Pharmaceutical computation increasingly depends on heterogeneous molecular, biological, experimental, clinical, and computational resources that are repeatedly transformed across data acquisition, preprocessing, modeling, interpretation, and decision-support activities. Although workflow systems, metadata standards, FAIR data practices, machine-learning reporting recommendations, and biomedical knowledge graphs address important parts of this landscape, they do not provide a shared language for describing how pharmaceutical evidence is produced, modified, interpreted, and reused. This article proposes a Common Pharmaceutical Provenance Language as an original ontological contribution for representing the entities, activities, agents, relationships, versions, decisions, and evidential boundaries that constitute computational pharmaceutical discovery. The proposed language distinguishes source observations from derived data, transformations from their outputs, computational predictions from interpretations, and evidence lineage from claims of mechanism, utility, or therapeutic validity. Its architecture organizes provenance across four linked analytical zones: acquisition, preprocessing, modeling, and interpretation. It further specifies minimum metadata requirements, typed relationships, version-aware transformation records, human-decision objects, and interfaces with FAIR data systems, executable workflows, and knowledge graphs. The contribution is conceptual rather than empirically validated. It does not establish that provenance completeness guarantees reproducibility, that transparent lineage ensures data suitability, or that auditable computation demonstrates biological mechanism, clinical value, or regulatory acceptability. Instead, it provides a structured vocabulary through which these distinctions can be recorded, evaluated, and communicated. Future research must test semantic coverage, cross-platform interoperability, annotation burden, governance processes, domain extensibility, and the relationship between provenance quality and reproducible pharmaceutical inference. A shared provenance language may thereby support more traceable computational research while preserving the limits of what computational records can establish.