On the acquisition of complex predicates: Introduction to the special issue

There may be two approaches to diversifying the languages cited in child language acquisition research: a theoretical canon approach and a language-specific approach.

The ‘theoretical canon’ approach recruits out-of-the-way languages into long-standing theoretical debates to either validate current assumptions using a new language and population, or utilize a feature of the new language as a wedge with which to separate two opposing theories (see Cutler, 1985, on this notion in psycholinguistics). For instance, passive constructions have been known to be produced late by English-speaking children at least since Brown (1973), but studies show that children learning K’iche’ (Pye & Poz, 1988), Zulu (Suzman, 1987), Sesotho (Demuth, 1990), and Inuktitut (Allen & Crago, 1996) produce them relatively early, implying that language-specific factors like formation and frequency in the ambient language must be at play, not anything inherent about passives.

The theoretical canon approach is important for honing theories, but it risks a self-referential circularity, by which the features of a handful of (spoken) languages (Crystal, 2014) set the research agenda for acquisition of all languages of the world. As we know from the work of Clifton Pye, Susan Suzman, Katherine Demuth, Shanley Allen, and colleagues, there are myriad interesting, important, and salient aspects of child K’iche’ Mayan, Zulu, Sesotho, and Inuktitut, beyond the timing of passive acquisition. Similarly, while Courtney’s (2006) study of Quechua child and adult relative clauses contributed new evidence against the relevance of the accessibility hierarchy (Keenan & Comrie, 1979) for relative clause acquisition in that language, relative clauses are in fact extremely rare in naturalistic Quechua child-directed and child discourse. Courtney’s (1998, 2002) other work offers clues to points of interest more salient in naturalistic Quechua discourse (e.g. variable case-marking).

Kelly et al. (2015) and Pye (2021) imply that we have reached a breaking point for the discipline: should we continue to pool resources toward ever-more-fine-grained understanding of acquisition of particular construction types found in English – or should we instead expand our lens to incorporate the urgent, far-reaching documentation of acquisition of under-studied, endangered languages – spoken and signed – around the world, regardless of how they fit into current theoretical niches? For instance, the Nungon language of Papua New Guinea (Sarvasy, 2017) lacks a passive construction, but does feature relative clauses. We could decide to use our limited resources to explore processing of subject and object relative clauses by Nungon-speaking children and adults, thus adding to the relative clause acquisition literature (e.g. Kidd, 2011). But in so doing, we could miss the fact that a more salient feature of Nungon for children is the distinction between ‘clause chain’ sentences (sentences including up to 20 or more clauses, each marked for switch-reference) and other types of complex sentences – and this distinction might be worthier of time and funding (Sarvasy, 2019, 2020; Sarvasy & Choi, 2020a).

English and other Western European languages lack clause chains. Clause chains are frequent in Japanese, Korean, and Turkish, but until very recently (namely, the contributions to Sarvasy & Choi, 2020a), there were few studies devoted to acquisition of clause chains, possibly because researchers into acquisition of these languages had oriented themselves toward existing acquisition literature – itself oriented toward features found in English. In fact, new research into clause chaining in these and other languages points to a potentially universal ‘two-clause stage’ in children’s clause combining (Sarvasy & Choi, 2020b). If researchers had not stepped outside the canon of English-like features studied by acquisition scholars, this stage would have remained unrecognized.

Thus, the ‘language-specific’ approach to diversifying the field of language acquisition studies targets the acquisition of special features of endangered and under-studied languages, in their own rights. Clause chains, switch-reference, classifiers, grammatical evidentiality, and myriad other un-English-like characteristics of languages around the world are not yet part of the canon discussed in language development textbooks (Pye, 2021). But they can still be studied: to learn, in the first instance, how their acquisition patterns, rather than how they relate to previous studies of the same phenomena. For that matter, acquisition of signed languages in general, including in Anglophone countries, also needs much more attention (Morgan et al., 2006).

The language-specific approach should contribute to acquisition theories, but this requires intellectual effort and creativity from fieldworkers and theorists alike. When I first described Nungon clause chains (Sarvasy, 2020) to a senior scholar of child language acquisition, her skeptical response was, ‘I don’t know what these are’ (in terms of syntactic theory). The implication was, without knowing how to analyze clause chains in a formal syntactic theoretical framework, she was at a loss as to how to fit them into any acquisition theories. The language-specific approach to acquisition of under-studied languages demands receptivity on the part of acquisition scholars to descriptions of the acquisition of linguistic features not found in English and similar languages. Let us be creative in thinking about how these could enhance existing models of language acquisition – often in the absence of a standing theoretical-syntax model that analyzes them in terms that can also apply to English. Evans and Levinson (2009: 430) called on field linguists to ‘look up from their fascinating particularistic worlds and engage with the larger theoretical issues in the cognitive sciences’. When field linguists heed this call and those of Kelly et al. (2015) and Pye (2021), inter alia, to report on child acquisition of under-studied languages, scholars of language acquisition should similarly be roused from their canon of passives and relative clauses (among other things, of course) to countenance the range of other features that children acquire in the world’s languages. In other words, the language-specific approach will necessarily expand the grammatical topics covered in acquisition research.

The present Special Issue stems from a Symposium at the 2017 IASCL conference in Lyon, France, on the acquisition of complex predicates in Indigenous languages of Australia and New Guinea. Languages of Australia are known for ‘coverbs’, elements combined with the verb root that contribute semantically to the predicate, but cannot stand alone. Papuan languages of New Guinea are known for serial verb constructions (Aikhenvald, 2018), in which more than one verb root contributes to the meaning of a single predicate, and also for ‘verb-adjunct’ constructions, in which light verbs combine with non-verbal ‘adjunct’ elements that, like the Australian coverbs, cannot function alone, to express what in many other languages would be denoted by a simplex verb (Pawley & Hammarström, 2017). Our Symposium explored child acquisition of these constructions in languages of that region.

This Special Issue zooms out from that initial narrow focus, to embrace the acquisition of ‘complex predicates’, writ large. Complex predicates are taken to be verbal predicates headed by more than one verb, or by a verb plus other lexico-semantic elements (Alsina et al., 1997; Amberber et al., 2010). Complex predicates are of theoretical interest because they may straddle the boundary between single lexical entries and/or single words, and lexically complex, or multi-word, phrases (Amberber et al., 2010). The term complex predicate itself has been criticized as too inclusive to be theoretically meaningful (Gunkel, 1998), with much tighter formal definitions extant for sub-types of complex predicates, like serial verb (Aikhenvald, 2018) or causative constructions (Comrie & Polinsky, 1993). That said, the studies in this volume show that different types of complex predicate may still raise similar questions for acquisition. The studies here, spanning five un-related languages from four nations, address several recurring themes: (a) the distinction between complex predicates with ‘grammatical’ versus ‘lexical’ functions, (b) whether verbs that can serve as auxiliaries in complex predicates are first attested in independent uses, and (c) whether complex predicates are produced later than simple predicates.

This Special Issue introduces several new areas of investigation for acquisition. Herewith, the first studies of child development of the Algonquian type of preverbs, complex verbs involving preposition incorporation in a Papuan language, multi-verb predicates in a Papuan language, the causative in Q’anjob’al, and ‘extended predicate’ constructions in Modern Hebrew.

The contribution by Brittain and Rose examines the acquisition of preverbs in Northern East Cree by a single child, aged 2 years 1 month to 4 years 3 months. Preverbs in this language are stand-alone morphemes that accompany verbs in a range of conditions and with a variety of functions: both grammatical (contributing meanings of tense, modality, and directionality) and lexical (expressing adjectival or adverbial meanings, such as ‘well’, ‘bad’). During the study period, the child produced almost exclusively grammatical preverbs, with only one token of a lexical preverb. But the authors attribute this predominance of grammatical preverbs in the child’s speech to the formal and syntactic predictability of that type of preverb, not to the grammatical preverbs’ function.

Hellwig’s article is an initial account of the acquisition of complex verbs in the Papuan language Qaqet, spoken by about 15,000 people in Papua New Guinea. Qaqet is unusual among Papuan languages in that it does not feature serial verb-type complex predicates. But like many other Papuan languages, Qaqet can be analyzed as having a relatively small number of lexical verb roots, to which other elements are added to attain greater diversity of meanings. Unlike with many other Papuan languages, Qaqet achieves this by adding different prepositions to verbs. Hellwig describes the occurrence of such complex root-plus-predicate verbs in the speech of eight children, aged 1 year 11 months to 5 years 10 months, and also discusses variability in preposition application to arguments of verbs. Overall, complex verbs are rare in the speech of these children; even the oldest group use complex verbs in fewer than 3% of all utterances.

The article by Lustigman examines ‘extended predicate’ constructions in Modern Hebrew child and child-directed speech, based on data from three children, aged 1 year 3 months to 2 years 5 months. Three verbs with modal meanings, ‘want’, ‘be able’, and ‘need’, can combine with lexical verbs in an infinitive form. While similar constructions in adult Modern Hebrew have been analyzed as underlyingly biclausal (Berman, 1978: 304; Zuckermann, 2006: 84–85), Lustigman analyzes the group here as essentially ‘modal operators’ that combine with lexical verbs into a single predicate. This question may be moot in relation to the analysis here, which centers on child and adult use of the modal verbs alone and in extended predicate constructions, over time. The three modal verbs are first attested alone in child speech, without the accompanying infinitive, in utterances like ‘I want the ball’. Lustigman suggests that this relates to meaning, not to syntax: that expression of the concrete here precedes expression of the abstract, in the form of a complex predicate (‘I want to go’). Lustigman shows that the discourse functions of these constructions evolve considerably over time in both child and child-directed speech.

The article by Mateo Pedro investigates the acquisition of causative constructions in Q’anjob’al Mayan, by three children, aged 1 year 9 months to 4 years 0 months. Mateo Pedro gives a full account of causative productions in child and caregiver Q’anjob’al, and compares this to causative development for two other Mayan languages, K’iche’ and Tzotzil. Q’anjob’al speakers have two options for causative constructions: morphological and periphrastic. Children acquire both the morphological and periphrastic causatives before age 3, though they produce more periphrastic causatives overall. One impetus for this study is that previous research showed that acquisition of causatives patterned very differently in the Mayan language K’iche’ from the Mayan language Tzotzil. This difference was explained previously by morphological differences between those two languages, but the Q’anjob’al results also point to distributional patterns in child-directed speech.

The article by Sarvasy surveys productions of multi-verb predicates – possibly analyzable as serial verb constructions – in child and adult Nungon speech. Three children, aged 1 year 1 month to 3 years 3 months, were studied. There are two major types of serial verb constructions: lexical, or ‘symmetrical’, where all verb roots contribute roughly equally to the predicate’s semantics, and grammatical, or ‘asymmetrical’, in which one verb has grammatical (here, aspectual or modal), rather than lexico-semantic, function in the construction (Aikhenvald, 2018). Sarvasy examines five types of Nungon multi-verb predicates in the speech of children and adults, and finds that there is no clear pattern by which lexical or grammatical status alone determines order of acquisition or frequency in child speech. Rather, phonological and morphological form, semantics, and distribution in child-directed speech all can be analyzed to combine with construction type to influence acquisition order. Although the child studied for the longest period produces simple, one-verb predicates before she produces multi-verb predicates (beginning at 2 years 4 months), her earliest multi-verb predicates precede her production of one of the Nungon tense inflections on simple verbs.

Arguments can be made for later acquisition of complex predicates including elements with grammatical, rather than lexical, components, since presumably the grammatical components are less imageable and may also require a more advanced capacity to conceive of tense-aspect-mood distinctions. But the articles on Northern East Cree and Nungon show that the grammatical versus lexical distinction among complex predicate constructions does not transfer neatly into early versus late acquisition. Both authors conclude that the formal and syntactic characteristics of individual morphemes or construction types, along with, for Nungon, distributions in child-directed speech, could explain developmental trajectories.

Causative constructions are already relatively well-represented in the acquisition literature. Indeed, Mateo Pedro frames his study of the Q’anjob’al causative constructions in terms of previous work on acquisition of causatives in other Mayan languages, using the comparative method of child language acquisition (Pye, 2017). His study confirms for the Mayan causative constructions that children acquiring related languages can still differ markedly in their linguistic development. In this case, the Q’anjob’al children’s early acquisition and preference for the periphrastic causative construction – unlike children learning K’iche’ or Tzotzil – seems to be rooted, at least in part, in distributions in child-directed speech, since morphology or syntax alone cannot explain it.

The collection here also highlights diversity among features of languages of the same region. Nungon and Qaqet are both among the 600 to 800 spoken languages of Papua New Guinea, but beyond that, they share few grammatical commonalities; in fact, Qaqet prepositional verbs are structurally similar to those of West Germanic languages, with similar developmental trajectories.

The contribution on Modern Hebrew shows that children first produce the three modal verbs alone, as simple predicates, before they combine them with infinitives. While this is similar to Nungon, where simple, one-verb predicates are produced before multi-verb predicates, an expectation that simple predicates should precede complex predicates may not apply to a polysynthetic language like Northern East Cree.

In the vein of language-specific acquisition research, inspired by calls for an increased focus on acquisition of lesser-studied and endangered languages, then, this collection aims to enrich the scope of empirical language acquisition research by adding new features, and new languages, to the grammatical features canon of this field.