In this paper, we examine the acquisition of relative clauses (RCs) in Colloquial Jakarta Indonesian, based on the analysis of naturalistic data collected at the Max Planck Institute's Jakarta Field Station.

In the adult language, relative clauses are preceded by the marker yang, coded as REL in our data. When present, the head NP corresponds to some constituent within the relative clause, but in Colloquial Jakarta Indonesian, RCs are very often "headless":

(1)	Headed RC (10% of the adult RCs in our data)
	Pesawat yang dibuat tadi begini
	airplane REL DI-make earlier like this
	'The plane that I made is like this' [adult is drawing picture for the child]
(2)	Headless RC (90% of the adult RCs in our data)
	Buat yang terbang di atas
	for REL fly LOC on
	'Draw the one that flies above' [experimenter asks the child to draw an airplane]

As discussed in Sneddon (1996), in addition to the verbal RCs shown above, RCs can be derived from other clause types (all examples are from the Jakarta database):

(3)	Prepositional RC
	Cuma kan yang dari Tante Gaby gede
	only KAN REL from aunt Gaby big
	'But the one from Aunt Gaby was big' [referring to the chocolate given to the child]
(4)	Adjectival RCs
	Nih, Tante bikin yang gede
	this auntie make REL big
	'Look, auntie draw a big one' [referring to the fish the experimenter is drawing for the child]
(5)	RC with determiner (demonstrative)
	Bukan, monyet yang ini
	NEG monkey REL this
	'No, the monkey is this one' [experimenter correcting the child and pointing to the right picture]

As described in Sneddon, verbal RCs can contain any constituent occurring in an independent clause except for the subject (and in our data base, in some rare cases the object). The resulting gap is identical to the null or lexical head of the RC. In addition to verbal RCs, Sneddon also lists yang+adj constructions as RCs and remarks that this is a very frequent type. It may be claimed then, that in the adult grammar, all of the examples we presented with yang should have a unified analysis. Under this analysis, these are all headless RCs (or free relatives); yang is a complementizer which precedes an RC, the RC itself can be of any type (verbal, or a small clause containing an adjective, a demonstrative or a PP). We will call this the unified RC hypothesis (URC).

In this study we would like to examine whether in the course of language acquisition all purported RCs are acquired at the same time. If the URC is the correct grammar for the child, the child should show no preference for one type of RC over another, once the item yang and lexical items like adj, dem, prep and V are acquired. On the other hand, it could be that children have a much simpler grammar for the yang+adj/dem combinations, in which yang between an NP and a modifier (the adj or dem) is a connector, or actually replaces an NP (meaning something like "the one").

Our data came from the longitudinal study of three children ages 1;07- 3;06 (two boys and one girl). The adult input to each child is also analyzed in order to determine whether children simply follow frequency tendencies from the adult input. We chart the development of the construction and conclude that children show evidence for the following stages:

(a)	An early stage (before age 2;0) at which they use adjectives and determiners with nouns, but do not use yang at all.
(b)	a second stage in which children start using yang only with yang+dem and yang+adj. This period can be very brief (2-3 months for some children) but also lasted up to 4 months for one of the children. This is a stage in which the child's grammar may be simpler than the adult's grammar, since all of the child's uses are compatible with an analysis in which yang is either a connector or means "the one that".
(c)	After this brief period, the construction is used productively with verbs. We claim that once children use yang with verbs (which have argument structure), a gap is being posited, and the child now has the adult analysis of RCs.

We can also show that the early acquisition of yang+adj/dem is not simply due to frequency of these combinations in the input. From our data, the most frequent use of yang for adults is in verbal sentences (about 40%). However, the child does not start out with the construction which is the most frequent subtype for the adult, but rather, starts with the ones which are simpler; constructions in which the child does not need to posit any complicated argument structure or gaps due to movement. Moreover, one can show that the child productively uses verbs at this stage, but the same verbs only show up inside RCs a few months later. In other words, it is neither the frequency found in the adult input, nor the child's own frequency of use of certain verbs which accounts for the delay in acquisition. The only explanation is that children start out with the structures which are amenable to a simpler analysis, fitting the developmental stage the child is in.