When morphology is not enough: The acquisition of voice in monolingual Greek children and bilingual children with Greek as a heritage language

ABSTRACT

The acquisition of voice in Greek remains understudied, especially in heritage populations. Voice in Greek poses a challenging acquisition task for children due to its syncretism, marking various verb classes as well as passives. The present study explores the acquisition of anticausatives, reflexives, and passives in 6-to-8-year-old monolingual Greek and Greek-German bilingual children with Greek as their heritage language. This age coincides with the immersion of children in the education system. A Sentence Picture Matching task tested participants’ interpretation preferences when ambiguity arises in reflexives and (optionally) (un)marked anticausatives (reflexive/anticausative vs. passive interpretation); it also tested accuracy in passives compared to actives-transitives (passive vs. active interpretation). All children exhibited adult-like performance in reflexives. In anticausatives, despite their lower performance, they were qualitatively similar to adults, exhibiting knowledge of verb classes. Children’s preferences were predicted by age (active-morphology/ACT) and vocabulary (non-active/NACT), confirming the role of the mental lexicon in the acquisition of NACT anticausatives. The only difference between bilingual and monolingual children was in optionally (un)marked anticausatives, both in ACT and NACT, that was marginally predicted by age and grammar production respectively, suggesting that optionality affects bilingual more than monolingual children. In passives, children were less accurate than adults despite ceiling performance in actives-transitives. Bilingual children performed less well than monolingual children and performance was predicted by grammar and exposure. These findings reflect the difficulties of monolingual children with passives and their late acquisition which further affects bilingual development due to reduced input in heritage Greek.

Author contributions

All authors conceived and designed the study and oversaw the implementation of the study. A. Paspali created the experiment of the study and wrote all sections of the manuscript with major help and feedback from T. Marinis and A. Alexiadou during the writing process. A. Paspali collected, transcribed, and coded the data. A. Paspali conducted the statistical analysis and interpreted the data with fundamental contribution by T. Marinis and A. Alexiadou.

Acknowledgments

Parts of the present study have been presented and published in the Proceedings of 41^st Annual Meeting of the Department of Linguistics, Aristotle University of Thessaloniki (May 13-15, 2021). We thank Despina Papadopoulou, Georgia Fotiadou, Dimitris Michelioudakis, and the audience of the 41^st Annual Meeting of the Department of Linguistics (A.U.Th) for their feedback during various stages of this work. We also thank the student assistants Fotini Karkaletsou and Nikolas Tsokanos for their help in data collection.

We would like to thank our illustrator Konstantinos Roungeris who drew the pictures for the present study under our own guidelines, detailed description for the content of the pictures, and supervision for corrections and clarity relevant to each sentence item. Any other studies that might use these pictures must refer to the present work and Konstantinos Roungeris and ask for our permission.

Declaration of interests

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.

Ethics statement

Adult participants and children’s parents gave written informed consent (children gave oral informed consent) in accordance with the Declaration of Helsinki. The study has been reviewed by the Ethics Committee of the Humboldt University of Berlin and has been given a favorable ethical opinion for conduct.

Data availability

The data of this study are available on request to the corresponding author.

Notes

1 We use Voice for the syntactic category, and voice for the corresponding morphological category.

2 We follow Tsimpli (2006) for the notion of “verb class” and how it is used in the present study, i.e., verbs that participate in the causative/anti-causative alternation (anticausatives of the present study) and verbs which favour the reflexive reading (reflexive verbs of the present study) are argued to belong to the lexically determined class of inherently reflexive verbs (Alexiadou & Anagnostopoulou 2004). As Tsimpli points out, the grammar does not restrict its options depending on verb class. Thus, the term is used to refer to the preferred reading that adult monolingual native speakers have for these verbs.

3 Recently, it has been argued that the external argument of the verb is in spec (little) v in Greek (Angelopoulos et al. 2020). As an extra layer VoiceP is included in that analysis as well for the creation of the passive, we think there is no difference in predictions with respect to the complexity/uniqueness hypotheses. However, Angelopoulos et al. (2020) do not make any claims with respect to reflexives or marked anticausatives, though arguably the latter could be analyzed as lacking Spec,vP in their system. If that is indeed the case, marked and unmarked anticausatives would be equally complex.

4 Except for deponent verbs (which are not examined in the present study), as an anonymous reviewer pointed out, see Alexiadou & Rizou (to appear) for some discussion on the creation of novel deponents in Heritage Greek.

5 Furthermore, for the representation of Greek Voice, they propose that NACT morphology involves a Middle Voice head (instead of the Pass and Voice functional heads as in other languages) because of the lower productivity of passive verbs in Greek (i.e., passivization) compared to languages such as English or German. This head introduces the external argument of the verb.

6 Reflexivity is also expressed via the analytic form containing the reflexive pronoun, which is subject to Principle-A, as can be seen in the following:

O Kostas xtenizei ton eafto tu.

the Kostas comb.3SING the self his

‘Kostas combs himself.’

7 As Alexiadou et al. (2015:138) explain “Naturally Reflexive Verbs (NRVs) come from a number of semantic subclasses which all represent events that carry . . . inherent in their meaning [ . . .] the lack of expectation that the two semantic roles they make reference to will refer to distinct entities . . .” (Kemmer 1993:58). Grooming verbs like wash, dress, etc., are central instances of NRVs. Note that the reflexives verbs of the present study are all NRVs.

8 See, also, Fotiadou (2022) on the online processing of anticausative verbs in Greek adult native speakers, in which the interaction between frequency and processing/acceptability judgements is explored.

9 For a recent syntactic analysis claiming that Voice is expletive in optionally marked anticausatives, see Alexiadou et al. (2015) and for experimental evidence supporting this view see Tsiakmakis et al. (2023).

10 Although this is not a processing experiment, we decided to include fillers to have a balance between the frequency of verbs in ACT and NACT across the experiment. Thus, we designed more fillers with ACT compared to NACT so that the total number of ACT and NACT verbs across the experiment is equal. Finally, notice that since we have different datasets (one for anticausatives, one for reflexives, one for passives and one for actives-transitives) which target different structures/materials (i.e., the anticausatives, the reflexives, the passive, the active-transitive respectively), the items of each dataset already act as fillers for the other datasets plus the fillers we created to keep the balance between the frequency of ACT and NACT verbs that each participant will be exposed to.

11 An anonymous reviewer points out that, previous research shows that in present tense there may be a conflict between anticausative/middle reading (the second includes an agent) which is not easily distinguishable in the pictures. We do not agree with this view. Notice that a) during piloting none of the participants denoted/claimed/implied that a middle reading is available to them when looking at the pictures. This means that the pictures and their visual context were designed appropriately so that they “block” a possible middle reading, something which, we admit, would not have been possible in a task without pictures, b) during the data collection, none of the participants mentioned anything related to a potentially accessible middle reading of the pictures during the debriefing phase (i.e., in the end of the session): instead, during the debriefing phase some participants tended to use the ¨by itself¨ criterion for anticausatives in ACT and NACT (in their effort to explain to the researcher what they had heard and saw in the pictures and to recall examples and some indicative pictures they had selected and how/why they selected them), c) we did not use adverbials which could potentially bias participants toward a middle reading, as was the case in Tsimpli (2006) with the item “to pukamiso afto leronete amesos, “This shirt gets dirty immediately/very soon”, and d) there was an agentive condition across all anticausative items which was significantly and systematically dispreferred across all groups. Thus, we have no reasons to believe that in the present experiment with the current design and pictures, middle reading (which includes an agent) was the reason participants systematically preferred the anticausative interpretation since the other condition with the agentive interpretation was always dispreferred. For the Greek middle structure, please see Alexiadou et al. (2015), Lekakou (2005), Manney (2000), Sioupi (1998), and references therein.

12 Agentive (passive) reading: We systematically use this term throughout the manuscript to have a constant term for the second picture of the picture triplet in Anticausatives. In this way, we denote an agentive reading for Anticausatives in ACT and a passive reading in Anticausatives in NACT.

13 This was an empirical decision since we observed this tendency in some children during the piloting stage. From a psycholinguistic and methodological point of view, the gold standard is that (paired) pictures are balanced and carefully designed when including the same number of referents, objects, same colours, sizes, etc. (Pliatsikas & Marinis 2020).

14 The production (and not the comprehension of morphosyntax) was selected because children exhibited more difficulties in the production of morphosyntax and the heritage bilingual children also significantly differed from monolingual children (see the Results of the Baseline tasks), while the comprehension of morphosyntax exhibited close to ceiling performance for both groups.

15 An anonymous reviewer asks whether “gremizo” is considered to belong to class III and whether it exhibits an ACT form. This is how it is theoretically categorized (see Alexiadou et al. 2015 and Tsiakmakis et al. 2023). To further explore this, since we admit that this verb is much better with NACT compared to ACT, we removed the item (gremise) from the dataset of Class III in ACT but the pattern of results does not really change:

• Anticausative preference with gremizo: adults 86.3%, monolingual children 79.4%, heritage bilingual children 64.3%

• Anticausative preference without gremizo: adults 88.2%, monolingual children 79.3%, heritage bilingual children 65%

16 An anonymous reviewer points out the verb “arrest” is of much lower frequency compared to the rest of the verbs used in the passives-dataset. Below as the pattern of results indicate, the mean accuracy of participants with and without the verb “arrest” does not change.

• Accuracy in passives including the verb arrest: adults 93.2%, monolingual children 85.5%, heritage bilingual children 71.9%

• Accuracy in passives without the verb arrest: adults 92.2%, monolingual children 84%, heritage bilingual children 72%

17 An anonymous reviewer points out that “kano volta” is periphrastic (an expression), thus very different from others and that a comment should be added. Yes, this is true. However, it also denotes an activity. Since all groups performed at ceiling (i.e., across all items), the inclusion of this item does not change the ceiling pattern of results in this dataset.

Additional information

Funding

This work was funded from DFG AL 554/7-1; AL554/7-2: Acquisition of Voice to Artemis Alexiadou.