GORDON i~. WOOD 
REFINEMENTS IN TABULAR. MODELS OF VAR.IATION 
IN REGIONAL AMER`ICAN ENGLISH * 
I. INTRODUCTION 
Models of the geography of American regional English have been 
greatly influenced by the work of researchers for the proposed Linguistic 
Atlas of the United States and Canada. Among the representative pub- 
lications one should note H. KtnlATH'S A Word Geography of the Eastern 
United States (1949) and E. B. ATWOOD'S The Regional Vocabulary of 
Texas (1962), particularly for their use of tables which place selected 
words in a designated locality and give an indication of relative abun- 
dance there. It is these two works which set the pattern for the present 
computer assisted analysis of the presence of 1,000 regional words in 
the southern United States. Thus there are resemblances among these 
models, but there is an essential difference also in that the newer model 
advances claims about relationships among word choices and thus is 
implicationally more dynamic than are its predecessors. 
2. FIRST STAGES IN MODEL FORMATION 
The present study began in the 1940's when the computer was 
first being praised as the machine for sorting vast bodies of data. At 
that time the Linguistic Atlas project had not gone forward in many 
parts of the South, but it had produced a printed questionnaire which 
asked persons to make choices among synonyms. Here were tools 
which could be used together as a means of discovering gross features 
of vocabulary distribution within the relatively unexplored states of 
Tennessee, Florida, Alabama, Mississippi, Louisiana, Arkansas and 
Oklahoma. Georgia was included as an eastern anchor, even though 
the state had been surveyed in part for the Atlas (R`. I. McDAVID, 
* This study was supported in part by Southern Illinois University, Edwardsville. 
344 GORDON R. WOOD 
1958). Texas was excluded because Atwood's survey was almost ready 
for publication at that time. 
When the questionnaires were returned, many were rejected be- 
cause the answerer's biography showed that he did not meet certain 
standards. From the chosen questionnaires the coded data was trans- 
ferred to punch cards. There were codes for the person's age, sex, educa- 
tion, race, place of birth and longest residence, places of birth and res- 
idence of his parents and grandparents; occupation was not coded 
simply because of the vast range of identification needed. Then the 
response to each set of synonyms was entered. If, for instance, in item 
125, VEHICLE FOR A SMALL BABY, the synonym baby cab had 
been selected from the synonym set baby buggy, baby cab, baby carriage, 
baby coach, then it would be punched as the next entry on the card 
after the biographical data. 
The original intention was that the computer would sort the data 
first under age, then under education, and then under sex. For each 
sort the computer would print a county by county table showing the 
actual number of local responses according to a selected heading. Where 
H. KURATrI'S tables (1949, p. 12) showed "regular " or "fairly com- 
mon " or "rare " in Ohio, New York State, and New England respec- 
tively, the computer list would show "0" or "2" or some other 
number for baby cab in Hamilton County, Tennessee, and so through 
each county for each word of the synonym set. 
In order to make some comparison with Atwood's tabulation of 
responses according to age, the first sort of all data was by three age 
groups - those above seventy, those between thirty and seventy, and 
those below thirty. A county by county computer listing of 1,000 words 
as chosen by 1,000 people is enormous. Its very mass made impossible 
the publication of such a document even for those dialectologists who 
might want to look at the raw data in its most detailed form. A byprod- 
uct, hand drawn maps made by plotting the presence of selected words, 
permitted a redefining of the technical terms " Midland " and " South- 
ern " dialect as they apply to this region (H. KtraATa, 1949;: G. 1L. 
WOOD, 1963, 197t). 
3. REFINEMENTS FOR BOOK PUBLICATION 
The computer was next instructed to make total counts of all choices 
of each synonym within four main divisions of the respective states as 
TABULAR MODELS OF VARIATION IN REGIONAL AMERICAN ENGLISH 345 
well as for the state itself, and for the entire eight state region. The 
grand total for each word in the questionnaire was machine sorted 
according to increasing numerical frequency of response. This computer 
printout became the source of a table photocopied and placed in a book 
(G. R.. WOOD, 1971, pp. 297 ff.). As a table, it seems at first glance to 
say what those in Kurath and Atwood also say: here is a regional word 
count. It shows that roughly 1,000 persons chose or rejected certain 
words and that their choices were somewhat mixed with regard to 
the rest of the vocabulary. Actually the claim 'is different. This is a 
model of a whole vocabulary as a closed system in which for an entire 
region the place of each synonym within the system is fixed at a given 
point on the scale. Obviously a search at a different time or with a 
different questionnaire or among different people would shift the en- 
tries, but it would not alter the systemic design. No such ordered 
conclusions can be gained from the Kurath and Atwood models since 
the Linguistic Atlas interview procedure has no way of testing an 
informant for his range of synonyms. 
Given these gross counts, the next stage was to design a table that 
would display the presence or absence of each word in a smaller geog- 
raphic setting. It became apparent that statements of percentages 
would be more informative than would be the presence of raw numbers. 
Hence a model emerged which listed all of the words in a synonym 
set and placed them proportionately in each of the eight states. It com- 
bined the features of locality found in the Kurath model with the fea- 
tures of relativity in the Atwood one, but it excluded the gradations by 
age. Furthermore, it refined the statement of relationships by breaking 
the data into four sub-patterns which answered these questions: What 
are the relationships in choice of the same word in a section of adjacent 
states such as those situated east of the Mississippi? Within any state and 
its major subdivisions what is the relationship of a given response to 
the largest number of possible responses there? Next, if the statewide 
response is regularly set at 100 per cent, what are the relationships 
between the responses in the north to those in the south, or in the east 
to those in the west? And finally, in each locality what are the relative 
preferences for each member of the synonym set? 
The computer printout of these percentile calculations became the 
photocopied text of a book (G. R.. WOOD, 1971, pp. 64 ft.). An extract 
from one will suggest the format. The synonyms given in the question- 
naire for the concept "a woman who helps at childbirth" were god- 
mother, granny, granny woman, and midwife. Consider two choices within 
346 GORDON R. WOOD 
the states of Tennessee and Georgia, and their western-eastern (W E) 
northern-southern (N S) halves. In the example below, the left column 
stresses the local relative importance of each word; the state (ST) per- 
centage is not printed but is set at 100 per cent. The right hand column 
shows the relative preference among synonyms within the full set. 
GRANNY WOMAN 
MID WIFE 
TENNESSEE GEORGIA 
W 31 4 N 53 6 
E 69 10 S 47 6 
ST 14 12 
W 45 30 N 63 45 
E 55 37 S 37 27 
ST 68 71 
For the word geography of the United States, the refinement call- 
ed a closed model here has certain advantages. The text of the question- 
naire can and should include standard words such as midwife alongside 
those synonyms which have been identified as having regional limits, 
words like godmother, a standard word with a special regional meaning, 
and granny woman, an entirely regional word. The computations, then, 
show among other things that the regional word and the standard 
word considered by themselves have variations in regional preference. 
It is evident that east Tennessee has stronger preference for granny 
woman and midwife than does west Tennessee; a similar pattern of major 
preferences appears in north Georgia (53 and 63 per cent versus 47 and 
37 per cent in the tabular extract given above). Obviously the calcu- 
lations confirm that among synonyms the general word is greatly pre- 
ferred over a more regional one. In addition they will show what had 
not been formally identified up to the time of calculation: general 
words have variations in regional preference as well. The degree of 
variation leads to the idea that the dialectology of American English 
must consider this characteristic along with the more easily identified 
characteristics such as the presence of distinctively local words such 
as granny woman. 

References

E. B. ATWOOD, The Regional Vocabulary 
of Texas, Austin, 1962. 

H. KURATH, A Word Geography of the 
Eastern United States, Ann Arbor, 1949. 

tL. I. McDAwD Jr., American English 
Dialects, and (with Vn~gINIA M¢- 
DAvn~) Appendix: Maps, in W. Nri- 
son FaANCIS, The Structure of Amer- 
ican English, New York, 1958. 

G. 1L. WOOD, Dialect Contours in the 
Southern States, in, American Speech ,, 
xxxvla (1963), pp. 243-56. 

G. IL. WOOD, Vocabulary Change: A 
Study of Regional Words in Eight of the 
Southern States, Carbondale and Ed- 
Wardsville, 1971. 
