Exploring language diversity: A commentary on Kidd and Garcia (2022)

Language diversity is the dark matter of the linguistic universe – better than four fifths of it remains hidden to language acquisition researchers. Even unseen, language diversity exerts an inexorable force on acquisition theory. It formed the initial attractor for Chomsky’s (1965) theory of language acquisition. The acquisition problem is hard precisely because of the diverse ways that languages warp linguistic space-time. Prosody, grammatical relations, and semantic concepts are relative to specific languages and times.

In the target article, Kidd and Garcia (2022) explore diversity in language acquisition research by analyzing articles published over the past 45 years in the Journal of Child Language, First Language, Language Acquisition, and Language Learning and Development. The articles in their survey describe the acquisition of just 103 languages out of the world’s 7,000 languages. They report that the articles are skewed toward English and other well-studied European languages. Articles on the 64 non-Indo-European languages are skewed toward seven national languages: Hebrew (90), Mandarin (83), Japanese (63), Finnish (30), Korean (29), Cantonese (28), and Turkish (28). Subtracting these national languages from the sample leaves just 172 articles for the remaining 57 lesser-studied languages or an average of three articles per language. There is only 1 article for 27 of these languages, and 2 articles for 11 of them. It is therefore no exaggeration to claim that we have little idea of how children acquire language.

The target article offers a casual look at the impact that research on lesser-studied languages has had on past theoretical debates. A recent issue of the Journal of Child Language titled ‘Testable Theories of Core First Language Acquisition’ (vol. 48, no. 5, 2021) affords the opportunity to assess the impact of language diversity on six current theoretical perspectives. Using Kidd and Garcia’s analysis of publication frequency (their Table A1), I assessed the diversity of child languages referenced in the six theory articles. I counted languages with 5 or fewer articles as lesser-studied languages and counted languages with more than 5 articles as better-studied languages. I only included overtly named languages with child data in the articles. I did not include languages referenced in descriptions of the adult languages or in studies of adult input, statistical modeling or predictive behavior. The 6 theory articles reference child language data from just 16 languages (Table 1).

OPEN IN VIEWER

Table 1.

Six theories of language acquisition and the languages they reference.

Article	Theory	Lesser-studied languages	Better-studied languages
Snyder (2021)	Parameter Setting	0	English, Spanish, Japanese
Grinstead (2021)	Interface Delay	0	Spanish, Catalan, English, German
Pearl (2021)	UG + Statistics	0	English, Chinese, Italian, French, German, Dutch
Arnon (2021)	Starting Big	K’iche’*, Navajo, West Greenlandic, Chintang	English, Quechua, Tamil, Sesotho, Japanese
Behrens (2021)	Constructivist	0	English, German, Dutch
Ramscar (2021)	Discrimination	0	English

*

The article lists this language as ‘Quechua’ but references an article on K’iche’.

The number and treatment of the lesser-studied languages in these six articles is telling. Several of the chapters reference children’s language without mentioning specific languages. The articles by Snyder, Grinstead, Pearl, Behrens and Ramscar do not reference child data from any lesser-studied languages. Arnon’s article only references child data from lesser-studied languages in a section on the acquisition of agglutinating and polysynthetic languages. This small-sample study illustrates one use of the results in the target article and suggests that language diversity has little impact on current theoretical debates.

Language diversity demands that acquisition theorists use a representative set of languages to evaluate their proposals (Pye, 2017, 2021) The ACQDIV approach is one response to the problem of language diversity (Jansco et al., 2020), but its utility is limited by the method used in selecting the languages for its database. Until we have a better understanding of the size and scope of language diversity, as Kidd and Garcia state ‘we run the risks of making incorrect generalizations on the basis of a limited and skewed data set’. I recommend using the Mayan language Mam (England, 1983) as a corrective for acquisition theories. None of the theories reviewed in the six articles offer a satisfactory account for the acquisition of Mam’s verb complex, extended ergative marking, person enclitic, weight-sensitive prosody, relational noun phrases, or five passive voice constructions (Pye, 2017; Pye et al., 2013; Pye & Pfeiler, 2019).

The publication record that Kidd and Garcia analyze is evidence that acquisition theories command inordinate attention in the field (Sarvasy, 2021). Textbooks on language acquisition detail theory but not language diversity. Journals place a premium on theory testing rather than diversity, regardless of whether the theory only applies to a handful of languages. The Optional Infinitive hypothesis (Wexler, 1998), for example, does not apply to languages that lack infinitival verb forms without seriously misinterpreting the theory or misanalyzing a language (Pye, 2001). Classes on language acquisition would ideally devote their first two weeks to teaching language diversity. (See Merrifield et al’.s, 2003 workbook on morphology and syntax for an accessible introduction to language diversity). The time spent on reviewing theories of language acquisition would be better invested in learning about the diversity of human language.

Kidd and Garcia observe that ‘It is incumbent on us as the custodians of the discipline to move the goals of the field into a direction that will more easily allow and reward research on child language acquisition in understudied languages’. There is an urgent need to find new approaches that are capable of documenting the acquisition of thousands of languages over the next three decades. Current practices are not capable of processing a hundred hours of language samples for 7,000 languages. Assuming that each hour of a child language sample requires roughly 2 weeks to process yields an estimate of 28,000 years. It is unrealistic to believe that we can train even a thousand investigators to document the acquisition of a new language in one decade.

The best way forward would be to reduce the 100-hour sample per language to a more manageable 4 or 5-hour sample. The production of acquisition sketches (Defina et al., 2021), as Kidd and Garcia suggest, is one potential response to the processing bottleneck. Acquisition theorists would then need to identify the best way to use such sketches in accounting for a diverse set of child languages. The publication of acquisition sketches requires the development of outlets that prioritize empirical discoveries over transitory theoretical preoccupations. Documenting the acquisition of thousands of lesser-studied languages and archiving the results requires the creation of new modes of acquisition research.

A focus on documenting the diversity of child languages helps to return the results of acquisition research to the children’s communities. It is far easier to train speakers of lesser-studied languages to write acquisition sketches than teach them the finer points of acquisition theories. Focusing on the needs of language communities rather than the needs of acquisition theorists opens new avenues of partnership between researchers and communities. Teaching speakers from the children’s communities how to use computers to transcribe child language samples enhances the prestige of their language by demonstrating that it can be written and studied like any national language. Indigenous language investigators can serve as resources in their communities and advocates for their languages in schools and other settings.