The ETHNOGRAPHIC ATLAS in its own words

Mike Salovesh (t20mxs1@CORN.CSO.NIU.EDU)
Wed, 10 Apr 1996 04:02:42 -0500

I've been waiting for several days for somebody to pick up on the real
Ethnographic Atlas in response to the thread that has been nibbling around
its edges. I guess it takes an old fart like me to remember the thing
clearly enough to be able to distinguish it from HRAF, to which it has
only a vague resemblance.

Since nobody else has taken on the task, I'll try to describe both the
Atlas and the HRAF. I'll begin by saying that both approaches to
establishing databases turned me off long since. I even published what I
thought of as a major critique of the approach way back in the 60's, in a
paper I originally presented at a AAA meeting under some such title as
"Living with the wife's father under a rule of patrilocal residence:
obligatory alternatives meet the Ethnographic Atlas". I'm proud to say
that in the refereeing process that applied to meetings at the time,
Murdoch saw my abstract and asked that my paper be included in a session
he chaired. He knew damned well that I was setting him up as one of two
straw men -- Floyd Lounsbury was the other -- on my way to saying
something else, and he even saw to it that Lounsbury gave the paper after
mine in that session. I'll always be grateful for what Murdock did for me
on that occasion. I always respected the unbelievable breadth of his
knowledge about everything, anyhow. (Of all the anthropologists I ever
met, only Fred Eggan knew more about anything and everything than Pete
Murdock.) Nonetheless, my description will not be entirely sympathetic. I
hope it will be essentially accurate anyhow.

What the ATLAS and HRAF shared was the intellectual drive of George Peter
Murdock. He was an inveterate note-taker and jotter-downer and classifier.
(One of his least-known products, typical vintage Pete Murdock, was a
collection of printed filecards bearing more or less standardized data and
listings of sources for the ethnography of just about everybody in Africa
who had been looked at by anthropologists. I've still got my copy around
here somewhere, and it still comes in handy once in a while.)

HRAF was presaged by some of his activities before WW II. (I hasten to
add that I knew nothing about it at the time; I was in grade school when
Pearl Harbor was bombed.) What he wanted at the start was some kind of
index that could be used to standardize information about all the known
cultures of the world. He spent a lot of time devising two tools for the
job. One of them led to a coding list that identified cultures, the
OUTLINE OF WORLD CULTURES; the other became a much more complicated index
of things that might be said about those cultures in some ethnography or

Somehow, Murdock's efforts became part of the "war effort" in WW II. It
turned out that the U.S. simply had no available military intelligence about
most of the world. (That wasn't surprising, since sometime in the 1920's
the U.S. Secretary of State had dismantled that part of his service that
had been devoted to cracking codes and intercepting messages. What he
said about that, allegedly, was that "gentlemen don't read each other's
mail". The statement deserves preservation along with "Let them eat cake"
and other classic illustrations of suicidal bons mots.) Many anthro-
pologists were recruited into a massive effort to fill the gap, largely
through a series of studies ostensibly sponsored by the Smithsonian
Institution. (Ruth Benedict's "The Chrysanthemum and the Sword" is a
revision of one of those studies, an attempt to describe Japanese culture
whose original purpose was to give direction to U.S. war policy and war
aims regarding Japan.) Murdock's attempts to sort data about all the
world's cultures was a natural candidate for massive support, and that's
what really got HRAF--The Human Relations Area Files--off the ground.

The original HRAF really was a filing system for arranging data in
real filing cabinets. The model was to devote one drawer in a filing
cabinet to a single monographic description of one culture (but a really
long monograph might need more than one drawer). The first slot in the
drawer was occupied by a photographic reproduction (it was too early for
such things as Xerox copies) of the monograph itself. The rest of the
drawer represented the product of a massive coding effort.

Coders were handed a book, or some section of a book, and asked to
classify every statement they read according to Murdock's index of
cultural categories. They'd make marginal notes that the sentence "Then
the family went to another place and built a new house" is somehow
relevant to kinship, to residence, to travel, and to construction *by
entering the relevant numbers from Murdock's code*. Essentially, the
GUIDE TO CULTURAL MATERIALS was the code-book used for this task. Since I
don't keep a copy handily near this computer, let me invent some fake
number/letter combinations to show how it worked. "Kinship" in general
might be category K, and "family" could be something like K sub 1, so one
of the codes to be entered in the margin next to my invented sentence
might be "K sub 1". "Manufacture" of any sort might be M; building
construction M.1; residences, M.1.3; single-family residences, M.1.3.1, so
the margin next to my sentence would include M.1.3.1. Never mind the
actual details, which are easy enough to look up if you want to. The
essence was that Murdock tried to design an overall outline of anything
that anybody might do or say or whatever in a culture so that everything
would have a number and a place in his scheme. Every aspect of every
statement in a monograph would get its coding number.

Once the coders were through with a monograph, and their codings were
cross-checked for accuracy and adherence to Murdock's principles and
guidelines, each page of the monograph was duplicated as many times as
necessary to provide one copy for every marginal note. Then those copies
of individual pages were sorted and placed in file folders. You could go
to file section "M" and pull out the file labeled M.1.3.1 and find all
references to the construction of single-family houses that appeared
anywhere in the whole monograph. Each of the references would give you
the full context of the page on which it appears, and you could then
backcheck that page by going to the first file and pulling out the whole
monograph. If all you wanted to know about was everything any monograph
in the collection had to say about building such houses, all you had to do
was go to file M.1.3.1 in each and every file drawer in the whole
collection and you would "KNOW".

The basic collection grew and grew. I remember going to the room that
housed the full-size cabinets full of the HRAF files in the mid-50's and
being awed by the seemingly endless rows of file cabinet after file
cabinet in what was a pretty huge room in the first place. (This was at
the U of Chicago Libraries, one of several depositories that maintained
full sets of the HRAF materials as they were produced.) It was an
unbelievable effort when you realize that all the sorting was done by
hand, with no effective computer assistance at all. Actually using the
files, in retrospect, was awkward beyond all reconstruction in this day
and age when we're used to such things as an AltaVista search that can
come up, after only a few seconds, with 20,000 references in a concordance
covering a database of millions of records. But just imagine how much of
an improvement Murdock's filing scheme was in a world that had never seen
anything as systematic before!

Of course, with the passage of the years technology advanced, and the
basic HRAF files came out on microfilm and microfiche and took up a
helluva lot less space. They also gained a lot in convenience of use. A
mere handful of microfilm reels, not to mention the more convenient
microfiches, meant that you could pull anything you wanted to know about X
out of several hundred monographs while sitting at a single reader and
never getting out of your chair. (Nowadays they're getting the whole
thing up on CD-ROM, which was mind-bogglingly beyond any possibility of
imagining when HRAF got its start.) But the basic product still was the
result of that painstaking hand-coding and hand-sorting, page by page and
monograph by monograph. There had to be a better way!

Once he set the basic framework, Murdock himself proceeded to show how all
the sorted data of HRAF could be *used*. The first such product was his
SOCIAL STRUCTURE (1948), in which he applied simple statistical measures
(mostly Chi-square tests) to world-wide compilations of data about kinship
and social structure. He built a grand descriptive edifice on the range
of possibilities in kinship and social structure exactly on that base.
The statistical tests he chose were weak at best, and I'm not sure they
prove anything much that I want to know, anyhow. But the book is still an
extremely useful reference, and a beautiful reflection of Murdock's
classifying mind. No matter what you run across in the world of cousin
terminology, for example, somewhere in this book there's a label for that
kind of terminology and a way of fitting it into a general scheme for
classifying cousin terms in general. I may not buy into any of his
allegedly causal explanations, but I sure do love the efficiency of being
able to say "this terminology makes a bi-Guinea system, as defined in
Murdock (1948:nn)".

The experience that came out of assembling/testing/writing SOCIAL
STRUCTURE moved Murdock to try to set things up so that anybody could test
virtually any proposition with similar statistical tools. As a step in
that direction, he devised something he called "The World Ethnographic
Sample" (as published in an AA article in 1957). It presented, in tabular
form, data about thirty "cultural characteristics" as they appeared in a
"sample" of 100 cultures selected to represent the range of variation
across the world. The tables actually do present a summary of all that data
-- in just twelve pages!

To show how he worked this miracle, I'll point to just one slot in the
table. (It repeats once for each culture, so that "one slot" actually is
100 boxes into which you can pour selected data.) Let's take Column 14,
Social Stratification. The tables actually have only fifteen columns,
but look what he packs into a column. Here's a condensation of part of it:
(The actual materials are considerably longer than the extracts I give!)

A Formal age grades.
C Complex stratification, 3 or more social classes
H Hereditary aristocracy different from ordinary freemen.
O Absence of significant social stratification among freemen.
W Wealth distinctions of importance, no hereditary classes.
h Hereditary slavery.
i Incipient or nonhereditary slavery.
a Absence or near absence of slavery.
s Slavery reported but no indication of whether hereditary.

Column 14, in other words, potentially provides a space for indicating two
different kinds of information, by the use of one capital and one lower
case letter in that slot. The list of possibilities within one column is
supposed to be logically exhaustive and mutually exclusive, so whatever is
reported for a single culture can be listed somehow. And if the base
monograph on that culture simply doesn't say anything about a particular
subject, Murdock indicates that, too, with a standard symbol used across
all categories.

Sometime after devising the world ethnographic sample and publishing that
original article, Murdock retired from Yale and HRAF and moved to the
University of Pittsburgh. When he got there, he founded a journal:
ETHNOLOGY. Volume I number 1 appeared with the date of January 1962.
That issue contained the first part of the Ethnographic Atlas, an
attempt to do the same kinds of things Murdock had already been doing
with HRAF and the World Ethnographic Sample. This time, however, he set
things up to use leading-edge technology: IBM punch cards. Here is an
extract from the original announcement of the Atlas (Ethnology I: 113 ff)

============ Long quote begins =========================

The editors of this journal feel that ethnological science has long stood
in need of some coordinated means by which the vast accumulation of
ethnographic knowledge can be synthesized, classified, and subjected to
cumulative reanalysis and revision. They have decided to devote to this
purpose a special section in each issue. The pattern of organization
which will be followed in this and subsequent installments comprises
three elements:

1. The definition of codes for the classification of particular types
of ethnographic information.

2. One or more tables in which the data are presented in codified or
symbolic form for a number of societies.

3. A section of "Notes" clarifying the identification of each society,
presenting information on its geographical position and population,
offering occasional comments on moot points, and listing a few of
the most important bibliographical sources on each culture covered.

A primary purpose of this first installment is to present a geographical
classification of the peoples of the world. This is derived . . . from
that employed by Murdock in his "World Ethnographic Sample" (AA 59:
664-687, 1957), which divides the world into six major ethnongraphic
areas and subdivides the latter into 60 cultural and geographical regions.
By selecting either one or two societies from each region . . . a total
of 100 societies has been arrived at, representing a fairly even
distribution throughout the world.

A second objective is to refine and improve the codes used in the above-
mentioned article for the cvlassification of data on social organization.
In addition, a new code is proposed for indicating the relative
dependence of each society on the several major types of subsistence economy.

Third, among the criteria of selection, prominence has been given to the
inclusion of societies on which correspondents have supplied important
addenda and corrigenda to the entries in the "World Ethnographic Sample",
as well as on societies on which important new ethnographic information
has subsequently become available . . .

<<big snip>>

Since some readers may wish to enter the codified data on punch cards for
machine calculations, the system of classification has been adapted for
this purpose. The numbering of columns, for example, is spaced so as to
allow for all necessary punch-card positions. Thus the second column is
numbered 3 since two positions are required to enter the data from Column
1, and the third column is numbered 7 since four positions are required
for Column 3.

=================== End of long quote =======================

As the years went on, data codified and published in the continuing
Ethnographic Atlas covered more and more societies and ever-increasing
numbers of coding categories. (The inherent limitations on how many
columns you can use on an IBM card put a firm ceiling on how far that
could go, of course.)

And what was it all good for? Well, with a set of punch cards carrying
all the data in the Atlas, you could cross-correlate any feature coded in
the Atlas with any other, for one thing. In fact, as I recall Al Coult
and a colleague actually published a tabulation that was, in essence, a
printout of all possible two-way correlations of all the entries in the
Ethnographic Atlas, plus some kind of report of the statistical
significance of each correlation. I suppose if you hypothesized some wild
connection (matrilocal residence with 7-tone musical scales, say) and both
your categories happened to be coded in the Atlas, you could just look up
how the two correlate with each other (or don't). Think of all the cross
checking time you'd save! Look at the pre-calculated statistical
significance! Wow! It's an automatic hypothesis checker!!!

Of course, there are one or two little assumptions that might not check

The trouble begins with Galton's problem, originally raised when Tylor
first presented a cross-cultural statistical comparison in support of a
hypothesis. The easy way to state Galton's problem is to ask "Should you
count the Tlingit and the Haida as one case or two? How about the Apache
and the Navajo?" The next level of sophistication asks about the
"sampleness" of ANY so-called cross-cultural sample. How do you know what
universe the sample is drawn from? To what universe (beyond your
"sample") can you extend any generalizations you come up with? Is there
any justification at all in using statistical methods that are valid only
for drawing generalizations from a statistically random sample when there
is no way of knowing whether ANY ethnographic sample is, or can be, truly

There are proposed solutions to these problems, of course, notably in the
life work of Raoul Narroll, among others. But I am not convinced that
you will find valid use of those solutions in the Ethnographic Atlas as
such. All of which once led me to write a verse to my song, "The
anthropologist", sung to the tune of "Jesse James". It went like this:

Old Pete Murdock while at Yale
Got statistics by the bale --
At the numbers game he's never been exceeded.
With just IBM to aid him,
Understanding will evade him:
It's ideas, not numbers, that are needed.


I'm an anthropologist and I've got a good trait list
With the structure and the function well in view
I have got a lot of poop on the nature of the corporate group
And multilineal evolution too.

Ohmigawd, look at the time. Say goodnight, Mike.

All right. Goodnight, Mike!

mike salovesh, anthropology department <>
northern illinois university PEACE !