How does one person use data from so many languages?

A question that often comes to mind about typological research, especially outside the

field of linguistics, is how can a single typologist or a small group of collaborators

acquire the data from the large number of languages in a sample? It is impossible for a

single linguist to know directly a hundred or a few hundred languages, or to consult with similar number of native speaker consultants. For that reason, typological research is necessarily built on the foundations of sound documentation of the world’s languages by native speakers or by field linguists and their native speaker consultants (Munro, this volume). Many typologists have also done fieldwork in particular languages, language groups or language areas. The severe endangerment of the vast majority of languages in the world has long been of great concern to typologists.

Typologists have used a variety of resources to obtain the data for typological

research. Perhaps the most important is a reference grammar of a language. A

surprisingly large percentage of the languages of the world have reference grammars,

though they vary substantially in breadth and depth of coverage and in quality. A

reference grammar is quite different from a pedagogical grammar: it is designed not for learning a language, but as a reference source for the major grammatical constructions of the language. A reference grammar is organized around families of constructions (and also the sound structure and morphological structure of words in the language), although this organization varies from grammar to grammar. A reference grammar presupposes to a greater or lesser extent a knowledge of basic linguistic terminology and a knowledge of the language itself (for example, many reference grammars give examples only in the script of the language).

In order to use data from a language that one does not speak or has not studied in detail, a linguist must be able to analyze the structure and the meaning of phrases or

sentences in the language. The meaning is generally provided by a translation of an

example into English (or French or Spanish, or whatever is the language that the reference grammar is written in). The structure of the example is ideally provided by a

separate line by the author of the reference grammar, sometimes referred to as the interlinear morpheme translation (IMT). Examples of IMTs can be seen in examples (4)- (9) above. An example of a sentence from Mokilese, an Austronesian language of

Micronesia, with an IMT is given in (10).

(10) mine woaroa-n woal-o war

exist CLF-3SG man-that canoe

‘That man has a canoe.’

The last line, the gloss, gives the meaning of the Mokilese original sentence as a

whole: from the gloss, we can see that this is an example of the predication of possession, to use the technical linguistic description of the function of the sentence. The middle line, the IMT, glosses each morpheme of the original, including affixes and other meaningful internal word changes, following relatively widely used conventions. The IMT therefore provides the structural analysis of the sentence: the order of elements (e.g., the verb comes first in the sentence); which elements are independent words and which are not (e.g. ‘that’ is a suffix on ‘man’); and the presence of special grammatical elements (e.g. the possessive classifier for the man’s canoe, and the person-number inflectional suffix on the classifier). The IMT also shows how differently the function of predication of possession is expressed in the language: in this case, a more literal translation of the Mokilese sentence would be ‘That man’s canoe exists’.

The presence of IMTs allows a linguist to rapidly recognize the grammatical structure

of the language in the description even if the linguist has not previously studied the

language. IMTs are used widely in contemporary reference grammars. Many grammars, especially older ones, do not use IMTs, and therefore a linguist must study the reference grammar (and a dictionary if available) in greater detail in order to analyze the structure of the constructions of interest in the language.

Reference grammars vary significantly in the constructions that they cover, and the

depth in which they cover them. A typologist may discover that some (or many) of the

reference grammars do not describe the construction s/he is interested in. One may need to accept that there will be gaps in the data in the sample (or one will replace the

language in the sample with another language possessing a grammar with the relevant

description). More recently, typologists have developed questionnaires to obtain data for particular constructions which are typically poorly described in reference grammars. One of the first questionnaires used a set of 200 sentence contexts to elicit tense-aspect constructions; the questionnaire was distributed to native speakers or specialists in 65 different languages (Dahl 1985). More recently, experimental elicitation techniques have been used to avoid the problem in questionnaires of translation from the language of the questionnaire. For example, a set of 71 pictures of different spatial scenes was used to elicit constructions for spatial relations in nine diverse languages (Levinson et al. 2003), and a set of 61 video clips of different cutting and breaking scenes was used to elicit different verbs of cutting and breaking in 28 languages (Majid et al. 2007).

Questionnaires and experimental elicitation techniques have the advantage of

allowing a linguist to directly elicit the constructions expressing the function s/he is

investigating. However, designing questionnaires and elicitation strategies is extremely difficult. Also, questionnaires and elicitation, while valuable in themselves, can benefit from a broader description of the language than just the constructions under investigation.

Many times a typologist discovers an unexpected correlation of constructions. For

example, one might not consider the English conditional construction in (11) and the

topic construction in (12) to be grammatically related:

(11) If he comes, I will stay.

(12) (As for) me, I will stay.

But in a number of languages, including Hua, a Papuan language of New Guinea, the

antecedent of a conditional and a topic are marked grammatically in the same way (in the case of Hua, with the suffix -ve; Haiman 1978: 570-71):

(13) e-si- ve baigu-e

come-3SG.FUT- ve will.stay-1SG

‘If he comes, I will stay.’

(14) dgai-mo- ve baigu-e

I.EMPH-CONN- ve will.stay-1SG

‘(As for) me, I will stay.’

This grammatical connection is at first surprising, although an explanation is

forthcoming: the situation described by the antecedent of a conditional provides the

framework for evaluating the consequent that follows it, and the referent described by a topic phrase constitutes the framework for the following discourse (Haiman 1978: 577-86). This grammatical and semantic connection would probably not have been detected if one constructed a questionnaire for conditional constructions; but it might be observed if one consulted reference grammars for a wide range of languages, and noticed the

similarity of form. Thus reference grammars continue to have great value for typological research.