LINGUIST List 15.2452

Fri Sep 3 2004

Review: Language Acquisition: Santelmann et al. (2004)

Editor for this issue: Naomi Ogasawara <naomilinguistlist.org>


What follows is a review or discussion note contributed to our Book Discussion Forum. We expect discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for review." Then contact Sheila Dooley Collberg at collberglinguistlist.org.

Directory

  1. Alicia Munoz Sanchez, Annual Review of Language Acquisition

Message 1: Annual Review of Language Acquisition

Date: Wed, 1 Sep 2004 19:56:08 -0400 (EDT)
From: Alicia Munoz Sanchez <amunozcsusm.edu>
Subject: Annual Review of Language Acquisition

EDITORS: Santelmann, Lynn M; Verrips, Maaike; Wijnen, Frank; Levelt, 
Claartje
TITLE: Annual Review of Language Acquisition
SUBTITLE: Volume 3 (2003)
PUBLISHER: John Benjamins
YEAR: 2004
Announced at http://linguistlist.org/issues/15/15-1639.html


Alicia Mu�oz S�nchez, California State University San Marcos

The third volume in the Annual Review of Language Acquisition series
presents four articles that reflect the work of dissertations finished
in 2002 and 2003. Their common theme notwithstanding, these articles
also illustrate different experimental and theoretical approaches used
in phonology, morphosyntax and syntax for the investigation of
language acquisition. As a result, this collection provides a useful
survey of numerous current issues in the field.

1. Optimality Theory and phonological acquisition. Paul Boersma and
Claartje Levelt

This paper presents an overview of Optimality Theory (OT) and its
application in the field of language acquisition. The paper also
discusses computational models that have been developed to test the
learnability of this theory.

In the first section of the paper the authors introduce the
fundamentals of Optimality Theory and describe the ranking mechanisms
that OT uses to explain the acquisition of grammar. It is also
explained how underlying representations are selected.

The second section provides examples of recent acquisition research on
syllable and prosodic structure to illustrate the acquisition of adult
grammars. The article shows that the development of final grammars
occurs through reranking of the constraints in developmental grammars.
In addition, the authors illustrate how the theory addresses aspects
of the grammar which are specific to child phonology.

The learnability issue is discussed at length in the third section, as
the authors provide a review of the different learning algorithms that
have been developed for OT. Two of the main problems with all these
computational models are that they give access to fully structured
surface forms and that they have not yet incorporated all the features
of the language learning process. Some models have have taken steps to
remedy these deficiencies: e.g. the Error Driven Constraint Demotion
model (Tesar & Smolensky 1998, Tesar 1995) does not contain coded
underlying forms; and the Gradual Learning Algorithm (Boersma 1997,
Boersma and Hayes 2001) shows variation in the choice of surface
forms. These models are valuable because they show that certain
elements of the theory are learnable.

However, there are still a number of prerequisites for a good grammar
model which are not met by any model. For example, no model can
account for the processes of both production and
comprehension. Similarly, we lack models which can distinguish the
three types of phonological forms commonly assumed in language
acquisition work: i.e. the underlying, surface and overt/perceived by
the child forms. Consequently, issues such as the acquisition of
underlying forms, and the mapping between surface and underlying forms
remain unexplored. Other tasks awaiting models are learning from
positive evidence (overt information), the impositions of the grammar
on forms (covert information) and finally the emergence of
language-specific constraints.

The last section discusses future directions in OT. A survey of the
researchers in the field reveals two main topics that need to be
addressed in future studies: first, the relationship between
computational models of learnability and empirical studies of
acquisition, and, second, the relationship between perception and
production -- i.e. are they supported by separate grammars or a single
one?

Evaluation

This article is extremely valuable for all linguists and cognitive
scientists who are interested in OT and how this framework is applied
to language acquisition data. The article combines the explanation of
the inner workings of OT with examples of recent research, while
answering many of the questions that should be asked about the theory
in an insightful manner. The detailed discussion of the computational
models that have been developed for this framework is another
strength. As the authors point out, the models still do not
incorporate all the aspects of realistic language learning, but they
do show that certain aspects of the theory are learnable. The authors'
interest in pushing the field towards models that incorporate the
elements of a 'realistic' language learning is commendable, as most
research in the past has overlooked the relationship between
perception and production and their implications for a theory of
language learning. In sum, this article provides an honest appraisal
of OT, and is a must read for anyone interested in language
acquisition research.

2. Early foreign language education and metalinguistic development: A
study of monolingual, bilingual and trilingual children on noun
definition tasks. Krassimira Dimitrova Charkova

The second article in the volume presents the result from a study on
how multilingualism affects the ability to define words. Data was
collected from Bulgarian children from three schools with parents of
similar educational background. Bulgarian monolingual, bilingual
(Bulgarian, English) and trilingual (Bulgarian, English, Russian)
children were tested on their ability to define words in their first,
second and third language. The children were between the ages of 10
and 11, which is when they typically start developing supeordinate
definitions of the type 'X is Y that Z'. The aims of the study were
twofold. First, explore whether bilingual and multilingual subjects
were better than monolingual children at defining nouns (concrete and
abstract) in their first language (Bulgarian). Second, the study
investigated whether the ability to define nouns in a second (English)
or third language (Russian) was correlated with the typological
closeness of these languages and L1.

The results show that early foreign language education improves
metalinguistic awareness in L1, since bilingual and multilingual
children are better at definining words. In addition, the study shows
that definition abilities in a third language improve if the first
language is typologically related, as is the case with Russian and
Bulgarian -- both Slavonic languages. In this case, the level of
competence in the third language is not a good predictor of definition
abilities.

No differences were found between the groups in their ability to
define abstract nouns, but very few such nouns were actually presented
to the children. As the author admits, 4 (out of 32) nouns may not
have provided enough data.

Evaluation

The main merit of this study is that it provides evidence of the
cognitive benefits of multilingualism using new data collected from
trilingual, bilingual and monolingual children. No previous studies
have looked at the effects of either a third language or typologically
related languages on the ability to define words. The author developed
a sound methodology suitable for the categorization and scoring of
definitions that would be beneficial to others trying to do research
on the acquisition of word definitions.

The author speculates that one of the possible causes for the
advantage of bilingual and trilingual children in defining words is
the fact that these children have more exposure to definitions. It
would be interesting to see whether the study can be replicated with a
bilingual and trilingual population that has learnt their second and
third languages at home. It may be the case that the advantage is
actually a combination of definition exposure as well as a more
natural context- driven demand to categorize words.

As mentioned earlier, the study has also found that multilingual
subjects are better at selecting semantically appropriate words in
their definitions. Even though it is not presented in the results
section with sufficient detail, I think this result is worth
noting. It suggests that all children have acquired the syntactic
structures appropriate for creating definitions, but the monolingual
subjects have more trouble choosing the semantically appropriate
terms.

Finally, this study has shown that the cognitive advantage in a third
language may be more related to typological similarity to the first
language than overall language proficiency. The correlation between
definition abilities and L2 competence was much higher in the
bilingual group (Bulgarian, English). This suggests that vocabulary
size is a better predictor than language competence of L2 and L3
definition abilities (Carlisle et al., 1999). If this is indeed the
case, it would be appropriate to test subjects in future studies for
both vocabulary size as well as competence.

3. Language convergence and bilingual acquisition: The case of
conditional constructions. Ee San Chen

This paper examines the acquisition of conditional constructions in
Singapore Colloquial English (SCE) and Singapore Colloquial Mandarin
(SCM) by bilingual children between the ages of 2;0 to 6;0. These
varieties of English and Mandarin exhibit similar patterns in the
structures they use for conditional constructions -- they range from
head marked (HM) to dependent marked (DM), with some having absent
marking (AM) or double marking (DdM). The similarities in the patterns
of SCE and SCM show a convergence of the conditional constructions in
this bilingual setting. In order to see whether such convergence is
also taking place in child language (both at the individual and
societal levels), the researcher conducted two different experiments
that involved imitation tasks. The first experiment tested convergence
at the societal level by presenting children sentences of all possible
types in SCE and SCM, and asking the children to replicate what they
heard to a toy puppet. The second experiment presented the children
with the most common constructions in SCE and SCM, i.e. DM 'if' and HM
'jiu' respectively, in a story setting and asked the children to tell
the experimenter what they just heard.

The results from the first experiment show that children experience
the least difficulty imitating dependent marked constructions with
'afterward' and 'dengyixia', followed by structures with absent
marking. Head marked structures showed the lowest accuracy rates with
the exception of SCM head marked 'dengyixia'. Statistically
significant differences were only found in the performance of the
subjects for HM constructions with 'dengyixia' and 'wait'. In general,
lower accuracy rates were found for the English conditional
constructions, so there appears to be a preference for the Chinese
structures.

In examining the SCE data, the author found that the head marker
'then' is preferred in substitutions of AM and DM. Substitution
preferences in SCE are for AM followed by HM constructions. The author
argues that this is evidence for the convergence of SCE with SCM due
to the strong influence of the Chinese substrate. He also shows,
though, that English monolingual children have a preference for AM
constructions. In addition, the high number of accurate responses
(73%) for the innovative SCE construction suggests that there is a
strong influence of SCM on SCE. The influence of SCE on SCM seems to
be minimal and is restricted to DM 'ruguo' conditionals.

The second experiment elicited data from 8 children between the ages
of 4;7 and 6;2. Each subject produced between 5 and 6 conditional
sentences. The results show that individual children have particular
preferences in the way they express conditionals in each
language. This in turn suggests that there need not be convergence at
the individual level.

Evaluation

This paper has some value for those interested in bilingual language
acquisition in contact situations. The paper presents very interesting
data on the acquisition of conditional constructions by bilingual
Singapore Colloquial English and Chinese children. However the
methodology used in this study makes it difficult to interpret the
results because the experimental setup biased the children to
replicate what the researcher produced. As a result, some of the
utterances by the children may be just repetitions, and thus they
would not represent the acquisition of the particular structures or
even a preference for them. Looking at the overall accuracy rate in
this case may also be misleading, again because we do not know what
percentage represents simple imitation and what percentage shows
preference for a particular structure.

It would have been interesting to look at the individual data for each
subject to detect those children that are better imitators than
others. In addition, the data on the failure by the children to
imitate certain conditional sentences could have been more informative
than the repetition accuracy rate. For example, HM 'then' conditionals
tend to be substituted by AM or other HM sentences, which shows that
children are more at ease with these types of conditionals. However, I
would have liked to see more of the children's choices for the DM
constructions, because the experimental setup may have been prompting
the children to substitute the DM structure with another DM and
likewise a HM structure for another HM.

The range in the ages of subjects also makes difficult to interpret
the results altogether, given that some of the conditionals, specially
those with AM, are usually the preferred pattern in the initial stages
of acquisition. This makes it hard to prove that the preference for AM
constructions is mostly due to the influence of Chinese.

In addition, it is not clear from the study what the adult preferences
are in using the conditional constructions. No frequency data is
presented for the conditional sentences in each language. The author
mentions that the Chinese influence is very pronounced, but given the
variety of Chinese constructions it would be useful to know which
constructions are more frequent and which are marginal.

The results of the second experiment are easier to interpret, because
the children were only presented one type of construction. So, their
substitution patterns do show their personal inclination for certain
constructions. Sadly, the number of tokens that each subject produced
was insufficient to carry out a statistical analysis.

The fact that children prefer the AM and HM constructions which are
more common in Chinese suggests language convergence is taking place
at the societal level. In future studies of this new data, it will be
interesting to look at the acquisition patterns for each language in
different groups of subjects (e.g. monolinguals, SCE dominant, SCM
dominant, true bilinguals) at different ages to see which
constructions are acquired similarly and which differently.

4. The acquisition of inflectional prefixes in Nairobi Swahili. Kamil
Ud Deen

The final paper in the volume describes a study of the acquisition of
verbal morphology in Nairobi Swahili by children aged 1;8 to 3;0.
Nairobi Swahili is an agglutinative language which has verb affixes
for subject agreement, for tense/aspect agreement, for object
agreement and for mood. These affixes tend to be omitted. In this
paper, the author discusses the acquisition of subject agreement
markers as well as tense markers and some mood markers. He concludes
that the Agreement-Tense Omission Model described by Sch�tze & Wexler
(1996) accounts for the omission facts in acquisition.

The paper is divided in six sections. The first three sections present
data on Swahili morphosyntax with particular attention to the omission
of the subject agreement verb suffix in certain discourse
contexts. The methodology used for the collecting and coding of the
data is described, and a summary is given of the different theories of
the acquisition of inflection. The next three sections present the
results. The children show very frequent omission of subject
agreement, similar to the level observed in adults. Unlike adults,
however, children also omit tense very frequently. The author
evaluates the predictions of the various theories against the data and
concludes that the ATOM model of Sch�tze and Wexler is most compatible
with the Swahili facts. However, he also points out that this model
does not have mechanisms to account for some developmental data in the
Swahili, such as the early acquisition of tense before subject
agreement.

Evaluation

This study is a model for those who aim to improve language
acquisition theories based on partial language data. The data
collected from four children between the ages of 1;8 and 3;0 is quite
valuable since there is only limited acquisition data available for
Nairobi Swahili.

The main merit of this paper is that is shows that some current
morphological theories cannot account for language acquisition data
from Nairobi Swahili without further modification or revisions. In
addition, the paper reminds us of some important desiderata for
morphological theories, such as the need to include mechanisms that
explain developmental grammars and the need for less restrictive
mechanisms to account for different language data.

OVERALL EVALUATION

Overall, this volume provides four well written papers with a healthy
mix of experimental data and theoretical work for all linguists
interested in the field of language acquisition. They share a big
emphasis on the collection of original language acquisition data, with
new facts being presented in three of the papers on typologically
diverse languages: Swahili, Singapore English and Chinese, Bulgarian,
English and Russian. Two of the papers also stress an important theme:
i.e. theories need to incorporate realistic language learning
components. The experimental papers also provide good examples of how
to overcome the challenges of collecting specific language data from
children.

REFERENCES

Boersma, P. (1997). 'How we learn variation, optionality, and
probability' Proceedings of the Institute of Phonetic Sciences, 21,
43- 58. University of Amsterdam

Boersma, P. & Hayes, B. (2001). 'Empirical tests of the Gradual
Learning Algorithm' Linguistic Inquiry, 32, 45-86.

Tesar, B. (1995). 'Computational Optimality Theory' PhD Dissertation,
University of Colorado. [ROA 90]

Tesar, B. & Smolensky, P. (1998). 'Learnability in Optimality Theory'.
Linguistic Inquiry, 29, 229-268.

Carlisle, J. F., Beeman, M., Davis, L.H. and Spharim, G. (1999).
'Relationship of metalinguistic capabilities and reading achievement
for children who are becoming bilingual' Applied Psycholinguistics, 20
459-478.

Sch�tze, C. and Wexler, K. (1996). 'Subject case licensing and English
Root infinitives' In A. Stringfellow, D. Cahana-Amitay, E. Hughes and
A. Zukowski (Eds), Proceedings of the 20th Annual Boston University
Conference on Language Development (pp. 670-681). Somerville, MA:
Cascadilla Presss

ABOUT THE REVIEWER

Alicia Mu�oz S�nchez is an Assistant Professor at California State
University San Marcos, where she teaches Spanish Language and
Linguistics. Her research interests are in the areas of language
acquisition, phonetics and phonology.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue