Language and Linguistics
Ashok R. Kelkar


The Scope of a Linguistic Survey

The word ‘survey’ may go either a synchronic viewpoint (as in A Survey of farming methods in contemporary India or in 16th century Europe) or with a diachronic one (as in A Survey of English Literature from Chaucer to Milton).  A linguistic survey can likewise be of either kind—G.A.  Grierson’s Linguistic survey of India (11 volumes in 19 parts and 3supplements, Calcutta 1903-28) is an example of the first kind, while Ernst Pulgram’s The Tongues of Italy: Prehistory and history (Cambridge, Mass., 1950) is an example of the second kind.  In the present context of the proposed survey of the contemporary dialects of Panjabi at Punjabi University we are obviously concerned with a synchronic survey.  The adjective ‘synchronic’ defines the survey where the dimension of space is the one that counts.  ‘Space’ is to be interpreted rather liberally here to include both geographical area and social span.


            A diatopic linguistic survey then answers the following question in some detail: Which speech habits are found in whom in a given population?  Let us spell this out further.


            In the first place we must remember that a survey is never chiefly interested I individuals. It is more like a census.  Indeed, as we shall see later, one phase of it is simply a census of a rather specialized kind-a census at tells us how many there are of speakers in whom we find more-than-chance correlation between the speech habit in question and other characteristic such as geographical location, age and generation, literacy, sex, caste and class, and area of language use.  The speech habits isolated for the purpose of finding out this correlation may be either whole systems or specific traits.  In a gross survey we handle large language units (like standard Panjabi or Bangru or South Dravidian family) in the census.  Alternatively, in a trait survey we look for the distribution of phonological or grammatical or lexical traits like the use of tones or certain postpositions or a high percentage of English borrowings.


            So we can rephrase our question as follows: Which language or language traits are found well-correlated with which speaker characteristics or which areas of use in how many speakers out of the population under survey?  The whole survey operation will therefore resolve itself into three-sub-operations:

1.      Preanalysis of language isolates (whole systems or traits):

2.      Isoglosing or locating the isolates in the geographical area or social span; and

3.      Census or counting heads.

The results to obtained may in part be put in publishable form—say, maps or tables—or stacked away for future reference in the for of tapes, microfilms, and the like.  So we can add two more sub-operations:

4.      Presentation  or mapping and tabulating; and

5.      Archiving.

          Let us examine these five phases serially.

  1. Preanalysis of Language Isolates

In a gross survey the speech habits isolated may be of the following kinds:

(i)     The use of a given whole linguistic system L1 (a particular homogeneous dialect or variety);

(ii)    The use by the speaker of any one of a group of family of whole linguistic systems L1 or L2 or L3 ( a particular dialect family or language-family or language-group); or

(iii)  The use by the same speaker of each one of a group of linguistic systems L1 and L2 and L3 (typically the standard variety plus one non-standard variety or the prevailing language or link language).


An ordinary census in which dialect-families centring around certain standard varieties are returned is thus a very gross linguistic survey.


            In a trait survey (very often called a dialect survey rather than a linguistic survey when limited to a single speech community) the speech habits isolated may be of the following kinds:

(i)                  The use of specific feature T1 of phonology or morphology or syntax or vocabulary (as, a particular system of tones);


(ii)                The use by the speaker of any one of a group of competing linguistic features T1 and T2 and T3 (as the use of any tone system or the general preference for periphrastic tenses over tense inflections); or

(iii)                     The use by the same speaker of each one of a group of correlated linguistic features T1 and T2 and T3 that need not be all phonological or all grammatical or all lexical (as, the use of a specific tone system combined with the non-use of voiced aspirates or the non-use of honorific plural combined with the low frequency of dignified synonyms).

It will be clear that before the central operation of locating the social or geographical isolosses or correlations can be started the linguistic surveyor must have a clear notion of what he is going to correlate.  To this end the surveyor must

(i)                       familiaze himself with any available diachronic survey of the field and with the history of the phonology, grammar, and vocabulary of the languages involved;

(ii)                      draw lessons from any previous diatopic survey of the field and even of neighboring areas (it would follow from this that coordination and cooperation between the various linguistic surveys under way in India will be a highly desirable thing);

(iii)                    supplement this knowledge by undertaking a pilot survey of his own to determine which the likely directions are in looking for correlations;

(iv)                    familiarize himself with any available phonological, grammatical, and lexical descriptions of the languages and dialects in the field; and

(v)                     supplement this knowledge by undertaking a monographic survey of his own so as to have at his disposal a series of descriptive sketches in accordance with modern methods of a representative selection of language varieties (some linguistic surveys do not proceed beyond this preliminary or monographic phase).

At the conclusion of the preanalysis language isolates there will be ready the following:

(i)                 a list of language varieties with their genetic relationships and their mutual contacts thorough bilingualism duly noted;

(ii)                a list of linguistic traits to be embodied in the questionnaire and the formulation of suitable questions for this purpose (naturally such traits will tend to come from the overall pattern minus the common core shared by all the related varieties and from areas of structural uncertainty and freedom of choice rather than the basic and obligatory patterns in the language); and

(iii)               a list of the other characteristics of the speaker to be elicited in the questionnaire such as the geographical locations(s) of settled, migratory, and nomadic speakers, the age group (juvenile, adult, old), literacy (with the script and language in question also noted), sex, caste and ethanic group, religion, class status, and finally the area of use or function of the language variety or language trait (intimate circle, school, public life and mass media).


2.             Isoglossing of Correlations

The questionnaire is now ready in its final shape: preferably it has been preteste for its  for its efficiency and practically, and has been calibrated with questionnaires in neighbouring surveys for comparability of results.


            What goes into the questionnaire will be determined by the over all aim and limits of the survey-which may be a gross survey of a large population or a trait survey of a small population or a comprehensive survey.  The question may be direct, e.g.  which  variety do you use at home and among close friends?  Do you say [khatir] or [xatir]?  Or they may be indirect, e.g. Will you read this letter?  Can you make out what this man is saying this tape playback?  How would you name this object?  (This last may be a test for a phonological trait no matter what the respondent may think!)


            The exact shape and size of the questionnaire will be determined by the exigencies of administering the questionnaire in a satisfactory manner.  Thus the length and number of questions will depend on the number of respondents, the number of fieldworkers, and available time and money.  The knowledgeableness, ability, willingness, an prejudices of the fieldworker as well as the respondent have to be taken into account.  (It must be remembered that revenue officials and school teachers have been pressed into service as fieldworkers and as such are apt to do the responding for the “ignorant” speaker.) some of the questions will obviously have to be answered by the fieldworker himself, e.g. about the speaker’s sex or attitude to the language situation.


            The mode of administration of the questionnaire (which may include open question like: Tell me a story or Retell the story of the North Wind the Sun or of the Prodigal Son or Say the Lord’s Prayer) may vary.  The fieldworker can carry the questionnaire with him, conduct an interview with more or less strict adherence to it, and record the proceedings, i.e. replies as well stage directions (the respondent’s laughter or resentment, etc.) in his fieldnotes with or without tape recording. (It is never wise to depend on the recorder alone; notes are always a help even if they are in ordinary spelling and not systematized transcription.)  The so-called postal method employs the local man affiliated to the survey as a fieldworker who receives and returns the questionnaire by post.  The former method is costly and reliable.  The latter has obvious dangers, but may be forced by circumstances: the advantage is that a much larger sampling of different speakers can thereby be afforded.  The size of the sample is sometimes referred to as the “grid of the survey” (one respondent per so much population or area, for example).


3.            Census Determination of the Strength of the Correlation

             At one extreme we have the vast house-to-house government census which amounts to a very gross linguistic survey: the inclusion of bilingualism in the Census of India counts relatively as a refinement.  At the other extreme there may be the mini-survey of his class by a linguist under training.  The normal survey lies somewhere in between making its own compromise between level of analysis (gross and trait), depth of analysis (the amount and refinement of the data sought), and the size of the sample.


            Ideally after all the collecting sorting has been done the surveyor should be in a position to come up with statements in accordance with the following scheme:

(i)                 Language variety L1 or linguistic trait T1 is found co-occurring,

(ii)                (a)   With functional situation S1 (home, etc.) and/or

(b) With geographical location GL1 (district, point of the grid, etc.)       and/or

(c) With social location SL1 (age group, sex, literacy, etc. or a   combination of these)

(iii)              In so many respondents (out of the total so many sampled out of the group of speakers defined by (ii) (b) and/or (ii)(c).

When the representative character of the sample is questioned because of the size

or for some other reasons, this kind of statement yields information on the strength of correlation only in a limited way (known to be compatible not own to be compitable.


4.             Presentation

            The presentation of results has to be done with some care.  It can take many forms some of which may be fruitfully combined-maps, map-like diagrams and charts (for social location), tables, and sample responses illustrating linguistic varieties or traits.


            Errors that have crept into the designing of the questionnaire itself  cannot of course be rectified in the presentation.  Such errors may arise from the absence of any theory for that matter): pre-phonemic surveys thus often manage to pile a large amount of phonetic data without providing the answers to even simple phonologically interesting questions.  Or errors may arise from an insufficient prenanalysis: the decision of the Census of India to limit questions about a second language to languages of the Indian origin was thus an unfortunate one.  A careful pre-testing of the questionnaire in a small-scale pilot survey should be of great help in avoiding such errors.

            Errors can of course creep in at the time of the administration of the questionnaire.  A pilot survey can uncover deficiencies in the fieldworker’s training, unavoidable personal equations, and peculiar pitfalls in the field being covered that the fieldworker should beware of.


            Errors can creep in at the collection and sorting stage too.

            The least that a good presentation will accomplish is not to add its own quota of errors in transferring figures an data and not dressing up unrectifiable errors handed over from the previous stages.


Scientific honesty demands that the reader should be duly forewarned about possibilities of error (by a proper sprinkling of question marks, for example).


            Pre-testing of maps, charts, tables with potential readers for their clarity and perspicuity may be a good idea.  Maps can take a lot of money and be nice to look at but very painful to decipher.


A good presentation has to strike a balance between two opposite and somewhat conflicting requirements; it must leave the reader free to draw his own conclusions an apply his won statistical techniques an at the same time must not saddle him with mere  detail.  The do-it-yourself principle can be carried too far!   Enough should be left buried in the archives.


5.             Archives       

            The importance of preserving the records in a form accessible to any bonafide future worker            cannot be stressed enough in a country like India so ridden with red tape and so light-hearted about history.  Modern technology has provided us with the recorded tape and the microfilm or microcard as relatively cheap modes of preservation and multiplication serves both as an insurance against destruction and as a means to ready accessibility.


            SO FAR we have spoken the surveyor, as if the organization of a survey is onbe-man affair.  And there have been one-man surveys and good ones too-Andre Martinet did a survey in rather unusual circumstances as a prisoner of the Germans during World Wa II (La Prononciation  du francais contemporin : Tèmoignages recueillis en 1941 dans un camp dofficers prisonniers, Paris: Droz, 1945).


            The practical organization of a bigger survey has to contend, among other problems, with pressures arising out of linguistic and other loyalties that militate against the scientific objectivity and validity of the results of the survey.  The machinery can divide itself along the following lines:

(i)                 Bibliography and library;

(ii)                Historical monographs and surveys;

(iii)              Descriptive monographs;

(iv)              Designing of questionnaire and pilot survey;

(v)               Administration of questionnaire and collection and sorting of results;

(vi)              Preparation of press copy and seeing it through the press;

(vii)            Archiving; and

(viii)           Administration, publication, and public relations (with other surveys and with the public and authorities in the field under survey).

The foregoing discussion of the philosophy of a linguistic survey will have a

bearing on the working of each of these departments.



            This was presented at a seminar on Linguistic Survey Project, Patiala, July 1969 and published in Pàkha sanjam (Punjabi University) 1:1:5-11, 1969.  It was subsequently reprinted in Language Surveys in developing nations: Papers and reports on sociolinguistic surveys, Siriapi Ohannessian, Charles A. Ferguison, Edegar C. Polem, Centre for Applied Linguistic, 1975, p7-12.