Poznan09



Agreement and its failures
Omer Preminger


In this course, we will examine issues of agreement in natural language.

While the mechanism of agreement has been exploited in the accounts of a wide range of phenomena, we will restrict ourselves to the kind of phenomena it was originally intended to account for—namely, agreement in phi-features (e.g., number, person, gender). Thus, we will be concerned with instances in language where the phi-features of a particular noun-phrase are (morphophonologically) represented not only on the noun-phrase itself, but also on some other nearby element—such as a lexical verb, an auxiliary verb, or a tense/aspect marker.

While agreement in phi-features has often been lumped together with other syntactic mechanisms that adhere to a feature-checking logic—where failure to check features results in a "crash", and hence in ungrammaticality—we will discuss evidence from Icelandic, Hindi-Urdu, Basque, Hebrew, and Uyghur, that shows that this cannot possibly be right for phi-agreement. In all of these languages, phi-agreement is certainly not optional: there are plenty of sentences where the omission of the morphosyntactic signs of agreement would indeed result in ungrammaticality; but crucially, there are plenty of examples (at least in these languages) where phi-agreement appears to have failed, but the result is nonetheless a grammatical utterance.

Some of the issues we will discuss:

• possible causes of failed phi-agreement (e.g., syntactic intervention, domain restrictions)
• an important morphosyntactic distinction between the presence of a default agreement-morpheme, and the absence of an agreement-morpheme altogether
• whether the aforementioned instances of failure to agree can be subsumed under agreement with some other element bearing default phi-features (SPOILER: no!)

Finally, we will sketch an alternative theory of phi-agreement as pure feature-valuation—i.e., literally, the transmission of a feature value from one syntactic locus to another, with no additional machinery assumed or implied—which fares much better with respect to the data under consideration.



home