33
A Developmental Model of Syntax Acquisition in the Construction Grammar
Framework with Cross-Linguistic Validation in English and Japanese
Peter Ford Dominey
Sequential Cognition and Language Group
Institut des Sciences Cognitives, CNRS
69675 Bron CEDES, France
dominey@isc.cnrs.fr
Toshio Inui
Graduate School of Informatics,
Kyoto University,
Yoshida-honmachi, Sakyo-ku, 606-8501,
Kyoto, Japan
inui@kyoto-u.ac.jp
Abstract
The current research demonstrates a system
inspired by cognitive neuroscience and
developmental psychology that learns to
construct mappings between the grammatical
structure of sentences and the structure of their
meaning representations. Sentence to meaning
mappings are learned and stored as
grammatical constructions. These are stored
and retrieved from a construction inventory
based on the constellation of closed class
items uniquely identifying each construction.
These learned mappings allow the system to
processes natural language sentences in order
to reconstruct complex internal representations
of the meanings these sentences describe. The
system demonstrates error free performance
and systematic generalization for a rich subset
of English constructions that includes complex
hierarchical grammatical structure, and
generalizes systematically to new sentences of
the learned construction categories. Further
testing demonstrates (1) the capability to
accommodate a significantly extended set of
constructions, and (2) extension to Japanese, a
free word order language that is structurally
quite different from English, thus
demonstrating the extensibility of the structure
mapping model.
1 Introduction
The nativist perspective on the problem of
language acquisition holds that the <sentence,
meaning> data to which the child is exposed is
highly indeterminate, and underspecifies the
mapping to be learned. This  poverty of the
stimulus is a central argument for the existence of
a genetically specified universal grammar, such
that language acquisition consists of configuring
the UG for the appropriate target language
(Chomsky 1995). In this framework, once a given
parameter is set, its use should apply to new
constructions in a generalized, generative manner.
An alternative functionalist perspective holds
that learning plays a much more central role in
language acquisition. The infant develops an
inventory of grammatical constructions as
mappings from form to meaning (Goldberg 1995).
These constructions are initially rather fixed and
specific, and later become generalized into a more
abstract compositional form employed by the adult
(Tomasello 1999, 2003). In this context,
construction of the relation between perceptual and
cognitive representations and grammatical form
plays a central role in learning language (e.g.
Feldman et al. 1990, 1996; Langacker 1991;
Mandler 1999; Talmy 1998).
These issues of learnability and innateness have
provided a rich motivation for simulation studies
that have taken a number of different forms.
Elman (1990) demonstrated that recurrent
networks are sensitive to predictable structure in
grammatical sequences. Subsequent studies of
grammar induction demonstrate how syntactic
structure can be recovered from sentences (e.g.
Stolcke & Omohundro 1994). From the
 grounding of language in meaning perspective
(e.g. Feldman et al. 1990, 1996; Langacker 1991;
Goldberg 1995) Chang & Maia (2001) exploited
the relations between action representation and
simple verb frames in a construction grammar
approach. In effort to consider more complex
grammatical forms, Miikkulainen (1996)
demonstrated a system that learned the mapping
between relative phrase constructions and multiple
event representations, based on the use of a stack
for maintaining state information during the
processing of the next embedded clause in a
recursive manner.
In a more generalized approach, Dominey
(2000) exploited the regularity that sentence to
meaning mapping is encoded in all languages by
word order and grammatical marking (bound or
free) (Bates et al. 1982). That model was based on
34
the functional neurophysiology of cognitive
sequence and language processing and an
associated neural network model that has been
demonstrated to simulate interesting aspects of
infant (Dominey & Ramus 2000) and adult
language processing (Dominey et al. 2003).
2 Structure mapping for language learning
The mapping of sentence form onto meaning
(Goldberg 1995) takes place at two distinct levels
in the current model: Words are associated with
individual components of event descriptions, and
grammatical structure is associated with functional
roles within scene events. The first level has been
addressed by Siskind (1996), Roy & Pentland
(2002) and Steels (2001) and we treat it here in a
relatively simple but effective manner. Our
principle interest lies more in the second level of
mapping between scene and sentence structure.
Equations 1-7 implement the model depicted in
Figure 1, and are derived from a
neurophysiologically motivated model of
sensorimotor sequence learning (Dominey et al.
2003).
2.1 Word Meaning
Equation (1) describes the associative memory,
WordToReferent, that links word vectors in the
OpenClassArray (OCA) with their referent vectors
in the SceneEventArray (SEA)1. In the initial
learning phases there is no influence of syntactic
knowledge and the word-referent associations are
stored in the WordToReferent matrix (Eqn 1) by
associating every word with every referent in the
current scene (a = 1), exploiting the cross-
situational regularity (Siskind 1996) that a given
word will have a higher coincidence with referent
to which it refers than with other referents. This
initial word learning contributes to learning the
mapping between sentence and scene structure
(Eqn. 4, 5 & 6 below). Then, knowledge of the
syntactic structure, encoded in SentenceToScene
can be used to identify the appropriate referent (in
the SEA) for a given word (in the OCA),
corresponding to a zero value of a in Eqn. 1. In
this  syntactic bootstrapping for the new word
 gugle, for example, syntactic knowledge of
Agent-Event-Object structure of the sentence
 John pushed the gugle can be used to assign
1 In Eqn 1, the index k = 1 to 6, corresponding to the maximum
number of words in the open class array (OCA). Index m = 1 to 6,
corresponding to the maximum number of elements in the scene event
array (SEA). Indices i and j = 1 to 25, corresponding to the word and
scene item vector sizes, respectively.
 gugle to the object of push.
WordToReferent(i,j) = WordToReferent(i,j) +
OCA(k,i) * SEA(m,j) *
max(a, SentenceToScene(m,k)) (1)
2.2 Open vs Closed Class Word Categories
Our approach is based on the cross-linguistic
observation that open class words (e.g. nouns,
verbs, adjectives and adverbs) are assigned to their
thematic roles based on word order and/or
grammatical function words or morphemes (Bates
et al. 1982). Newborn infants are sensitive to the
perceptual properties that distinguish these two
categories (Shi et al. 1999), and in adults, these
categories are processed by dissociable
neurophysiological systems (Brown et al. 1999).
Similarly, artificial neural networks can also learn
to make this function/content distinction (Morgan
et al. 1996). Thus, for the speech input that is
provided to the learning model open and closed
class words are directed to separate processing
streams that preserve their order and identity, as
indicated in Figure 2.
Figure 1. Structure-Mapping Architecture. 1. Lexical categorization.
2. Open class words in Open Class Array are translated to Predicted
Referents in the PRA via the WordtoReferent mapping. 3. PRA
elements are mapped onto their roles in the SceneEventArray by the
SentenceToScene mapping, specific to each sentence type. 4. This
mapping is retrieved from Construction Inventory, via the
ConstructionIndex that encodes the closed class words that
characterize each grammatical construction type.
2.3 Mapping Sentence to Meaning
Meanings are encoded in an event predicate,
argument representation corresponding to the
SceneEventArray in Figure 1 (e.g. push(Block,
triangle) for  The triangle pushed the block ).
There, the sentence to meaning mapping can be
35
characterized in the following successive steps.
First, words in the Open Class Array are decoded
into their corresponding scene referents (via the
WordToReferent mapping) to yield the Predicted
Referents Array that contains the translated words
while preserving their original order from the OCA
(Eqn 2) 2.
n
i 1
PRA(k,j) = OCA(k,i) * WordToReferent(i,j)
=
 (2)
Next, each sentence type will correspond to a
specific form to meaning mapping between the
PRA and the SEA. encoded in the
SentenceToScene array. The problem will be to
retrieve for each sentence type, the appropriate
corresponding SentenceToScene mapping. To
solve this problem, we recall that each sentence
type will have a unique constellation of closed
class words and/or bound morphemes (Bates et al.
1982) that can be coded in a ConstructionIndex
(Eqn.3) that forms a unique identifier for each
sentence type.
The ConstructionIndex is a 25 element vector.
Each function word is encoded as a single bit in a
25 element FunctionWord vector. When a
function word is encountered during sentence
processing, the current contents of
ConstructionIndex are shifted (with wrap-around)
by n + m bits where n corresponds to the bit that is
on in the FunctionWord, and m corresponds to the
number of open class words that have been
encountered since the previous function word (or
the beginning of the sentence). Finally, a vector
addition is performed on this result and the
FunctionWord vector. Thus, the appropriate
SentenceToScene mapping for each sentence type
can be indexed in ConstructionInventory by its
corresponding ConstructionIndex.
ConstructionIndex = fcircularShift(ConstructionIndex,
FunctionWord) (3)
The link between the ConstructionIndex and the
corresponding SentenceToScene mapping is
established as follows. As each new sentence is
processed, we first reconstruct the specific
SentenceToScene mapping for that sentence (Eqn
4)3, by mapping words to referents (in PRA) and
2 Index k = 1 to 6, corresponding to the maximum number of scene
items in the predicted references array (PRA). Indices i and j = 1 to
25, corresponding to the word and scene item vector sizes,
respectively.
3 Index m = 1 to 6, corresponding to the maximum number of
elements in the scene event array (SEA). Index k = 1 to 6,
corresponding to the maximum number of words in the predicted
referents to scene elements (in SEA). The
resulting, SentenceToSceneCurrent encodes the
correspondence between word order (that is
preserved in the PRA Eqn 2) and thematic roles in
the SEA. Note that the quality of
SentenceToSceneCurrent will depend on the
quality of acquired word meanings in
WordToReferent. Thus, syntactic learning
requires a minimum baseline of semantic
knowledge.
n
i=1
SentenceToSceneCurrent(m,k) =
PRA(k,i)*SEA(m,i) (4)
Given the SentenceToSceneCurrent mapping
for the current sentence, we can now associate it in
the ConstructionInventory with the corresponding
function word configuration or ConstructionIndex
for that sentence, expressed in (Eqn 5)4.
ConstructionInventory(i,j) = ConstructionInventory(i,j)
+ ConstructionIndex(i)
* SentenceToScene-Current(j) (5)
Finally, once this learning has occurred, for
new sentences we can now extract the
SentenceToScene mapping from the learned
ConstructionInventory by using the
ConstructionIndex as an index into this associative
memory, illustrated in Eqn. 65.
n
i=1
SentenceToScene(i) =
ConstructionInventory(i,j) * ConstructinIndex(j) (6)
To accommodate the dual scenes for complex
events Eqns. 4-7 are instantiated twice each, to
represent the two components of the dual scene. In
the case of simple scenes, the second component of
the dual scene representation is null.
We evaluate performance by using the
WordToReferent and SentenceToScene knowledge
to construct for a given input sentence the
 predicted scene . That is, the model will
references array (PRA). Index i = 1 to 25, corresponding to the word
and scene item vector sizes.
4 Note that we have linearized SentenceToSceneCurrent from 2 to
1 dimensions to make the matrix multiplication more transparent.
Thus index j varies from 1 to 36 corresponding to the 6x6 dimensions
of SentenceToSceneCurrent.
5 Again to simplify the matrix multiplication, SentenceToScene
has been linearized to one dimension, based on the original 6x6
matrix. Thus, index i = 1 to 36, and index j = 1 to 25 corresponding to
the dimension of the ConstructionIndex.
36
construct an internal representation of the scene
that should correspond to the input sentence. This
is achieved by first converting the Open-Class-
Array into its corresponding scene items in the
Predicted-Referents-Array as specified in Eqn. 2.
The referents are then re-ordered into the proper
scene representation via application of the
SentenceToScene transformation as described in
Eqn. 76.
PSA(m,i) = PRA(k,i) * SentenceToScene(m,k) (7)
When learning has proceeded correctly, the
predicted scene array (PSA) contents should match
those of the scene event array (SEA) that is
directly derived from input to the model. We then
quantify performance error in terms of the number
of mismatches between PSA and SEA.
3 Learning Experiments
Three sets of results will be presented. First the
demonstration of the model sentence to meaning
mapping for a reduced set of constructions is
presented as a proof of concept. This will be
followed by a test of generalization to a new
extended set of grammatical constructions.
Finally, in order to validate the cross-linguistic
validity of the underlying principals, the model is
tested with Japanese, a free word-order language
that is qualitatively quite distinct from English.
3.1 Proof of Concept with Two Constructions
3.1.1 Initial Learning of Active Forms for
Simple Event Meanings
The first experiment examined learning with
sentence, meaning pairs with sentences only in the
active voice, corresponding to the grammatical
forms 1 and 2.
1. Active: The block pushed the triangle.
2. Dative: The block gave the triangle to the
moon.
For this experiment, the model was trained on
544 <sentence, meaning> pairs. Again, meaning is
coded in a predicate-argument format, e.g.
push(block, triangle) for sentence 1. During the
first 200 trials (scene/sentence pairs), value a in
Eqn. 1 was 1 and thereafter it was 0. This was
necessary in order to avoid the effect of erroneous
6 In Eqn 7, index i = 1 to 25 corresponding to the size of the scene
and word vectors. Indices m and k = 1 to 6, corresponding to the
dimension of the predicted scene array, and the predicted references
array, respectively.
(random) syntactic knowledge on semantic
learning in the initial learning stages. Evaluation
of the performance of the model after this training
indicated that for all sentences, there was error-free
performance. That is, the PredictedScene
generated from each sentence corresponded to the
actual scene paired with that sentence. An
important test of language learning is the ability to
generalize to new sentences that have not
previously been tested. Generalization in this form
also yielded error free performance. In this
experiment, only 2 grammatical constructions were
learned, and the lexical mapping of words to their
scene referents was learned. Word meaning
provides the basis for extracting more complex
syntactic structure. Thus, these word meanings are
fixed and used for the subsequent experiments.
3.1.2 Passive forms
The second experiment examined learning with
the introduction of passive grammatical forms,
thus employing grammatical forms 1-4.
3. Passive: The triangle was pushed by the block.
4. Dative Passive: The moon was given to the
triangle by the block.
A new set of <sentence, scene> pairs was
generated that employed grammatical
constructions, with two- and three- arguments, and
active and passive grammatical forms for the
narration. Word meanings learned in Experiment 1
were used, so only the structural mapping from
grammatical to scene structure was learned. With
exposure to less than 100 <sentence, scene>, error
free performance was achieved. Note that only the
WordToReferent mappings were retained from
Experiment 1. Thus, the 4 grammatical forms
were learned from the initial naive state. This
means that the ConstructionIndex and
ConstructionInventory mechanism correctly
discriminates and learns the mappings for the
different grammatical constructions. In the
generalization test, the learned values were fixed,
and the model demonstrated error-free
performance on new sentences for all four
grammatical forms that had not been used during
the training.
3.1.3 Relative forms for Complex Events
The complexity of the scenes/meanings and
corresponding grammatical forms in the previous
experiments were quite limited. Here we consider
complex <sentence, scene> mappings that involve
relativised sentences and dual event scenes. A
37
small corpus of complex <sentence, scene> pairs
were generated corresponding to the grammatical
construction types 5-10
5. The block that pushed the triangle touched the
moon.
6. The block pushed the triangle that touched the
moon.
7. The block that pushed the triangle was touched by
the moon.
8. The block pushed the triangle that was touched the
moon.
9. The block that was pushed by the triangle touched
the moon.
10. The block was pushed by the triangle that touched
the moon.
After exposure to less than 100 sentences
generated from these relativised constructions, the
model performed without error for these 6
construction types. In the generalization test, the
learned values were fixed, and the model
demonstrated error-free performance on new
sentences for all six grammatical forms that had
not been used during the training.
3.1.4 Combined Test
The objective of the final experiment was to
verify that the model was capable of learning the
10 grammatical forms together in a single learning
session. Training material from the previous
experiments were employed that exercised the
ensemble of 10 grammatical forms. After
exposure to less than 150 <sentence, scene> pairs,
the model performed without error. Likewise, in
the generalization test the learned values were
fixed, and the model demonstrated error-free
performance on new sentences for all ten
grammatical forms that had not been used during
the training.
This set of experiments in ideal conditions
demonstrates a proof of concept for the system,
though several open questions can be posed based
on these results. First, while the demonstration
with 10 grammatical constructions is interesting,
we can ask if the model will generalize to an
extended set of constructions. Second, we know
that the English language is quite restricted with
respect to its word order, and thus we can ask
whether the theoretical framework of the model
will generalize to free word order languages such
as Japanese. These questions are addressed in the
following three sections.
3.2 Generalization to Extended Construction
Set
As illustrated above the model can accommodate
10 distinct form-meaning mappings or
grammatical constructions, including constructions
involving "dual" events in the meaning
representation that correspond to relative clauses.
Still, this is a relatively limited size for the
construction inventory. The current experiment
demonstrates how the model generalizes to a
number of new and different relative phrases, as
well as additional sentence types including:
conjoined (John took the key and opened the door),
reflexive (The boy said that the dog was chased by
the cat), and reflexive pronoun (The block said that
it pushed the cylinder) sentence types, for a total of
38 distinct abstract grammatical constructions. The
consideration of these sentence types requires us to
address how their meanings are represented.
Conjoined sentences are represented by the two
corresponding events, e.g. took(John, key),
open(John, door) for the conjoined example above.
Reflexives are represented, for example, as
said(boy), chased(cat, dog). This assumes indeed,
for reflexive verbs (e.g. said, saw), that the
meaning representation includes the second event
as an argument to the first. Finally, for the
reflexive pronoun types, in the meaning
representation the pronoun’s referent is explicit, as
in said(block), push(block, cylinder) for "The
block said that it pushed the cylinder."
For this testing, the ConstructionInventory is
implemented as a lookup table in which the
ConstructionIndex is paired with the corresponding
SentenceToScene mapping during a single learning
trial. Based on the tenets of the construction
grammar framework (Goldberg 1995), if a
sentence is encountered that has a form (i.e.
ConstructionIndex) that does not have a
corresponding entry in the ConstructionInventory,
then a new construction is defined. Thus, one
exposure to a sentence of a new construction type
allows the model to generalize to any new sentence
of that type. In this sense, developing the capacity
to handle a simple initial set of constructions leads
to a highly extensible system. Using the training
procedures as described above, with a pre-learned
lexicon (WordToReferent), the model successfully
learned all of the constructions, and demonstrated
generalization to new sentences that it was not
trained on.
That the model can accommodate these 38
different grammatical constructions with no
modifications indicates its capability to generalize.
This translates to a (partial) validation of the
hypothesis that across languages, thematic role
assignment is encoded by a limited set of
38
parameters including word order and grammatical
marking, and that distinct grammatical
constructions will have distinct and identifying
ensembles of these parameters. However, these
results have been obtained with English that is a
relatively fixed word-order language, and a more
rigorous test of this hypothesis would involve
testing with a free word-order language such as
Japanese.
3.3 Generalization to Japanese
The current experiment will test the model with
sentences in Japanese. Unlike English, Japanese
allows extensive liberty in the ordering of words,
with grammatical roles explicitly marked by
postpositional function words -ga, -ni, -wo, -yotte.
This word-order flexibility of Japanese with
respect to English is illustrated here with the
English active and passive di-transitive forms that
each can be expressed in 4 different common
manners in Japanese:
1. The block gave the circle to the triangle.
1.1 Block-ga triangle-ni circle-wo watashita .
1.2 Block-ga circle-wo triangle-ni watashita .
1.3 Triangle-ni block-ga circle-wo watashita .
1.4 Circle-wo block-ga triangle-ni watashita .
2. The circle was given to the triangle by the
block.
2.1 Circle-ga block-ni-yotte triangle-ni watasareta.
2.2 Block-ni-yotte circle-ga triangle-ni watasareta .
2.3 Block-ni-yotte triangle-ni circle-ga watasareta .
2.4 Triangle-ni circle-ga block-ni-yotte watasareta
.
In the  active Japanese sentences, the
postpositional function words -ga, -ni and  wo
explicitly mark agent, recipient and, object
whereas in the passive, these are marked
respectively by  ni-yotte, -ga, and  ni. For both
the active and passive forms, there are four
different legal word-order permutations that
preserve and rely on this marking. Japanese thus
provides an interesting test of the model s ability to
accommodate such freedom in word order.
Employing the same method as described in the
previous experiment, we thus expose the model to
<sentence, meaning> pairs generated from 26
Japanese constructions that employ the equivalent
of active, passive, relative forms and their
permutations. We predicted that by processing the
-ga, -ni, -yotte and  wo markers as closed class
elements, the model would be able to discriminate
and identify the distinct grammatical constructions
and learn the corresponding mappings. Indeed, the
model successfully discriminates between all of the
construction types based on the ConstructionIndex
unique to each construction type, and associates
the correct SentenceToScene mapping with each of
them. As for the English constructions, once
learned, a given construction could generalize to
new untrained sentences.
This demonstration with Japanese is an
important validation that at least for this subset of
constructions, the construction-based model is
applicable both to fixed word order languages such
as English, as well as free word order languages
such as Japanese. This also provides further
validation for the proposal of Bates and
MacWhinney (et al. 1982) that thematic roles are
indicated by a constellation of cues including
grammatical markers and word order.
3.4 Effects of Noise
The model relies on lexical categorization of
open vs. closed class words both for learning
lexical semantics, and for building the
ConstructionIndex for phrasal semantics. While we
can cite strong evidence that this capability is
expressed early in development (Shi et al. 1999) it
is still likely that there will be errors in lexical
categorization. The performance of the model for
learning lexical and phrasal semantics for active
transitive and ditransitive structures is thus
examined under different conditions of lexical
categorization errors. A lexical categorization error
consists of a given word being assigned to the
wrong category and processed as such (e.g. an
open class word being processed as a closed class
word, or vice-versa). Figure 2 illustrates the
performance of the model with random errors of
this type introduced at levels of 0 to 20 percent
errors.
Figure 2. The effects of Lexical Categorization Errors (mis-
categorization of an open-class word as a closed-class word or vice-
versa) on performance (Scene Interpretation Errors) over Training
Epochs. The 0% trace indicates performance in the absences of noise,
with a rapid elimination of errors . The successive introduction of
categorization errors yields a corresponding progressive impairment in
learning. While sensitive to the errors, the system demonstrates a
desired graceful degradation
39
We can observe that there is a graceful
degradation, with interpretation errors
progressively increasing as categorization errors
rise to 20 percent. In order to further asses the
learning that was able to occur in the presence of
noise, after training with noise, we then tested
performance on noise-free input. The interpretation
error values in these conditions were 0.0, 0.4, 2.3,
20.7 and 33.6 out of a maximum of 44 for training
with 0, 5, 10, 15 and 20 percent lexical
categorization errors, respectively. This indicates
that up to 10 percent input lexical categorization
errors allows almost error free learning. At 15
percent input errors the model has still
significantly improved with respect to the random
behavior (~45 interpretation errors per epoch).
Other than reducing the lexical and phrasal
learning rates, no efforts were made to optimize
the performance for these degraded conditions,
thus there remains a certain degree of freedom for
improvement. The main point is that the model
does not demonstrate a catastrophic failure in the
presence of lexical categorization errors.
4 Discussion
The research demonstrates an implementation of
a model of sentence-to-meaning mapping in the
developmental and neuropsychologically inspired
construction grammar framework. The strength of
the model is that with relatively simple  innate 
learning mechanisms, it can acquire a variety of
grammatical constructions in English and Japanese
based on exposure to <sentence, meaning> pairs,
with only the lexical categories of open vs. closed
class being prespecified. This lexical
categorization can be provided by frequency
analysis, and/or acoustic properties specific to the
two classes (Blanc et al. 2003; Shi et al. 1999). The
model learns grammatical constructions, and
generalizes in a systematic manner to new
sentences within the class of learned constructions.
This demonstrates the cross-linguistic validity of
our implementation of the construction grammar
approach (Goldberg 1995, Tomasello 2003) and of
the  cue competition model for coding of
grammatical structure (Bates et al. 1982). The
point of the Japanese study was to demonstrate this
cross-linguistic validity  i.e. that nothing extra
was needed, just the identification of constructions
based on lexical category information. Of course a
better model for Japanese and Hungarian etc. that
exploits the explicit marking of grammatical roles
of NPs would have been interesting  but it
wouldn t have worked for English!
The obvious weakness is that it does not go
further. That is, it cannot accommodate new
construction types without first being exposed to a
training example of a well formed <sentence,
meaning> pair. Interestingly, however, this
appears to reflect a characteristic stage of human
development, in which the infant relies on the use
of constructions that she has previously heard (see
Tomasello 2003). Further on in development,
however, as pattern finding mechanisms operate
on statistically relevant samples of this data, the
child begins to recognize structural patterns,
corresponding for example to noun phrases (rather
than solitary nouns) in relative clauses. When this
is achieved, these phrasal units can then be inserted
into existing constructions, thus providing the basis
for  on the fly processing of novel relativised
constructions. This suggests how the abstract
construction model can be extended to a more
generalized compositional capability. We are
currently addressing this issue in an extension of
the proposed model, in which recognition of
linguistic markers (e.g.  that , and directly
successive NPs) are learned to signal embedded
relative phrases (see Miikkulainen 1996).
Future work will address the impact of
ambiguous input. The classical example  John
saw the girl with the telescope implies that a
given grammatical form can map onto multiple
meaning structures. In order to avoid this violation
of the one to one mapping, we must concede that
form is influenced by context. Thus, the model
will fail in the same way that humans do, and
should be able to succeed in the same way that
humans do. That is, when context is available to
disambiguate then ambiguity can be resolved. This
will require maintenance of the recent discourse
context, and the influence of this on grammatical
construction selection to reduce ambiguity.
5 Acknowledgements
This work was supported by the ACI
Computational Neuroscience Project, The
Eurocores OMLL project and the HFSP
Organization.

References
Bates E, McNew S, MacWhinney B, Devescovi A,
Smith S (1982) Functional constraints on
sentence processing: A cross linguistic study,
Cognition (11) 245-299.
Blanc JM, Dodane C, Dominey P (2003)
Temporal processing for syntax acquisition.
Proc. 25th Ann. Mtg. Cog. Science Soc. Boston
Brown CM, Hagoort P, ter Keurs M (1999)
Electrophysiological signatures of visual lexical
processing : Open- and closed-class words.
Journal of Cognitive Neuroscience. 11 :3, 261-
281.
Chang NC, Maia TV (2001) Grounded learning of
grammatical constructions, AAAI Spring Symp.
On Learning Grounded Representations,
Stanford CA.
Chomsky N. (1995) The Minimalist Program. MIT.
Crangle C. & Suppes P. (1994) Language and
Learning for Robots, CSLI lecture notes: no. 41,
Stanford.
Dominey PF, Ramus F (2000) Neural network
processing of natural language: I. Sensitivity to
serial, temporal and abstract structure of
language in the infant. Lang. and Cognitive
Processes, 15(1) 87-127
Dominey PF (2000) Conceptual Grounding in
Simulation Studies of Language Acquisition,
Evolution of Communication, 4(1), 57-85.
Dominey PF, Hoen M, Lelekov T, Blanc JM
(2003) Neurological basis of language and
sequential cognition: Evidence from simulation,
aphasia and ERP studies, Brain and Language,
86, 207-225
Elman J (1990) Finding structure in time.
Cognitive Science, 14:179-211.
Feldman JA, Lakoff G, Stolcke A, Weber SH
(1990) Miniature language acquisition: A
touchstone for cognitive science. In Proceedings
of the 12th Ann Conf. Cog. Sci. Soc. 686-693,
MIT, Cambridge MA
Feldman J., G. Lakoff, D. Bailey, S. Narayanan, T.
Regier, A. Stolcke (1996). L0: The First Five
Years. Artificial Intelligence Review, v10 103-
129.
Goldberg A (1995) Constructions. U Chicago
Press, Chicago and London.
Hirsh-Pasek K, Golinkof RM (1996) The origins of
grammar: evidence from early language
comprehension. MIT Press, Boston.
Kotovsky L, Baillargeon R, The development of
calibration-based reasoning about collision
events in young infants. 1998, Cognition, 67,
311-351
Langacker, R. (1991). Foundations of Cognitive
Grammar. Practical Applications, Volume 2.
Stanford University Press, Stanford.
Mandler J (1999) Preverbal representations and
language, in P. Bloom, MA Peterson, L Nadel
and MF Garrett (Eds) Language and Space, MIT
Press, 365-384
Miikkulainen R (1996) Subsymbolic case-role
analysis of sentences with embedded clauses.
Cognitive Science, 20:47-73.
Morgan JL, Shi R, Allopenna P (1996) Perceptual
bases of rudimentary grammatical categories:
Toward a broader conceptualization of
bootstrapping, pp 263-286, in Morgan JL,
Demuth K (Eds) Signal to syntax, Lawrence
Erlbaum, Mahwah NJ, USA.
Pollack JB (1990) Recursive distributed
representations. Artificial Intelligence, 46:77-
105.
Roy D, Pentland A (2002). Learning Words from
Sights and Sounds: A Computational Model.
Cognitive Science, 26(1), 113-146.
Shi R., Werker J.F., Morgan J.L. (1999) Newborn
infants’ sensitivity to perceptual cues to lexical
and grammatical words, Cognition, Volume 72,
Issue 2, B11-B21.
Siskind JM (1996) A computational study of cross-
situational techniques for learning word-to-
meaning mappings, Cognition (61) 39-91.
Siskind JM (2001) Grounding the lexical semantics
of verbs in visual perception using force
dynamics and event logic. Journal of AI
Research (15) 31-90
Steels, L. (2001) Language Games for
Autonomous Robots. IEEE Intelligent Systems,
vol. 16, nr. 5, pp. 16-22, New York: IEEE Press.
Stolcke A, Omohundro SM (1994) Inducing
probabilistic grammars by Bayesian model
merging/ In Grammatical Inference and
Applications: Proc. 2nd Intl. Colloq. On
Grammatical Inference, Springer Verlag.
Talmy L (1988) Force dynamics in language and
cognition. Cognitive Science, 10(2) 117-149.
Tomasello M (1999) The item-based nature of
children’s early syntactic development, Trends in
Cognitive Science, 4(4):156-163
Tomasello, M. (2003) Constructing a language: A
usage-based theory of language acquisition.
Harvard University Press, Cambridge.
