The effect of givenness on word order in ditransitive structures in Croatian child language

This research investigates how monolingual Croatian children implement different givenness values in their ditransitive structures, more precisely how the givenness of an object influences the ordering of the two objects. According to previous research, Slavic languages, and thus also Croatian, follow the given before new principle (Kučerová, 2012; Siewierska, 1993) and other studies on ditransitive structures have shown that givenness influences object order and yields DO-IO with given themes, and IO-DO with given recipients in a variety of languages in both adult and child language (Anderssen, Rodina, Mykhaylyk, & Fikkert, 2014 for Norwegian ; Bresnan, 2005; de Marneffe, 2012 for English; Kizach & Balling, 2013 for Danish; Mykhaylyk, Rodina, & Anderssen, 2013 for Russian and Ukrainian ). However, when it comes to child data and givenness, given elements do not necessarily precede new ones (Narasimhan & Dimroth, 2008). This is also the case for Croatian children: Velnić (Under review) has found both given>new and new>given orders in children’s utterances, but only given>new in child directed speech in the Kovačević corpus contained in the CHILDES database (MacWhinney, 2000). However, the naturalistic setting in that study provided very little variation of givenness contexts, thus we have decided to analyze this more thoroughly in an experimental setting.

A total of 58 monolingual Croatian children (ages 3;8-5;2, mean: 4;4) completed the task. We have divided the children into two age groups based on whether their age was below or above the mean. The task consisted in a puzzle board methodology inspired by Eisenbeiss (2011). The puzzle board acted as a divisor between the experimenter and the child so the child was the only one to see the images, thus incentivizing the use of NPs over other forms. We used three sets of images that depicted ditransitive actions of transfer between animals and objects and were given in a sequence that contained four givenness conditions: nothing given, subject and DO given, subject and IO given, and all given. In between each action image, the child was presented with an image of the participant that was going to bear the givenness value in the next action image; this helped to establish what is given in that context and to have a short conversation abut the image and thus avoid self-priming of the structure used with the previous action image. Two sets of images had an inanimate DO, and one had an animate DO; the subject and the IO were animate in all the sets. This allowed us to check whether the object order preferences changed with relation to animacy as well as givenness.

The results indicate that, as in corpus data, children have a preference for IO-DO structures and use it 69% of the time. We have looked separately into the results of the sets based on the animacy of the DO. We have established a word order baseline for each age group and DO-animacy condition, based on the IO-DO and DO-IO proportion in the no given and all given conditions and observed how that proportion changes based on which object is given. In the DO inanimate condition the children from the younger group used IO-DO 76% of the time in All-G and No-G conditions. Surprisingly, the children significantly preferred IO-DO in the DO-G condition than in IO-G (84% of IO-DO in DO-G, 66% in IO-G, p<0.05). This suggests that there is a preference for new>given. The older children did not seem to be affected by the givenness manipulation of the DO, and equally prefer IO-DO across all conditions. Therefore, there is no new>given preference, but also no given>new. In the DO animate condition there are much more errors and a portion of the data was non-applicable: 21% in the younger group and 7% in the older group. This might be due to the non-prototypical setting of three animates in a ditransitive action (e.g. the turtle is sending the parrot to the snail). In this condition we see a considerable rise of the use of the DO-IO order which relates to the animacy of the DO, especially in the younger group: in the DO-G condition the IO-DO order is at only 38% thus making DO-IO the more attested word order; in the IO-G condition IO-DO order is attested 52% of the time, almost at chance level. So, the DO-IO order is used considerably more, with what seems to be a givenness effect within the DO-G condition, but its use is also higher across all conditions including IO-G. The older group used the IO-DO order in the same proportion in the DO-G and IO-G conditions: 68% and 66% thus indicating that there is no effect of givenness. We have observed a rise in the usage of DO-IO, but the IO-DO is still predominantly used.

From these data, we can conclude that Croatian children do not have the given>new principle in place, but that there is a significant change in between the two age groups since younger children tend to put new>given in the DO-given condition and the older group shows no preference related to givenness. Also, the children from the younger group seem to be more sensible to animacy than the children in the older group as we can see from a considerable rise of DO-IO instances. The younger group shows a slight preference for new>given in the DO inanimate condition but there is some evidence for given>new when the DO is animate within the same group of participants. This indicates that for 3-year-olds animacy is more relevant than givenness and givenness influences object order when there is no interference from animacy. This however does not explain their preference for new>given within the DO-inanimate condition. The older group does not show any particular preferences related to givenness, but we can observe an animacy effect since there is a reduced use of IO-DO once the DO is also animate. In order to check the progress of the given>new principle we should test 6-year-olds and see whether they prefer a given>new order, and consequently observe a progression from new>given preference (3-year olds) through no preference (4-year olds) ending in a given>new preference (target).  

 Table 1: percentiles of word orders in DO-inanimate in Age group 1

Table 2: percentiles of word orders in DO-inanimate in Age group 2

Table 3: percentiles of word orders in DO-animate in Age group 1

Table 4: percentiles of word orders in DO-animate in Age group 2