Chapter 12: Sensitivity to plat presentation


Every principal-part analysis is based on a set of data encoding the inflectional properties of each member of a set of inflection classes. This information, which we represent in a plat, is subject to various manipulations that aim to simplify the presentation without losing data. However, these manipulations often influence the resulting analysis. We present examples that elucidate this problem. We show examples of manipulations based on introducing rules of sandhi, separating stem segments, introducing morphophonemes, representing phonetic instead of spelled surface forms, distinguishing multiple stem contexts, including syntactic information, and omitting problematic inflection classes. These manipulations tend to reduce the number of distinct inflection classes and the number of distillations. In doing so, they also affect all our measures of complexity: principal-part analyses, predictability/predictiveness measures, and entropy-based measures. Several of the manipulations raise questions of linguistic and psycholinguistic motivation, which must be resolved on a case-by-case basis.