It is possible to state verb agreement in Hindi in terms of word order and surface case, with no reference to grammatical relations: the verb agrees with the leftmost phonologically null instance of case marking (Saksena 1981: 468). It is questioned whether this is, however, the correct statement of the rule. Other Indo-Aryan languages, closely related to Hindi, present interesting variations on the case marking of subjects and direct objects, and in all such variations it turns out that the grammatical relations, shared with Hindi, take precedence over case marking. While Hindi verb agreement can be stated as above, it is doubtful whether this is indeed a simpler rule, in the sense of a rule preferred by speakers of a natural language, than one stated primarily in terms of grammatical relations with subsidiary reference to case marking: in the imperfective, the verb agrees with its subject unless this is overtly case marked; in the perfective, agreement is with the absolute (intransitive subject, transitive direct object) unless this is overtly case marked.
A recent upsurge of interest in the general linguistic problem of agreement -I write these words between a session on agreement at the Eastern States Conference on Linguistics (Ohio State University) and a conference on agreement at Stanford University -has led me to reconsider the analysis of verb agreement in Hindi presented by Saksena (1981). 1 In Hindi, the verb can agree with at most one of its arguments, in some combination of gender, number, and person. Thus an intransitive verb can agree with its subject, while a transitive verb can agree with either its subject or its direct object. It is also possible for a verb, transitive or intransitive, to agree with none of its arguments; in this instance, the verb stands in the default form, third person masculine singular. (Since the third person masculine singular is ambivalent in Hindi, reflecting either agreement with a third person masculine singular argument or no