A favorite method, especially of linguists gathering data about a little-known language in fieldwork, but also by questionnaires, is the example-translation method. Here, the linguist produces an example sentence in the lingua franca (the language he shares with the informant), the informant translates it into the target language.

While this method has produced too many useful results in the history of linguistics to be discarded once for all,1 it is beset with all kinds of methodological problems that cannot be discussed in full here. One fundamental methodological point must be clear: The question cannot possibly be how the informant can imitate, in his language, some structural trait of the lingua franca. Instead, the purpose of the method is to find out how he would conceive, structure and communicate a certain idea and whether this involves dedicated grammatical categories of the language. For the method to work above some sensible threshold of reliability, it must meet the following requirements:

  1. The example-sentence is not isolated, but put in a natural context. This may be a linguistic context, i.e. a set of sentences preceding the target, or a non-linguistic context, i.e. a situation in which the target is uttered. This method was first elaborated in the form of questionnaires on tense, mood and aspect (s. below).
  2. The research question does not depend on structural properties of the lingua franca. Quite on the contrary, it is safe not to use these in the presentation of the example. Instead, it is the specification of the context which nails down the semantics of the category sought.
  3. Needless to say, suggestive questions of the linguist (“and isn't this alternative version possible, too?”) are illicit.
  4. All the reactions of the informant are tape-recorded. The linguist is too liable to focus just on that reaction which he expects or happens to be currently interested in.
Requirements #1 and #2 are met by the questionnaires on tense, mood and aspect. In the English version of the questionnaire, the category would appear in its English grammatical manifestation, but that is suppressed by presenting its host word as a mere vocable, without any grammatical categories and, in particular, without any hint to the grammatical category being tested. In the domain of Design of Situations, subdomain of Telicity, the questionnaire checks the expression of imminential progressive. In the example of Q3, this is presupposed as a functional category, and it is asked which structural category the target language uses to express it.

Q3. Progressive, imminential meaning (Dahl (ed.) 2000:813)
S56.Hurry up! The train leave.
S57.The old man die, but finally they found the right medicine.

In the same functional domain, the question may be whether the language makes a grammatical distinction between such dynamic situations which are bounded and such which are unbounded. The following examples illustrate an attempt to determine tense and aspect use in temporal clause constructions whose taxis and time reference is set by the preceding context:

Telicity (Dahl 1985: 202)
(101)[Last year, the boy's father sent him a sum of money.]
When the boy get the money, he buy a present for the girl.
(102)[The boy used to receive a sum of money now and then.]
When the boy get the money, he buy a present for the girl.
(103)[The boy is expecting a sum of money.]
When the boy get the money, he buy a present for the girl.

If this method detects an aspectual category related to boundedness, it may then be named ‘perfective vs. imperfective’ or ‘plain vs. progressive’ or ‘completive vs. incompletive’, corresponding to diverse shades in function and descriptive traditions.

It is clear that just tense and aspect are categories which can only be understood in suprasentential contexts. However, the same goes, to a greater or lesser degree, for all linguistic signs. Even seemingly simple lexical items of concrete meaning do not have a direct translation equivalent. S. the page on semantics-driven data collection.

1 “If what we have said here is true, one may ask how anyone could be so stupid as to choose translations as a basis for an investigation of language use. The simple answer is that this is the only realistic method for large-scale data collection in typologically oriented linguistic research. We simply have to accept that it is unreliable and try to use the data with the necessary care.” (Dahl 1985:50)