Form and distribution
Given the clear ‘atomistic’ tendencies in Bloomfield’s own work, it is not entirely plausible to claim him as a precursor of contemporary WP approaches. Nevertheless, the analysis of form into sub-meaningful contrasts and the imposition of operational criteria on the analysis of meaning determine a strikingly convergent perspective. The morphological variation exhibited by a system can be described in terms of contrasts within and across the two observable dimensions of variation in a system, form and distribution classes.
Despite the importance imputed to semantics in the Post-Bloomfieldian tradition, grammatical analysis remains as independent of issues of semantic ‘content’ as it was during Bloomfield’s time. The main semantic component of grammatical analysis is the ‘same-different’ discrimination invoked to distinguish phonemes from allophones. The specific ‘meanings’ assigned to a pair of words play no role in determining the phonemic structure of a language; all that matters is that speakers can determine whether they are the same or different. Any more fine-grained meanings required for the analysis of a grammatical system can again be expressed in terms of distributional vectors.
Grammatical features, properties and categories can likewise be interpreted as proxies for form classes, distribution classes or some combination of the two. In this way, morphological terminology that misleadingly implies an associated semantics can be reduced to robustly observable dimensions of form variation. The overloading of grammatical terms is often most pronounced in the classification of verb forms in terms of ‘tense’ features that imply a temporal interpretation, and ‘mood’ and ‘aspectual’ features that imply meanings connected to notions like factivity, perfectivity, etc. The nomenclature for nouns can be less semantically loaded, as reflected by descriptions of the German case system that eschew Latinate terms in favour of a simple enumeration of forms into erster/zweiter/dritter/vierter Fall ‘first/second/third/fourth case’. But these categories are all essentially taxonomic; to the extent that the labels have any function at all it is as mnemonics for the meanings conventionally associated with a form or even with the most salient use of that form. Particularly in realizational models, the labels may have no semantic effect but merely encode form contrasts that are ‘spelled out’ by realization rules, in the process subverting the basic feature-form ‘separation’ (cf. Beard 1995).
-  See Ramscar and Port (2015) for discussion of the problems that arise on a view of categorizationon which features are taken to associate items with categories.
-  The mnemonic character of feature labels is largely a symptom of the fact that they, like most ofthe classificatory apparatus of modern linguistics, are simply inherited from earlier, mainly classical,traditions. One solution involves clarifying the correspondence between features and observable formand distribution classes, as proposed in Blevins et al. (2016a). Alternatively the implicit semantics canbe specified, for example, by tracing definitional paths for labels that make their semantics explicit, asproposed in the domain of tense and aspect by Lieb (2005) and Viguier (2013).