• TABLE OF CONTENTS
HIDE
 Front Cover
 Main
 Back Cover














Title: Sampling reliability in ethnological field work
CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00091716/00001
 Material Information
Title: Sampling reliability in ethnological field work
Physical Description: Book
Language: English
Creator: Honigmann, John and Irma
Publisher: Southwest Journal of Anthropology
Copyright Date: 1955
 Record Information
Bibliographic ID: UF00091716
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.

Table of Contents
    Front Cover
        Front Cover 1
        Front Cover 2
    Main
        Page 282
        Page 283
        Page 284
        Page 285
        Page 286
        Page 287
        Page 288
    Back Cover
        Page 289
        Page 290
Full Text










.. 1 / ,
_. *1JALA~ FIEL 'W R -, 5, '.
S'. .u .K l q w u j % l 1 1 .:. -. .; '






.. -i. .- -. '._ -.!
~ ~ ~ ~ ~ ~ ~ ~ ~ erne from ". -;... -% .. 1_--: .
.SO TH ES ER JO R A OF ANTHROPOLOGY :,,,

.: : V .t
. -~ ;,_. .. !.%I... 1










. .., : -. .



~~~~ ... .$4.4

... ~ ~ 4. .. ...........
j,.'.5 3 cg lP %-& *' : .
V .%.2 .$j. s C # W -' .4a ,:. .,.
1 H:.. ,
.. ..,4 L54 .17. ... .
;1. 1,*' '.44. A I:- 4455 .I-%... ,. I.






















,'...j 49. 7 .1 L 4 ....- .:1,I . p L.1 .1.. '
;? ~ .) ,~1 : ... I % r. 'F f i. .1..

) .4i Jqy_, a '! ,,.L is--, .. ; ... .. 2 I .. '.I.. .- ,+, :4-....0.- i .. .
"m 4 .j : -'., .
.. .! .t .I : ~ I i;, :- -' ... . .i 'r,- 1I.- "I I.. ,- .W
.V~ C !$,%i. !p I.I .. '4, :. ,. ; :' ':-,%1. ,
.. 1. -. ;4: :. L i. I.4 .J;. .i U A ;'. ;t j. r
.~. b... ~ I ;

1~~~~....:., .:- ':1 1 jLt 4'4gnf q, y '. I. '


Z" 11I "', ,. X4 t 5 :4 1 4 : I : .. .I, :i .
V,4 ;, 6{. k r: 4-4. *. 4 ~ .j
.3 _-..',,, : I I *. .-' i, ,


.. ... -. ... ...--f ... .., ... q- I. .., ic.... 4 .4 %. I ,













SAMPLING RELIABILITY IN ETHNOLOGICAL FIELD WORK
JOHN AND IRMA HONIGMANN

THE PROBLEM
S THE METHOD of sampling customarily followed in ethnological field
work reliable? It seemed possible to throw some light on this question during
a research visit to West Pakistan in 1952.1 The basic object in the trip was to learn
the reactions of rural audiences to American informational films presented in the
Urdu language by the United States Information and Educational Service. Find-
ings were to be related to cultural and other factors in the situation. We worked
in three villages, numbered in Table 1 as follows: (1) Tciho, Sind Province;
(2) Tordher, Northwest Frontier Province; and (3) Chak 41-MB, Punjab Prov-
ince. To ascertain the composition of the audience and the degree of understand-
ing with which people saw the films, respondents were interviewed. A different
Pakistani interpreter served as assistant in each of the three localities.
Always in previous field work we had employed the traditional method of
ethnological sampling-using mature, willing informants who, perhaps, spoke
English and were willing to describe relevant aspects of their cultures. The problem
of how safely the information received from such sources can be generalized for a
whole community has received increasing attention from anthropologists in recent
years at the same time that more rigorous methods of sampling have appeared in
field work. In Pakistan we determined to select both an opportunistic sample-as
is customary in ethnology-as well as a random sample (stratified where possible).
We would compare the categories in terms of responses given in the interviews.
The present paper presents the results of the experiment.
Opportunistic sample subjects were secured in various ways. Sometimes the
ethnologist walked through a village with an interpreter seeking out otherwise
unoccupied respondents. Also we subjected men who visited our home to inter-
viewing and sent out assistants to find respondents. Often, when visiting a court-
yard or the shady grove overhanging a Persian wheel in search of a sample subject,
the investigator or his wife encountered other people who agreed to answer ques-
tions. (Female respondents remained few and were randomly drawn only in the
Punjab village.) Random samples were devised to insure that every eligible person
1 Field work was conducted under the direction of the Institute for Research in Social Science
of the University of North Carolina, Dr Gordon W. Blackwell, Director. Results have been pub-
lished in a report currently out of print: see J. J. Honigmann, Information for Pakistan (Chapel
Hill, Institute for Research in Social Science, University of North Carolina, 1953, mimeographed).
282


VOL. 11, 1955










SAMPLING RELIABILITY


in the community of whom we had information (i.e., males over eighteen, although
in a Punjabi village females as well) would have an equal chance of being selected.
The person whose name had been drawn was then assiduously sought for. Refusal
to be interviewed occurred practically never. In several cases, however, randomly
selected subjects remained absent from the village throughout the period of field
work and in two villages a small proportion could never be located. Substitutes
were drawn for those cases.
The principle hypothesis being tested follows:

The opportunistic sample in each village and in the combined villages will be
different significantly (in composition and response) from the random sample.

From the analysis we hoped to be able to hypothesize further that the oppor-
tunistic sample differed significantly from the random sample with respect to such
variables as might indicate that persons with certain characteristics in the total
population tended to attract the anthropologist's attention or were attracted to his
company.
Actually, we first tested the null hypothesis that the random and opportunistic
samples in each village (as well as in the combined three villages) were not dif-
ferent, i.e., they consisted of persons similar in respect to the variables. For this
we calculated the chi squares for all variables. Differences found between the
samples, according to this hypothesis, should be due to chance alone. If the dif-
ferences between the two categories were found to be significant (p at the .05 level
or lower), then the differences could not be ascribed to chance but, we assumed,
indicated that the samples were different. In that event our initial hypothesis would
be confirmed.
RESULTS
Relatively few differences at the .05 level of probability or lower obtain between
random and opportunistic samples. Audiences saw eighteen films and only in one
village, in the case of a single picture, did the difference between the samples
reach the .05 level. For this reason we have omitted any statement of p values for
the films from this paper. Do the random and opportunistic samples behave simi-
larly from one village to another with respect to comprehension? It may be claimed
that should this be the case, then we have further assurance of the reliability of
our finding that the categories tended to be homogeneous with respect to under-
standing the films. In Tciho and Tordher (Villages 1 and 2 in Table 1), where
Sindhi and Pushto are spoken respectively, "insignificant comprehension" occurs
with modal frequency in both the random and opportunistic samples. In the third
village, where the language, Punjabi, is closely related to Urdu (the language of










SOUTHWESTERN JOURNAL OF ANTHROPOLOGY


the film sound tracks), responses are more evenly distributed between insignificant
and good comprehension. Here, too, the samples differ considerably from each
other in respect to comprehension. Consistency among the three villages, there-
fore, is limited, throwing doubt on our conclusion of homogeneity. (It must be
pointed out that when working with audience response to individual films the
number of cases often remained quite small and cells in the tabulations for chi
square calculation several times contained numbers below five. Working with
such small numbers and using the chi square test warrant no claims of great
preciseness.2)
Table 1 includes the variables for which significant differences did appear be-

TABLE 1
Number of Respondents and Significance of Differences between
Opportunistic and Random Samples in Three Pakistani Villages8

Combined
Village 1 Village 2 Village 3 Villages
Variable N p N p N p N p

1. Attendance at more
performances 50 .01 85 .001 61 .02 196 .001
2. Land ownership 74 .10 44 .20 118 .20
3. Length
of schooling 74 .01 42 .70 116 .05
4. Social category
("class") 40 .95 75 .01 42 .90 157 .05
5. Admitted
understanding
of Urdu 75 .01 41 .98 116 .02
6. Familiarity with
radio programs 39 .90 -
7. Awareness of
government pre-
senting the films 48 .05 84 .01 61 .40 193 .01
8. Comprehension
of Urdu titles 31 .05 -

2 For great help in suggesting appropriate statistical procedures and directing analysis we
are indebted to Dr Daniel O. Price. Both he and Dr Grant Dahlstrom reviewed the manuscript
before publication.
3 Dash indicates absence of data.










SAMPLING RELIABILITY


TABLE 2
Variables Distinguishing Random from Opportunistic Sample in Three
Pakistani Villages

Random Sample Opportunistic Sample
Variable N % N %

3. Length of schooling
0-4 years 65 79 21 62
5-8 16 19 8 23
9-14 1 1 5 15
Totals 82 99 (P =.01) 34 100
4. Social category ("class")
Upper 10 9 4 17
Middle 64 60 11 46
Lower 33 31 9 37
Totals 107 100 (P =.70) 24 100
5. Admitted understanding of Urdu
None 34 42 7 20
Poor 19 23 8 23
Good 13 16 3 9
Excellent 15 19 17 49
Totals 81 100 (P = .01) 35 101

tween the two samples. Contrary to what we discovered in the case of comprehen-
sion of the films, with respect to background factors (including attendance and
awareness) the samples tended to be different from one another and in Tordher
(Village 2) to such a degree (except for variable 2) that there is less than one
chance in a hundred of the differences being due to chance. Table 2 tests the
hypothesis that in all the villages the ethnologist (or his assistants) tended to
secure respondents who differed significantly from the general population in quali-
ties that made them more attracted or attractive to a professional stranger and
college-educated interpreter (variables 3, 4, and 5).4 This hypothesis is supported
with regard to length of schooling and admitted understanding of Urdu (mani-
festly related characteristics of a person) but not with respect to membership in
4 Such differences scarcely appear in the third village, which is a planned resettlement of
Muslim refugees from what is now East Punjab (India). This village lacks a formally educated
landlord class. All but half-dozen men are farmers and practically every adult male possesses
about the same degree of sophistication.










286 SOUTHWESTERN JOURNAL OF ANTHROPOLOGY
social categories. Inspection of our original data reveals that differences between
villages were again not consistent with regard to the three variables of Table 2.
For example, in Tordher, a densely settled locality of over 8,000 population, the
opportunistic sample contained twice as many people with 5 to 14 years of school-
ing as the random one. But in the more homogeneous Punjab village, both samples
were weighted with unschooled persons. In Tciho we drew a stratified random
sample. However, here as well as in Tordher the opportunistic category included
more "lower class" persons than the random one and in Chak 41-MB the samples
do not differ with regard to class. More than half of the randomly selected sub-
jects in Tordher claimed to know no Urdu but more than half the opportunistic
sample reported "excellent" grasp of Urdu, a finding in the expected direction.
In the Punjab both samples were about evenly matched for language ability.

SUMMARY AND CONCLUSIONS
Analysis of random and opportunistic samples interviewed in three villages of
West Pakistan does not bear out the hypothesis that an opportunistic sample, such
as is frequently selected in ethnological field work, differs significantly from a
random one.
The experiment revealed that the samples tended to differ significantly only in
certain related background characteristics (like length of schooling, social category
membership, and admitted understanding of Urdu) but not consistently so from
one village to another. They also differed with respect to attendance at the film
performances and, whether attending or not, in regard to accurate knowledge of
the government which had come to show the motion pictures in the villages. Those
persons in each sample category who attended the performances practically never
differed significantly in their understanding of what they saw and heard. This
allows the deduction that both samples were homogeneous with respect to ability
to comprehend the motion pictures, although the level or understanding is not con-
sistently different between the samples from one village to the other.
There is some indication that members of the opportunistic sample tended to
be selected (unconsciously) from among those persons in the villages who were
among the most sophisticated and hence more likely to attract, or be attracted to,
a professional stranger and college-educated interpreter.
The inconsistent statistical results and variation from one village to another
suggest caution in applying these findings. Obviously a single experiment of this
kind allows little generalization or advice for field workers who may work with
problems far removed from motion picture effectiveness. At best one can be
warned not to assume that an opportunistic sample will approximate a random
sample in all respects. Meanwhile, more use of sampling experiments in field










SAMPLING RELIABILITY 287

work is desirable in order to clarify relationships between the mode of selecting
informants and the kind of information that is acquired. For best results future
research should work with variables pretested to insure their diagnostic value for
distinguishing between two samples. It is likely that rural Pakistani villagers on
the whole were about equally able to understand the films. Hence, questions bear-
ing on the content of the motion pictures possessed little diagnostic value.

UNIVERSITY OF NORTH CAROLINA
CHAPEL HILL, NORTH CAROLINA




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs