features to be independent. The empirical results reported here offer support
for the latter rather than the former analysis.
A point to bear in mind, however, is that the analysis of Di Domenico is
based mainly on Italian, whereas the one of Picallo is based mainly on
Catalan, although both are claimed to apply to Spanish. Given that Italian, on
the one hand, and Spanish and Catalan, on the other, differ morphologically
in the sense that gender and number share a morpheme in Italian but are
realized by different morphemes in Spanish, it is possible that the degree of
independence of the two is indeed different in the two languages. It would be
interesting in this respect to conduct the same experiment in Italian.
A gender effect—a difference between masculine and feminine genders—
was found for the correct response category. Although masculine gender has
been considered the default gender in theoretical accounts (Harris 1991), a
difference between the genders was not necessarily expected from a psycho-
linguistic perspective, because both genders can be thought to be specified (as
opposed to having one gender left unspecified—the default—and the other
marked, as is the case with number). In fact, no asymmetry in gender marking
was found in most empirical studies on gender (in Italian [Vigliocco and
Franck 1999]; in Spanish [Igoa, Garcı´a-Albea & Sa´nchez-Casas 1999; Anto´n-
Me´ndez 1999, experiment 1]), nor in the analysis of spontaneous speech
errors in Spanish (Igoa, Garcı´a-Albea & Sa´nchez-Casas 1999). But an effect
of gender has been reported for French (Vigliocco & Franck 1999).
The gender result is puzzling because the direction of the asymmetry is
opposite to that found in studies of number agreement with respect to the
default number. This may be because the default number is unmarked, but the
default gender does not seem to be unmarked. Masculine gender is considered
the default because it is applied to new nouns, it is more common than
feminine agreement, and, furthermore, it is the agreement choice in unclear
cases—that is, when the subject is not specific and is left unmentioned, as in
fue azaroso (‘it was hazardous.
MASC’, where ‘it’ can be the adventure, the day,
etc.), or mixed cases, such as when there are two conjoined heads with
different genders, as in el barbero y su mujer parecı
´
an enojados (‘the.
MASC
barber.MASC and his wife.FEM looked angry.MASC.PL’). Therefore, what may be
happening in these cases where a gender effect was found is not so much the
result of having asymmetrically marked genders, as of speakers’ tendency to
impose masculine agreement whenever in doubt. Another possibility, given
that the item sets with masculine and feminine nouns were different, is that the
sentences with feminine heads were more difficult in some way that we have
not been able to detect, which would be consistent with the fact that the
miscellaneous responses also showed a gender effect.
But why is this effect so variable, being found in some response categories
and not others? And why was it not found in Italian? One possibility is that the
effect is a small one and will only be found when the number of responses in a
given category is large enough; it is possible that, in some experiments, the
effect was not found because of a floor effect. This explanation is supported by
The Relation between Gender and Number Agreement Processing 23
ß Blackwell Publishers Ltd, 2002