Explaining the Ditransitive Person-Role Constraint: A usage-based approach

Abstract

In this paper, I propose a frequency-based explanation of the Ditransitive Person-Role Constraint, a cross-linguistic generalization that can be formulated as follows: “Combinations of bound pronouns with the roles Recipient and Theme are disfavored if the Theme pronoun is first or second person and the Recipient pronoun is third person.” This constraint is well-known from Spanish and other Romance languages: Alicia se lo mostrar . 'Alicia will show it to her.' (3rd theme, 3rd recipient), but * Alicia te lemostrar . '...you to her.' (2nd theme, 3rd recipient).
The theoretical literature offers a number of explanations of this constraint (e.g. in terms of structural positions, a clash of positional alignment requirements, or markedness), but none can account for the fact that it is both widely found in the world’s languages, independently of morphosyntactic factors like case-marking peculiarities, and non-universal (some languages are shown to violate the constraint).

My own proposal starts out from the observed correlation between allowed grammatical patterns and frequency in language use. In languages that lack bound pronouns and therefore cannot be subject to the constraint, we see a significant skewing in the frequencies of various person-role combinations. Combinations with 1st/2nd Recipient and 3rd Theme greatly outnumber combinations with 3rd Recipient and 1st/2nd Theme, although the latter do occur occasionally. This performance-grammar correspondence can be explained on the basis of a principle of language change that I call the Frequency Condition on Entrenchment in Grammaticalization. It says that when a loose combination of expressions becomes entrenched and is conventionalized as a separate construction, which particular  elements may figure in the construction often depends on their frequency of occurrence. I go on to explain in semantic-pragmatic terms why certain combinations of Recipient and Theme are rare, and I show that grammaticalizations of these usage preferences are more widespread than current discussions of the Ditransitive Person-Role Constraint imply.