1.

From Sampling to Publication

1.1 Selective Sampling

1.11 Introduction

Selective sampling in accordance with a stated plan is a seientific method devised for the purpose of achieving a general view of a complicated situation within a reasonable time. It is used to some extent in all scientific research, as in topographical and geological surveys, in ecological studies of animal and plant life, in delineating economic areas, in testing public opinion, and so forth.

In the study of living speech it was first employed on a large scale by G. Wenker (1876ff.), who sent his 40 test sentences to all the public schools in Germany (ca. 40,000) with the request that they be translated by a native speaker of the local dialect. Two decades later (1897) J. Gilliéron sent E. Édmont out into the field with a set questionnaire to sample the folk dialects spoken in 639 communities of France. With certain modifications and refinements, the later large-scale surveys of Italy, Spain, Romania, the United States, and England follow Gilliéron’s plan of “direct” recording by properly trained observers.

The problems connected with the planning and the execution of a survey of this type are taken up below. Here we shall briefly consider its purpose, character, and potentials.

By choosing a limited body of linguistic items for investigation in a limited number of carefully selected communities, each of them represented by a single speaker belonging to a certain social class, or by one for each of two or more social levels, the area linguist hopes to obtain a general view of the dialectal structure of the total area within a relatively short time. He has no illusion about achieving a complete coverage of usage. He knows that linguistic features other than those that he selected may also vary, that the inclusion of additional communities may turn up unobserved variants, and that his informants are only approximately representative of their communities or of a social or age group living there. He is fully aware of the fact that the adequacy and the reliability of his sample is proportionate to the relative complexity of the linguistic and sociocultural situation in all or in parts of the area surveyed. Thus he will reserve judgment and recommend further investigation when the linguistic variants exhibit a complicated and apparently erratic dissemination as in certain transition zones or relic areas.

The area linguist will cheerfully grant that the method of systematic selective sampling has certain limitations. But he is on solid ground when he claims that no other method of gathering reliable information on usage is equally efficient and productive. Properly planned, a sampling survey will provide a general view of the dialectal structure of an area on the basis of comparable data recorded in a consistent fashion. Areas of uniform and of divided usage, clear lines of demarcation, and transition belts emerge to furnish the basis for establishing major and minor dialect boundaries, which reveal the dialectal structure of the total area surveyed.

It is of some importance to emphasize the fact that the findings of a linguistic survey carried out on this plan constitute a census of usage for a given period of time, however incomplete it may be. As such it cannot be superseded by a later survey any more than the Census of the United States for 1970 has replaced that of 1960 or 1900. Any new survey, say a generation later, will reveal changes that scholars will be able to trace in considerable detail by relying upon the data furnished by the earlier investigation.

A first sampling survey has further values and functions. It provides a framework for dealing with the usually uneven or fragmentary evidence of earlier studies and of unconventional spellings in better perspective, and it points to the need for more detailed investigations of certain areas or of specific problems. For such studies, the incomplete or insufficient data presented by the general survey inevitably contribute to the formulation of a well-oriented and economical plan.

1.12 Constructing the Questionnaire

A set questionnaire, one that prescribes the recording of specific items and the manner of securing responses, is an indispensable tool of systematic selective sampling. Comparable data from community to community and from speaker to speaker, on the basis of which the dialectal structure of the area must ultimately be determined, cannot be secured in any other way. Constructing a questionnaire adequate for the purpose in hand, neither too skimpy nor too full, and so arranged and styled as to facilitate the interview, is a complicated and time-consuming task. But care and time spent on its preparation are richly rewarded in the end.

Questionnaires can of course be set up for limited purposes, such as the determination of regional or social differences in the lexicon, in verb forms, in phonemic contrasts, or in the articulation of shared phonemes. Choosing promising items and presenting them for the convenience of the investigator and the informant always presents certain problems. Here our concern will be with procedures used in constructing a general questionnaire intended to cover the regional and/or social differences in the lexicon, the morphology, and the phonology in a given language area, the type of questionnaire used in the surveys of France, Italy, Spain, Romania, Switzerland, England, and the United States.

The choice of items to be included in a general questionnaire rests upon available information, however fragmentary or ambiguous, concerning regional or social differences within the area to be surveyed.

As a first step, variations in vocabulary, in grammatical forms, and in pronunciation reported in dialect dictionaries, regional word lists, dialect grammars, phonological studies, etc., are assembled in separate files. Then the lexical items are grouped by semantic fields, the morphological by grammatical categories, and the phonological on the basis of a tentative phonemic analysis of one of the better known dialects or from the diachronie point of view. As a third step, additional examples to illustrate known morphological and phonological differences are chosen from the vocabulary of everyday life, especially such as more or less naturally fall into the semantic fields represented by the lexical items. This is of special importance in phonological matters, since phonemes shared by all or most of the dialects in the area are rarely reported in the earlier literature. The survey must of course cover this aspect of phonology as well as the reported or anticipated variants. Moreover, words whose phonemic shapes are invariable are needed to exhibit phonic differences in shared phonemes. Taped free conversations from a number of well distributed points in the proposed network of communities can furnish invaluable leads for the selection of representative phonological items.

After an ample collection of “promising” items has been assembled on slips and tentatively arranged by semantic fields, the task of producing an effective questionnaire of manageable size can be faced. This involves (1) a rough forecast of the extent and kinds of linguistic diversity within the area, (2) an estimate of the time required to find and interview an informant, and (3) a rough calculation of the number of communities and informants needed for effective regional and social representation. Last but not least, a hard-headed estimate (4) of financial requirements and resources for carrying out the survey, and (5) of available man power for completing the fieldwork within a reasonable time must be brought to bear upon this problem.

In view of all these uncertainties, a questionnaire that is both adequate and not unduly long is hard to achieve, especially for a first survey. Nevertheless, maximum coverage by the shortest possible questionnaire must be the aim. If this is to some extent conjectural—a leap into the dark, if you will—֊such a questionnaire will bring to light much information that future investigators can use in constructing more effective questionnaires, and will prepare the way for detailed investigation of subareas revealed by the initial survey.

Before the questionnaire is put into final shape, it should be tested in a number of key points within the area. Even limited field experience will disclose unproductive items and some that can be elicited only with difficulty. These should be eliminated. It may also suggest some important additions. Above all, it will provide a measure of the time required for completing a single interview and give some indication of the informants’ endurance and toleration of prolonged questioning. Allowing for a speedup of fifty percent that may result from field experiences, the questionnaire should then be trimmed down to manageable size.

In addition to the normal questionnaire, a reduced version containing the more important phonological or morphological items is sometimes useful for recording the usage of auxiliary informants, especially those that cannot spare the time for a full interview.

Supplementary lexical questionnaires dealing with regionally restricted enterprises, such as cotton farming, viticulture, or dairying, can be set up to supplement the lexical fields covered by the normal questionnaire.

The arrangement of the items in the questionnaire is of crucial importance for the speedy progress of the interview and for the purpose of securing trustworthy responses. It is not the linguist’s interest in the problems confronting him, but the situation in which the informant finds himself that is the controlling factor in achieving a satisfactory arrangement.

The speaker of a dialect is willing to talk about what he knows and does. He is usually not interested in what he calls a thing or an activity. Hence the questionnaire must be so arranged that the informant can talk consecutively about such topics as the dwelling, farming, the weather, the lay of the land, plants and animals, the family, social activities, and so forth. The investigator must do his best to keep the attention of the informant focused on the subject matter, even when he asks questions that must seem trivial or trying to a native speaker.

There will always be some morphological and phonological items of importance that do not fit neatly into any of the lexical fields covered by the questionnaire. If not observed incidentally in the course of the interview, responses should be elicited after the informant’s interest has been aroused and his confidence gained.

In a questionnaire arranged by semantic fields, the very sequence of the items serves to narrow the choice of possible responses to the several items. Further definition of the response sought can be achieved by providing a lexical context or suggesting a specific situation verbally or by gesture. For some items illustrations or photographs can be used to cue responses. Some questionnaires prescribe a fixed phrase or sentence into which the informant is asked to insert his response; but most of them leave the precise method of eliciting trustworthy responses to the ingenuity of the trained investigator. Either method has its advantages and drawbacks.

Examples of the types of comprehensive questionnaire described above will be found in Jaberg-Jud 1928: 145–74 (Italian), Kurath-Bloch 1939: 147–58 (American English), Orton 1962: 49–101 (English Dialects), Hotzenköcherle 1962 C: 1–78 (Swiss German). These publications also contain comments on the problems involved in constructing such questionnaires. See also Pop 1950: 1136–41.

1.13 Conducting the Interview

Successful fieldwork presupposes not only adequate training in linguistics, ranging from phonology and morphology to semantics, and a general knowledge of the culture of the area in its varied manifestations, but also a sympathetic understanding of people in all walks of life. Unless the investigator has the knack of dealing with personalities of all kinds so as to gain their confidence, of guiding them gently through the inevitable “dry” stretches of his questionnaire, of humoring them when their interest lags, he will experience serious difficulties and disappointments. Even the best-trained linguist can be a failure in fieldwork.

The interview is channeled by the questionnaire from topic to topic. However, the investigator is free to start the interview with any of the semantic fields provided for: with housekeeping and cookery, with farming, with social activities, etc., depending upon the informant’s special interests or competence. Numerals, names of the days of the week and the months, and some other lexical fields that are apt to strike the informant as too trivial to bother with sometimes get the interview off to a bad start. Though they offer an excellent opportunity for “sizing up” the phonological characteristics of the speaker, it is often better to do them after his interest has been aroused.

The questionnaire is normally drafted in the standard language, or a widely used dialect, of the area to be investigated, supplemented by regional terms as needed. In the Western world this language or dialect usually serves also as the medium of eliciting responses from the informants, most of whom are bidialectal or at least understand the standard language, even if they have no facility in using it actively. There is some danger in this practice, unavoidable as it may be, especially if the dialect is close to the standard (as in American English), or if the informant is given to “mixing.” But an experienced fieldworker can rather easily detect such intrusions and query the speaker. In view of the well-known fact that in Europe and in America regional expressions of folk speech have constantly been replaced by terms current in the standard language, particularly since the establishment of public schools, such “intrusions” often turn out to be fully established in the speaker’s dialect. Antiquarian interest on the part of the investigator must not be allowed to obscure such facts. Often enough the informant will report that his parents “used to say” so-and-so, but that he himself has given up his childhood usage.

Peculiar circumstances may favor the use of a regional dialect as the cue language, whether that of the informant or not. Thus Hotzenköcherle [1962-A: 30-31] states, in substance:

A similar situation may obtain in other parts of Europe, say in Scotland or Ireland. Even in parts of the eastern United States some fieldworkers found it advantageous to adapt their usage more or less to that of the informant, but without mimicking him outright.

The primary task of the fieldworker is, of course, the accurate observation and recording of the responses he elicits from the informant. Responses given freely, without hesitation or qualification, are taken down without comment, whenever the investigator feels confident that they represent normal usage. For responses offered hesitatingly, with amusement or reserve, or with any significant comment, as well as for those secured under special circumstances, the investigator must provide appropriate information. Such comments, invaluable to the editor, should be entered in a separate column set aside for this purpose.

The fieldworkers of the Linguistic Atlas of the Eastern United States were instructed to identify all such “limited” and “qualified” responses in ways that proved effective [Kurath-Bloch 1939: 143- 45]. Since they may be helpful to others, the abbreviations and labels used for this purpose are listed below:

c.: an expression observed in conversation, esp. when it differs from that elicited by questioning

cr.: a spontaneous correction of the first response, whether real or an attempt at elegance

r.: a response repeated at the fieldworker’s request

f.: a forced response, i.e., secured by repeated questioning

s.: a suggested response, i.e., a term or form actually pronounced by the investigator and recognized by the informant as his own

(:) preceding the recorded response marks hesitation, whatever the reason

(!) registers amusement, whatever the cause

(?) reports the observer’s doubt concerning the trustworthiness of the response

(†) marks an old-fashioned or remembered expression

(→) identifies a recent expression if actually used by the informant

(┴) marks an expression reported as heard by the informant in his community but not used by him

(*) marks an expression offered by an auxiliary informant, usually a member of the family or a local friend who happens to be present during the interview.

The identification of responses not secured by direct questioning and the indication of their status in the informant’s usage furnish important information to the editor of the collections and to the scholar who undertakes to trace shifts in usage.

1.14 Recording the Speech Sounds

A decision regarding the method of recording the sounds of the dialects has to be made at the very beginning of the field survey. Both scientific and practical matters have a bearing upon this problem.

As to the scientific aspect of the problem, it is important to keep in mind that the dialects may differ (a) in the system of phonemes, (b) in the lexical incidence (etymological distribution) of shared phonemes, or (c) in the phonic manifestation of shared phonemes from place to place and from class to class (regional and social diaphones). Moreover, (d) the phonemes of any given dialect may exhibit positionally or prosodically conditioned allophones.

Though of uneven value from the structural point of view, phonemic, incidental, diaphonie, and allophonic phenomena all have their relevance in dialect research, whether synchronic or diachronic. Without adequate phonic data no single dialect or idiolect can be adequately described; nor can its historical relation to its sister dialects or its parent dialect be traced in realistic fashion [Kurath 1961].

If that be granted, as it must be, the question is simply this: How can a full record of relevant phonological information be best obtained under the conditions confronting the fieldworker charged with the task of eliciting and recording a large number of lexical, morphological, and phonological responses in a limited time? Could he reliably, or even tentatively, establish the phonemic inventory and the phonic range of the individual phonemes, say, in the first two hours and thereafter record purely phonemically? But by the time he could start phonemic recording most dialect speakers would have bid him “goodbye.”

Under the circumstances, strictly phonic recording imposes itself and has been adopted in nearly all major dialect surveys since the turn of the century. Its feasibility cannot be questioned. The requirements are (1) a finely graded and flexible system of phonic notation adapted to anticipated variations within the area, such as those used in Italy [Jaberg-Jud 1928: 24-36], in the United States [Kurath-Bloch 1939: 122-43], and in German־speaking Switzerland [Hotzenköcherle 1962-B: 79-86]; and (2) intensive training of the fieldworkers in the application of this system, preferably under field conditions, real or simulated.

The task of recording phonically, i.e., of writing down all audible shades of sounds by means of a finely graded system of notation, is not as demanding upon the investigator or as cumbersome as a casual inspection of the records might suggest to the uninitiated. Any experienced fieldworker can testify to this fact, and add that the very fullness and flexibility of the notation system often relieved him of time-consuming decisions.

The obligation to record all responses phonically does not prevent the field observer from commenting upon actual or potential structural implications of the phones he records, of pointing out the real or apparent existence of certain phonemic contrasts in connection with specific examples. He should, indeed, be encouraged to report such potentially structural phenomena in his general characterization of the speech of each informant after completing the record.

The task of establishing the phonemic structure of idiolects and dialectal types, and of describing the phonic range of the phonemes, can safely be entrusted to others. If the record is accurate and ample, there is no difficulty in making reliable decisions. If it is inadequate in one respect or another, a decision may have to be postponed. In any event, careful tabulation and leisurely analysis of the phonic field data are more apt to lead to reliable determinations than improvised decisions in the field. To the analyst, field experience is of course a helpful asset in working out the phonemic system(s).

The procedure to be followed in establishing the phonemic system(s) on the basis of phonically recorded field data is illustrated in Kurath-McDavid 1961 (see especially pp. 1–9 and 31–100). Its feasibility and effectiveness cannot be questioned, as W. G. Moulton has demonstrated in an impressive series of studies based upon the wealth of phonic data furnished by the Sprachatlas der deutschen Schweiz [Hotzenköcherle 1962]. After establishing the phonemic system of idiolects and regional dialects and describing the phonic range of the several units of the systems, he undertakes (1) to construct maps of regionally varying phonemic systems [1960, 1967 1968]; (2) to show that phonic and phonemic maps of the same underlying phonic data may exhibit divergent regional patterning [1963, 1964]; (3) to demonstrate that for dealing with diachronie problems a phonemic analysis of the phonic data is an essential prerequisite [1960, 1961]; (4) to show that the phonic range of a phoneme depends upon the neighboring units in the system [1960, 1962]; and (5) to test A. Martinet’s theory [1955] that asymmetries of vowel systems tend to be mended by inner drives [1961].

Quasi-phonemic field recording was used by G. Bottiglioni in his survey of Corsica [1933-42]. Rejecting “impressionistic,” i.e., phonic, recording outright, his aim was to establish “mean” or “average” usage in phonological as well as morphological and lexical matters. Probing the informant’s usage repeatedly and even consulting other speakers in the community, he hoped to achieve an “objective” determination of “normal” local usage. Since he knew the area well and could spend from four to eight days with each informant, his findings would seem to have considerable validity, though involving a subjective judgment on his part. To what extent this method can be adapted to limited regional surveys—especially with the necessary refinement in phonological matters—is an open question. See S. Pop’s description and criticism of Bottiglioni’s procedures [Pop 1950: 539–40, 546–47, 553–55].

It goes without saying that for specific purposes the phonemic approach is not only possible but convenient, if not imperative, in field recording. The existence or absence of specific phonemic contrasts or of positional allophones in other dialects can be traced, if the phonemic structure of one of them is known. Morphological and lexical studies call for a speedy progression from phonic via quasi-phonemic to phonemic recording. Conversations and texts should of course be taken down phonemically as soon as a reliable analysis has been worked out.

1.15 Choosing Communities

In a far-ranging general survey, observation posts must be carefully selected to achieve a general view of the dialectal structure of the area within the limits imposed by available resources. It stands to reason that all available information concerning the living conditions of the population, the social structure, and the sociocultural history of the area should be brought to bear upon the problem of establishing a network of communities. Our accumulated knowledge about language behavior in the vicinity of dominant cultural centers and about the effects of barriers to communication furnish additional leads. A well informed choice of listening posts will yield a rich harvest. The principles guiding the choice of communities should, of course, be stated, so that the findings can be evaluated critically.

A general survey should include the larger cities, communities of medium size, and villages; old and derivative centers and settlements; points along highways and waterways as well as secluded hamlets and homesteads. Since the selection is made in reference to the sociocultural characteristics of the area under investigation, the choice of types of communities will vary from area to area.

Until recently large-scale surveys have been deliberately restricted to folk speech, especially to that of the countryside. A. Griera [1923 ff.] was the first to include the folk speech of urban centers in his survey of Catalonia, and this practice was followed by Jaberg-Jud in their Italian atlas [1928: 187-88]. In the Linguistic Atlas of the United States, all population centers of any importance are regularly included [Bloch 1939: 39–40] and, in principle, all social levels are represented [Kurath-Bloch 1939: 39–44].

Practical considerations inevitably limit the number of communities that can be investigated. Financial resources, available manpower, and time limits of one kind or another always impose restraint. Nevertheless, any well-designed network of communities will bring the main features of the dialectal situation into relief and prepare the ground for properly oriented intensive investigations in any part of the area.

In practice, the organizer of a survey plans the network of communities, assigns priorities to desirable listening posts, and makes his final selection after he knows how much time he must allow for the completion of an interview. The number of communities, the number of informants per community, and the length of the questionnaire must be brought into balance for optimum results of a sampling survey.

Though the network of communities should be planned beforehand, the fieldworker can be given a choice between certain communities to facilitate his labors. Often enough he can secure helpful information locally to make a wise choice; again and again he will find easier access to informants in one or the other of the alternate communities.

Each community represented in the survey should be briefly characterized in the manual accompanying the published records. Its natural setting, contacts with other communities (especially with market towns), size and population trends, chief occupations of the inhabitants (farming, industries, etc.) should be pointed out. Information concerning the history of each community, with bibliographical references to works consulted in making the selection, should be provided for the convenience of those who will undertake to evaluate and to interpret the linguistic records in their sociocultural context. For examples of such sketches see Jaberg-Jud [1928: 37–139] and Kurath-Bloch [1939: 159–240].

1.16 Choosing Informants

Language is the chief medium of communication between the members of a community. To fulfill the function of communication, the usage of the individuals constituting the community, or a social class within the community, must be relatively uniform. This well known fact underlies the assumption that any native speaker’s usage is in a large measure representative of the speechways of a social or age group in his community. The area linguist shares this assumption with all students of language who base their descriptions upon the usage of one or several speakers. This practice can hardly be questioned.

A further consideration has a bearing upon the essential reliability of the informant’s usage as representing that of his class in a given community. What the investigator observes and records are ingrained habits of speech of which the speaker is largely unaware. If he offers opinions about the way he talks, the observer will note them as such. He will have ample opportunity to determine whether the informant practices what he preaches. All in all, the sampling of speech habits is simpler than the sampling of opinions in social science and yields more reliable results.

The informant chosen to represent his community, or a social or age group in it, should have certain personal characteristics, some of them indispensable, others desirable. First of all he must be intelligent (even if illiterate), communicative (but not talkative), and tolerably well informed on the topics covered by the questionnaire. He must have enough interest, patience, and endurance to put up with prolonged questioning. He should, of course, be free of speech defects and hear well enough to understand the questions readily. Set ways and a degree of self-confidence are definite assets, as is community pride.

Since the informant is chosen to represent his community, or one of its social or age groups, biographical data must be secured from him or from reliable consultants living in the community. He should be a native of the community (preferably also his forebears) and have lived there all or most of his life. Schooling confined to the locality or district is a requirement for middle class and folk speakers.

This information must be presented in the manual accompanying the published records, along with data on the informant’s age, sex, schooling, occupation, membership in religious and other social organizations, standing in the community, special interests, and reading habits. Moreover, the field investigator should be encouraged to comment on general characteristics of the informant’s speech, such as tempo, precise or slack articulation, stability or fluctuation in usage, his attitude concerning the speech of others, inclination to “improve” his usage, etc. For examples of such sketches see Jaberg-Jud [1928: 37–139], Kurath-Bloch [1939: 159–240], and Hotzenköcherle [1962–B: 97–174].

Finding suitable informants to represent the chosen communities calls for skill, tact, and sound judgment on the part of the fieldworker. He may have some information about the character of the community before he arrives, even about some of the prominent families living there. He may have a letter of introduction to a local historian or some other person who may be willing and able to help him find the kinds of informant(s) he needs. He may need a letter of identification to present to local officials. After some experience in the field, he may discover that informal contacts in the general store, barber shop, or local tavern can provide him with useful leads. The approach must, of course, be adapted to regional or local customs and sensibilities. Appeal to local pride or reference to sponsorship of a learned organization may be helpful in some communities and with some informants. Assurance that the results of the interview(s) will be held in confidence must at times be given. Procedures and experiences connected with the search for suitable informants have been discussed by two experienced field investigators, P. Scheuermeier [1932: 99–105] and B. Bloch [1935: 4–6].

The problem of selecting and securing suitable informants is essentially the same whether the sampling survey is limited to one social level or includes several of them. In Europe the practice has been to confine the survey to the speechways of the folk, and to give prominence to the oldest living generation in rural communities A predilection for historical problems, the hope of shedding light on processes of linguistic change by observing the linguistic behavior of the folk, and admiration for the soil-bound “ethos” or “world view of “natural” people have been the motives and the justification offered for this practice. Much has been accomplished within this limited perspective since the turn of the century. To advocate the application of selective sampling to the usage of the middle class and the cultured does not detract from the value of the work that has been done. It rather implies the recognition of the effectiveness of this method in dealing with living speech of any kind.

Deliberately deviating from European practice, the Linguistic Atlas of the United States included the investigation of the speechways of several social levels from the very beginning (1930). Representatives of the cultured, the middle class, and the folk are systematically included in the survey, the last two types in nearly every community chosen for investigation, the cultured in about one fifth of them (chiefly in cities of some importance). This plan rather imposed itself owing to the fact that, except for some of the old cities on the Atlantic seaboard, American society lacks clearcut social classes. There is, instead, a gradation from level to level resulting from the social mobility of the American people. By sampling the usage of the social extremes—the cultured and the folk—and that of the middle group(s), a broad conspectus of cultivated, middle class, and folk speech is secured region by region. Identification of the sociocultural status of the informants is, of course, provided. A tabulation by social levels and age groups [Kurath-Bloch 1939: 41–44] enables the student of the field records to choose for investigation or cartographic presentation the usage of any type of informants or to trace social differences from community to community.

It would be unwise to assume that a full coverage of the social dissemination of variants is to be found in the records of the American atlas. The author has never indulged in such an illusion. However, enough is revealed by the sampling of social levels in American English to set the stage for properly oriented research in social dialects, one of the major tasks of the future. See Kurath 1964–A: 135–43.

1.17 Concluding Remarks

The method of gathering dialect data described in this chapter would seem to be the best way to secure reliable information on usage within a reasonable time. The effectiveness of selective sampling as a scientific procedure cannot be questioned. Direct observation in the field by properly trained linguists is surely superior to information furnished by correspondents, and not only in phonological and morphological matters: quantity is no substitute for quality, although some scholars seem to think so.

This does not mean that, with careful planning and under favorable circumstances, valuable data cannot be secured by the “indirect” method. Thus most of the data for Netherlandish are furnished by translation into the local dialects of 131 test sentences, five short word lists, and five sets of morphological cue forms on the part of carefully selected observers [Blancquaert 1948]. In 1939, W. Mitzka sent his list of 188 Standard German cue words to about 50,000 schools with the request that the local synonyms be furnished by a speaker of the local dialect under the supervision of a teacher. His Deutscher Wortatlas [1951 ff.] presents the findings [rev. by Kurath 1958]; and an imposing series of studies for which the Wortatlas forms the point of departure has appeared under the title Deutsche Wortforschung in europäischen Bezügen [Schmitt 1958ff.]. In these investigations, word histories are traced in their sociocultural setting with full philological documentation [rev. by Kurath I960]. For the lexical part of his Survey of Scottish Dialects, A. Mcintosh has adopted this practice [Mcintosh 1952: 70–84; Catford 1957: 113–17], which has also been widely used in Sweden [S. Benson in Germanische Dialektologie 1968: 364–67].

The manner in which lexical, morphological, and phonological data are secured inevitably slants them in one way or another. For this reason the student of English dialectology must be aware of certain differences in the methods employed by the American, the English, and the Scottish surveys, when he undertakes to compare the data or to trace the European sources of features of American English.

For the fieldwork in America and in England, the lexical questions are arranged by topics (semantic fields), so as to provide a focus for the informant and to guide his choice of terms, and the observer is in a position to judge the ease and assurance with which he responds. Hesitation or uncertainty on the speaker’s part leads to further probing to establish usage. When cue words are presented by correspondence, whether topically arranged or not, the written responses have to be accepted without the benefit of gaging the correspondent’s behavior. Lacking this information, the lexical survey of Scotland and Northern Ireland, and that of Germany, rely upon the mass of responses for weeding out probable errors and identifying unsure or unsettled usage.

Although the American and the English surveys agree in providing topical settings for determining word usage, they differ sharply in another respect. In the United States the fieldworker is free to choose his own way of eliciting the desired response; in England the informant is asked to “fill in” the usual term in a sentence framed in Standard English [Orton 1962]. The English way sharply narrows the choice, whereas the American approach may produce variants that are not strictly synonymous. Either method has advantages and drawbacks of which scholars must be aware.

In securing phonological data, the survey of Scotland agrees with the atlases of England and the United States in relying upon direct observation of the informants and in recording the syllables phonically (“impressionistically”) within the framework of the “cardinal vowels” of Daniel Jones. Under the circumstances, students of English are provided with readily comparable phonic data. Nevertheless, allowance must always be made for somewhat different subdivisions of the phonic scale—a multidimensional spectrum—on the part of the several observers. Such differences should be pointed out by the editor [Kurath-Bloch 1939: 50-52, 126-27]. They rarely prevent the abstraction of the phonemic system, as Kurath-McDavid [1961] have amply demonstrated.

W. G. Moulton’s recently expressed pessimism concerning phonic recording [1968: Lang. 44.461–65] is hardly justified. Even Ladefoged’s experiment shows—contrary to Moulton’s interpretation—that all fifteen phoneticians trained to D. Jones’ scale of cardinal vowels identified ten phonemic entities in a language unfamiliar to them (Gaelic), although they differ considerably in recording the relevant phones. Ladefoged himself admits that in the state of our present knowledge and techniques neither an articulatory nor an acoustic identification of vowels is feasible. He concludes: “Consequently the traditional rigorous training in the performance and use of known reference points will remain for some time an essential for all who wish to make useful phonetic statements about vowel sounds” [Ladefoged 1960: 396]. To let the tape recorder do the work that fieldworkers now perform, as Moulton suggests (464), is illusory because it overlooks the fact that only trained phoneticians can report to us what the tapes contain. On the other hand, I agree wholeheartedly with Moulton that fieldworkers should be fully aware of the potential systemic relevance of phonic entities and make appropriate comments.

Although adhering to phonic recording, the Scottish survey has introduced a method of gathering the phonological data in such a way that the phonemic entities can be more easily abstracted, speaker by speaker. Instead of embedding the phonological test words in a topically arranged questionnaire, the nearly 1,000 items are presented in isolation, arranged in sequences that seemed to promise a more or less automatic discovery of phonemic entities and their sources in the parent language (Middle English or Middle Scots). The informants respond to cue words pronounced in Standard English and utter their Scottish dialect equivalents as “sentence words,” i.e., stressed and with sentence final intonation [Catford 1958: 117–21]. Under these prosodie conditions, the phonic character of the syllabics may be expected to be clearer, if not exaggerated, than under weaker stress and without a pitch contour. There is no doubt that this method leads to a reliable—and quasi-automatic—phonemic analysis of the dialects, if a sufficient number of trustworthy informants can be found to submit to the ordeal of matching 1,000 semantically isolated words uttered in Standard English with their local equivalents.

In the United States such a procedure is out of the question. Here the phonological test words must be included in a topically organized questionnaire that prevents self-conscious adaptations of the pronunciation. Since in this country the lexical and morphological items are also recorded phonically, the evidence furnished by the phonological test words is buttressed by a wealth of supplementary data. Moreover, prosodically conditioned allophones of the phonemes, deliberately eliminated in the Scottish survey, are amply documented.

1.2 Editing the Field Data

Editing the materials gathered in the field interviews is a formidable task that has to be handled with care and circumspection. To lighten the burden of the editor, the field notes should be recorded on an open page where each item has its separate “niche” identified by number. Responses to the question, comments of informant and fieldworker, and incidental materials should be entered in a separate column to forestall misunderstandings. There is much to be said in favor of recording in duplicate in a looseleaf field book, so that one copy—properly stamped with the community number—can be filed, page by page, with all the other field records, and the other copy rebound and filed by informants. The editor will want to consult the field record of an informant whenever the response to a particular question seems to call for comment, especially in phonological matters.

The manuscript is best prepared on a set of sheets on which the community numbers are printed in sequence, to match the order in which the field notes are filed. Everything that is to be entered on the map (or published in lists) is taken down in one column. Queries and comments of informant, fieldworker, or editor are tentatively jotted down in another column to form the body of notes on which the introductory remarks of the editor will be based. When an item is especially complicated, the original decision on what will go on the map, and how, may have to be modified in the light of insights gained in the process of editing the mass of material.

The first draft of the manuscript of simpler items can often be done by an assistant in accordance with general instructions. But even here an experienced editor will have to assume responsibility for reviewing it and approving it for publication. Utmost accuracy and reliability of the published record is expected of him.

Careful editing of the data furnished by the fieldworker involves not only the preservation of the recorded responses along with significant comments of informants and observers on the status of the expressions offered and recorded, but also the elimination of obvious slips and the appraisal of apparent errors of judgment on the part of the field observer.

Whether the editor should do more than that has been a subject of debate. Some scholars hold that “interfering” with the recorded data—i.e., judging their reliability—must be avoided; others advocate an expression of opinion on his part.

It is the present writer’s conviction that the editor should present all relevant information at his disposal for the convenience of those who will inspect the record of usage more or less casually and for the orientation of those who will undertake to interpret it in their own linguistic or sociocultural studies. He is best qualified to perform this important task, which otherwise would have to be done over and over again, often with more limited resources. The stand he takes on debatable issues is, of course, subject to scholarly criticism in the light of additional or conflicting evidence.

The special qualification of the editor to provide the essential background information for the record of usage concerning the items provided for in the questionnaire derives from his intimate knowledge of the linguistic and social conditions under which the fieldworkers do their job and from his experience in planning the editing item by item. More often than not, he has also participated in designing the survey and done some fieldwork himself. The former gives him insight into the general linguistic and sociocultural situation in the area; the latter gives him a “feel” for the conditions under which usage is elicited and recorded.

Presenting the relevant information, item by item, for proper orientation does not impose an additional burden upon the busy editor, since he must in any event accumulate this information for his own purposes. Except for the simpler items for which a general editing plan is an adequate guide, he faces the task of designing effective and economical procedures for the presentation of the more complicated items, each of which confronts him with special problems. Among the factors that he may have to take into account in planning his presentation are the following:

(1) The range of meanings of lexical items;

(2) the dissemination of synonyms, or near-synonyms, regionally, socially, or by age groups;

(3) the situational, stylistic, and emotive connation of words (neutral, euphemistic, jocular, derogatory, technical);

(4) the choice of terms as reflecting differences or changes in fashion;

(5) the informant’s familiarity with the “thing-meant”;

(6) the status of morphological and phonological variants as reflected in social dissemination and the informants’ reaction and comments;

(7) the fieldworker’s judgment concerning the relative trustworthiness of certain responses or comments;

(8) differences between fieldworkers in eliciting responses or in phonic notation.

Since the editor must take all of these factors into consideration in designing effective editing plans for the more complicated items, there is no reason whatever for withholding this information from others. Far from it. He owes it to the reader to say how he planned his presentation of the data at his disposal. Those who consult his work must, of course, keep in mind that the evidence of usage presented in this manner is deliberately restricted to the data furnished by the sampling survey. Extraneous sources of information are not admitted.

This editorial procedure is illustrated below by several examples adapted from the Linguistic Atlas of New England [Kurath-Bloch- Lowman 1939-43].

Further information about the way in which certain factors that condition usage have been dealt with in the Linguistic Atlas of New England can be gathered from the maps referred to below.

(1) The effect of changing fashions: 339 bureau, 340 chest of drawers.

(2) The status of folk customs: 409 serenade, 410 a dance, 414 social gathering.

(3) Neutral, polite, jocular, derogatory, and technical expressions: 190 bull, 354 privy, 450 a rustic.

(4) The great variety of expressions for emotionally charged activities or states: 404–7 courting, jilting, 463–64 awkward (person) 472 angry, 479–82 tired, exhausted.

(5) The social standing of morphological variants and resulting sensibilities: 642 climbed, 646 ate, 644 drank, 674 (he) isn’t (going), 688 (he) doesn’t (care), 718 double negation.

1.3 Publication of the Findings

The edited findings of a sampling survey can be published in full or selectively.

For full publication two different methods are used: (1) overprinting on a base map or (2) printing in the form of lists. In either case the communities represented in the survey are identified by numbers. Overprinting on a base map has been employed in the atlases of France [Gillieron-Edmont 1902–10], Italy [Jaberg-Jud 1928–40], Rumania [Puscariu-Pop-Petrovici 1938 ff.], Corsica [Bottiglioni 1933–42], and New England [Kurath-Bloch-Lowman 1939–43]. The less costly method of giving the data in lists was introduced for presenting the phonological data for French-speaking Switzerland [Gauchat-Jeanjaquet-Tappolet 1925], is being employed in the survey of the dialects of England [Orton 1962ff.], and will be used by R. I. McDavid, Jr., in publishing the materials of the Linguistic Atlas of the Middle and South Atlantic States. Either method has its advantages and drawbacks.

Overprinting on a map is costly, especially because it requires the services of a skilled draftsman to letter the hundreds of entries in their proper locations, map by map. For quick orientation it has great advantages, particularly if the base map, printed in a light tint, contains the major political boundaries as well as the more important physiographic features of the landscape (rivers, mountain chains, etc.) that channel or hinder travel and traffic. Even a casual inspection of such a map is apt to reveal the regional dissemination of variants and thus challenge the scholar’s interest. When the usage of several social levels or age groups is offered on one and the same map, as in the Linguistic Atlas of New England, the more consistent social and/or age differences in usage are also immediately apparent, if the data are properly presented.

Publishing the findings in the form of lists is decidedly more economical, especially now that typewriter composition is feasible. The disadvantage is that only those that are rather intimately acquainted with the area can readily visualize the geographic dis-semination of the variants, even with the help of a foldout map showing the location of the numbered communities, an aid with which each volume must be provided. However, publishing the findings in this fashion also has its good points. The scholar who pursues specific problems in phonology or morphology, undertakes to establish dialect boundaries, or wants to determine the characteristic traits of a speech area, has to assemble pertinent evidence from the storehouse of data published in one form or another. He does this by charting items relevant to his purpose. There is no question but that this necessary operation can be carried out more conveniently and more speedily by drawing upon listed rather than mapped data, as the present writer can testify from ample experience.

Selective publication of field data in more or less simplified form is highly rewarding. When full publication is not feasible or must be delayed for financial reasons, selective publication is imperative as a means of showing the scientific implications of the project and as a service to other scholars in the field. Even when full publication is in progress, it is often advisable to present some of the data in simplified form in order to create a sound perspective for those less intimately acquainted with the data and the emerging problems.

While publication in full involves the presentation of all the data recorded in response to each of the items investigated—e.g., phonological along with lexical or morphological variants—selective publication is focused on specific problems for which the field data provide the material item by item. Thus one can undertake to assemble evidence for the phonological reflexes of a group of phonemes of the parent language, gather evidence for recurring dissemination patterns in regional vocabulary, or compile relevant data concerning the social behavior of certain morphological features. Such operations provide insight into the linguistic situations within the area; they are also a necessary preliminary step toward historical interpretation.

The variants of the selected items can be effectively displayed on outline maps by contrasting symbols—circles, triangles, squares, in outline or solid. Placed near the numbers showing the location of the communities, or simply in the relative position of the communities, geographic dissemination patterns of the variants stand out clearly. Areas of uniform usage, transition zones, enclaves, scattered items, effectively make their appearance. This technique of presenting variants of selected items has been widely used in the last two decades.

The Word Geography of the Eastern United States [Kurath 1949] contains 163 full page maps of this type (of which two are reproduced below), 43 maps displaying heteroglossic lines based upon them, and a map [figure 3] showing the speech areas of the Eastern States as established by the lexical heteroglosses. In the preparation of these maps the unpublished field records from the Middle and the South Atlantic States as well as the fully published material for New England are drawn upon. The companion volumes on verb forms [Atwood 1952] and pronunciation [Kurath-McDavid 1961] follow the same general plan. So does Atwood’s Regional Vocabulary of Texas [1962]. This method has also been used in Europe: by Kloeke [1950], Hotzenköcherle [1962], and Kolb [1966].

An important application of this procedure, involving a notable departure from earlier practices, is to be found in the linguistic atlases of French-speaking Belgium [Remacle-Legros 1953ff.] and German-speaking Switzerland [Hotzenköcherle 1962ff.]. In these two works, the volumes included for publication of the phonological, lexical, and morphological materials in full are preceded by a selective presentation of typical phonic reflexes of parent phonemes in their areal dissemination. Hotzenköcherle devotes two large volumes—375 pages, including about 300 maps—to this task, following a procedure briefly described in his preface to volume I. In doing this, he orients the future users of the atlas and prepares the way for a structural interpretation of the complicated phonic data. See my review in Language 30. 515-20 (1963).