American Anthropology and Company: Historical Explorations

Chapter 2: The Manufacture of Linguistic Structure

22 2 The Manufacture of Linguistic Structure The human intellect, from its peculiar nature, easily supposes a greater order and equality in things than it actually finds, and while there are many things in Nature unique, and quite irregular, still it feigns parallels, correspondences , and relations that have no existence. —FRANCIS BACON, Novum Organum (1620: L:xlv) The observer who sets out to study a strange language or a local dialect often gets data from his informants only to find them using entirely different forms when they speak among themselves. . . . An observer may thus record a language entirely unrelated to the one he is looking for. —LEONARD BLOOMFIELD, Language (1933:497) Classic sociolinguistic work on variation and the permeability of speech community boundaries (Gumperz 1962, Labov 1972) made comprehensible what earlier generations of linguists regarded as “error” deviating from clearly delineated structures of a language. To properly use texts elicited and recorded by hand—in many cases the only records of languages no longer spoken—requires hermeneutic attempts to recover the intent of the recorders (linguists) and informants (not all of whom were native speakers) in the original elicitation context, as well as the problematic(s) motivating the work and the theories through which data were filtered and dictations were heard. The kinds of caution needed to use work based on classical elicitation procedures can be clarified by considering what was typical in the best elicitation-based (white room) fieldwork, such as that of Li Fang-Kuei (1902–87). The fundamental lessons from the first generation of sociolinguistics (see Murray 1994b, 1998)—that context matters to the patterns of language use and that the world is not split into hermetically sealed units— calls into question the empirical status of linguistic structures derived Manufacture of Linguistic Structure 23 from classic elicitation from a single informant when those analyzing unwritten languages believed that “text reveals structure” and that any fluent speaker would do, because there was a single structure to each language (and structural descriptions approximated “God’s truth” rather than the “hocus pocus” of alternative structures, to use a famous distinction made by Fred Householder [1951]). First, before the advent of portable sound-recording equipment, linguists necessarily slowed down their informants’ speech to a pace at which they could write in phonetic script what informants said. This unnaturally slow pace and the focus on the transcriber hearing distinctly led to careful pronunciation. The unnaturally slow pace also provided informants opportunities to reflect on how they should speak—in contrast to how they usually spoke unselfconsciously. At a pace somewhere between a morpheme at a time and truncated clauses, distinctions not usually made may be made (Samarin 1967:54; Blom and Gumperz 1972:415) and syntax and phonology may be “cleaned up,” because extraordinary attention is being paid to how the informant is speaking (see the systematic differences found by Labov 1972). Second, as the codifier of American structuralist linguistics noted, “The difficulty or impossibility of determining in each case exactly what people belong to the same community is not accidental, but arises from the very nature of speech communities” (Bloomfield 1933:54–55; also see Gumperz 1962; Blom and Gumperz 1972); although he, and still more his followers, set out to chart idealized structures of particular, clearly bounded languages. When Chomsky (1965:3) asserted that the task of linguistics was to consider the language of an ideal speaker-hearer of/on a perfectly homogenous, monolingual speech island, he was not suggesting anything new but making explicit what was common practice in Geneva , Prague, Copenhagen, or New Haven structuralisms. Many of his predecessors and followers dealt in similar idealizations, while writing as if they were describing observable realities (“God’s truth” analyses). The Descriptivist Paradigm Franz Boas attempted to reconstitute linguistic work on Native American languages with inductivist fieldwork done by those under his supervision and mostly trained by him. Boas was a “splitter” rather than a “lumper.” He was especially eager to leave behind any talk of “primitive language” in general, stressing diversity between languages. Boas paid 24 Manufacture of Linguistic Structure lip service to description being preliminary to comparison. Although the languages described in the first volume of the Handbook of American Indian Languages (Boas 1911) were chosen “to facilitate comparison of different psychological types in a single broad geographical region,” the time for comparison had not arrived in 1911 (or by 1943, when Boas died), and “the...