Content uploaded by Anna Gavarró
Author content
All content in this area was uploaded by Anna Gavarró on Mar 04, 2018
Content may be subject to copyright.
Language typology, and the acquisition of bare noun/DP contrasts
1. Introduction
If compositionality (the assumption that the meaning of an utterance is determined by the
meaning of its parts, and the way they combine) is an inherent property of the human language,
we expect children to attribute meaning to even small differences in sentence form. The input
contains minimal contrasts as in (1) whose interpretation may not be easily distinguishable in the
experience. How do children acquire these minimal distinctions?
(1) a. Someone is wearing gloves. b. Someone is wearing the gloves.
Formally, these cases differ in the presence of a determiner against a bare form. Referentially,
(1a) includes (1b), so for every context where (1b) is true (1a) is also true. The definite DP has a
distinct signature in the context, in terms of frequency at second mention of a referent (Sneed
2008). The bare noun is less specified, and thus less falsifiable. We propose to explore the
acquisition of nominal mapping in terms of the semantic competition between bare nouns and
DPs, in a comparative study of two language: Catalan, where these nominal contrasts occur in
restricted positions, and English, where the contrast is instantiated in both subject and object
position (English). To this purpose, we replicated, in English, Gavarró, Pérez-Leroux and
Roeper’s (2006) study on the acquisition of bare/DP objects in Catalan.
2. DP/bare noun contrasts
The now familiar typology in Chierchia (1998) states languages vary according to how nominal
categories are mapped into semantic types. In his Nominal Mapping Parameter (NMP), nouns
are associated with two independent features that determine their denotation. The feature
[+arg(ument)] indicates N is kind-denoting and has a mass syntax; the feature [+pred(icate)]
determines that a N is predicate-denoting and has a countable syntax. In Chinese, and other
classifier languages, Ns are mapped as [+arg, –pred] and thus can be arguments in their own
right. In Romance, Ns are mapped as [–arg, +pred] and need a determiner to function as
arguments. In Germanic languages, Ns are mapped as [+arg, +pred]: when N are [+pred] they
require a D like Romance count nouns; when [+arg] they act like mass nouns. In Romanian,
Catalan, Spanish and Italian, bare objects in the immediate domain of a P or V head. This is
productive for mass and plural count nouns, and lexically restricted for bare singular count nouns
(Bosque 1996, Dobrovie-Sorin, Bleam & Espinal 2006). Chierchia’s approach analyzes the bare
objects in (2) differently despite their comparable distribution and interpretation.
(2) a. Children drink milk. b. Els nens beuen llet.
The-pl children drink milk
Milk in (2a) is treated as a bare noun, while llet in (2b) is a DP with a null determiner head.
3. Acquisition of bare/DP contrasts.
Research on the acquisition of DPs has identified typological variation in the timing on the
acquisition of determiners. Lleó and Demuth (1999) showed that young Spanish children
achieved adult levels of article production earlier than German-speaking children, and proposed
that prosodic differences across languages determine children’s rates of determiner insertion.
Guasti, Gavarró, de Lange & Caprin (2008) examined determiner production in Catalan, Italian
and Dutch children under the age of 3, divided in groups according to utterance complexity
rather than age. They found no difference in rates of articles or protoarticles used across
languages, but a clear contrast in the pace of development. At the initial stage, Catalan and
Italian children had article omissions in roughly half of the NPs produced, and these dropped
substantially under 16%, by the second stage. In contrast, Dutch started with higher omission
rates (92%) in the first stage and retained substantial omission rates in the later stage (39%).
Guasti and colleagues note that Dutch is statistically different from Catalan and Italian, and
conclude that differences in acquisition rates for the various language types exist beyond the
effect of prosodic factors and should be attributed to setting of the NMP. Bilingual children
learning Germanic and Romance languages exhibit comparable asymmetries (Kupisch 2006).
Less is known about children’s interpretation of bare/determined NP contrasts. Several
studies in English examine the role of language in children’s development of generic concepts.
By comparing bare plural generic subjects, and determined (non-generic) subjects the literature
contributes indirectly to our understanding of morphosyntactic development. Gelman and Raman
(2003) tested whether children aged 2-4 differentiated bare and definite plurals (Do birds/the
birds fly?) when presented with a representation of atypical birds (e.g., penguins). The results
consistently demonstrate that, from the outset, children are sensitive to the linguistic contrast
between generic and nongeneric subjects. Additional support for children’s sensitivity to
morphosyntactic cues has been elicited through comparisons with plural demonstratives (pagons
vs. those pagons (Chambers, Graham & Turner 2008). In Pérez-Leroux, Munn, Schmitt and
DeIrish (2004), children mastered the contrast between bare and definite plurals, but made errors
with definite plurals unlike children in Gelman and Raman who made errors with the bare plurals
instead. Overall, the studies demonstrate sensitivity to the determiner in subject position in
English. Further evidence for English comes from Burns and Soja (2000) and Pérez-Leroux and
Roeper (1999) who found discrimination for bare nominal idioms (go home/go to church).
Gavarró, Pérez-Leroux and Roeper (2006) directly examined the interpretation of
bare/DP objects. They tested 3 to 5-year-olds’s comprehension of bare and full DPs in object
position in Catalan (the girl needs shoes/the shoes). They found an initial stage of non-
discrimination of the contrast follow by development: at the age of 4 the effect of the determiner
absence was significant; and discrimination levels continued to increase by 5. Children generally
preferred the specific reading of DPs and bare nouns. This was taken to suggest that children
prefer inclusive readings to bare nominals (i.e., if someone needs specific shoes, he or she needs
shoes in general). Finally, comparing to data in Guasti et al.’s (2008) suggests that children may
acquire the distribution of bare noun before the interpretation of their semantic contrasts.
4. Study
4.1 Questions
The present study examines linguistic variation in the acquisition of bare noun/DP contrasts. As
discussed in Section 2, Chierchia’s theory provides different analyses to bare objects in English
and Catalan. The literature on early article production supports the derived prediction that there
will be different patterns of development, while leaving open the question of interpretation.
Do we expect variation in semantic acquisition as well? If children treat the distribution of NP
types locally (V +NP vs. V+DP), there should be no differences between the two languages since
the meanings’ contrasts are in fact fully comparable. Alternatively, we could assume children
learning Catalan are taking the overall semantic distribution of NP types without factoring in
syntactic position. If so, they should be influenced by the extension of definite to mass/generic
interpretation in subject positions, and acquire the semantic contrast in object position later.
Last, if the predictions of the NMP extend to comprehension, we predict children learning
Romance languages discriminate between NP types earlier than children learning Germanic
languages. These contrasting hypotheses are summarized in (5):
i. Local distribution hypothesis. Given similarities of bare/definite object contrasts in
English and Catalan, children will acquire these in comparable timing.
ii. Statistical frequency hypothesis. Since bare/definite contrast appear in subjects and
objects in English; acquisition of the semantic contrast should be earlier in English.
iii. Typological difference hypothesis. If convergence with adult grammar is earlier in
Romance-type languages (as predicted by the NMP), acquisition of the semantic mapping
contrasts should also take place earlier in this language type.
4.2 Methods
As in Gavarró, Pérez-Leroux and Roeper (2006) in English. Children were presented with stories
depicting two characters, one with specific needs that are presented in the story, and one with
generic needs that are visually indicated but not explicitly referenced, as in Figure 1.
There is a little girl who wants to dress up as
Cinderella and needs special shoes for that
(point to little girl, dressed as a princess). Her
older sister (point to taller, barefoot girl)
helped her look for them in the trunk, and finds
them.
Figure 2. Older sister hands little Cinderella the special shoes
The experimenter then presented the child with a question with either a definite DP or a bare
noun as in (3). Question (3a) targets the little sister, who needs the specific shoes that go with the
Cinderella outfit; (3b) refers to an individual needing any shoes: in this case, the barefoot big
sister, who needs shoes but is not interested in the special, smaller pair.
(3) a. Who needs the shoes? b. Who needs shoes?
Children heard four stories, followed either by bare nouns or DP questions, and question type
was counterbalanced across stories. The pictures were left in front the child while s/he was asked
and answered the question. Children’s responses were mostly verbal, but a small proportion of
accepted responses consisted of clearly pointing at one of the two characters in the picture. We
translated Catalan stories from Gavarró et al. into English (needs shoes/needs soap). Items that
did not yield a grammatical translation because they contained a bare singular count noun
(*needs chair/ *needs hat) were replaced by new sets of materials (need pants/need gloves).
4.2 Participants
Thirty-three monolingual English-speaking preschoolers and twelve adult controls were recruited
in Toronto. Children were split into age groups: three-year olds (mean= 3;04, SD=3.6, n=11),
four-year olds (mean= 4;04.9, SD=4.1, n=10) and five-year olds (mean= 5;06.8, SD=4.9, n=12).
Our recruitment targeted children matching in age to participants in Gavarró et al.’s study (mean
3;02.7, SD=4.4, n=11; mean=4;04, SD=3.9, n=11; mean=5;03.3, SD=4.9, n=12). English
participants were slightly older than their Catalan counterparts (by 1.3, .9 and 3.5 months for the
three age groups, respectively), but this difference was not statistically significant (language
F1,61=3.316, p=.074; language by group interaction F2,61=.617, p=.543).
4.3 Results
English-speaking children and adults provided few generic responses to DPs, and more generic
responses to the bare nouns. However, their level of discrimination was not the same. Adults
give over five times as many generic responses to bare nouns, whereas children only have a
small difference between the two syntactic conditions, as shown in Table 1.
Table 1. Proportion of generic responses to bare nouns and DPs by English-speakers
Group
DP
Bare noun
Children
.15
.30
Adults
.08
.54
Since the distribution of the scores was skewed, we employed an arcsine transformation. A
repeated measures analysis of variance on these scores supports the observation that there is
development in the magnitude of the contrast. We observed a highly significant effect of
determiner (F1,43=18.989, p<.000), no main effect of group (F1,43=1.071, p=.306), but a
significant group by determiner interaction (F1,43=5.730, p=.021), indicating changes in the
contrast. To this point these results appear at least partially comparable to the Catalan results in
Gavarró et al. (2006). However, our question is whether rates of development of these contrasts
in comprehension are the same or different across the two languages. To test this, we submitted
individual scores on proportions of generic responses per condition for a comparison across
languages, matching the data available from Gavarró et al. on Catalan to our current data in
English, as shown in Figures 2 and 3.
Figure 2. Proportion of generic responses to
bare nouns in Catalan-speaking Children
Figure 3. Proportion of generic responses to bare
nouns in English-speaking Children
A repeated measures anova on the (arcsine) transformed proportion scores for the two NP
conditions indicates a highly significant effect of NP type (F1,61 = 19.539, p < .000), a significant
effect of language (F1,61 = 5.797, p = .019), a significant NP type by group interaction (F2,61 =
5.429, p = .007). The most relevant finding for our purpose is a three-way interaction of NP type
by language by group (F2,61= 4.804, p=.022), which suggest that the development of the
grammatical contrast is not the same across languages. English-speaking three-year olds, unlike
their Catalan counterparts, show some discrimination at the early phase, but no improvement in
the level of differentiation over the preschool years, with differences between syntactic
conditions changing from a 14% to a 17% difference. In contrast the Catalan children
demonstrate a clear pattern of development starting from no differences in three-year olds, to
30% difference in 4 year-olds and 60% differences in 5 year-olds.
5. Conclusion
Our results suggest that the typological contrast between generalized determiner type (Romance)
and mixed type (Germanic) languages predicts not just the acquisition of determiner use, but
also of the acquisition of the semantic NP/DP contrast. We compared children’s sensitivity to
these contrasts in object position in English and Catalan. Our data indicates that, although the
meanings are the same, children behave differently across these two languages. Our results fail to
support a frequency-based hypothesis based on robustness of cue in the input, given that in
English the NP/DP contrast is present in both subject and object position, where in Catalan the
contrast is neutralized for subjects. Contrary to this prediction, children show strong patterns of
development and semantic attainment in Catalan, but little evidence of development or clear
attainment in English. These findings are only compatible with the parametric proposal, based on
Chierchia’s parameter, that there will be an advantage in the acquisition of nominal mapping in
Romance when compared to Germanic languages.
Classical parameter-setting scenarios under scrutiny in both syntax, learnability theory and
language acquisition. Current theory places variation at the level of features and lexical items, so
that large-scale binary parameters are seen as surface consequences of small scale featural and
lexical facts. In the case of nominal typologies and the nominal mapping parameter, many
factors are not yet well integrated into the theory of semantic variation (Dayal 2004), etc. Our
comparative findings are then specially striking because, unlike production data, they cannot be
explained in terms of either local input at the level of direct objects, nor at global cue strength of
the NP/DP comparison. This latter point supports the view presented in Sneed (2008) that
assigns a key role to clausal architecture in solving mapping problems in the nominal domain.
Sneed’s analysis of the input for generic bare noun subjects shows that the input is robust when
NP distribution is processed in terms of its position in the clause, and that children are highly
sensitive to the structure of predicates (the individual/stage level distinction) in making decisions
about the interpretation of indefinite subjects.
These results provide a new extension to a well-established parametric theory, but they
leave two important questions unanswered. As shown in Gelman and Raman’s results, English-
speaking children master the NP/DP contrast for bare plural generic subjects as soon as they can
complete these experimental tasks. Why should bare objects be different? We propose the
reason lies is the behavior of objects at the semantic interface. Objects can be introduced into a
semantic derivation in two different ways. One is standard argument saturation, where a nominal
expression identifies the argument variable of a predicate. Indefinite objects have one additional
mode of composition, predicate restriction, which leaves the argument unsaturated, but
contribute to event modification. This is the case of incorporation structures (Chung and
Ladusaw 2004). In English, we see this pattern in verbal compounds, which require an additional
non-incorporated object to saturate the predicate valence (e.g., to baby-sit the baby). We also see
other cases where it is clear that the DP contributes to event restriction rather than introduce a
semantic argument in English non-referential definites such as in made the decision. In an
experimental task, preschool aged-children consistently fail to distinguish between tnon-
referential NPs and referential DPs (made the decision vs. liked the decision) (DeVilliers and
Roeper 1995). The same children failed the restrictions on binding and long distance extraction.
treating referential DPs as ambiguous. This observation supports the view that children allow DP
objects to function as predicate restrictors rather as participating in argument saturation. We
interpret our English results to mean that some children are not yet using the structural NP-DP
difference in interpretation. Eventually, they restrict their interpretation of these sentences to one
where a referential argument is introduced, and the theme argument variable becomes saturated.
Our last question pertains to the observed advantage of Catalan children in learning these
facts of nominal interpretation. We would like to propose the following speculation: reliance on
the distribution of bare nouns in predicative contexts may be a relevant element in the input.
Predicate nominals are DPs in English but bare nouns in Catalan (He is a doctor vs. és metge, lit.
‘is doctor’). The source of this parametric difference in the distribution of determiners remains a
puzzle in syntactic theory (Munn and Schmitt 2001), but its relevance to the acquisition problem
is clear. English-speaking children are exposed to predicative, non-referential uses of determined
NPs. Their input is thus ambiguous with respect to the modes of composition of determined
nouns. If this is correct, our results become less surprising, and provide further support for the
generative assumption that what counts as input to the child is not necessarily in the immediate
analogical neighborhood of a structure.
The current study provides support for linguistic approaches to language acquisition,
because it demonstrates that problem of syntactic development is not uniformly formulated as
local learning problems set in terms of general-domain learning theories involving frequency,
cue-strength, contrast, and so forth. Rather, it seems specific to the architecture of the learning
domain. Grammar acquisition appears to involve inferences from grammatical properties that are
unrelated in the surface to the target structure, but linked parametrically.
References
Bosque, Ignacio, 1996. Por qué determinados sustantivos no son sustantivos determinados. In El
Sustantivo sin determinación. La ausencia de determinante en la lengua española, Ed. I.
Bosque, 13–120. Madrid: Visor Libros.
Burns, Tracy and Nancy Soja. 2000. Children's acquisition of NP-type nouns: Evidence for
semantic constraints on productivity. Language and Cognitive Processes 15: 45 – 85.
Chambers, Craig, Susan Graham and Juanita N. Turner. 2008. When hearsay trumps evidence:
How generics guide preschoolers inferences about unfamiliar things. Language and Cognitive
Processes 23: 749-766.
Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6:
339–405.
Chung, Sandra and William Ladusaw. 2003. Restriction and saturation. Cambridge, Mass.: MIT
Press.
Dayal, Veneeta. 2004. Number marking and (in)definiteness in kind terms. Linguistics and
Philosophy 27:393–450.
de Villiers, Jill G and Thomas Roeper. 1995. Barriers, binding and acquisition of the DP/NP
distinction. Language Acquisition 4, 73-104.
Dobrovie-Sorin, Carmen, Sonia Bleam and M. Teresa Espinal. 2006. Bare nouns, number, and
types of incorporation. In Svetlana Vogeleer, Liliane Tasmowski (ed.), Non-definiteness and
plurality, 51-80. Amsterdam, John Benjamins.
Gavarró, Anna, Ana T. Pérez-Leroux and Thomas Roeper. 2006. Definite and bare noun contrasts
in child Catalan. In The Acquisition of Syntax in Romance Languages, ed. Vicens Torrens and
Linda Escobar, 51–68. Amsterdam/Philadelphia: John Benjamins.
Gelman, Susan A., and Lakshi Raman. 2003. Preschool children use linguistic form class and
pragmatic cues to interpret generics. Child Development 24: 308-325.
Guasti, Maria Teresa, Anna Gavarró, Joke De Lange and Claudia Caprin. 2008. Article omission
across child languages. Language Acquisition 15(2): 89-119 .
Kupisch, Tanja. 2006. The acquisition of determiners in bilingual German-Italian and German-
French children. München: Lincom Europa.
Lleó, Conxita and Katherine Demuth. 1999. Prosodic constraints on the emergence of
grammatical morphemes: crosslinguistic evidence from Germanic and Romance languages. In
Proceedings of the 23rd Annual Boston University Conference on Language Development,
Greenhill, A. et al. (eds.), 407–418. Somerville: Cascadilla Press.
Munn, Alan and Cristina Schmitt. 2001. Bare nominals and the morphosyntax of number. In
Current Issues in Romance Linguistics: Selected Papers from the XXIXth Linguistic
Symposium on Romance Languages (LSRL), D. Cresti, T. Satterfield and C. Tortora (eds.),
217-231. Amsterdam: John Benjamins.
Pérez-Leroux, Ana T., Alan Munn, Cristina Schmitt and Michelle DeIrish. 2004. Learning
definite determiners: genericity and definiteness in English and Spanish. Proceedings
supplement to the 28th Boston University Conference on Language Development.
http://www.bu.edu/linguistics/BUCLD/supp.html
Pérez-Leroux, Ana T., and Thomas Roeper. 1999. Scope and the structure of bare nominals:
evidence from child language. Linguistics 37: 927-960.
Sneed, Elisa. 2008. Input in the acquisition of genericity. Doctoral dissertation, Northwestern
University.
