Arild Hestvik and William Phillip
Experimental Evidence for the Difference between Syntactic and
Logophoric Reflexive Binding
The binding theory developed by Reinhart and Reuland (1993)
(R&R) states that any reflexive not bound by a co-argument is an
instance of logophoric binding. This raises the question of what the
properties of logophoric binding really is, and whether those cases
that R&R's theory delegate to logophoric status indeed have those
properties. R&R gives as evidence, for example, (i) the ability
of logophoric reflexives to pick up antecedent in the discourse, and
not via syntactic binding, and (ii) the ability to be in
non-complementary distribution with pronouns. However, the evidence
is not unequivocal. One diagnostic for logophoric binding is the
ability to take a non-c-commanding antecedent (probably related to
the discourse binding ability). But compare (a) and (b):
(a) Clinton's car carried a picture of himself on its top.
(b) ??Clinton's car was parked behind himself
Whereas a reflexive in a picture- NP easily takes a
non-commanding antecedent, a reflexive in a locative PP does not.
Concerning the non-complementarity, this is not an empirical
argument, but a theoretical one. It is easy to design a binding
theory that predicts non-complementary distribution in specific
circumstances (cf. Huang 1983, Hestvik 1991), specifically, the
BT-compatibility theory of binding domains from Chomsky (1986) can be
made to predict that binding of a reflexive into a locative PP is a
"core" case of binding, as opposed to binding into picture
NPs.
Are there other sources to decide on the
"grammatical-or-logophoric" binding issue? Avrutin and Cunningham
(BUCLD 1996) presented experimental evidence that English-speaking
children assigned antecedents to reflexives in precisely contexts of
locative PPs in a way that supported the logophoric analysis of this
type of binding, i.e. children accepted non-c-commanding discourse
referents as antecedents. Their argument is that logophoric binding
requires knowledge of discourse and pragmatic conditions of use,
commonly assumed to be missing in pre-school children. Consequently,
the child data provides a unique window of evidence for the
grammatical/logophoric binding distinction.
However, the purpose of this talk is to bring forth new evidence
from acquisition of Norwegian binding that contradicts these
findings, i.e. which shows that binding of a true long-distance
reflexive like Norwegian seg into a locative PP and binding of
a possessive reflexive in Spec of NP/DP are indeed NOT cases of
logophoric binding, contradicting the consequences of the R&R
theory that must consider these two cases of non-co-argument binding
as logophoric binding.
The evidence comes from a truth-value judgment experiment carried
out in Bergen, Norway with around 30 children (mean age 5.10). The
experiment was constructed as a game with pictures, in which the
children would see a picture depicting some action involving two
participants, and the experimenter, who would not see the picture,
had the task of purportedly "guessing" what the pictures showed, by
asking the child questions and having the child answer yes or no as a
confirmation about whether the guess was correct or not.
We found that children performed virtually adult-like with
sentences like (c), with the Norwegian LD reflexive seg and
(d) with the possessive reflexive sin ,
(c) Satte jenten stolen bak seg
"Did the girl put the chair behind SEG"
(d) Loftet gutten hatten sin?
"Did the boy lift his hat?"
According to the R&R theory, these are cases of logophoric
reflexives, and one would therefore expect children to allow
discourse reference in these cases, just as in the Avrutin &
Cunningham experiment. However, when the adult answer to these
questions were "yes", the children would virtually always also answer
"yes", i.e. they did not allow the accusative or possessive reflexive
to corefer with the other potential (but non-c-commanding) discourse
referent---even though with accusative and possessive PRONOUNS they
would do this significantly more often, displaying the typical "delay
of principle B"-effect (cf. Hestvik & Philip, NELS 1996) here.
However, in cases of binding into picture NPs, as in (e),
(e) Viste hesten gutten et bilde av seg?
"Did the horse show the boy a picture of SEG?"
children accepted the non-subject antecedent (i.e. the non-adult
reading) roughly 50% of the time.
This finding strongly suggests that the Norwegian pre-school child
grammar treasts these three cases of binding differently. Binding
into a locative PP and of a possessive reflexive pronoun do NOT
pattern with pronoun binding (where the "delay of principle B"-effect
appears); but binding of a reflexive in a picture NP does pattern
with pronouns, displaying the delay of principle B-effect.
An alternative binding theory, where these cases of reflexives are not logophors, but X-0 reflexives that move into a local binding configuration with their antecedent at LF, can however accomodate these facts. The main conclusion, however, is that a simple equation between coargument vs non-coargument surface binding and grammatical vs. logophoric binding is not supported by the acquisition data from Norwegian.