Read an Excerpt

CHAPTER 1

Comparing Languages

1.1 The Monolingual Approach to Crosslinguistic Research

The modern era of language acquisition research began with the recognition that human language ability has more in common with human locomotion than with the ability to write (Chomsky 1965). Normally developing children learn to walk and talk without special instruction and in a wide variety of cultural contexts. Children only learn to write after years of training, and even then many children struggle to match written characters with the language that they speak. The difference between the tacit learning evident in learning to walk and speak and the conscious effort necessary to learn to write is evidence that oral or signed language is an ability that is built into the human genome. Children have been acquiring language without special instruction for at least two hundred thousand years.

Despite the evidence that children are able to acquire any human language, language acquisition research has failed to develop a systematic method for comparing language acquisition in different languages. In her summary article on crosslinguistic research Ruth Berman (2014: 33) notes "an explicit and generally applicable articulation of what is universal and what particular in first language acquisition ... would greatly advance the field of child language research as a whole." Crosslinguistic research on language acquisition resembles a giant jigsaw puzzle in that we know more than ever before about the acquisition of individual languages, but we lack the means to assemble all of these pieces into a coherent picture of the acquisition process as it unfolds in diverse human languages.

Since the first baby biographies of Darwin (1877) and Taine (1876), acquisition research remains focused on the acquisition of single languages. The monolingual focus of language acquisition research is reflected in the methods, descriptions, and theoretical orientation of the field. Researchers are trained to investigate how children acquire a single language rather than comparing how children acquire different languages. We know more than ever before how children acquire individual languages, but we lack an explicit procedure for comparing results across different languages. Language acquisition research lacks a framework for systematic crosslinguistic investigation that would fit results from individual languages into a comprehensive picture of children's language abilities. Understanding how children acquire specific languages does not address the more general problem of understanding children's ability to acquire all languages.

Textbooks on language acquisition reflect this single language approach. With few exceptions, textbooks take students through the basic descriptions of language development in English (Slobin 2014). Students learn that babies can perceive all of the sounds in the world's languages, but they only learn how children acquire phonemes in one language. Ambridge and Lieven acknowledge this limitation in their textbook and attribute this situation to "the fact that most of the theory development and empirical research has been conducted in countries where these languages [cp. European] are spoken" (2011:139). The textbooks written by David Ingram (1989) and Barbara Lust (2006) have sections that discuss the acquisition of other languages, but these discussions only show how different results can be in other languages rather than build a comprehensive understanding of children's ability to acquire all languages.

My claim that we lack a method of crosslinguistic research may appear to be extreme in the face of the growing number of crosslinguistic studies on language acquisition, some of which I have contributed to myself. Multiplying studies of individual languages does not address the problem of comparing results in different languages. One indication of the single-language orientation of crosslinguistic research is that the term "crosslinguistic" is even applied to research on a single language when its focus is not English. Research on the acquisition of Turkish is presumed to be crosslinguistic, while research on English is not. This usage reflects an assumption in the field that research on the acquisition of English provides a standard for crosslinguistic comparison, and that English presents a typical set of acquisition problems. While we certainly know more about the acquisition of English than about the acquisition of any other language, there are many reasons why English is a spectacularly poor choice for a standard of acquisition research. For example, English has an odd vowel system, little morphology, and a rigid word order. No rational linguist would choose English as a model of human language.

Crosslinguistic research on language acquisition typically focuses on the study of individual languages. Most of the chapters in the five-volume series The Crosslinguistic Study of Language Acquisition edited by Dan Slobin describe the acquisition of individual languages. Two summary chapters by Slobin (1985, 1997) attempt to draw common threads from all of these studies, but the chapter by Melissa Bowerman (1985) reminds readers that exceptions to broad generalizations are easy to find. Another example comes from a series of studies directed by Wolfgang Dressler (cf. Bittner, Dressler, and Kilani-Schoch 2003). These studies examine the impact of typology on the development of different linguistic categories such as inflection in a variety of languages (almost entirely European). Both of these research programs provide a wealth of findings on the acquisition of individual languages, and yet neither project addresses the general problem of comparing children's ability to acquire languages with different grammatical structures.

Experimental research on language acquisition is especially prone to a single-language orientation. This orientation does not stop investigators from drawing universal conclusions from the study of one language. Soja, Carey, and Spelke's study of children's use of count and mass nouns (1991) is a classic example of the single-language approach to experimental investigation. Soja et al. presented two-year-old children acquiring English with novel objects that had the characteristics of individuated objects or masses and asked the children to extend a novel label to another instance of the objects and masses. They found that their subjects distinguished between objects and masses in a statistically significant fashion and concluded "young children have an innate knowledge of the difference between mass and count nouns."

Imai and Gentner (1997) replicated Soja et al.'s study with groups of children and adults acquiring English and Japanese. Japanese does not distinguish between mass and count nouns in the English manner. Instead, Japanese treats virtually all nonhuman nouns as mass terms, and then requires numeral classifiers to count individual objects. Imai and Gentner replicated Soja et al.'s results for their subjects acquiring English, but they found significant differences between the Japanese and English children and adults. Their study showed that the children's count/mass distinction was learned rather than the product of an innate ontology, but they did not discuss the results relative to count/mass distinctions in all languages.

Jakubowicz (1996) presents another example of the single-language approach to crosslinguistic investigation. She outlines several acquisition studies that focus on the distinction among strong pronouns, weak pronouns, and clitics. She references a linguistic literature on pronouns that established this distinction among person markers in the Germanic and Romance languages (e.g., Corver and Delfitto 1993; Haegeman 1998; Kayne 1975). In this, like other monolingually oriented investigations, Jakubowicz generalizes a distinction from work on a few European languages to all languages without checking to see whether the generalization is valid. A more comprehensive survey of person markers in the world's languages reveals a continuum from strong pronouns to pronominal affixes and even zero marking (Siewierska 2011).

Jakubowicz outlines an experimental test of whether children produce strong forms before weak forms in French and German. Jakubowicz claims that German uses strong pronouns as subjects and weak pronouns as objects and that French uses weak pronouns as subjects and pronominal clitics as objects. She omits any discussion of person marking realized by the verb agreement suffixes in French and German, which artificially limits the hypothesis to pronouns and clitics rather than recognizing the full spectrum of person marking forms. According to the hypothesis that Jakubowicz tests, children should produce pronominal clitics before pronominal affixes. The extension to pronominal affixes strengthens the original hypothesis by recognizing the full spectrum of person forms in the languages.

Jakubowicz reported that both French and German children aged 2;2 to 2;6 produced more strong pronouns than weak pronouns. She acknowledged that this result is confounded by the use of strong forms as subjects in both languages and suggests that further research would resolve whether children acquire strong pronominal subjects earlier than weak pronominal subjects. She neglected to test whether German children produce more strong subject pronouns than the French children produce weak subject pronouns. This prediction also follows from the logic of Jakubowicz's hypothesis, but it requires a between-language comparison rather than a within-language comparison. The same comparison could be made for the weak object pronouns in German and the object clitics in French. French and German children would be expected to produce subject agreement on verbs at the same rate and less frequently than their production of either the subject or object forms.

This last observation underlines the monolingual nature of Jakubowicz's approach to crosslinguistic research. Her experiment assumes linguistic categories such as weak pronouns and clitics that are not informed by a survey of the world's languages. She does not acknowledge the possibility that her results reflect the shared history of French and German rather than a more general feature of all languages. She neglects the agreement affixes that French and German also use to mark person because her theoretical framework imposes an artificial distinction between pronominal clitics and pronominal agreement. Finally, she maintains a within-language design even though her hypothesis allows a between-language test. She does not discuss the difficulty of testing the hypothesis in non-European languages.

The parametric approach to language study represents the best known attempt to develop a framework for crosslinguistic research. The parametric approach was proposed by Chomsky as a way to account for differences between languages. He thought that differences between languages would fall into neat categories such as languages with verb-initial and verb-final word orders. The pro-drop parameter received the most attention in language acquisition research (Hyams 1986). The pro-drop parameter divides languages between those with obligatory subjects as in English, and those with optional subjects as in Spanish. Baker (2001) proposes a general list of parameters based on an analysis of 45 languages.

The parametric approach developed out of the monolingual approach to crosslinguistic research and retains most of its limitations. The pro-drop parameter, for example, was based on differences among English, Spanish, and Italian, rather than on a survey of the world's languages or even a survey of Italian dialects (Kayne 2000). The parametric approach only envisions a categorical distinction between languages with and without pro-drop. It does not explain the variation in subject use across languages, most notably in the variation among strong pronouns, weak pronouns, clitics, and affixes (Siewierska 2011). It is not surprising to find that linguists eventually realized that the parameters did not extend beyond the initial sample of languages, and Italian turned out to have dozens of regional variants that do not parameterize neatly (Newmeyer 2005).

A common element of the monolingual approach to crosslinguistic research is to take a linguistic feature from one language, usually English, and impose it on other languages as a basis for comparison. This approach assumes that features in different languages realize the same linguistic category. For example, children's use of infinitive verb forms in different languages remains a topic of interest despite the absence of a definition for the infinitive verb form that would apply to all languages (Wexler 1998). Whether researchers investigate verbs, subjects, pronouns, passives, causatives, or bilabial stops, they make no attempt to survey such forms in the world's languages in order to establish whether they have the same properties in all languages. Researchers do not acknowledge the degree of crosslinguistic variation of basic features of language and the difficulties involved in crosslinguistic comparison (Stassen 2011).

I am acutely aware of crosslinguistic differences in linguistic categories through my acquaintance with the Mayan languages. The passive is a good example of a construction that many researchers refer to as if it were a single linguistic category despite the variation that exists between "passives" in different languages (cf. Ambridge and Lieven 2011). The Mayan language K'iche' has two passive constructions. The first passive adds the suffix -x to derived transitive verbs (e.g., verbs with a causative suffix) and lengthens the vowel of root transitive verbs. The resulting verb stem has the intransitive forms of the status suffix. The first passive only allows third-person agents to be expressed in an oblique phrase, but not first- or second-person agents (Larsen 1988).

The second passive adds the suffix -taj to both root and derived transitive verbs. The resulting verb stem also takes the intransitive forms of the status suffix. There is a subtle semantic distinction between the two passives. The second passive emphasizes the resulting state of the patient or the successful completion of the action. It allows agents of all persons to be expressed in an oblique phrase, including first- and second-person agents.

England (1983:202–209) describes five passive forms for the Mayan language Mam. The -eet passive is used with most Mam verbs. Agents of all persons can be expressed in oblique phrases with the -eet passive of root transitive verbs, but not with the -eet passive of derived transitive verbs. The -njtz passive is only found on some root transitive verbs. The -njtz passive indicates that the agent has lost control of the action. It only allows third-person agents to be expressed in an oblique phrase. The -j passive has a function that is similar to the njtz passive, but is more productive. The -j passive is restricted to root transitive verbs. Agents of any person can occur in an oblique phrase with the -j passive. The -b'aj passive is used with both root and derived transitive verbs, but it only allows third-person agents to be expressed in oblique phrases. The -b'aj passive requires the use of a movement verb in order to express the motion taken by the agent to accomplish the action. The fifth Mam passive does not have an overt suffix and only allows third-person agents to be expressed in an oblique phrase. It is used when the agent is unknown or when the focus is on the patient.

The "passive" constructions of K'iche' and Mam illustrate the difficulties inherent to a crosslinguistic study of passives or any other linguistic object. The passives of both languages reference the distinction between root and derived transitive verbs, but even this distinction differs between K'iche' and Mam. A root transitive verb in K'iche' may correspond to a derived transitive verb in Mam and vice versa. The K'iche' and Mam passives also differ in the expression of agents as well as their contexts of use. The cover term "passive" confers a false sense of identity to crosslinguistic investigations of passives and their acquisition. I will present many examples of such correspondence failures in the following chapters. Categories such as "passive," "subject," "pronoun," and "bilabial stop" do not provide an objective basis for crosslinguistic research.

One of the major goals of crosslinguistic research on language acquisition has been to identify universals in the acquisition process. Knowledge that children can acquire any human language motivates the search for universal properties that make all languages accessible to children. The search for language universals is another manifestation of the monolingual orientation of the field. A universalist approach attempts to fit all acquisition results into one monolingual basket and disregards any results that do not fit the underlying assumption. The universalist perspective provides investigators with an excuse for ignoring results from other languages because they can claim a universal result by studying how children acquire a single language. Universalists take comfort in the belief that eventually discordant results from other languages will be reconciled with results from their own languages.

(Continues…)

---

Excerpted from "The Comparative Method of Language Acquisition Research"
by .
Copyright © 2017 The University of Chicago.
Excerpted by permission of The University of Chicago Press.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

---