With the adultlike and nonadultlike responses in view, it becomes evident that the children
exhibit a steady progression toward adultlike use of SUBJ in the nonassertion condition. By
analyzing responses other than the expected SUBJ response, we found that, although it appears
that the 6–7-year-olds are less adultlike than the 4–5-year-olds in that their use of SUBJ decreases,
the 6–7-year-old group looks more adultlike based on their increased use of the va a + infinitive
construction. In fact, the 6- and 7-year-olds actually produce considerably more adultlike forms (i.e.,
the SUBJ or IND va a + infinitive) with no creer to express Nonassertion than the 4–5-year-old
group. By 9–10 years of age, almost all of the children’s alternative responses (with the exception of
two responses) were adultlike involving va a + infinitive. Thus, this alternative response analysis
provides good evidence that with age, children become more adultlike in terms of the types of
responses that they entertain.
The alternative response analysis is corroborated by a one-way ANOVA that excludes SUBJ responses
that are nontarget following no creer (i.e., uses of SUBJ preceded by negation), F(3,75) = 5.671, p=.001,
η
p
2
= .19. Pairwise comparisons using the Bonferroni correction show that in this analysis including only
target SUBJ use, the 4–5-year-old group (37% target SUBJ use) does perform significantly differently
from the adult group,
X
1
–
X
2
= 1.60, SE = .577, p = .042, which was not the case in the original statistical
analysis counting all SUBJ verb forms. The 6–7-year-old group (29% target SUBJ use) also clearly
remains significantly different from the adult group in this analysis,
X
1
–
X
2
=1.93,SE=.582,p = .009.
Since the 9–10-year olds (66% target SUBJ use) did not produce any nontarget SUBJ responses with
negation, they continue to look adultlike in this analysis, with no significant differences in their target
SUBJ use from that of adults (77% target SUBJ use),
X
1
–
X
2
= .44, SE = .559, p = 1.000. By examining the
alternative responses exhibited by the participants, it becomes clear that the children exhibit a steady
progression toward adultlike performance in the Nonassertion condition by exhibiting a progressive
retreat from nontarget responses.
Having described the general path of acquisition in the nonassertion condition, we can now discuss
various factors that may be conditioning the acquisition of the SUBJ in this context. First, the negation
nonadultlike alternative response could derive from syntactic complications revolving around the
negation with negative polarity SUBJ. For example, although the experimenter produced negation in
the matrix clause, the younger children may have produced it again because they have yet to master
negation raising. That is, it could be that the younger children may not have learned that negation has
raised from the lower clause, which is where negation is proposed to have originated in the case of
SUBJ in negated epistemic contexts (Rivero 1971, see also Harrington & Pérez-Leroux 2016 for a
discussion).
4
Additional evidence for children’s difficulty with negation comes from the finding that
English-speaking children can overextend the negative interpretation of the main predicate into the
complement predicate and that this difficulty continues until 6;00–7;11 (Hopman & Maratsos 1978).
Alternatively, the negative responses exhibited by children could be further evidence for the difficulties
processing complementation discussed previously. If the child does not process the sentential comple-
ment as inside the scope of the matrix clause and thus does not evaluate the complement from the
perspective of the subject in the matrix clause, they will likely restructure the complement clause as an
independent clause. This would predict an IND response and, if the child intends to preserve the
negation presented in the matrix clause, a negated IND response.
Second, factors relating to the semantic character of mood selection with nonassertion could
condition acquisition of SUBJ in this context. As discussed in section 2, IND and SUBJ mood
selection in the Nonassertion condition with no creer depends on whether the speaker chooses to
evaluate the embedded proposition as true or false in relation to the default epistemic model of the
speaker (IND) or in relation to the epistemic model of subject in the matrix clause (SUBJ). Since
mood selection depends on the speaker’s semantic intention, SUBJ use in this context is not
4
It is unclear, though, how negation raising could account for the fact that some children also produced negation with IND in the
lower clause, where negation raising is not expected to occur. Additionally, Blake (1980:146) fails to find a global effect of
negation in mood selection when comparing negative with positive polarity matrices.
116 M. DRACOS ET AL.