Variationist typology: Shared probabilistic constraints across (non-)null subject languages

A key parameter in received classifications of language types is the expression of pronominal subjects. Here we compare variation patterns in conversational data of English – considered a non-null-subject language – and Spanish – a well-studied null-subject language. English has a patently lower rate of expression (approximately 3% unexpressed 1sg and 3sg human subjects vs. 60% in Spanish). Despite the stark difference in rate of expression, the same probabilistic constraints are at work in the two languages. Contrary to popular belief, VP coordination is neither a discrete nor a distinguishing category of English. Instead, a shared constraint is linking with the preceding subject, a refinement of accessibility to include, alongside coreferentiality, measures of structural connectedness – both prosodic and syntactic. Other shared constraints on unexpressed subjects are coreferential subject priming (a tendency to repeat the form of the previous mention) and lexical aspect (reflecting the contribution of a temporal relationship to subject expression). Where the languages most differ is in the envelope of variation. In English, besides coreferential-subject verbs conjoined with a coordinating conjunction, unexpressed subjects are limited to prosodic initial-position in declarative main clauses, a restriction that is absent in Spanish. We propose that the locus of cross-language comparisons is the variable structure of each language, defined by the set of probabilistic constraints but also the delimitation of the variable context within which these are operative.