indefinite subject constructions

[Matthew S Dryer]

A few comments on the recent discussion of "indefinite subject"
constructions.

First, the discussion ignores that there are two possible types of such
clauses, ones which are semantically intransitive ("someone was singing")
and ones which are semantically transitive ("someone ate the pie").  While
the latter may resemble agentless passives functionally, the former does
not (ignoring instances of passives of intransitives in languages which
allow such - "it was danced").  Languages differ as to whether they have
a single construction covering both of these.

Second, referring to such clauses as "passive" runs the risk of
Eurocentrism.  Prototypical instances of such clauses in languages which
have them are grammatically active and transitive, with a structure like
English "someone ate the pie", except that the "someone" is expressed by a
pronominal affix on the verb rather than with an independent pronoun: in
prototypical cases, the "patient" is grammatical object, the "agent" is
expressed by the indefinite pronominal affix occurring in a morphological
slot associated with subjects, and the "agent" cannot be expressed by a
independent noun phrase.  As Spike suggests, such constructions may
sometimes get reanalyzed as passive constructions, though my impression is
that more often than not, they behave "schizophrenically", behaving like
indefinite subject constructions in some ways and like passive
constructions in other ways.  It is not clear whether this "schizophrenic"
stage is only an intermediate stage in a reanalysis to a passive
construction, or a state in which languages can happily remain
indefinitely.

Indefinite subject constructions appear to be much more common in North
America than in most areas of the world, though the constructions often
deviate from the prototype and the conditions in which they are used are
rarely discussed.  There is a long history of debate among Algonquianists
as to whether the construction in question in Algonquian languages should
be viewed as a passive or as an indefinite subject construction (see
Hockett's preface to Bloomfield's grammar of Ojibwa, where he takes issue
with Bloomfield's characterization).  Boas points out that the apparent
indefinite subject construction in Tlingit occasionally occurs with an
independent expression of the agent.  And I have been working on a
construction in Kutenai which is clearly a passive morphologically and
syntactically, but which is used in texts in ways that are unusual for a
passive, but which are shared with the intransitive indefinite subject
construction, showing that at the level of text, the passive must be
viewed as being a kind of transitive indefinite subject construction.

Matthew Dryer