Semantic Feature Analysis

Start with minimal features that clearly distinguish the members of the domain. Add features only if they are necessary (i.e., two items differ in a feature) or systematic (the feature applies to multiple items). Seek confirmation from native speakers: are the proposed distinctions meaningful to them? Features should be motivated by linguistic contrasts, not arbitrary theoretical choices.

A semantic feature is one that is explicitly marked in language—through words, affixes, or grammatical distinctions. A cultural category may be important to the community but not lexically marked. For example, Westerners distinguish 'cousin' by gender (cousin vs. cousin), but in English, cousin is unmarked for gender, reflecting different cultural priorities than some other languages (which distinguish paternal vs. maternal cousins).

Yes, but with caution. Abstract domains are messier and features may be less clear-cut. Build feature matrices for emotion or personality terms, but validate them against native speaker intuitions, corpus data, and contextual usage. Abstract domains often show more metaphor, polysemy, and context-dependence, so a feature analysis may be less complete than for concrete domains.

Feature analysis assumes sharp category boundaries: items either have or lack a feature. Prototype theory allows fuzzy boundaries and graded membership based on similarity to a prototype. Feature analysis is more rigorous for clear-cut domains; prototype theory is more flexible for natural categories where boundaries are soft. Some analyses combine both approaches.