For two centuries, pathology practice has woven precise quantitative observations with rich qualitative analysis. Pathologist language is spun from this web, encoded in a unique parlance that efficiently communicates both measurable features and their clinical implications. While current machine learning approaches excel at specific quantitative tasks, their attempts to fully capture this expertise reveal a deeper challenge. Whether focusing on human-interpretable measurements or abstract deep-learning features, these approaches often miss the sophisticated interplay between observation and interpretation that characterizes pathology practice. Through tracing the evolution of histologic feature analysis, a path toward computational methods may emerge: one that truly complements pathology workflow - preserving the nuanced language pathologists use while enhancing their ability to derive insights from both qualitative and quantitative observations.