Subjects in L1 Attrition: Evidence from Greek and Italian Near-Native
Speakers of English
*
Ianthi Tsimpli1, Antonella Sorace2, Caroline Heycock2, Francesca Filiaci2, and
Maria Bouba1
1
University of Cambridge and 2University of Edinburgh
1.
Background assumptions and the basic hypothesis
In this paper we present some results from a study that we have been
conducting into the effects of syntactic attrition. Syntactic attrition is a
consequence of long-term contact of the native with the second language, and is
expected to be found primarily in near-native speakers of the second language; it
could be described as a change in the steady-state of L1 as found in monolingual
speakers.
Within the Principles and Parameters framework, the change in the S s (L1)
is predicted to affect aspects of a parameter for which the L1 and L2 assume a
different value (e.g. the Null Subject parameter in Italian vs English). However,
syntactic attrition cannot be argued to change (or “unset”) the L1 value of a
given parameter as this would imply that the syntax of the native language can
be lost when a second language is acquired.
In Minimalist terms (Chomsky 1995), morpho-syntactic features differ in
terms of their interpretability at LF. Parameters are associated with the value of
non-interpretable features only. By assumption, parameter-setting in L1
acquisition remains constant and cannot be affected by syntactic attrition (for
reasons to do with the modularity of the language system). In L2 acquisition and
language impairment, the distinction between interpretable and noninterpretable features seems to affect learnability (Tsimpli 1996, 2001).
Our hypothesis is that syntactic attrition primarily affects morphosyntactic
features that are interpretable at the LF interface (Sorace 2000). Being an
interface level, LF is not modular and allows for changes in the interpretation of
morpho-syntactic features in the context of L1/L2 contact resulting in syntactic
attrition. Through attrition the affected features become unspecified, giving rise
to optionality.
2.
Aims and hypotheses of the study
Our aim in this study was to investigate syntactic attrition in the domain of
grammatical subjects in Greek and Italian near-native speakers of English. More
*
This research was supported by ESRC grant R000239158.
specifically, the areas of investigation included the use of null and overt
subjects, the positioning of subjects pre- or postverbally, the interpretation of
overt subjects in these two positions as discourse-old or discourse-new, and
subject extraction out of declarative complement clauses.
2.1. Null vs overt subjects
Greek and Italian are [+null subject] languages, English is [-null subject]:
(1) a.
b.
c.
Efije
left-3S
‘He/She left.’
E
partito
is-3S gone-M
‘He left.’
*(He/she) left.
In Greek and Italian, the option of a null or overt subject is conditioned by the
[topic-shift] and the [focus] feature.
(2) a.
b.
O Janisi prosvale ton Petrok otan proi / aftosk ton plisiase.
the Janis insulted the Petro when pro / he
him approached
‘Janisi insulted Petrok when hei/k approached him.’
O Janisi prosvale ton Petrok otan AFTOSi/k ton plisiase.
‘Janis insulted Petro when HE approached him.’
The prediction that follows from our hypothesis with respect to null and
overt subjects is that attrition will not affect the null subject option of Greek and
Italian (given that this is a parametric option regulated by a non-interpretable
feature). On the other hand, attrition is predicted to affect the distribution and
interpretation of overt pronominal subjects in L1 Greek and Italian (given that
this is regulated by the interpretable [topic-shift] and [focus] features). In
particular, we expect the features to become unspecified. As a result, overt
pronominal subjects in attrited Greek and Italian will not necessarily be
interpreted as shifted topics or as foci.
2.2. Preverbal and postverbal subjects
Postverbal subjects are possible in null subject languages as a consequence
of a positive setting for the [null subject] parameter (Rizzi 1982, 1986). The
choice of postverbal over preverbal subjects in Greek and Italian one-place
predicates is further regulated by (i) the definiteness of the subject (cf. Belletti,
1988) and (ii) the thematic properties of the verb (Pinto 1997):
(3) a.
Irthan
kapji fitites.
arrived-3P some students
b.
I fitites irthan.
the students arrived-3P
(4) a.
Edho dulevi o
Yanis.
here works the-NOM Yanis
b. # Edho o
Yanis dulevi.
here the-NOM Yanis works
(5) a.
Sto telos tis
tenias o
Yanis jelase.
at the end of the movie the-NOM Yanis laughed
b. # Sto telos tis
tenias jelase o
Yanis.
at the end of the movie laughed the-NOM Yanis.
Roughly, Pinto argues that verbs like work with an intrinsic locative argument
(whether expressed overtly or not), prefer postverbal subjects whereas verbs like
laugh that lack such an argument prefer preverbal subjects (cf. (4) and (5)).
There is a further distinction between preverbal and postverbal subjects:
preverbal indefinite subjects are interpreted as “old” information (topic) whereas
postverbal subjects may be interpreted either as “old” or “new” information.
(6)
a.
b.
I gitonisa mu ston trito orofo apektise dhidhima.
`My neighbour on the third floor had twins.’
Xtes vradhi ena moro ekleje.
last night one baby was-crying (= one of the twins)
Xtes vradhi ekleje
ena moro.
last night was-crying one baby (= one of the twins OR some other
baby)
Finally, it should be noted that Greek differs from Italian with regard to
postverbal subjects: Greek is a VS(O) language, the postverbal option being
unmarked with respect to its interpretation, whereas Italian reserves the
postverbal subject position for new information focus (Belletti, 2001, Roussou
& Tsimpli, 2002).
The prediction that follows from our hypothesis here is that the effects of
attrition are unidirectional: attrition will result in preverbal subjects being
preferred over postverbal subjects, overriding differences in the thematic
properties of the verb. Thus, the frequency of production of postverbal subjects
will decline, but when used, they will be used correctly. Further, attrition will
result in possible ambiguity for preverbal indefinite subjects between “old” and
“new” interpretation.
3.
Subjects and Methodology
Four groups of subjects participated in the study: one group of Italian nearnative speakers of English (n=20), one group of Greek near-native speakers of
English (n=19), one group of Italian controls (n=20) and one group of Greek
controls (n=20). Subjects in the experimental groups (Greek and Italian nearnative speakers of English) had a minimum of 6 years of residence in Britain,
used L1 and L2 in everyday contexts and their level of English (L2) is nearnative according to White & Genesee’s (1996) criteria. Subjects in the control
groups live in Italy and Greece respectively and have minimum or no
knowledge of English.
The study consisted of four tasks testing production and comprehension of
null and overt NP and pronominal subjects, and a grammaticality judgement
task. This paper is restricted to the presentation of relevant data from one
production task (“Headlines”), and one comprehension task (“Picture
Verification”). For lack of space, we present some data from the Greek speakers
only, and some from the Italian speakers only.
3.1 The Headlines Task
In this production task the subject was asked to produce a sentence
consisting of a given verb, an NP and an adverbial expression presented to
him/her as scattered phrases on the computer screen. The phrases were
presented together with a picture depicting the story meant to be described by
the phrases. The subject was asked to produce a sentence starting with the
phrase Did you hear that…. In some cases the subject was indefinite (e.g. …a
building collapsed because of an explosion); in some definite (… the Wimbledon
tournament started in bad weather).
The aim of the task was to test the use of pre- and postverbal subjects in this
all-focus context. The postverbal position was expected to be preferred by the
control group to a greater extent than by the experimental group. The effect was
predicted to be stronger for the Greeks than for the Italians.
3.2 The Picture Verification Task
In this task there were both test items and “fake fillers.” For the former, the
subject was presented with a sentence and a set of three pictures (presented in
randomized order for each subject) and was asked to indicate which of the three
pictures matched correctly the meaning of the sentence. The sentence consisted
of two clauses, one main and one subordinate. The subject of the main clause
was always an NP; the subject of the subordinate clause could be either an overt
pronoun or a null subject. The main clause always had an animate complement.
The subordinate clause could either precede or follow the main clause (forward
or backward anaphora):
(7) a.
Mentre leik/l/proi si
mette il cappotto, la madrei da’ un bacio
while she/pro REFL puts the coat
the mother gives a kiss
alla bambinak.
to the daughter
b.
La madrei da’ un bacio alla bambinak mentre leik/l/proi si mette il
cappotto
‘The mother kisses the daughter while she/pro is wearing her coat.’
The aim of the task was to identify the preferred interpretation for the referent of
the subject of the subordinate clause.
The “fake fillers” had the same presentation as the test items but in this case
there were two sentences in each item. The first sentence provides a possible set
of referents and an event, and the second includes a singular indefinite subject in
either preverbal or postverbal position (see (6) above). The aim of the task was
to identify a preference for the “old” or “new” referent interpretation, or the lack
of preference (ambiguity).
4. Results
4.1 Headlines Task (Greek)
As Figure 1 shows, there is a significant difference between the two groups
of Greek speakers in their production of preverbal definite subjects in the
Headlines task (sig=.002). There is also an effect of subject type/definiteness for
the experimental group (Subj type*Group: sig=.034).
70%
61%
% of sentences
60%
50%
42%
40%
34%
29%
30%
Control
Experimental
20%
10%
0%
Indefinite Subject
Definite Subject
Figure 1. Headlines Task: production of preverbal subjects (Greek)
4.2. Picture Verification Task: interpretation of indefinite subjects (Greek)
Figures 2 and 3 show the results from the “fake filler” items for the Greek
speakers: the task here concerned the interpretation of indefinites in pre- and
post-verbal position as “old” or “new”. Subjects were permitted to pick more
than one picture, as reflected in the figures for “both”; completely irrelevant
answers were classified as “wrong.” Figure 2 gives the interpretations that were
associated with preverbal indefinites, and Figure 3 those that were assigned to
postverbal indefinites.
49%
50%
% of answers
40%
30%
36%
35%
23%
21%
17%
20%
8%
10%
Control
Experimental
11%
0%
New
Wrong
Old
Both
Reference
Figure 2. PVT: Interpretation of preverbal subjects (Greek)
Figure 2 shows significant differences between the control and experimental
groups in how often they interpreted a preverbal subject as “new” (f=7.043,
sig=.012), and how often they picked both the old and the new interpretation
(f=8.083, sig=.007).
In Figure 3 below, which shows the interpretations given to postverbal
indefinite subjects, the difference between the groups in the frequency of an
“old” interpretation is significant (f=6.949, sig=.012).
Figure 3. PVT: Interpretation of postverbal subjects (Greek)
4.3 Picture verification task: interpretation of null and overt pronouns
(Italian)
Figures 4–7 show how the two Italian groups interpreted null and overt
subject pronouns in situations of possible forward and backward anaphora.
Figure 4 shows the interpretations given for a null pronoun in a situation of
possible forward anaphora. That is, a sentence of this type is (8):
(8)
Quando pro attraversa la strada, l’anziana signora saluta la ragazza.
when
crosses
the street the old woman greets the girl
`When (she) is crossing the street, the old woman greets the girl.’
% of answers
The three referents that could be picked for the null pronominal in the initial
adverbial clause were the subject of the main clause (the old woman), the object
(the girl) or some other person not mentioned in the sentence.
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
85% 87%
Control
Experimental
4%
5%
New
11%
8%
Complement
Subject
Referent chosen
Figure 4. PVT: Forward anaphora with null pronoun (Italian)
There was no significant difference here between the behavior of the control and
experimental groups with respect to the referent of the null pronoun in this
condition (Main effect for Group: F(1,38)=1; p=.324; Interaction
Referent*Group: F(2,76)=.373; p=.690).
Figure 5 shows the results for the same type of sentence but with an overt
pronoun (lei ‘she’) rather than the null pronominal:
(9)
Quando lei attraversa la strada, l’anziana signora saluta la ragazza.
when she crosses the street the old woman greets the girl
`When she is crossing the street, the old woman greets the girl.’
Figure 5. PVT: Forward anaphora with overt pronoun (Italian)
In this condition there is a significant difference between the behavior of the
control and experimental groups (Referent*Group: F(2,76)=3.877; p=.025).
Figure 6 shows the referents attributed to a null pronominal subject in a
condition of backward anaphora, as in (10).
% of answers
(10)
L’anziana signora saluta la ragazza quando pro attraversa la strada
the old woman greets the girl
when
crosses the street
‘The old woman greets the girl when she is crossing the street.’
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
70%
51%
44%
Control
Experimental
25%
5%
5%
New
Complement
Subject
Referent chosen
Figure 6. PVT: Backward anaphora with null pronoun (Italian)
There is a significant interaction between Referent and Group here (F(2,
76)=8.433; p=.003).
Finally, Figure 7 gives the figures for the referents associated with an overt
pronominal in the same condition of backward anaphora, as in (11)
(11)
L’anziana signora saluta la ragazza quando lei attraversa la strada.
the old
woman greets the girl when she crosses the street
‘The old woman greets the girl when she is crossing the street.’
Figure 7. PVT: Backward anaphora with overt pronoun (Italian)
Here again there is a significant interaction between Referent and Group
(F(2,76)=3.404; p=.049)
5.
Discussion and Conclusions
First, let us consider the production task reported here (the “Headlines”
task). As shown in Figure 1, the Greek experimental group prefers the preverbal
subject position for both definite and indefinite subjects to a greater extent than
the controls. The difference between the experimental and the control groups in
the production of preverbal definite subjects is significant in this task. This
result supports the original hypothesis, i.e. that preverbal subjects will be
preferred by the experimental group over postverbal ones.
A definiteness effect is also found in the within-groups analysis in the data
from the experimental group. Definites are preferred in preverbal position and
indefinites in postverbal position. This is probably associated with the default
interpretation of a definite subject as a topic, assumed to be a universal
(semantic/pragmatic) condition, hence not affected by syntactic attrition. Given
that Greek does not use definiteness to determine the distribution of preverbal
and postverbal subjects, the performance of the experimental group indicates
that, in language attrition, semantic (interpretable) features, like definiteness,
may be imposed on the affected grammar to minimize unconstrained optionality.
The controls, on the other hand, show a syntactic preference for a
postverbal subject which overrides the definiteness/topichood condition that
governs the behavior of the experimental group.
With respect to the data concerning the interpretation of indefinite preverbal
and postverbal subjects (the “fake fillers’” in the Picture Verification
comprehension task), the Greek experimental group showed an increased lack of
preference (choosing “both” old and new reference) in their interpretation of
both preverbal and postverbal indefinite subjects compared to the controls
(Figures 2 & 3). With respect to preverbal indefinite subjects both groups show
a similar preference for the “old” referent interpretation (45% vs 47%);
postverbal subjects are sometimes interpreted as “new” and sometimes as “old”
by the controls whereas the experimental group shows ambiguity in half of the
cases. We assume that there is a difference between choosing either “old” or
“new” on the one hand, and choosing “both” options on the other. Presumably
for both groups the postverbal position is not associated with a particular
interpretation, as pointed out in section 2.2.. Thus, we assume that the control
group decides on the “new” or “old” interpretation on the basis of contextual or
lexical preferences, given that the syntax does not impose interpretative
restrictions on postverbal subjects. If this is the case, then the fact that the
experimental group shows increased ambiguity possibly indicates indeterminacy
in their performance.
Turning then to the data from the Italian subjects in the comprehension task
(Picture Verification): in sentences of forward anaphora with a null subject in
the subordinate clause such as (8), the preferred choice of referent is the Subject
for both groups (85% vs. 87%), as shown in Figure 4. This result confirms our
original hypothesis: the interpretation of null subjects is not affected by attrition
since the parametric choice determining the availability of null subjects in the
language remains the same.
In contrast, Figure 5 shows that the control group differs significantly from
the experimental group when the subject of the subordinate clause is an overt
pronoun, as in (9). The controls strongly prefer a “new” referent for the
interpretation of the subject pronoun whereas the experimental group do not
show any strong preference for any of the choices (that is, it can refer to some
person not otherwise mentioned, or to the subject or complement of the matrix
clause), indicating multiple ambiguity in the interpretation of the overt subject
pronoun. This result also confirms our original prediction, i.e. that the
interpretation of overt pronouns will show attrition effects.
Figures 6 and 7 present results from sentences with backward anaphora
such as (10) and (11). In the null pronoun case (see Figure 6), the controls seem
to allow for either the matrix subject or the complement as a possible referent
whereas the experimental group strongly prefer the subject interpretation (51%
in the controls’ responses vs. 70% in the experimental group’s responses): this is
an unexpected result. We hypothesize that the experimental group may treat the
subordinate clause as a non-finite one (as in the English sentence The old
woman greets the girl when crossing the street) in which case the matrix subject
is necessarily the controller of the null subject in the subordinate clause.
When the embedded subject is an overt pronoun (see Figure 7), the two
groups show a similar strong preference for the matrix complement to be the
referent of the pronoun (82% and 72% respectively). Crucially, however, the
difference in the choice of the matrix subject as a possible referent is significant
between the control and the experimental groups (8% vs. 21% respectively).
This result is consistent with our original prediction that the interpretation of
overt pronouns will show attrition effects: the experimental group are showing a
significantly greater tendency to allow an overt pronoun to be interpreted as a
continued topic.
Overall, the results presented here support our initial hypothesis concerning
the nature of syntactic attrition. As indicated, much further data still remains to
be explored in our continuing work on this topic.
References
Belletti, Adriana. 1988. The Case of Unaccusatives. Linguistic Inquiry 19, 1-34.
Belletti, Adriana. 2001. ‘Inversion’ as focalisation. In A. Hulk and J-Y Pollock (eds)
Subject Inversion in Romance and the Theory of Universal Grammar. Oxford
University Press, 60-90
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Pinto, Manuela. 1997. Licensing and interpretation of inverted subjects in Italian. PhD
dissertation, University of Utrecht.
Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris.
Rizzi, Luigi. 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17,
501–555.
Roussou, Anna and Ianthi Tsimpli 2002. Postverbal subjects and microparametric
variation. Paper presented at GLOW, Amsterdam.
Sorace, Antonella. 2000. Differential effects of attrition in the L1 syntax of L2 nearnative speakers. In BUCLD Proceedings 24, Somerville, Mass.: Cascadilla Press.
Tsimpli, Ianthi. 1997. Resumptive Strategies and Second Language Acquisition: a
Minimalist Account. In BUCLD Proceedings 21, 639-655 . Somerville, Mass:
Cascadilla Press.
Tsimpli, Ianthi. 2001. LF-Interpretability and Language Development: A Study of Verbal
and Nominal Features in Greek Normally Developing and SLI Children. Brain
and Language 77, 432-448.
White, Lydia and Fred Genesee. 1996. How native is near-native? The issue of ultimate
attainment in adult second language acquisition. Second Language Research 12,
233–265