irregular polysemic processes to which terms as lexical units are subjected in discourse. Definitions and explanatory contexts are retrieved from a specialized text using metalanguage patterns. Data are analyzed quantitatively and qualitatively. The conceptual equivalence axis is used as an instrument for the semantic analysis of in-context term behavior. Conceptual variation represented by regular polysemic processes comprises 81% of the sample, whereas conceptual variation represented by irregular polysemic processes makes up 14% of this. Mixed variation occurrences account for 5% of the sample. Criteria are specified to distinguish among various meanings of a same concept represented by a term or its lexical-semantic variants. Results will provide a basis for future research on the disambiguation of in-context terms.