Re-binding and the Derivation of Parallelism Domains

Here, there emerges a clear preference for the strict reading hit John. What might account for this remarkable fact? In this paper, I will suggest that this rebinding puzzle reflects a fundamental fact about the computational system which links syntactic structure with meaning. I will propose that this system operates in a monotonic fashion: that is, meaning representations are constructed as early as possible during a bottom-up derivation, and the resulting meaning representations cannot be revised later. In what follows, I begin with the original account of the re-binding puzzle, due to Sag (1976), who first observed the phenomenon. I argue that Sag’s account is unsuccessful, and I continue with a proposed Re-binding Generalization: namely that re-binding is possible only when necessary to satisfy parallelism. I next show that this generalization follows if the syntax-semantics interface operates monotonically. At this point, I describe a survey I performed which supports Sag’s original observations concerning re-binding—this is important, because the observations are subtle, and have been questioned in the literature. Next, I turn to a puzzle concerning focused pronouns. I argue that the Monotonic Derivation system makes available a