A large-scale cross-linguistic investigation of the acquisition of passive

ABSTRACT

This cross-linguistic study evaluates children’s understanding of passives in 11 typologically different languages: Catalan, Cypriot Greek, Danish, Dutch, English, Estonian, Finnish, German, Hebrew, Lithuanian, and Polish. The study intends to determine whether the reported gaps between the comprehension of active and passive and between short and full passive hold cross-linguistically. The present study offers two major findings. The first is the relative ease in which 5-year-old children across 11 different languages are able to comprehend short passive constructions (compared to the full passive). The second and perhaps the more intriguing finding is the variation seen across the different languages in children’s comprehension of full passive constructions. We argued, based on the present findings, that given the relevant linguistic input (e.g., flexibility in word order and experience with argument reduction), children at the age of 5 are capable of acquiring both the short passive and the full passive. Variation, however, stems from the specific characteristics of each language, and good mastery of passives by the age of 5 is not a universal, cross-linguistically valid milestone in typical language acquisition. Therefore, difficulties with passives (short or full) can be used for identifying SLI at the age of 5 only in those languages in which it has already been mastered by typically developing children.

Notes

¹ In what follows, we will not make a distinction between theme and patient theta roles. We refer the arguments of a transitive verb as internal and external arguments, leaving aside the type of theta role they may bear.

² Cypriot Greek is a dialect of Greek spoken in Cyprus. While it shares most language features with Standard Modern Greek spoken in Mainland Greece, it also differs in specific aspects of phonology, morphology, and syntax, which makes Cypriot Greek a distinct dialect from Standard Modern Greek (see Kambanaros & Grohmann [2010] for discussion on the differences between the two dialects).

³ The type of construction that was used in Estonian is called “impersonal construction.”

⁴ In our study, we used the periphrastic passive in both Catalan and Danish.

⁵ On the other hand, some lexical items used in the passive construction may be used in constructions other than the verbal passive: The past participle is used in verbal and adjectival passives in, e.g., Catalan, English, Dutch, and German, among others; the past participle is also used in perfective aspect in, e.g., English, Dutch, German, and others; in Hebrew, the morphology for passives and resultatives take the same morphological form in the present tense; the form of past tense and past participle are identical in English; and so on.

⁶ The term sister and girl are used, depending on the form that was found to be more appropriate for testing in each language.

⁷ There are two types of passives in Catalan. One is the periphrastic passive, used in this experiment, and the other is a passive using a clitic es, which is homophonous with a reflexive clitic, and because this type of passive does not allow the external argument to be overtly expressed, we chose the periphrastic passive for this experiment.

⁸ Although it has been observed that passive sentences are more demanding to process (longer reaction time and more mistakes) compared to active sentences (Hirsch &Wexler 2006), it is unclear how we can control for language variations following this hypothesis. We therefore do not address this approach here.

⁹ The question remains of whether there exists a challenge with transferring the subject’s theta role to the by-phrase in all passive constructions or whether this operation is only a challenge in the particular cases in which there is a conflict in theta roles, but this issue is beyond the scope of the present article.

¹⁰ This was established using ANOVAs with the Games-Howell post hoc test.

¹¹ This was established using the independent-samples t-test.

¹² This was established with the binomial test.

¹³ Due to an administrative error, the verb photograph was substituted for the verb make up in the active condition of the English full passives experiment.

¹⁴ The chance level was calculated for three pictures rather than for four, as this criterion was a stronger one for the hypothesis that children answered randomly. The fourth picture, the neutral/control picture (presenting no action but only three protagonists not performing any activity), could be eliminated by children only because there were no action at all, and still the children’s choice could remain random. See Section 5.3.2. for details.

¹⁵ In this and all subsequent analyses we report actual results unadjusted for chronological age. While significant between-group differences in chronological age did emerge (see Methods section), correlations between age and performance were weak or negligible (all r values lower than .2).

¹⁶ One-sample t-test was used to examine this. Performance of the Catalan group was at chance whether the chance level was assumed to be 25%—t(23) = 1.353, p = .189—or 33%—t(23) = 0.102, p = .919. For the assumption beyond those “lenient” and “strict” chance level, see Section 5.3.2.

¹⁷ We thank an anonymous reviewer for bringing this possibility to our attention.

¹⁸ The Hebrew preposition, though being unique to the passive, is phonologically similar to (and historically derived from) the locative preposition meaning ‘next to.’

¹⁹ Catalan also has a passive form that can be interpreted as reflexive, but this form of passive is not the one chosen for Catalan in the present study.

²⁰ One reviewer suggested that the exclusive use of actional verbs made this possibility more likely. First, in many languages, adjectival or resultative passives are easier with actional verbs. While it is true that this is not the case for all verbs that can be seen in our results, there is an overarching generalization that should not be ignored. Second, in numerous languages, there are additional forms of the passive that seem to be consistent in their noncanonical behavior. The English get-passive, as well as its counterparts in German, French, Dutch, and Swedish (and perhaps adversative passives in East Asian languages and their counterparts in a variety of Austronesian languages) are each restricted to highly actional, or affected, contexts and participles. Thus, it is conceivable that actionality and the use of the present tense could serve as powerful cues to children of an alternative structure for short passives. In full passives, the presence of the external argument phrase may make such an alternative less feasible. This needs to be tested by complementary experiments testing the same structure in the past tense and with nonactional verbs.

²¹ The present study tested only actional verbs that take an agent and as such could not test Fox & Grodzinsky’s (1998) proposal for difficulties with theta-role transmission in nonactional passives, which is one of the mechanisms that may be involved in constructing the passive construction.

²² We are grateful to one of the reviewers for this suggestion.

Additional information

Funding

The materials were obtained with the help of the Danish Agency for Science Technology and Innovation through the NASUD project led by Kristine Jensen de López (grant 273-07-0495). The Polish study was partially supported by public funding from Faculty of Psychology Warsaw University (grant no BST 184724/09) and from National Science Centre/Ministry of Science and Higher Education (grant no 809/N-COST/2010/0 and grant no N-N106-223538).