three centuries of women´s dress

ANTHROPOLOGICAL RECORDS

5:2

THREE CENTURIES OF WOMEN'S DRESS FASHIONSA QUANTITATIVE ANALYSIS

BY

JANE RICHARDSON AND A. L. KROEBER

UNIVERSITY OF CALIFORNIA PRESSBERKELEY AND LOS ANGELES

1940


BY


ANTHROPOLOGICAL RECORDSVol. 5, No. 2

ANTHROPOLOGICAL RECORDSEDITORS: A. L. KROEBER, E.W GIFFORD, R. H. LowlE, R. L. OLSON

Volume 5, No. 2, PP. 11-154, 11 diagramsTransmitted April 12, 1939Issued October 5, 1940

Price, 50 cents

UNIVERSITY OF CALIFORNIA PRESS

BERKELEY, CALIFORNIA

CAMBRIDGE UNIVERSITY PRESS

LONDON, ENGLAND

The University of California publications dealing with anthro-pological subjects are now issued in two series.The series in American Archaeology and Ethnology, which

was established in 1903, continues unchanged in format, but isrestricted to papers in which the interpretative element outweighsthe factual or which otherwise are of general interest.The new series, known as Anthropological Records, is issued in

photolithography in a larger size. It consists of monographs whichare documentary, of record nature, or devoted to the presentationprimarily of new data.

MANUFACTURED IN THE UNITED STATES OF AMERICA

CONTENTS

I. The problem . . . .The measures . .Sources . . . .Match of old and new data

II. Measurement data . . . . . . . . . . . .

III. Descriptive history of the proportions ofWidth of skirt . . . . . . . . . . . .

Length of skirt .Position of the waist .Diameter of the waist . . . . . . . .Decolletage . . . . . . . . . . . . .Width of decolletage . . . . . . . . .

IV. Periodicity . . . . . . . . . . . . . . .

Skirt length . . . . . . . . . . . . .Skirt width . . . . . . . . . ... . .

Waist length . . . . . . . . . . . . .Waist width . . . . . . . . . . . . .Decolletage . . . . . . . . . . . . .

Decolletage width . . . . . . . . . .Comparison . . . . . . . . . . . . . .

dress* * a

* * v

* * X

V. Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

VI. Interrelations of dimensions ........................

VII. Variability and stability of style . . . . . . . . . . . . . . . . . . .Year-to-year variability .........................Variability within the year .......................Causality of change ...................... ...

VIII. Conclusions .................................Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

TABLES

1.2.3.4.5.6.7.8.9.

10.11.12.13.14.15.16.17.18.19.20.21.22.23.24.25.26.27.28.29.

Comparability of observers ........................Comparability of sources .........................Ratio of dress diameters to height of figure: year-by-year means, 1787-1936Year-by-year means, 1605-1786 . . . . . . . . . . . . . . . . . . . . . . .

Ten-year averages, 1605-1936 . . . . . . . . . . . . . . . . . . . . . . .Five-year averages, 1788-1936. .. .............. ......Ten-year averages, 1788-1936, by decades ending in 6 . . . . . . . . . .Five-year moving average of dress diameters, 1788-1934 . . . . . . . . . .Dates of maximum and minimum skirt length . . . . . . . . . . . . . . . . .Summary of periodicity in skirt length . . . . . . . . . . . . . . . . . .

Dates of maximum and minimum skirt width ............Summary of periodicity in skirt width. ..... .............Periodicity in waist length ........ .. .. .... ... .....Periodicity in waist width .......... .......... ....Periodicity in decolletage . . . . . -Periodicity in decolletage width .......Comparison of six periodicities .............Interrelations of certain dimensions, by five-year periods, 1788-1936 .Periodicity of dimension interrelations . . . . . . . . . . . . . . . . .Percentage deviations of actual year means from trend, 1788-1934..Five-year averages of annual deviations from trend, 1788-1934 . . . . .Ten-year averages of annual deviations from trend, 1788-1934 . . . . .Frequency of fluctuation units, 1788-1934.............Percentage sigmas of annual means, 1787-1936 . . . . . . . . . . . . . . .Five-year averages of percentage sigmas, 1787-1936 . . . . . . . . . . .Ten-year averages of percentage sigmas, 1787-1936. . . . . . . .Distribution of size of variability coefficients among the six measures . .Maxima of five-year averages of coefficients of variability . . . . . . . .Extremes of dimension and variability by period . . . . . . . . .

[iii]

Page... 111

. * 112. 113. 113. 117

127127129129129130130130130131132132132134134135136137137141147148151

113116117120122123124125131131132132132133133134134136136137139139141141143143143144147

0 0

. 0

0 0

0 *

iv ANTHROPOLOGICAL RECORDS

FIGURES Page1. Vertical dimensions, 1787-1936, moving-average and year values. .. . . . . . . . . . . . 1142. Horizontal dimensions, 1787-1936, moving-average and year values. .. . . . . . . . . . . 1153. Dimensions 2 and 3, by ten-year means, 1605-1936. .. . . . . . 1224. Dimensions 4 and 5, by ten-year means, 1605-1936 .............1225. Dimensions 7 and 8, by ten-year means, 1605-1936 . . . . . . . 1246. Skirt width and waist width, 1787-1936. . . ... . . . . . . . . . . . . . . . . . . . . 1287. Frequency of deviations from trend, by fluctuation units, 1788-1934 . . . . . . . . . . . . 1408. Variability and amplitude of dimension, skirt, 1787-1936. .. . . . . . . . . . . . . . . 1449. Variability and amplitude of dimension, waist, 1787-1936 .... . ..... . . o. . . . 145

10. Variability and amplitude of dimension, decolletage, 1787-1936. .. . . . . . . . . . . 14511. Clustering of extremes related to variability, 1605-1936 . ... . . . . . . . . . . . 146


BY


I. THE PROBLEM

This study is an attempt to define stylisticchanges in an objective and quantitative man-ner.1 That dress fashions were chosen for in-vestigation, rather than works of pure or in-dustrial art, is due to the greater ease ofobtaining material which is not only fairlyabundant but strictly comparable from decade todecade and even from century to century. Thisdesideratum enforces that the range of materialbe narrow: one cannot compare landscapes withgenres, atmospheric treatments with portraits.Also the material must not be too utilitarian:chairs in one period may be primarily objectsof a certain degree of state, in another theymay deliberately consider comfort or serve forlounging. Women's evening or formal dress hasfulfilled a fairly constant function for severalcenturies. At the same time it is about as freefrom utilitarian motivation as dress can wellbe. Furthermore, for well over a century it hasbeen designed and published in fashion plates,which have often been preserved where most ofthe costumes themselves have long since perishedor become inaccessibly scattered. In short,formal dress, as a topic for investigation,possesses the advantages of representing an artwhich while not of the highest order is rela-tively free and self-sufficient; relativelylittle limited or warped by considerations ofexternal utility; specific and uniform enoughto be comparable from one period to another; ofa nature which precludes complete repetitivecrystallization and stand-still; and on which,with reasonable industry of search, there canbe accumulated fairly adequate information overa long span of time.

It is for these reasons that this type ofcostume has been chosen for study, rather thanbecause of any special importance or interestwhich it may possess in itself. In other words,it provides a convenient and promising set of

1For their kindness and generosity in offer-ing their collections of original fashion platesas source material for this survey, the authorswish to express their appreciation to the De-partment of Decorative Art, and the late Mr.Robert P. Utter, of the University of California;to Mr. John Howell, Mrs. Morton Gibbons, and theS. & G. Gump Company, of San Francisco; and toMrs. August Ericson, of Berkeley. Assistance inthe preparation of these materials was furnishedby the personnel of Works Projects Administra-tion Official Project No. 665-08-3-30, Unit A-15.

data for a study of the problem of how stylisticor aesthetic changes prove to take place whenthey are examined quantitatively instead ofthrough subjective intuition or feeling. It can-not of course be asserted that the change be-havior of women's evening dress would follow thesame patterns as style changes in painting ormusic or even in some other type of dress. Butany findings will presumably have some signifi-cance for the wider problem of how aestheticstyles change in general; to which in turn wemust have some answer before we can hope to in-quire fruitfully why they change.

The investigation had its beginning in a briefarticle by Kroeber in 1919.2 The techniques ofexamination there developed--which will be ex-plained in a moment--are now applied to a muchlarger body of material. The assembling of thisnew material was the first contribution ofRichardson. Whereas the earlier article coveredthe seventy-six years from 1844 to 1919, thepresent study carries on to 1936 as well as backto 1787, continuously except for two years (1822,1833) for which no data were encountered. Thisdoubles the span for continuous data. Back of1787, contemporary portraits and pictures hadlargely to be substituted for pre-wear fashionplates, and they run fewer; but a fair set ofspecimens was assembled back to 1605. Our totaltime range is thus three hundred and thirty-twoyears--longer, we believe, than in the over-whelming majority of statistical studies in eco-nomics. To be sure, the seriation is badly brokenbefore 1787. The decade 1631-40 yielded twenty-one available illustrations, the double decade1691-1710 none at all; 1711-20, twenty, but 1721-30 only three. Our pre-1787 findings are there-fore far less significant and reliable than thosesince 1787. Our more detailed analysis is ac-cordingly based wholly on the last one hundredand fifty years. But the findings made there,projected backward, and supported by the inter-mittent materials over the preceding one hundredand eighty-two years, allow some tentative con-clusions for the whole span of three hundred andthirty-two years.

To the figures computed and plotted year byyear, we have added a five-year moving average,

2On the Principle of Order in Civilizationas Exemplified by Changes of Fashion, Amer.Anthr. 21:235-263. Cited hereafter as Kroeber.

[111]


which of course smoothes out the mere annualvariations and gives a much more vivid pictureof the trend of fashion at any one time. On theother hand, the deviation or fluctuation of eachyear's style from the average for five years isalso much more clearly brought out by this newdevice. This annual fluctuation is obviously ameasure of the stability of the style.

Another type of variability is that withinthe year. How different are the several dressesof one year from one another, as expressed bytheir "sigma" or standard deviation from theirmean? The sigmas as compared over a period ofyears express the changes in variability.

In short, we have worked out quantities whichexpress the extremes of certain features ofwomen's dress style; the times of these extremesand the intervals between them; the rapidity andconsistency of the trends of change; and the de-gree of homogeneity or stability of the styleboth in a given year and over longer spans.

THE MEASURES

The traits or features of dresses dealt withnumber six. These comprise three vertical andthree horizontal diameters: of the skirt ordress as a whole, of the constricted middle orwaist, and of the decolletage or cut-out at theneck. We are really examining the dimensions ofthe silhouette of the whole dress. There aremany other features of probably equal signifi-cance, and of which fashion is perhaps even moreconscious: trains, sleeves, girdles, flounces,yokes, and so on. All these however come andgo. They are never permanent, but sooner orlater disappear completely for a time. Thismeans that only short-range comparisons can beinstituted for them. The skirt and waist di-ameters, however, and in full dress the decol-letage, cannot be escaped, as long as the funda-mental style of women's wear remains at all. Itis this permanence of the six silhouette dimen-sions that has led to our confining attentionto them.

All measurements were made on fashion platesor other pictures with calipers and ruler inmillimeters. To render them comparable, theyhad to be reduced to a common standard. Forthis the "total length of figure" was chosen andrecorded as measurement No. 1. The six dimen-sions were then converted into percentage pro-portions of this. It is these percentages thatare presented and dealt with throughout. Itseems useless to publish the raw or absolutemeasures; but they have been preserved. Actu-ally, the basic measurement No. 1 is not thewhole length of figure, but the length up to themiddle of the mouth. The top of the head doesnot answer, because of varying increments ofcoiffure and headdress.

All six of the dimensions are maximum diame-ters. They are as follows:

No.No.No.No.No.No.

2,3,4,5,7,8,

length of skirt or dress.length of waist.length (or depth) of decolletage.width of skirt.width or thickness of waist.width of decolletage.

Originally another measure was made: No. 6,maximum width of skirt if this width occurredabove the hem. This was soon dropped as too ir-regular in occurrence, and it is mentioned onlyto account for the gap in the numbering; thoughstylistically, as in recent years, this diametermay be of importance.

In detail, the measurements were executed asfollows:

No. 1, or base: Total length of figure fromthe center of the mouth to the tip of the forwardtoe.

No. 2: Distance from the mouth to the bottomof the center front of the skirt.

No. 3: Distance from the mouth to the minimumdiameter across the waist. The girdle, or thelower edge of the corsage part of the dress, maycoincide with this or lie above or below thisdiameter. The girdle and edge have been disre-garded because neither is a permanent feature.

No. 4: Depth or length of the decolletage,measured from the mouth to the middle of the up-per corsage edge in front.

No. 5: Diameter of the skirt at its hem orbase.

No. 7: Minimum diameter in the region of thewaist. See comment under No. 3.

No. 8: Width of the decolletage across theshoulders.

Full-face or nearly full-face figures wereused so far as possible. If the cases availablewere few, profiles and near-profiles were in-cluded. A side view eliminates decolletagebreadth, and, if the forearm is held horizontally,as is frequent in some periods, one or both waistmeasures may also be lost. Otherwise, profiles,and especially semiprofiles, seem mostly to yieldin fashion plates results not very different fromfull-face views.

In the seasons covered by monthly fashionjournals, the winter months are of course theones in question for full dress. So far as pos-sible, illustrations were sought in the Januaryto March issues. If these did not suffice, theApril to June numbers of the journal were ex-amined, or preferably the December and Novemberissues of the preceding calendar year. For in-stance, January-March of 1850 plus November-December of 1849 have been counted as 1850.

An absolute requirement was that each particu-lar figure be dated in a specific year. Recon-structed "typical" fashions, even if for a givenyear, such as abound in most histories of costume,were of course of no use. Moreover, no approxi-mate datings were included. The only exceptionoccurs in the case of a few Van Dyck and Watteauportraits, whose dates are known to fall within

112

RICHARDSON AND KROEBER: THREE CENTURIES OF WOMEN'S DRESS FASHIONS

periods of several years. As the pre-1787 datahave been used only in ten-year blocks, theseapproximations would not matter unless theperiod lapped over from one block into another.Such overlaps have not been used statistically,except that a group of ten Watteaus dated 1710-16 has been included in the 1711-20 average.

The old Kroeber measurements for 1844-1919were limited to ten figures for each year.Generally some of these ten failed to show oneor more of the six diameters. Richardson soughtas many examples as possible, so that therewould not be less than ten cases for each meas-urement. This was not a hard task for the lastdecade and a half. Back of 1844, however, tenladies per year became a rarity; ten full-faceviews hardly ever occurred. Therefore it wasnecessary to get perhaps six dimensions from onefigure, content oneself with two from the next,and hope that in the end none of the dimensionswould be wholly unrepresented for the year.Nevertheless, there were richer years: 1799yielded twelve illustrations; 1809, fifteen;1841, twenty-six.

SOURCES

To fill out the years from 1920 to 1936,Vogue and Harper's Bazaar were consulted, and,for 1920, Costume Royal. Each of these Americanmagazines devotes a great deal of space to thecreations of Paris designers. If, however,there were not sufficient of these Paris modelsin a given year, the gap was filled by unsignedAmerican style plates.

Before 1844, it was necessary to go from onebroken set to another, to the few books of copi-ous dated illustrations (Price: Dame Fashion;Fischel and Boehn), and to the engravings, draw-ings, and portraits by fashionable painters offashionable women. In these different sourcesit was possible to find many plates from thePetit Courrier des Dames, previously used byKroeber for 1844-68, and other elegant litho-graphs issued in monthly or quarterly seriessuch as "Wiener Moden," the "Galerie des Modes,"

and the Ankerman engravings for the Ladies'National Magazine, of London.

Seasonal fashion plates are scarce, however,before the French Revolution, not only on ac-count of their age, but also because they werenot published to a great extent. Hence we be-came increasingly dependent on painters and en-gravers, such as Winterhalter, Debucourt, Rey-nolds, Moreau le Jeune, Chodowiecki, Boucher,Nattier, Fragonard, Hogarth, Watteau, Terborch,Codde, Velasquez, Van Dyck, and minor paintersof the Dutch and French schools.

The idealized lithographed fashion plates from1789 on, whether published in Paris, Karlsbad,or Vienna, are strikingly uniform. There arechanges in face and pose only with the advent ofthe wood-cut and the zinc-engraved ink drawingand photograph of recent years. The earlierpainters are less subject to conventionalizationthan the lithographers. Faces and attitudes areindividualized, waists are thicker, and the ex-cellent likenesses are often far closer to thephotograph of today than to the draftsman's orlithographer's formalized delineations. Thechange from lithograph or drawing to photographis comparable to the reversed change from paint-ing to lithograph. Thus we may say of our spanof three hundred and thirty-two years, that thedata are rather conventionalized for the hundredyears 1789-1889, but are tempered with realismbefore and after.

MATCH OF OLD AND NEW DATA

Although the measurements are easily taken,the question may arise as to comparability ofthe Kroeber data for 1844-1919, and those ofRichardson for the remaining years. In none ofthe six measures as tabulated or graphed, isthere any large offset between 1843 and 1844,and only one (No. 5) between 1919 and 1920. How-ever, to see if there were any personal equationof measurement, 1844-46 and 1919 were measuredon new data by Richardson, and a comparison withthe Kroeber measures is herewith appended (table 1).

TABLE 1Comparability of Observers*

1844 1845 1846 1919Dimen. K R K R K R K R

2 .... 97.910 98.223 97.510 97 311 98.29 98.010 84.210 89.418

3 .... 28.910 28.817 27.910 26.910 28.49 28.89 24.110 25.0184 .... 14.610 14.32 3 14.19 13.110 13.18 12.010 14.29 14.918

5 .... 57.0Q7 58.21'4 59.46 58.17 57.34 62.87 33.210 12.5187 .... 8.210 9.713 8.44 9.810 8.37 8.89 13.29 13.9188 .....20.310 19.22 19.78 19.29 18.76 1 19.38 12.98 10.218*Superior figures to the right are the number of pictures measured from which the per-

centage averages here given are derived.

113

I II

0

~~~~~~~~~~~~~~~~~~~~~~~~~~~100

0

0

00

.

0

0

0

0

:

0

N

0

0

0

0

0

0

0

.

0

0

0

0

0

0

.

**

_0t-o

_ O

q*- Z~~~~~~~U

-

_Uh

U)

,

-

- 9aU

-h-

~ >

N

0

N.

2-

It

N-

N-

* -,>

9 a a am aw wR

dwi r. 4M cr4 0 --I !4 I 'ti 9 1 % . '(. A m a g A A A R fl,

I I I I IIiJII II ItIII I I I I I I I I-

0~~~~~~~~~~~~~~~~~~~~~~~

N 0

N 0~~~~~~~~~~~~~~~~~~~~~~~~~~~

10- ~ ~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~

0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0ND- 0El--

0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 4

* 0~~~~~~~~~~~~~~~~~~~~~~~~~~f

* *~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0r

coN0000.00000. *-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-

0)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

0) *~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0

00~~ ~ ~ ~ ~~40

N- 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~c* 0)~~~~~~~~~~~~~~~~~~~~~r-


The two sets of data for 1844-46 seem similarenough. In only one case, dimension 5 for 1846,is the difference of the two sets of averagedmeasures more than 1.5.

The 1919 comparisons, however, show a start-ling difference of 20.7 in dimension 5: 33.2against 12.5. Here Kroeber's data were takensolely from the March number of Vogue.3 Fourof these dresses had trains, six did not. Hispercentages run (* denoting train): 33; 22; 134or *33; 10; 19 or *67; 16 or *60; 19 or *64;17; 17; 9. He used the higher train widths forhis average of 33.2. The Richardson series for1919 was taken from Vogue of December 1918, and

less fashion variability, hence the two sets ofmeasurements coincide more closely.

On the whole, it seems that the sets ofmeasurements of the two observers are suffi-ciently alike to be treated as parts of oneseries.

Another problem concerns the difference be-tween the signed French and the anonymous Ameri-can designs, as shown in any one magazine, andalso what difference might exist between themagazines themselves. Table 2 shows such varia-tions.

In only one case does a range as great as 4.3occur: No. 4, 1919, 17.5 and 13.2. This is about

TABLE 2

Comparability of Sources

1919 1919 1919 1920 1920 1920Vogue Costume Costume Vogue Costume Costume

Dimen. ord. Royal Royal ord. Royal Royalmodels ord. mod. French models models ord. mod. French models*

_ (Dec., 1918) (Apr., June) (Apr., June) (Feb., May) (Apr., June) (Apr., June)

2 ...... 89.7 91.0 87.7 83.3 83.0 79.73 ...... 24.9 26.0 25.1 26.4 26.1 27.44 ...... 17.5 13.2 14.0 16.4 15.3 13.75 ...... 12.6 10.0 14.9 15.2 17.6 15.77 ...... 13.8 13.0 14.8 14.5 14.9 14.68 11.1 9.0 10.5 12.2 12.5 13.0

*Includes one French model from Vogue.

from Costume Royal for April and June, 1919.All eighteen of these illustrations have notrains, and they average 12.5. The Kroeberaverage for the ten widths minus trains is17.5. This is much nearer the Richardson valueof 12.5, and it fits well between Kroeber 1918,20.3, and Richardson 1920, 16.7.

The other larger discrepancies for 1919 are

in skirt length: Kroeber 84.2, Richardson 89.4;and in decolletage width: Kroeber 12.9, Richard-son 10.2. These we cannot explain, unless itbe that since 1919 falls in a period of high in-dividual variability, as shown by the standarddeviations, any two samples of ten and eighteenplates, respectively, might well differ as much.The period adjacent to 1844-46 is one of much

3Kroeber, 244.4The smaller number given for the trained

dress is the width of the skirt exclusive of thetrain. These measures are from the manuscriptdata.

one and a half times the standard deviation forthe period 1916-22: 3.6, 0.8, 2.1, 2.0, 2.5, 3.7,4.6, mean 2.8. There is a suggestion that theFrench models of 1919 are closer to the ordinaryones of 1920: compare dimensions 2, 7. This mayreflect the fact that the French models used in1919 and 1920 are late spring modes, and are al-ready pointing the way to fashions of the follow-ing winter.

Our conclusion is that any difference betweenthe Kroeber and Richardson measurements, and be-tween models of different magazines, or dressesof French and American design, seems to be less,on the whole, than the fashion changes from yearto year. This also holds as regards the changein illustrations from drawings to photographs ofliving models in the second decade of the presentcentury, on which Kroeber has previously givensome sample data.5

5P. 244.

116

II. MEASUREMENT DATA

The data obtained from the several thousandmeasurements made by the two authors, respec-tively twenty years ago and more recently, willnow be given in the form of means for each di-mension or proportion in each year studied.

For instance, for 1806, eleven fashion platesor illustrations were found which showed lengthof skirt without impairment. These, in terms ofthe total length of figure as defined, ran to95.0, 97.5, 98.5, 97.3, 98.0, 97.1, 97.4, 98.5,99.5, 97.5, 98.4 per cent. (In other words, thefirst plate measured showed a "length of figure"[from mouth down] of 119 mm., a length to bottomof skirt of 113 mm.; the second is 95.0 per centof the first. The second plate happened also tohave the figure length 119 mm.; but the skirtlength 116, giving 97.5 per cent. And so on.)The mean of these eleven percentages is 97.7,

which is therefore the value assigned to dimen-sion No. 2 for 1806. Two of the eleven platesdid not show the full width of skirt; the ninethat did, yielded 25.2, 46.3, 13.0, 57.0, 35.6,50.4, 48.5, 35.1, 35.2 per cent, whose mean, 38.4,is the dimension No. 5 value for 1806. The indi-vidual raw and percentaged measurements, as herecited in illustration, are being preserved, andthe latter have of course been used in calcula-tion of the standard deviations or variabilitycoefficients presented and discussed in sectionVII; but they are not printed on account ofcost. Their means for each dimension each yearreplace them.

These means, on which all our other quantita-tive expressions rest, are given in tables 3and 4. The first of these two tables coversthe period from 1787 to 1936, in which almost

TABLE 3Ratio of Dress Diameters to Height of Figure:

Year-by-Year Means 1 7pV-1i 9Q

Year 2 3 4 5 78L.Sk. L.Wai. L.Dec. W.Sk. W.Wai. W.Dec.1787 98.73 27.43 7.23 56.33 11.21 7.911788 96.14 20.84 8.92 51.12 16.511789 .... 96.97 23.26 10.37 55.15 14.94 10.73

1790 .... 98.68 25.86 7.97 51.07 9.95 7.611791 ... 98.3 4 25.44 11.94 53.2 3 9.04 8.531792 .... 98.51 25.42 9.81 58.42 9.321793 .... 98.13 23.74 5.33 52.13 10.641794 .... 100.02 18.82 9.92 43.32 14.442 9.721795 96.98 20.17 12.97 52.66 12.35 7.021796 ... 98.2-5 20.64 10.24 43.8 3 18.011797 .... 99.43 20.72 13.32 49.33 14.92 8.011798 .... 98.68 19.99 11.58 44.410 10.59 10.381799 97.712 19.413 11.612 42.77 11.210 9.510

1800 .... 98.11" 18.11" 10.511 42.91" 12.08 11.961801 .... 96.47 17.96 15.54 42.26 16.02 11.921802 .... 100.02 19.42 16.31 59.61 11.32 14.911803 .... 97.712 19.012 12.610 38.58 12.88 13.081804 .... 95.110 19.410 12.38 44.79 13.55 14.341805 .... 97.312 18.311 14.410 35.07 14.85 14.361806 ... 97.711 17.911 12.59 38.59 14.3's 12.851807 .... 96.518 18.318 11.015 37.815 15.28 14.761808 ... 95.610 18. 97 11 .89 32.67 14. 13 13.8r'1809 95.815 17.415 11.714 24.813 15.113 13.713

1810 95.6 7 21.06 13.5 7 24.2 7 12.91' 12.341811 94.48 21.77 13.38 27.28 13.57 15. 051812 .... 92.74 20.94 13.03 29.44 13.42 17.021813 ... 88. 03 19. 23 11.13 28. 93 14.43 19.831814 94.2 4 19.84 13.2 3 31.34 12.84 17.2 3

1815 92. 95 17.3 5 11. 94 45.85 13.63 15.251816 .... 91.63 17.23 12.03 39.83 12.02 16.621817 94.72 20.12 16.52 42.92 13.51 9.621818 .... 93.93 20.63 11.32 38.63 11.32 16.421819 .... 90.11 15.51 13.21 41.1' 10.1' 16.3'

[117]

a _**vf-ovw

118

TABLE 3 (Continued)

Year L.Sk. L.Wai. L.Dec. W.Sk. |[W.Wai. |IW.Dec.

1820182118221823182418251826182718281829

1830183118321833183418351836183718381839

1840184118421843184418451846184718481849

1850185118521853185418551856185718581859

1860186118621863186418651866186718681869

1870187118721873187418751876187718781879

92.2396.43

96.9392.5295.5195.0292.8690.41093.93

93.7994.8490.8

89.4492.9995.91495.71197. 1O97.11497. 02296.92896.91797.32498.21097.51098.3998.41098.01097.910

97.81098.71097.61098.11097.91098.21098.3998.41099.610

100.010

99. 810100.01099.61098.71099.51099.81099.8997.91098.81L

100.010

99.1999.31099 ,1099.21099.29

100.01099. 21098. 71099.01098.710

21.4321.73

24.4321.5225.4125.2228.4526.71026.73

27.9923.7426.4

27.0428.0926.91327.31028.91028.713

28.42129.1 3

29.21329.02128.81027.91028.4928.9927.81028.710

28.61029.41027.01027.71027.01027.91027.7926.71026.81025.39

24.81024.91024.11024.9723.9722.8922.4921.21022.0921.810

22.2922.0922.91024.51022.21022.3923.61023.81024.81026.110

13.2210.12

12.1213.8210.9110.8212.2512.2812.21

12.2412.1312.9

14.0312.7712 .91413.4914.0814.813

14.32014.82714.51614.42214.41014.1913.1814.8913.41013.39

12.71013.91014.1912.81014.11013.31013.4713.91015.21014.410

12.31012.31013.21013.1913.5912.7912.81011.7912.7713.69

12.00813.0615.0613.8514.1714.1713.4713.5614.5613.36

37?.0342.73

46.3338.1248.6148.5253.2649.61055.63

48.7844.2454.4

64.3367.0970.01262.3870.9863.613

63. 91564.61 760.01460.116B58.2759.4657.3464.8459.6562.76

64.2661.3770.3670.267933783.0489.2486.25

100.36115.69

107.17104.3996.19

101.69100.11108.6199.8298.7988.41085.58

88.O674.9877.6884.81084.5679. 784.8876.41070.9862.05

10.0311.22

10.3812.4210.1110.4211.2610.0611.33

12.1813.5311.0

9.849.499.499.189.159.510

9.0208.8179.199. Pl9.1169.0108.448.378.868.588.48

8.388.498.3107.888.499.088.658.057.977.83

7.658.047.689.088.558.678.298.199.o89.04

9.249.03

11.0310.169.53

10. 49.548.768.968.88

17.2217.8318.3218.1222.5117.0223.2321.3921.42

24.7221.2123.8

23.9317.0717.91019.1619.9620.710

20.61819.31918.0820.71919.21019.8818.7619.6820.01020.09

20. 71021.2821.4921.2920.6821.01019.1919.61018.81018.38

18.1917.81017.9917.1918.1917.5818Q0916.7916.1816.0 7

18.5616.4516.7218.3315.0417.0o13.5213.7614.6815.08

--r--T-


TABLE 3 (Concluded)

Year 1__L.S.__Sk.L.Wai. L.Dec. W.Sk. W.Wai._J W . Dec.

1880188118821883188418851886188718881889

1890189118921893189418951896189718981899

1900190119021903190419051906190719081909

1910191119121913191419151916191719181919

1920192119221923192419251926192719281929

1930193119321933193419351936

98.81097.71096.61096.91096.41097.01095.8995.61095.71096.710

97,31097.31097.41098.81098.21098.71099.21099.91099.810

100.010

99.31099.710

100.010100.010100. 010100.01099.61099.61-099.31099.710

99.21098.71098.31092.6991.8991.11084. 31088.11085.31084.210

82.21783.12186.01490.71085.61377.01072.21169.71170.01071. 220

89.02095.11397.31098.31099.31098.21798. 825

27.61027.61026.01026.01026.21027.41027.31027.21027.61027.710

28.21028.31028.81027.01028.81027.41027.91028.91029.51029.710

30.51030.51030.11032.6932.31030.31028.81028.01025.4924.310

25.21026.11024.31025.51025.31024.4925.0924.3824.21024.110

26.51727. 92133.31430.91038.01335.51031.91137.91136.21032.520

28. 12026.81224.41025.01025.01025.21825.32 3

15.4614.2712.8812.8913.1914.0714.9912.9814.1913.71

14.1814.4913.4713.6814.0914.0914.31014.41014.71014.69

15.11012.51013.1915.2914.0914.61016.2513.3911.7914.67

13.3814.2613.4915.4714.4716.2616.4914.8913.7914.29

15.01713.42114.31413.31013.01312.61015.81113.01115.01013. 020

14.62014.91211.41011.2109.310

11. 11911.822

68.8852.3756.0854.7752.2956. O656.6950.9957.8851.59

50.21053.7951. 11055.01055.51060.7968.3860.01053.01065.310

52.51064.81058.91050.41056.51053.71056.01051.21049.01038.410

32.9923.21027.41033.7729.1746.11049.11055.7420.3833.210

16.81726.02126.41422.41021.31322.01016.51118.01121.21026.020

25.92026.01225.81025.91026.21038.71633.821

8.888.597.878.688.268.788.9108.388.3109.68

8.5109.2109.2109.399.2108.679.678.698.189.39

8.7109.4109.989.689.989.299.589,76

10.9812.89

11.7912.0913.2913.6913.91013.7912.6913.0913.41013.29

14.71715.22116.51415.41015.91315.61014. 21113.81114.71013. 42

16. 52013.01211.01013.11013. 7109.912

10.717

14.1814.7815.3717.3714.4615.2514.8814.2813.1913.29

13.5812.6614.3913.3814.0915.0715.2915.81011.9912.510

13.4913.31011.1613.0714.8715.3811.2612.3712.9812.110

13.0912.22611.5713.31015.2411.2512.3611.4710.6812.98

12.91714.72113.11414.81013.31312.61012.31111.21112.11011.110

11. 72012.81210.31013.21011.81011.4129.816

119


every consecutive year is represented and formost years the data are reasonably sufficient.(See figs. 1-2, pp. 114-115.) The second tablethe preceding one hundred and eighty-two yearsfrom 1605 to 1786, where the data are both dis-continuous and fewer. In both these tables, thesmall superior figures at the right indicate thenumber of cases measured.

Next follow combinations of these year-by-year means into the averages for longer blocksof years. Table 5 covers the whole three hun-dred and thirty-two years in ten-year periods.(See figs. 3, 4, 5.) This seemed a better in-terval than five-year averagings on account ofthe irregularity of the data before 1787. It-should be noted that these values, and all otheraverages of the year-by-year means, are un-

weighted for number of cases. For instance, forthe decade 1641-50, only four dated illustra-tions were found: three for 1641 and one for1647. In dimension 5, skirt width, the threefor 1641 averaged 72.2, the one for 1647 was78.0. The mean of these two values is 75.1,which is the one that will be found in table 5for this decade; whereas the weighted averagewould be 73.7.

Table 6 gives the five-year means for the

TABLE 4

Year-bv-Year Means

fuller period 1787-1926. Table 7 combines theseby pairs into ten-year means. It is, however,not a mere replica of the last part of table 5,because the absolute dates are different: wheretable 5 treats the years 1881-90 as a unit,table 7 lumps 1877-86.

Next comes the "trend" or moving average, intable 8. Here the value for any given dimensionfor 1840 is the average of the values for 1838,1839, 1840, 1841, 1842 as found in table 3; for1841, of 1839, 1840, 1841, 1842, 1843. In thisway, the "exceptionalness" of any single year isminimized by its being merged with the years on

either side, and a smoother curve results. Forinstance, from 1846 to 1855 skirt length was toall intents and purposes at a stand-still. Theindividual year 1852 comes out, in the measure-ments that happened to be made, somewhat lowerthan any other in that decade: 97.6 as againstnine others ranging from 97.8 to 98.7. It isdoubtful whether a fluctuation so small as thisis either significant or reliable. The movingaverage, which stays at 98.0, 98.1, 98.2 for thedecade, probably gives a truer picture of theevents. On account of its much greater steadi-ness, its variations are also much more readilygrasped, when they do become appreciable. Take

1605-1786

Year 2 3 4 5 7 8

1605 .... 100.0 23.5 7.0 76.41610 .... 97.12 19.42 3.62 68.1 11.62 5.82

1613 95.9 22.5 8.5 78.9 14.1 8.51617 .... 100.0 21.7 2.8 13.0 5.7

1622-27.. 100.02 25.22 3.72 13.6 7.91625 .... 26.5 16.3 16.31628 ... 96.4 28.6 2.4 53.5 14.3 4.81629 .... 97.44 25.9w 7.64 41.73 11.3 6.731630 .... 100.0 25.9 2.8 11.1 5.5

1631 .... 100.02 23.78 4.82 54.52 6.321633 .... 99.23 24.33 13.23 49.62 14.8 16.331634 97.83 22.63 6.73 50.43 7.631635 .... 100.04 24.14 9,.74 52.6 13.041636 .... 100.02 21.22 13.1 35.6 16.21638 .... 100.0 24.1 11.1 63.0 14.81639 .... 100.03 23.53 14.13 52.23 15.531640 .... 99.03 23.94 14.2 46.44 14.22

1641 .... 100.03 26.12 10.13 72.23 15.8 13.731647 .... 100.0 27.1 3.1 78.0 9.3

1656-57.. 100.0 36.6 10.0 87.5 17.51658 .... 100.0 27.4 6.8 82.2 11.7 10.91659 .... 100.0 29.2 9.0 126.2 16.2 22.21660 .... 100.02 29.72 12.22 88.92 18.82

1666 100.0 32.2 11.3 58.01667 .... 98.4 6.9 45.01668 100.0 30.2 15.1 75.5 22.6

120


TABLE 4 (Continued)

Year 2 HI3 4 5171678 ....

1679 ....

1680

1683 ....

16851686 ....

1710-16..1711 ....

1714 ....

1715 ....1716-18..171717191720

1721 ....

17261728

1731 ....

1734 ....

1735 ....

1738 ....17391740 ....

1742 ....

17431744 ....

1744-45..1749 ....

1750 ....

1751 ....1752 ....

1753 ....

1755 ....

1757 ....

1758 ....

1759 ....

17611762-64..1765 ....

1766 ....

1767 ....

1769 ....

1770 ....

1771 ....1772 ....

1773 ....

17751776 ....

1777 ....1778 ....

1779 ....

1780 ....

1781 ....

1784 ....

1786 ....

97.597.6296.2

97.32100.099.2

99.4996.597.52

100.0298.7296.4

100.0100.02

89.6100.0100.0

92.8298.0396.2296.4597.099.0

93.82

100.096.495.697.22

96.598.5293.5295.2291.497.298.0

89.592.394.5397.695.595.5

98.498.595.091.8796.5595.9395.4592.4292.3290.33

92.2293.3497.4

33.323.331.5

29.12

29.1

24.9729.025.3226.7225.7221.4

29.4

26.6

26.1328.8327.9528.228.6

25.6227.524.3

28.625.2

30.126.2226.9230.3327.531.230.2

25.025.027.33

31.42

22.130.828.924.8626.3227.9326.8433.5230.8226.63

23.924.4423.7

15.411.614.8

12.313.1

13.7515.310.0

12.114.3

15.02

18.310.5

12.915.3213.5216.4513.414.3

7.713.314.212.514.215.42

15.114.4210.916.714.712.217.4

16.815.6213.1314.911.515.22

11.619.216.511.8613.8516.0312.8518.4216.5215.53

16.411.038.8

33.440.0240.8

30.1263.748.7

52.6673.825.546.8458.3250.046.597.0

62.3

57.9

61.3276.466.654.24

64.3

134.5

70.091.0

102.9101.0

100.744.931.8259.O358.439.780.3

65.172.365.935.648.459.0

34.840.0

101.046.4757.6472.5277.0355. 3241.6257.93

57.550.0236.8

10.2

9.3

12.4211.9

13.6411.910.017.410.7

14.7

12.5212.816.713.413.4

10.814.211.4

12.9

12.314.310.013.0

14.5

14.512.7212.32

16.513.612.910.3210.7310.9

8.6

9.2

16.7

13.6314.510.013.0

14.7

15.0

13.213.810.9323.911.6

10.0

7.120.018.7

21.212.3

8.814.3

13.0

12.211.510.93

9.6

8.8

9.826.8

12.429.948.7

8.32

10.511.02

I I IA..A. _ _ _ _ _

I II I I I

121


for example the moving average of waist lengthfor 1923-32, which runs, slightly rounded fromtable 8: 33, 34, 35, 36, 35, 34, 33, 30, 28, 26,

IO0 210 0 610 810 1700 210 I 108020190 20 0 60 S0 190 20 jO

80-82-

8-86-88-90-

e2t-es-n-86-88-100-

16-

18-20-

22-

211-26-28-

30-

32-

_ 5-

_ 6-- 7-

- 8-

_ 9-- 10-

- 11-

_ 12-- 19-- 1t-

_ 18-16-

2

100-

90-

80-70-60-80-,10-80-20-

3

brings out 1926 as the peak of the curve. Inthe actual figures this is, however, low yearfor the series of,six from 1924 to 1929, and the

9002,0p O 610 810 171IX 0 ¶ GP08,06 I 00 ro610 9l 19&p00 o 0

4

5

Fig. 3. Skirt length and waist length, dimesions 2 and 3, by ten-year means, 1605-1936.

as against the year-by-year values of 30, 38,36, 32, 38, 36, 33, 28, 27, 24 of table 3. ThEsmoothed or idealized moving-average series

Fig. 4. Decolletage depth and skirt width,dimensions 4 and 5, by ten-year means, 1605-1936.

peak is left double and indeterminate betweena 1924 and 1927. We assume that there are statis-

tical techniques expressing the probability or

TABLE 5

Ten-Year Averages, 1605-1936*

Period 2 3 4 5 7 8

1605-10 .... 98.63 21.53 5.33 72.32 11.62 5.821611-20 98. 22.12 5.72 78.91 13.62 7.121621-30 .... 98.58 26.47 6.69 47.64 13.36 6.261631-40 99.521 23.422 10.918 50.517 14.81 13.0191641-50 .... 100.04 26.63 6.64 75.14 15.81 11.54

1651-60 .... 100.05 30.75 9.55 96.25 14.02 l7.451661-70 .... 99.53 31.22 11.13 59.53 22.611671-80 .... 97.14 29.43 13.93 38.14 9.82 16.711681-90 .... 98.84 29.23 12.72 47.54 12.231691-1700

1701-10 ....1711-20 98.620 26.116 13.411 56.315 13.19 13.271721-30 96.53 26.61 14.42 60.12 15.01

have been averaged,the values are av-tables 4 and 3.

*To 1780, all the percentaged means within a decadeirrespective of the year means in table 4. From 1781,eraged from the year means for each decade as given in

122

)n-_


TABLE 5 (Continued)

Period 2 3 4 5 7 8

1731-40 .... 96.614 27.9'13 14.312 60.79 13.88 14.771741-50 .... 96.67 26.26 12.97 99.95 12.34 14.04

1751-60 .... 95.810 28.911 14.58 60.210 12.8w 13.951761-70 .... 94.89 26.38 14.111 54.47 13.25 10.671771-80 .... 94.229 28.524 15.628 50.926 11.910 9.3121781-90 .... 96.229 24.225 10.124 51.121 12.312 9.581791-1800 .. 98.457 21.258 10.754 48.349 12.247 9.332

1801-10 .... 96.8104 18.898 13.287 37.882 14.057 13.6541811-20 .... 92.536 19.435 12.929 36.236 12.528 16.0271821-30 ....... 94.l~13 9 25.338 11.827 47.938 11.031 20.5281831-40 . ... 94.591 27.386 13.580 62.375 10.070 20.5831841-50 ........97.7136 28.6125 14.0130 61.187 8.792 19.6107

1851-60 .... 98.739 27.098 13.796 86.31l 8.270 19.9901861-70 .... 99398 23.090 12.885 97.186 8.562 17.4821871-80 .... 99.199 24.098 14.062 76.475 9.552 15.4521881-90 .... 96.696 27.1100 13.784 53.883 8.583 14. 6721891-1900 . . ~98.9 28.7100 14.391 57.596 9.089 13.885

1901-10 .... 99.7100 28.898 13.985 51.299 10.384 12.9741911-20 .... 89.7103 25.0103 14.888 33.5104 13.3100 12.4771921-30 .... 79.5140 33.2104 13.8140 22.6140 15.1140 12.7741931-36 ....I 97.885 25.381 11.683 29.479 11.97' 11.671

TABLE 6Five-Year Averages, 1788-1936

Period 213141 5 71788-91 ....1792-96 ....1797-18011802-06 ....

1807-11

1812-16 ....1817-211822-261827-31 ....1832-36 ....

1837-41 ....

1842-461847-51 ....1852-56 ....1857-61 ....

1862-66 ....

1867-71 ....

1872-76 ....1877-811882-86 ....

1887-91 ....1892-961897-19011902-06 ....1907-11 ....

1912716 ....1917-21 ....1922-261927-311932-36 ....

97.7*98.398.097.695.6

t91.993.595.093.192.3

96.797.698.298.0*99.6

99.599.099.498.6t96.5

t96.598.599.7*99.999.3

91.684.682.3t79.098.4

*High points.

*24.521.719.2

t18.819.5

18.919.924.126.727.1

28.5*28.7*28.727.525.7

23.6t21.823.126.026.6

27.828.029.9

*30.825.8

t24.925.4

*33.932.325.0

tLow points.

t9.29.6

12.5*13.612.3

12.212.9

tll.912.213.1

14.314.113.613.513.6

13.412.614.114.213.5

13.813.914.3

*14.613.4

15.214.213.814.1

tll.0

53.350.044.343.3

t29.3

35.040.545.450.363.9

65.159.062.578.4

*102.7

101.287.182.166.155.1

52.858.159.155.138.9

37.130.4

t21.723.430.1

12.312.912.913.3*14.1

13.211.210.811.69.9

9.18.88.58.4

t7.9

8.48.9

10.18.78.4

8.89.28.89.6

11.4

13.413.9

*15.514.311.8

8.78.410.313.913.9

17.215.519.0

*22.420.7

19.919.320.320.718.5

17.716.716.114.415.413.314.413.413.112.5

12.712.513.211.8

tll.3-- --- - - - - - - - - - - --- - - I -- - - I -- -

123


improbability of 1926 really being a year oftemporary recession near the climax of the curve,

or on the contrary of its low value being merelya by-product of the unrepresentativeness of thesmall number of pictures measured. But thesetechniques are cumbersome, and the issues in-volved are small. Whether the waist line over

a period of one hundred and fifty years reachedits lowest position (highest percentage) pre-cisely in 1926 or perhaps rather in either 1924or 1927, can be of no great moment. The essen-tial truth seems sufficiently expressed, andcertainly much more vividly, by the movingaverage.

For this reason we have graphed the movingaverages in figures 1 and 2 (pp. 114-115).These should be considered our basic diagrams.We have shown in these diagrams also the actualvalues for each year, from table 3; but thesevalues have been indicated by disparate pointsor dots, whereas the movi~ng-average values areconnected by a continuous line.

Earlier than 1787, the data are too scatteredfor a satisfactory moving average, and it hasnot been attempted. Its place is partiallytaken by the means of means in ten-year blocksin the first part of table 5.

lGpO 20o 61 O 1°0 17,OO)0 6 0 al1 1 210 306,08G)0 191Z0 31016-

1i-

it -

18-

12 -

11 -

10 -

9-

8-

23-

21 -

19-

17 -

16-

13-

11-

9-

7-

8-

7

8

II

Fig. 5. Waist width and decolletage width,dimensions 7 and 8, by ten-year means, 1605-1936.

TABLE 7

Ten-Year Averages, 1788-1936,By Decades Ending in 6

Period 2 3 4 5 7 8

1788-96 .... 98.3 23.1 9.4 51.7 12.6 8.5

1797-1806 .. 97.8 19.0 13.1 43.8 13.1 12.1

1807-16 .... 93.7 19.2 12.3 32.2 13.7 15.5

1817-26 .... 94.1 21.8 12.4 42.6 11.0 17.0

1827-36 .... 92.7 26.9 12.6 56.3 10.9 21.6

1837-46 .... 97.2 28.6 14.2 62.0 8.9 19.6

1847-56 .... 98.1 28.1 13.6 70.5 8.5 20.51857-66 .... 99.5 24.7 13.3 102.0 8.1 18.1

1867-76 .... 99.2 22.5 13.4 84.6 9.5 16.4

1877-86 .... 97.6 26.3 13.9 60.6 8.6 14.9

1887-96 .... 97.5 27.9 13.9 55.5 9.0 13.8

1897-1906 99.8 30.3 14.4 57.1 9.2 13.2

1907-16 .... 95.5 25.4 14.3 38.0 12.4 12.6

1917-26 .83.4 29.7 14.0 26.1 14.7 12.9

1927-36 .... 88.7 28.6 12.5 26.8 13.0 11.5

124

- - . ----- --- . --r-I ... -- --- -- -- --- ----- - . - -- --- -

Year 2 3 4 5 7 8L.Sk. L.6rai. L.Dec. W.Sk. W.Wai. W.Dec.

- - - I - - - - --.- I It

RICHARDSON AND KROEBER: THREE CENTURIES OF WOMEN'S DRESS FASHIONS 125

TABLE 8Five-Year MovinR Average of Dress Diameters. 1788-1934

1?881?89

1?9017911?921?931?941?9517961?9?17981?99

1800180118021803180418051806180?18081809

1.810181118121813181418151816181718181819

1820182118221823182418251826182?18281829

1830183118321833183418351836183?18381839

184018411842184318441845

97.597. 7

24.224.5

t8.69.2

50.153.3

12.512.3 8. 7

8.98.98.68.48.4

t8. 28.88.79.9

10.3

ii. 712.213.213.713.811.814.013.913.513.9

14.415.616.316.817.215. ?15.014.815.215.5

16.917.417.919.219.019.820.421.121.522.4

22.522.8

*23. 421.520.719.519.618.919.619.9

19.719.919.619.419.319.6

97. 798.1

*98. 798.498.398.598.698.298.498.0

98.298.097.597.397.696.996.496.696.295.6

95.893.393.092.4191.992.393.592.692.593.5

93.293.994.595.395.094.593.293.593.293.1

92.793.392.292.092.393.594.295.796.596.7

97.097.097.397.49?.69?. 9

24.1*24. ?23.822. ?21.720.820.020.119.719.2

18.918.818.818.818.818.618.618.218. ?19.5

20.020.020.519.818.918. ?19.0

t18. 119.019.9

19.820.822.323.324.125.025.426.527.026. ?

26.326.226.326.227.127.327.628.028.028.5

*28. 9*28. 9*28. 928.828. ?28.6

9.89.09.0

10.09.6

10.311.611.911.412.5

13.113.313.4

*14.213.612.612.412.312-.112.3

12.712.512.812.512.212.913.013.013.212.9

12.012.212.3

til. 711.912 ' 012.0

til. ?11.912.2

12.312.412.812.913.113.313.413.613.914.3

14.5*14.6.14.514.414.114.2

53.8*54. 051.651.950.048.246.746.644.644.3

46.445.245.644.943.338.937. 733.731.629.3

27.6t26.928.232.535.037. ?39.741.639.940.5

39.941.841.043.945.446.947.651.151.150.3

50.550.7.54.957.563.965.966.966.866.165.1

64.662.461.460.559.060.0

11.910.710.611.112.914.014.013.413.312.9

12.212.713.113.713.314.114.4

*14. 714.314.1

13.813.813.413.513.213..312.612.111.411.2

10. 710.411.011.010.810.910.810.611.011.6

11.612.011.610.99.99.49.49.39.29.1

9.19.19.08.98.88.7

*High points. tLow points.

Year L .Sk . L .Wai . L .De c. W .Sk . W .Wai . W .De c.

-----..-I- - - - t - - - - - - - - I-- - I~~~~~~~~~~~~~~~

126 ANTHROPOLOGICAL RECORDS

TABLE 8 ( Cont inued )

98.1980.098.198.2

98.098.098.098.198.098.298.598.999.299.6

*99.899.699.599.599.599.199.299.499.199.0

99.399.499.299.499.499.399.299.198.998.6

98.297.797.396.996.596.3

t96.196.296.296.5

96.997.597.898.198.599.099.299.599.699.7

99.899.899.8

*99 .9*99.9

28.428.328.528.7

28.328.327.927.827.527.427.226.926.325.7

25.224.824.524.123.623.022.522.421.9

t2l.8

22.222.722.822.823.123.323.324.125.226.0

26.426.726.726.626.626.827.127.427.627.8

28.128.028.228.128.028.028.528.729.329.8

30.130.7

*31.2*31.230.8

14.013.713.513.6

13.513.413.513.613.513.514.014.013.813.6

13.513.112.913.013.112.812.712.7

t12.6t12.6

13.313.513.614.014.113.813.913.814.014.2

14.013.713.713.413.513.513.813.913.913.8

13.913.813.913.913.914.114.314.414.614.3

14.014.114.013.914.6

59.960.861.762.5

63.665.769.172.878.481.687.694.999.7

102.7

104.7*104.9101.8102.1101.2101.899.196.292.187.1

82.982.282.080.282.181.979.174.672.666.1

62.058.856.854.255.154.155. 755.653.452.8

52.952.353.155.258.159.959.561.559.859.1

58.958.456.656.955.1

8.68.58.58.5

8.48.28.38.48.48.48.48.38.07.9

t7.88.08.18.38.48.58.58.68.78.9

9.49.79.8

10.010.19.69.49.38.98.7

8.68.58.48.48.48.58.58.88.q8.8

9.09.29.19.19.29.18.88.88.98.8

9.19.49.59.69.6

19.519.619.820.3

20. 720.921.021.120.7?20.319.819.418.818.5

18.217.817.817.717.717.517.316.917.116.7

16.717.217.016.716.115.514.814.814.214.4

14.715.315.215.415.415.214.314.113.813.3

13.313.413.513.814.614.714.414.113.813.4

12.412.713.113.513.1

1846 ..1847 ..1848 ..1849 ..

1850 ..1851 ..1852 ..1853 ..1854 *1855 ..1856 ..1857 ..1858 ..1859 ..

1860 ..1861 ..1862 ..1863 ..1864 ..1865 ..1866 ..1867 ..1868 ..1869 ..

1870 ..1871 ..1872 ..1873 ..1874 ..1875 ..1876 ..1877 ..1878 ..1879 ..

1880 ..1881 ..1882 ..1883 ..1884 ..1885 ..1886 ..1887 ..1888 ..1889 ..

1890 ..1891 ..1892 ..1893 ..1894 ..1895 ..1896 ..1897 ..1898 ..1899 ..

1900 ..1901 ..1902 ..1903 ..1904 ..

*High point s. tLow point s.


TABLE 8 (Concluded)

Year 2 3 4 5 7 8L.Sk. L.Wai. L.Dec. W.Skl. W.Wai. W.Dec.

1905 .... 99.8 30.4 14.7 53.5 9.6 13.31906 .... 99.7 29.0 14.0 53.3 9.8 13.31907 .... 99.6 27.4 14.1 49.7 10.4 12.81908 .... 99.5 26.3 13.8 45.5 10.9 12.31909 .... 99.3 25.8 13.4 38.9 11.4 12.5

1910 .... 99.0 25.1 13.4 34.2 12.1 12.31911 .... 97.7 25.1 14.2 31.1 12.7 12.41912 .... 96.1 25.3 14.1 29.3 12.9 13.01913 .... 94.5 25.1 14.7 31.9 13.3 12.71914 .... 91.6 24.9 15.2 37.1 13.4 12.71915 .... 89.6 24.9 *15.4 43.1 13.4 12.71916 .... 88.1 24.6 15.1 40.5 13.3 12.11917 .... 86.6 t24.4 15.1 41.3 13.2 11.71918 .... 84.8 24.8 14.8 35.4 13.4 12.01919 .... 84.6 25.4 14.2 30.8 13.9 12.5

1920 .... 84.2 27.2 14.1 24.5 14.6 12.81921 .... 85.2 28.5 14.0 25.0 15.0 13.71922 .... 85.5 31.3 13.8 22.6 15.5 13.81923 .... 84.5 33.1 13.3 23.6 *15.7 13.71924 .... 82.3 34.9 13.8 21.7 15.5 13.21925 .... 79.0 34.8 13.5 20.0 15.0 12.81926 .... 74.9 *35.9 13.9 t19.8 14.8 12.31927 .... t72.0 34.8 13.9 20.7 14.3 11.91928 .... 74.4 33.3 14.3 21.5 14.5 11.71929 .... 79.0 32.3 14.1 23.4 14.3 11.8

1930 .... 84.5 29.6 13.8 25.0 13.7 11.61931 .... 90.2 27.4 13.0 25.9 13.4 11.81932 .... 95.8 25.9 12.3 26.0 13.5 12.01933 .... 97.6 25.3 11.6 28.5 12.1 11.91934 .... 98.4 25.0 tll.0 30.1 11.7 tll.3

*High points. tLow points.

III. DESCRIPTIVE HISTORY OF THE PROPORTIONS OF DRESS

Each dimension of dress appears to have amore or less independent history. At least itcan be considered independently. It seems ad-visable to treat the histories of the six di-ameters in two ways: first, descriptively, here-with; and again quantitatively, in the followingsection on Periodicity.

The problem of the relations of the severaldimensions to each other, as they integrate intoa whole style of dress, or the structural skele-ton of a style, will be touched on in stillanother section, VI, on Interrelations of Dimen-sions; and again in the interpretation attemptedin section VII on Variability and Stability.

WIDTH OF SKIRT

The series begins with a fairly wide skirt,exemplified by the smooth, padded cone-shapedskirt of the Spanish fashions of the late six-teenth century, or the squarer Queen Elizabethdress. However, the year 1605 lies in the midstof a gradual tendency toward narrowing. The

farthingale (vertugarde, crinoline), which isvery wide about 1570,6 is completely out by 1625-30,7 when skirts attain a relative minimum ofwidth. Increasing diameters then follow untilabout 1660. This is reflected in the court stylespainted by Rubens under Marie de Medici, by Cal-lot, and in Van Dyck's regal portraits. Holland'swealthy bourgeois women in black satin with whitelace, painted by Codde, Rembrandt, and Terborch,are also typical of the scene. The Spanish Haps-burg portraits of Velasquez illustrate a courtdress of extraordinary uniformity and enormouswidth.

After 1660, a fine series of fashion platesby Bonnard shows a narrowing skirt. As early as1665 the farthingale is ousted from general wear,but not at court.8 For a while there is inFrance, as elsewhere, considerable discrepancy

6C. K'dhler and E. von Sichart, A History ofCostume (New York: G. Howard Watt, 1933), 237.Cited hereafter as Klh1er.

7Kbhler, 314.85K hler, 288.

127


between court regulations and the freer flowingcurrents of general fashion, but in the end itis court dress that has to yield to the pressureof the narrowing trend. In the latter part ofthe reign of Louis XIV, beginning about 1680,the farthingale starts a long slow recovery. InWatteau, then Hogarth, and later in the magnifi-cent Versailles galaxy including Boucher andNattier, we see a gradual approach to the maximumof skirt width reached about 1750. The Parisgowns that find their way to the American colo-nies corroborate, though with less luxury, thefashions seen in the portraits of Mme. de Pompa-dour.

After 1750 or so, the trend of the eighteenthcentury, though still hampered by court regula-tions, is a steady narrowing. A jog in thedecade 1771-80 coincides with Marie Antoinette'sreintroduction in 1774 of the wide flat farthin-gale, exquisitely depicted in the engravings ofMoreau-le-Jeune. In general, though, what islost in width is made up in the train. The worksof Chodowiecki, Reynolds, Romney, and Gains-borough, and the fine engravings of France andEngland of the later eighteenth century recallthe general picture. By the time of the FrenchRevolution the farthingale is discarded, and amere pillow at the back gives the necessary full-ness. One would have said that the imitation ofclassic dress during and after the Directoirewas a novel idea, symbolizing, perhaps, the be-ginning of a new and "natural" life. Instead wesee that the clinging skirts are merely the cul-mination of a drift that had its inception fiftyyears before. The years up to 1800-08 derivefullness from the trains that are occasionallyfound. 1810-11 is the bottom of the curve, witha trainless, short skirt.

Then the trend turns slowly to rise. Itspeak in 18b9-61 (figs. 2, 4, 6) was the resultof the adoption of the farthingale again, nowcalled crinoline. Afterward a gradual narrowingset in, destined to attain its limit of possi-bility in the middle twenties of our century.Since then, skirts seem to be gradually widening.A return, several decades hence, to a crinolinewider than the wearer might be difficult in thisday of automobiles, but the effect could beachieved with a train.

With peaks of fullness at about 1570, 1660,1750, 1860, and of narrowness around 1625, 1680,1810, and 1925, the full period is thereforeabout a hundred years. Oscillations withinthese huge curves now seem somewhat dwarfed.Many of them are due to the temporary introduc-tion of the train, which seems to have littleeffect on the general drift.

60 50 40 30 20 10 0 10 20 30 40 50 60

30-

20-

I0-

1900-

90-

80-

70-

60-

50-

40-

30-

20-

-30.

-20

-10

-1900

-90

-80

--70

-60

-50

-40

-30

-20

10-

1800-

790-

-10

-1800

-1790

Fig. 6. Skirt width and waist width, dimen-sions 5 and 6, showing generally inverse rela-tion, 1787-1936.

128


LENGTH OF SKIRT

This dimension is also analyzed in terms ofcycles, each being the time within which thehem leaves and returns to the ground. Theperiods of maximum length are the double decade1641-60, 1794, 1860, and 1902-05, with a near-approach in 1934-36. These suggest a diminish-ing wave length. This curve has several specialfeatures, too. The 1860 and 1905 maxima areseparated by a recession that remains rathersmall, and they may have to be counted together.A similar doubleness occurs in the minima. Itis difficult to decide whether 1813 or 1833represents the true extreme of skirt brevity inthe early nineteenth century. Moreover, theeighteenth-century minimum falls around 1780,only a decade and a half before the century'smaximum under the Directoire of 1794. Similarly,the recent minimum of 1927 is preceded by anear-minimum only eight years earlier. At thistime skirts nearly reach their upper limit ofpossibility, and probably our less definablelimits of decency. Glib explanations of theacting forces range through such things as war-time emergency, relaxing of sex morals, drivingof motor-cars, a passing tendency toward non-femininity to emphasize equality with men. Buteven a superficial consideration of the wholetrend, of dress-function and other aspects ofthe silhouette, reveals the inadequacy of thesephrases. The violent increase in variabilitywithin the years themselves in the late nineteen-twenties will be recalled, the result of theconflict between the knee-length dress and thelong "robe de style." Since then the styleshave been more homogeneous.

These peculiarities of the periodic curvefor the skirt length and its variability willalso be discussed more fully in the statisticalsections IV, V, and VII.

POSITION OF THE WAIST

Since the beginning of the intensive year-by-year record in 1787, there have been crests ofshort-waistedness in 1807, 1869, 1917, and 1936or still to be reached; and intervening maximaof low waists in 1842. 1903, and 1926. Theperiod thus shortens as time goes on. Thistendency holds also if the more intermittentpre-1787 data are included.

However, the extreme minima and maxima arefewer. The curve begins around 1605-10 with avery high waist, but sinks with the advent ofthe corseted V-waist. Its rise again to theapex in 1807 is slow except at the very end.This gives a period of about two hundred years.The pendulum then swings back. A slow drift

toward long-waistedness sets in. Our previouslyestablished swings prove but oscillations in agreater cycle. One hundred nineteen years after1807, in 1926, the lowest point is reached. Thetide has probably now turned into the second halfor upswing. One would have said that the longwaists of 1926 were another cultural abnormalityfollowing the war: they prove to be the normaltrough of a tremendous curve. Moreover, one re-calls with interest the similarly long-waisted,girdled frocks of 1450-1550, pictured for examplein the early Gobelins and in the court of Maxi-milian I.

DIAMETER OF THE WAIST

Inasmuch as it is impossible to perceive thetrend in this dimension from the 1844-1919 dataalone, Kroeber's former estimates prove to relatechiefly to minor irregularities.

In 1605 a sturdy corset9 achieves the tinywaist that for some time has been de rigueur onevery formal occasion. One may call to mind someof the portraits of Queen Elizabeth. This corsetgoes out gradually in the next few decades.Voluminous sleeves make it hard to get waist di-ameters for the seventeenth century. Our fewmeasurements are supported by an impressiongained from the Van Dyck and Dutch portraits thatwaists are amply wide up to 1660 or so. Thesharp drop soon after is the result of the re-introduction in 167010 of the corset. After 1680the trend changes. The considerable width formost of the eighteenth century is surprising, be-cause one has the feeling that waists are quiteslender. Though court dress is undoubtedly moreexacting in this respect, the effect of slimnessis achieved without real constriction partly bythe slenderizing V of the corsage, and partly byjuxtaposition to full skirts. A minimal extremeseems only to have been reached just before thefall of the French monarchy, after which diame-ters increase. A loose sash tied in the back istypical of the Revolution itself. With "Greek"dress as a pattern during and after the Direc-toire, the waist is wide. Following to a certainextent anatomical exigencies, though this we wellknow is not always so, waists increase in diameteras they shorten. The widest are those of Empiredresses, directly under the breasts.

There now starts a very gradual constriction.At the turn of the trend during the eighteen-fifties and sixties, the diameter approximatesQueen Elizabeth's famous edict (fig. 6). Thebreathing spell of 1872-75 is negligible. Notuntil the opening of the twentieth century doesthe-enlarging of the waist show itself in earnest.The maximum comes in 1923, just one hundred andsixteen years from the previous peak.

9K6hler, 320: "invented (by the Spanish) inthe first half of the 16th century."10Kohler, 320.

129


DECOLLETAGE

It is difficult at first to see major sig-nificant movements in this endlessly fluctuatingline. The longer span of three centuries re-veals a more significant form of the curve. Theruffed high neck of the Spanish sixteenth-cen-tury mode persists briefly in France. Frenchdecolletage is mostly low during the seventeenthand eighteenth centuries, under the particulariniLuence of court dress. The fichued throatof the French Revolution is the apex of height,then a new decline ensues. For 1844-1919, "themost striking event in the history of decolle-tage depth is its increase in recent years."11This reaches its maximum around 1915. A reverseculminates in the rather high necks of 1934-35.The evidence is not wholly satisfactory, how-ever, on which to base a periodicity.

WIDTH OF DECOLLETAGE

It is indeed true that "this trait appearsto have a very long periodicity.212 The highruffed necks of 1605 open the series at onelimit of possibility. Wide lace collars lead toa more open decolletage, and the outer limit ofpossibility, a point several inches down thearm, is reached about 1670. A gradual narrowingends in the close kerchief of the Revolution.It should be noticed that the width of decolle-tage of eighteenth-century dress and even courtdress is not great. It is depth that is empha-sized, so the general effect is that of a deepsquare. A rapid widening in the early nine-teenth century brings a peak around 1832. Sincethen for a hundred years the trend has beentoward less exposure of the shoulders.

IV PERIODICITY

It seems worth inquiring in how far there maybe any more or less constant duration to theswings of the fashion pendulum just described;whether there is any period of years within ornear which such swings tend to accomplish them-selves. There is, of course, no necessaryreason why even in one feature or dimension thetime for change from one extreme to the oppositeand back again should be constant over severalcenturies, nor why the rate of change should bethe same in separate features within a givencentury. At the same time there might be acause or causes tending to operate toward uni-formity; and in so far as there might be, thefirst step toward its recognition would be de-termination of the degree of uniformity whichexists, and of the time value expressing theuniformity.

On the whole, the style changes are so longin their range, and so progressive, that thereis no great difficulty, as soon as data are suf-ficiently ample, in determining recurrent maximaand minima. We call a full wave length the timeinterval from one crest to the next, or fromtrough to trough. For instance, skirts wereclearly at minimum width in 1811 and again in1926; at their fullest, not far from 1749 andabout 1860. This gives wave lengths of 115 and111 years respectively; or half wave lengths--one-way swings--of 62, 49, and 66 years. Forwaist width, correspondingly alternating maximaand minima fall at about 1780, 1807, 1860,1923,'3 giving wave lengths of 72 and 116 years,and one-way swings of 27, 45, 71 years.

11Kroeber, 257.12Kroeber, 256.13The moving-average figures are used.

Actually the situation is often less simple,because of minor crests and troughs. Say a di-mension of 20, the lowest in a century, changesgradually over fifty years to a value of 40,which is the highest reached in a century or so.If about halfway up the climb there is a reces-sion which in five or six years carries the valuefrom say 31 back to 28, after which the upwardmarch is resumed, this recession is obviously aminor fluctuation. It is no doubt stylisticallyand historically significant, but less so thanthe longer and larger swing from 20 to 40 onwhich it is superimposed. Suppose, however,that in twenty years the value of 35 has beenreached, that the recession lasts ten years andcarries down to 25, and that in the followingtwenty years the value mounts again and reaches40. What have we then? Still a secondary fluc-tuation, although an accentuated one? Or threeswings of 20, 10, 20 years duration and 15, 10,15 points amplitude? Obviously, if the answeris to be not wholly subjective or arbitrary, ithas to be rendered, so far as possible, in termsof all known values for the trait in question;also in comparison with the trend of durationsof undeniable swings; and in some degree withreference to high and low values and durationsin other traits. At that, there are likely toremain some cases difficult to adjudge, andothers dubious because of insufficient data;but enough may emerge clear-cut to justify theinquiry.

SKIRT LENGTH(Dimension 2)

The first half of the seventeenth century isa time of full-length skirts. From 1605 to 1640,half of the dress pictures measured have a length

130

RICHARDSON AND KROEBER: TEREE CENTURIES OF WOMEN'S DRESS FASHIONS

of 100; for only one of the years on which thereare data does the average fall below 96. From1641 to 1660 full length is even more unanimous:all of nine specimens available are full 100per cent. We can take the midpoint of thisdouble decade, namely 1650-51, as the moment ofclimax of length.

Thereafter there is slow shortening. From1661 to 1690, three years show 100; six years,values of 96 to,99. Then there is a twentyyears' gap in data. From 1711 to 1730 thelength is still about the same as before thegap.

From 1731, however, shortening increasesprogressively to about 1780: successive decadesaverage 96.6, 96.6, 95.8, 94.8, 94.2. The peakis evidently comprised in 1778-81: 92.4, 92.3,90.3, 92.2.

Now follows a rapid lengthening, expressed inhigher percentages. The 1784 value is up (skirtlonger), 1787 reaches nearly 99, 1790-9314 areall over 98, Directoire styles of 1794 touch100. The moving-average crests are 1792, 98.7,and 1796, 98.6.

This full length is only momentarily held,in contrast with the long persistence around 100the century before. The years 1795-1803 oscil-late between 97 and 100. Length decreases pro-gressively from 97.6 in 1804 to 91.9 in 1814.15It is, however, not certain that 1814 is thetrue crest of the wave, because a second peak ofshortness comes at 1833 with 92.0.18 Betweenlies a swell of partial lengthening of the hem.This whole double decade 1814-33 is something ofa unit: it has shorter skirts than any period inthe preceding two centuries.

At any rate, skirts lengthen progressivelyafter 1833. By 1836 the moving average hasreached 94; by 1841, 97; 1846, 98; 1858, 99;1860, 99.8. The first individual year to showfull 100 per cent length for every dress examinedis 1859; thereafter three more such years occursporadically until 1875; then no others until1899. Between 1875 and 1899 there is a minorshortening, reaching 96.1 in the moving averagefor 1886 and 95.5 actual mean in 1887. Thistemporary recession is steady from year to year,both up and down, but is modest. The real maxi-mum of length comes in 1903-04, with 99.9 in themoving average, and the four successive years1902 to 1905 showing actual averages of 100.

Since 1905, no year has shown a full 100.For about a quinquennium, the shortening was,timid and barely perceptible. But by 1911, the

14From here on, moving-average maxima andminima are used.

The lowest individual year is 1813, with87.9.

16No data for 1833; minimum year is 1834,89.3.

moving average has fallen below 98; by 1914, be-low 92; by 1919, below 85. This, the year inwhich Kroeber's first study terminated, seemedthe limit. However, by 1925 the moving averagefell below 80, and in 1927 dipped to the minimumof 72 (year figure 69.7).

Once this corner was turned, the lengtheningwas rapid: 1929, 79; 1930, 84; 1931, 90; 1934, 98.

If now we summarize, these are the maxima andminima:

TABLE 9Dates of Maximum

Years

and Minimum

Maximum

c. 1650 ...... 1001780.1792-96 ... 991814.1823 951833.1860 ......... 10018861903-04 ...... 1001927.? 1934 ....... 98

*Moving-average values

Skirt Length*

Minimum

90

92

92

96

72

after 1788.

It will be seen that two extremes have beendouble peaked: the short skirt of 1814 and 1833,the long one of 1860 and 1903-04. Passing overthe recessions within these as minor, we have:*1650, t1780, *1794, t1833, *1903, t1927. Theresulting wave lengths, crest to crest and troughto trough, are 144 (1794 minus 1650), 53, 109,and 94 years. The average of these four is 100years--the empirically found value of the skirt-length cycle in modern European history.

TABLE 10Summary of Periodicity in Skirt Length

165017801794183319031927

Max.Min.

Max. Max.-Max.: 144 yearsMin. Min.-Min.: 53 years

Max. Max.-Max.: 109 yearsMin. Min.-Min.: 94 years

Average of four waves: 100 years

SEIRT WIDTH(Dimension 5)

Skirt width shows three cycles in three cen-turies: four peaks of slimness, three of full-ness.

131


TABLE 11

Dates of Maximum and Minimum Skirt Width

Years1629 .........1651-60 ......1678-80 ......1749-51 ......1811 .........1861 ........1926 .........

Maximum

96 (1659, 126)

101

105 (1859, 116)

Minimum42

39

27

20

These convert as follows:

TABLE 12Summary of Periodicity in Skirt Width

Year

1629 .........1655±.........1679±.........1750±.1811.1861.1926 ........

Extremes Intervals,years

Narrow

Narrow 5095

132111

115

Full

FullNarrow

FullNarrow

36, or over a third. These are moving-averagevalues; the actual means for particular yearsare somewhat more extreme: 17 for 1809 and 1815-16, 38 in 1927. The several minima--or high-waist peaks--fall between 18 and 25; the maximaof low-waistedness between 29 and 36.

These are the indicated crests:

TABLE 13Periodicity in Waist Length

High Low Interval 8 yearswaist waist Min.-Min. Max.-Max.

1605-10 .... 211661-70 311711-20 ... 26 1081751-60 29 901817 ...... 18 1011842 29 861869 ...... 22 521902/03 31 601917 ...... 24 481926 36 24Average ... 77 65

The average is a fraction over 100 years.The histories of costume give 1570 as the

maximum of farthingale expansion. This date liesabout eighty-five years before the 1655 peak offullness. The inclusion of this earlier wavelength would reduce the average from 101.3 to98 years. So far as the periodicity of this.skirt width has changed during the last threeand a half centuries, it seems to have slowed.

WAIST LENGTH(Dimension 3)

From here on, we are concerned with smallermeasures--waist and shoulders--and our values,which are percentages of the body height, runlower. For our first dimension, length of waist,the periodicity is also somewhat less.

It must be remembered that this measure re-fers to the vertical height of the narrowestpart of the middle of the silhouette figure.The belt, or demarcation of blouse and skirt,may fall lower, especially in front, but hasbeen disregarded because it is not always pres-ent.

The minimum position found (lowest percentage,highest position on the body) is 18, or lessthan a fifth of the body height;17 the maximum,

17As body height has been calculated from themouth, the place of the waist line at its high-est would be a full fifth down from the realbody height at its highest position (lowest per-

centages); probably two-fifths when lowest onthe body (highest percentages).

The average of 77 and 65 gives 71 years asthe mean duration of a complete wave.

The decrease of wave length is steady andnotable.

WAIST WIDTH(Dimension 7)

Waist width or diameter fluctuates betweenabout 8 and 16 per cent of figure height, infashion plates. The duration of its swings--contrary to that of waist position--does not ap-pear to be shortening. Early data are unusuallyfew, on account of arm interference and art con-ventions. (See table 14 on opposite page.)

The mean full wave length is 93 years, withrather low variability from 72 to 116.

DECOLLETAGE(Dimension 4)

For the first thirty years of our record therewas no decolletage in the modern sense. For thenext thirty, it alternates with mere neck open-ing. Thereafter, we can properly speak of breastexposure as standard in court or formal dress.However, there is a long period from about 1670to 1780 in which shifts of trend are difficult todefine, although depth of cutout grows graduallygreater. Around 1784 there is a swift change:decolletage as such almost disappears, and staysout for a decade. 1794-96 bring it back, and amaximum is reached in 1803, with a moving-average

132


percentage of 14. Twenty years later, around1823-27, there is a minimum just below 12. By1834, the figure is back above 13;18 and for a

It must be remembered also that the dimensionsof skirt and waist length and breadth expressthe outline of the dress or figure as a whole,

TABLE 14Periodicity in Waist Width

Decades Extremes Extreme value Extreme Approx. Approx. Interval___________(10-yr.av.) of year year max. min.1641-50.15.8 c.1645 161671-80 ... 9.8 c.1675 101731-40 ........ 13.8 c.1735 14 901771-80 ........ 11.9 1780: 8.6 c.1780 9 1051801-10 ........ 14.0 1807:14.7 1807 15 721851-60 ........ 8.2 1860: 7.8 1860 8 801921-30 ........ 15.1 1923:15.7 1923 16 116Average .._....III |lll| 93 93

century, to 1931 inclusive, the moving average(and very few actual year values) does not fallbelow 12.5. During this century, the movingaverage several times goes above 14, with minorpeaks in 1841, 1874, 1879, 1898, 1905, and afinal and highest one in 1915 at 15.4. The in-tervening dips in the curve--decolletage short-enings--are, however, so slight that they can-not fairly be construed as basic waves. Theyare only oscillations in a slowly deepeningtrend, which climaxes around 1915. From this1915 climax the raising of the decolletage(smaller values) is gradual, with even a re-cession to more than 14 as late as 1928. With1932, however, the break comes: the decolletageis transferred to the back, the neck-line risessharply in front. A seeming minimum is reachedwithin three years: 1934, 11.

It is clear that after decolletage once wasaccepted (or reaccepted) as a feature of formaldress, in the early seventeenth century, ittended to remain established with only minorfluctuations, except for three sharp but tempo-rary crises. The first of these was in 1784-94,when decolletage proper simply went out. Thesecond was a longer period, marked in 1806-13,and even more so in 1820-28, with peak about1823, when decolletage remained in force, butpartially sacrificed depth in order to obtainan effect of breadth. The third was the recentyears, when the breast was covered to exposethe back.

The situation as regards this proportion, andthe factors at work in its changes, are there-fore of a somewhat different order from thoseconcerned in the proportions so far considered.

18In view of the tendency toward inverse re-lation between depth and width of decolletage,discussed in sections III and VI, it shouldhowever be noted that the minima and maxima ofthe two dimensions do not coincide in time.Depth, minimum in 1823-27; width, maximum in1832.

but decolletage really refers to a feature in-ternal to the figure. As regards dress, in theliteral sense, it is negative instead of posi-tive.

TABLE 15Periodicity in DecolletageI g

Apro. ApproxPeriod Value Year ApproxI. Approx ntervalsmax. mmn.

1605-10 5.41631-40 .... 10.8 1635± 111641-50 .... 6.6 1645+ 71771-80 .... 15.2 1'778 18 1431788 ....... 8.6 1788 9 1431803 ....... 14.2 1803 14 251823, 1827.. 11.7 1825+ 12 371841 ....... 14.6 1841 15 381868-69 .... 12.6 1869 13 441915 ....... 15.4 1915 15 741934 ........ 11.0 1934 11 65Average...II I 70 72

If we consolidate the second and third waveswith the fourth, as there is possible warrantfor doing because their extremes are less accen-tuated, we have left two crest-to-crest waves-of143 and 137 years respectively, and two trough-to-trough ones of 143 and 144--a mean of 142,instead of 71 as in the table.

These mean wave lengths do not signify toomuch. What we can also see is two long spans,1645-1778 and 1788-1915, in which decolletagegradually deepens to a maximum; punctuated byseveral brief reversals, c.1635-45, 1778-88,1803-23, 1841-69, 1915-34. This gives for theperiods of deepening of decolletage an averageof 140 years, for those of raising, 18 years.In other words, the wave profile is markedlyskew. The cutout creeps slowly downward, thenjumps back up rapidly.

133


DECOLLETAGE WIDTH(Dimension 8)

Decolletage proper does not exist in ourseries, for width any more than depth, before1630. After that date, it gets establishedrather more rapidly, however, and reaches apeak, with nearly full shoulder exposure or per-centage value of about 21, around 1659-68. Ourdata are blank between 1680 and 1710. Afterthat, the value declines gradually through theeighteenth century, at any rate during its sec-ond half; the trend in the first half is notquite clear. The minimum width is reached witha value of 8 in 1795, a few years after thesudden minimum of decolletage depth in 1788.From here there is a climb of nearly forty yearsto above 23 in 1832. Thereafter the record isone of slow narrowing to the present. The lastyear of the moving average, 1934, is lower, with11.3, than any year since 1800, 11.6. The lastactual year, 1936, shows 9.8. The long swingtoward narrowing of neck-opening may thus be notyet completed.

Th,ere are several temporary recessions in thecentury of drift: 1847-53, 1878-83, 1890-95,1900-03, 1917-22. But these add only one to twopoints, each time, while the century-long de-crease drift is from 23 to 11; so they clearlyare superficial oscillations. As with depth,there is a tendency toward slow creep--in thisinstance toward narrowing; broken by definitelybriefer and sharper-curve reversals.

TABLE 16Periodicity in Decolletage Width

Years

1659-68 ....

1795 .......

1832 .......

1936 .......

Value Intervals

21.28.323.411.3

168141+

It is evident that the wave profiles areagain skew. But decolletage width mounts rapid-ly to its maximum and then shrinks gradually,whereas decolletage depth is reduced suddenly

and then increases again for a long time. Thisinverse relation is reflected in the ratio ofthe two dimensions, as shown in table 18. Theratio (4/8) sinks rapidly for 30 years to 1827-31, then climbs slowly for 85 years to 1912-16.

COMPARISON

If now we bring our six sets of periodicityfindings together, we have the following:

TABLE 17Comparison of Six Periodicities

Dimensions Mean wavelengths, years

No. 2 Skirt length ............ 100No. 5 Skirt width ............. 100No. 3 Waist length ............ 71No. 7 Waist width ............. 93No. 4 Decolletage length ...... 71No. 8 Decolletage width ....... 154

Mean of six ................. 98

We are not wholly clear how much weight shouldbe attached to this clustering of the wave lengthsof change in our six dress dimensions around avalue approximating a century. The question isin how far the significance lies in the intrinsicfact of a century-value, or, on the other hand,in the nearly synchronous clustering of peaks incertain periods, which might be due to a commoncause. This problem is discussed further insection VII.

One thing, however, is certain--whether or notthe six mean wave lengths do or do not bear re-lation to one another--namely, that women's dressfashions change slowly, as regards the fundamentalproportions of the silhouette or contour. On theaverage, any one proportion is a half-centuryswinging from its extreme of length or fullnessto extreme of brevity or narrowness, and anotherhalf-century swinging back. This is more thanwould usually be supposed, in view of the civi-lized world's general assumption that women'sdress fashions are in their nature not only un-stable but capriciously and rapidly unstable.

134

V OSCILLATIONS

It also seems worth while to try to estimatethe average duration of minor fluctuations ortransient oscillations over and above the majorswings or trends so far considered. This wouldrequire first of all the reliable determinationof the long-time trends. The deviations fromthis of the actual averages for each of a se-ries of years would then give the periods forwhich the actual style, with respect to any onetrait, remained above or below its underlyingtrend.

As regards statistical execution, however,the matter is not so simple. On account of thesmall number of measures, no mean for a particu-lar year is very reliable; and its probable er-ror would have to be taken into account. More-over, the basic trend can be differently ex-pressed; for instance by three-, five-, nine-,or fifteen-year moving averages; or by moreelaborate methods. A moving average from itsvery nature produces certain apparent fluctua-tions in the actual annual means, even when thelatter are proceeding in a regular curve inrounding a peak. Suppose for instance succes-sive annual values run 69, 70, 71, 72, 73, 75,78, 82, 87, 82, 78, 75, 73, 72, 71, 70, 69.As the sharper rise begins, the moving averageclimbs ahead of the actual values; at the peakit is considerably lower; then falls more slowlyso that it is higher for a while. In otherwords, in the rounding of such a corner, we getthree departures of the two lines from eachother: that is, of the year-by-year graph andthe moving-average graph; though each is a per-fectly symmetrical curve. In short, three ap-parent fluctuations are due merely to the mathe-matical properties of the technique used.

Since the fewness and variability of theprimary measurements thus appear to render anelaborate treatment hardly worth-while, weshall proceed to see cursorily what the surfaceresults are.

Dimension 3 shows 38 full oscillations ofannual means from the five-year moving averagein the one hundred and forty-seven years from

1788 to 1934; that is, 38 periods when the ac-tual value moves from above the moving averageto below it and back above again (see tables 3,8). This gives an apparent mean of 3.87 years,0=4.11.19

Dimensions 4, 5, 7, 8 show 40, 38, 38, 41oscillations in the same one hundred and forty-seven years; the mean for these dimensions is3.77 years.

Dimension 2 cannot be directly compared withthese, because the measured values often reach100, but cannot pass it: a dress as worn maysurpass the distance to the ground, but in thefashion plate, as soon as the toe becomes in-visible, the measurement has to be read: skirtlength - 100. The range of variation is there-fore small when dresses are near the limit oflength, and the number of oscillations wouldrise to 55 and the mean duration fall to 2.5years. If we count only fluctuations passingfrom below 99.5 to above 100.5 of the movingaverage, or vice versa,20 the number of oscilla-tions is 40, and their average duration 3.68years.

The average length of oscillation, between3.5 and 4 years, is not far from the averageduration generally assumed for the business cy-cle. This is probably a coincidence. The valuewill scarcely be very significant until thereare more individual measurements available foreach year and until more technical statisticalconsideration is given the moving-average "trend"which forms one of the two variables whose rela-tion expresses the oscillations.

The size of deviation of the actual averagefor each year from the moving-average trend is,however, almost certainly significant for sta-bility of style, as discussed below in sectionVII.

19For one-way fluctuations (intervals betweencrossings of the two lines), M1=.95,c 2.13.

soThat is, the year-to-year line not onlycrosses the moving average, but crosses it witha motion of at least 1.0.

[135]

VI. INTERRELATIONS OF DIMENSIONS

We have computed about half the relations be-tween dimensions, and present them herewith insummary. (Table 18.) The variations come out verymuch like those for diameters considered alone.

with accentuation, in the ratios. Also, skirtlength varies so little, proportionately to theother dimensions, that any ratio into which itenters becomes largely a function of the other

TABLE 18Tnft.rrlnfa i rvnnra rnf npc.rfni n fimpn.in.n-. Ihv Fi VQ-YOcr PQr;i n1 1 788-1 9Q3

Period 5/2 7/3 4/8 4/3 7/5 8/5 7/8

1788-91 ...... 54.6 50.2 105.7 37.6 23.1 16.3 141.41792-96 ...... 50.9 59.4 114.3 44.2 25.8 16.8 *153.61797-1801 .... 45.2 67.2 *121.4 65.1 29.1 23.3 125.21802-06 ...... 44.4 70.7 97.8 *72.3 30.7 32.1 96.41807-11 ...... t30.6 *72.3 88.5 63.1 *48.1 47.4 101.4

1812-16 ...... 38.1 69.8 70.9 64.6 37.7 *49.1 76.71817-21 ...... 43.3 56.3 83.2 64.8 27.7 38.3 72.31822-26 ...... 47.8 44..8 62.6 49.4 23.8 43.9 56.81827-31 ...... 54.0 43.4 t54.5 t45.7 23.1 44.5 51.81832-36 ...... 69.3 36.5 63.3 48.3 15.5 32.4 47.8

1837-41 ...... 67.3 31.9 71.9 50.2 14.0 30.6 45.71842-46 ...... 60.5 30.7 73.1 49.1 14.9 33.7 45.61847-51 ...... 63.6 t29.6 67.0 47.4 13.6 32.5 41.91852-56 ...... 80.0 30.5 65.2 49.1 10.7 26.4 t40.61857-61 ...... *103.1 31.5 73.5 52.9 t7.9 18.0 43.8

1862-66 ...... 101.7 35.6 74.0 55.5 8.3 tl7.5 47.51867-71 ...... 88.0 40.8 75.5 57.8 10.2 19.2 52.91872-76 ...... 82.6 *43.7 87.6 *61.0 12.3 19.6 62.71877-81 ...... 67.0 33.5 98.6 54.6 13.2 21.8 60.41882-86 ...... 57.1 31.6 87.7 50.8 15.2 27.9 54.5

1887-91 ...... 54.7 31.7 103.8 49.6 16.7 25.2 66.21892-96 ...... 59.0 32.9 96.5 49.6 15.8 24.8 63.91897-1901 59.3 t29.5 106.7 48.0 14.9 22.7 65.71902-06 ...... 55.2 31.2 111.5 t47.4 17.4. 23.8 73.31907-11 ...... 39.2 44.2 107.2 51.9 29.3 32.1 91.2

1912-16 ...... 40.5 53.8 *119.7 *61.0 36.1 34.2 105.51917-21 ...... 35.9 *54.7 113.6 55.9 45.7 41.1 111.21922-26 ...... t26.4 45.7 104.5 40.7 *71.4 *60.8 117.41927-31.29.6 44.3 119.5 43.7 61.1 50.4 *121.21932-36 30.6 47.2 97.3 44.0 39.2 37.5 104.4

This is due largely to the fact that the sixseparate dimensions mostly show significant maxi-ma or minima in the decades around 1810 and 1920,so that these tend to be repeated, sometimes

measure. On the other hand, waist length, whichshows five crests instead of the usual threesince 1800, produces five crests in those ratiosinto which it enters. (Table 19.)

TABLE 19Periodicity of Dimension Interrelations

Horizontal and Vertical Ratio of Same Part of Dress:5/2: Minima, 1807-11, 1922-26 interval, 115 years)7/3: Maxima, 1807-11, 1872-76 interval, 65 years

Maxima, 1872-76, 1917-21 interval, 45 yearsMinima, 1847-51, 1897-01 interval, 50 years)

4/8: Maxima, 1797-01, 1912-16 interval, 115 years)Vertical Diameters inter se:

4/3: Max., 1802-06, 1872-76Max., 1872-76, 1912-16Min., 1827-31, 1902-06

(70 years40 years)75 years)

Horizontal Diameters inter se:7/5: Max., 1807-11, 1922-26 (115 years8/5: Max., 1812-16, 1922-26 (110 years7/8: Max., 1792-96, 1927-31 (135 years

[136]

JLiiUU-L-I-U.LCLU.LVlLO VI VVI-UQ.LIL 11-LIUVILDIVILOs Uff X.LVV-.LVCX.L JLIU L t WU-.L 4w %Jv

VII. VARIABILITY AND STABILITY OF STYLE

The question of when, under what circum-stances, and why traits of fashion are rela-tively stable and unstable is approached by usin two ways.

One is a year-by-year comparison of thestandard deviations of the means for each trait;that is, the variability inter se of the actualmeasurements which go into the annual average.This is probably the most satisfactory expres-sion of stability and instability.

The second method is to compare each annualaverage with the "trend" or moving average forthe same year. If the latter is held constantat 100, how many per cent above or below 100 isthe actual average for the year? Thus for skirtwidth, the moving average for 1801 is 45.2, theyearly mean 42.1, or 6.9 per cent less. For1802, on the other hand, the moving average hasgone up only to 45.6, but the year's mean is59.6, or 31 per cent higher. For 1803 the devi-ation is 10.3 per cent under. Obviously thisis a period in which the style for skirt widthwas highly variable from year to year, eventhough the trend is pretty consistently in onedirection for two decades. By contrast, theyears 1854-58, which also show a strong one-waytrend in this dimension, run 101.1, 101.7, 101.8,90.8, 100.3; and 1839-43, with the trend changemild, show 97.8, 99.1, 103.5, 98.2, 99.8. Itis plain that the year-to-year fluctuation wasmuch more marked in 1801-03 than in 1839-43 or1854-58. In other words the fashion, with re-spect to this trait at least, was much lessstable in its trend in the earlier period thanin the two later.

The objection which can be made against thissecond measure is that it expresses the rela-tion of an actual year average to a short movingaverage to which it contributes; also that themoving average, our base, possesses properties,in relation to the actual sequence of events,which vary according to the nature of the se-quence of events. It behaves somewhat'differ-ently when it is steadily progressing in onedirection and when it is turning a corner; andagain, in different parts of its curve aroundthe turn. It is for this reason that the seriesof simple percentaged standard deviations, orvariability coefficients, is probably sounder.However, these coefficients directly expressonly the variability or instability within oneyear: how much the several fashion plates for1801 differ from one another, for example; in-stability over several years must be inferred bycomparison. The annual deviations from the mov-ing average express variability within a spanof years directly. We therefore use this measurealso. On the whole, the two measures give re-sults fairly in agreement. Those by the methodof deviation from the trend will be presentedfirst.

YEAR-TO-YEAR VARIABILITY(Percentage Deviations of Annual Mean

from Moving Average)The percentages by which each annual mean de-

viates from the five-year moving average for thesame year--the basic data for this section--aregiven in table 20. More convenient are tables21 and 22, which express the same values averagedfor five- and ten-year periods respectively. Wehave thought it unnecessary to diagram these re-sults separately; in substance they are shown infigures 1 and 2 (pp. 114,115), where the linerepresents the moving average, and dots theannual means.

TABLE 20Percentage Deviations of Actual Year Means from

Trend. 1788-1934

Year 2 1 3 '4 [ 51 71 817881789 ..

1790 ..1791 ..1792 ..1793 ..17941795 ..1796 ..1797 ..17981799 ..

1800 ..1801 ..1802 ..1803 ..180418051806 .1807 ..1808 ..1809 ..

1810 ..1811 ..1812181318141815 .1816 .1817 .1818 ..1819 ..

1820 ..1821 ..18221823 .1824 .1825 .1826 .

1.4.8

.9

.2

.2

.31.71.6.4

1.2.2.3

.11.62.6.4

2.6.4

1.3.1.6.2

.81.2.3

4.82.5.7

2.02.31.53.6

1.12.7

1.72.61.11.9

14.05.3

7.12.86.74.4

13.43.43.03.01.01.0

4.24.83.21.13.21.63.8.5

1.110.8

5.08.52.03.04.87.59.5

11.08.4

22.1

8.14.3

4.710.81.6.8

3.512.0

19.432.28.947.03.1

25.212.111.8

.97.2

19.816.521.611.39.614.3

.810.62.54.9

6.36.41.6

11.28.27.87.7

26.914.42.3

10.017.2

3.416.09.2

10.0

2.03.4

5.21.5

13.2.4

13.49.16.25.8.4

3.6

7.56.6

30.712.53.2

10.02.1

12.23.2

15.4'12.31.14.3

11.110.621.5

.33.13.31.5

7.32.2

5.516.13.61.9A..---I- -

LA _ _ IL.-_

32.021.1

16.815.912.34.5

11.612.128.611.221.113.2

1.626.013.76.61.55.0.7

3.41.46.4

6.52.2.0

6.73.02.34.8

11.6.9

9.8

6.57.7

6.414.87.33.7

23.0

14.64.5

15.514.6

8.04.07.8

1.72.5

12.95.12.9

21.28.65.82.21.4

14.63.84.3

17.9.0

3.210.735.17.95.2

1.82.3

4.74.7

13.616.7

[137]

I I-r-'rI I

_I

138 ~~~~~~~~ANTHROPOLOGICAL RECORDS

TABLE 20 (Continued)

7.21.1.0

6.19.5

.4

.4

2.6

2.5

2.5

3.2

7

1.7

7

1.0

.3

2.4

.0

2.1

2.5

.0

1.1

3.9

3.2

.4

1.8

1.8

1.8

.7

1.9

1.6

1.6

.4

1.6

3.3

1.3

.9

.4

3.6

.5

.0

.0

3.1

.4

7.5

3.9

4.3

1.3

1.2

1.6

4

4.5

3.4

4.32.5.0

.82.4.8

6.9

4.5

3.7

1.5

7

3.5

1.4

1.4

.0

..02.1

.7

6.4

8.0

.7

2.2

5.9.3.7

4.4

5.9

4.4

1.5

4.3

.7

10.1

5.9

8.9

6.1

2.3

.8

3.1

.8

.8

7.9

.8

9.5

9.8

3.7

10.3

1.4

.0

2.2

3.6

2.2

3.6

6.3

10.0

3.6

4.72.910.5

3.612.82.8

.61.74.66.77.32.3

1.13.52.3

.71.41.04.36.63.4.3

.96.71.73.61.11.71.89.2.6

1-2.6

2.3.6

5.6.5

1.16.7

.72.64.01.8

6.28.95.45.72.93.57.22.42.36.2

11.011.1

5.79.12.6

4.312.55.2

1.0.0.0

2.21.14.4

1.13.31.12.22.33.4.3.53.5.0

1.2

1.22.4.0

7.1.0

7.12.43.61.21.3

2.6.0

6.28.41.21.23.55.83.41.1

2.17.2

12.21.05.98.31.16.5.0

1.1

2.3.0

TABLE 20 (Concluded)Year I 2 [ 3 [ 4 1 5 1 7 I 8

10.0.9

4.5

9.87.01.7

15.612.88.71.11.54.0

4.63.08.26.7

.51.04.1.0

1.01.5

.01.41.9

3.53.5

1.0.0

1.1

.5.0.6

3.42.3.0

4.01.25.84.2

10.84.6.6

9.66.89.78.87.42.84.2

4.13.9

18821883188418851886188718881889

1890189118921893189418951896189718981899

1900190119021903.190419051906190719081909

1910

1911

1912

1913

1914

1915

1916

1917

1918

1919

1920

1921

1922

1923

1924

1925

1926

1927

1928

1929

1930

1931

1932

1933

1934

.7

.0

.1

. 7

.3

.6

.5

.2

.4

.2

.4

7

.3

.3

.0

.4

1.2.3

.5

.1

.2

.1

.1

.2

.1

.0

.2

.4

.2

1.0

2.3

2.0

.2

1.7

4.3

1.7

.6

5

2.4

2.5

.6

7.3

4.0

2.5

3.6

3.2

5.9

9.9

5.3

5.4

1.6

.7

2.62.31.52.2

. 7

. 7

.0

.4

.4

1.1

2.1

3.9

2.9

2.1

2.1

. 7

. 7

.3

1.3

. 7

3.5

4.5

4.9

.3

. 7

2.2

3.4

5.8-

.4

4.0

4.0

1.6

1.6

2.0

1.6

.4

2.4

5.1

2.6

2.1

6.4

6.6

12.1

2.0

11.1

8.9

8.7

.6

5.1

2.2

5.8

1.2

.0

6.64.53.03.78.07.21.4

. 7

1.4

4.3

3.6

.2.2. 7

. 7

.0

.0

. 7

2.1

7.9

11.3

6.4

9.4

4.1

7

15.7

5.7

15.2

9.0

7

.0

5.0

4.8

5.3

8.6

2.0

7.4

0

6.4

4.3

3.6

.0

5.8

6.7

13.7

6.5

4.97.8

5.8

14.6

7.3

3.4

15.5

1.4.9

5.33.53.5.6.88.22.5

5.1

2.7

3.8

4

4.5

1.3.

14.8

2.4

11.4

10.5

10.9

11.0

4.1

11.4

2.5

.4

5.1

3.0

7.7

1.3

3.8

25.4

6.5

5.6

21.6

7.0

21.2

39.7

42.7

7.8

31.4

4.0

16.8

5.1

1.8

10.0

16.7

13.0

1.4

11.1

3.6

4

.8

9.1

13.0

7.12.42.42.44.75.74.69.1

5.6

.0

1.1

2.2

.0

5.5

9.1

2.3

9.0

5.7

4.4

.0

4.2

.0

3.1

4.2

3.1

6.7

.0

12.3

3.3

5.5

2.3

2.3

3.7

2.2

5.3

1.5

.0

5.0

7

1.3

6.5

1.9

2.6

4.0

4.1

3.5

1.4

6.3

20.4

3.0

18.5

8.3

17.1

. 712.36.5.0

3.5. 7

5.1.8

1.56.05.93.64.12.05.6

12.113.86.7

8.14.7

15.33.7

13.015.015.83.94.93.2

5.71.6

11.54.7

19.711.81.72.6

if. 73.2

.87.35.18.0.8

1.6.0

5.93.4~5.9.9

8.514.210.94.4

182718281829

18301831183218331834 .

1835 .

1836 .

183718381839

1840184118421843184418451846184718481849

1850.185118521853185418551856185718581859

18601861186218631864 .

1865 .

1866 .

1867 .

1868 .

1869 .

1870 .

1871 .

1872 .

1873 .

1874 .

1875 .

1876 .

1877 .

1878 .

1879 .

1880 .

1881 .

.73.0.6

1.11.61.5

3.1.6

1.8.0.5.4

.0

.1

.4

.1

.6

.4

.2

.4

.1

.3

.2

. 7'4.0.1.0.2.5.4.4

.0

.4

.1

.8

.0

. 7

.61.5.3

1.0

.2

.1

.1

.2

.2

.7

.0

.4

.1

.1

.6

.0

Year 2 -1 3 1 4 1 5 .1 7 1 8I II I I I I I I I I

138

6 0

* 0

0 0

6 0

0 0

* 6

* 0

i 4

* 0

6 6

0 0

0 6

0 6

60

00

00

&0

06


First of all, it is clear that the propor-tionate amount of deviation varies among the sixdimensions dealt with. On the whole, the largediameters have low variabilities. Thus, skirtlength, the absolutely largest dimension, has5.9 as its highest five-year-mean percentagedeviation (table 21), while 19 out of 30 values

TABLE 21Five-Year Averages of Annual Deviations

from Trend. 1788-1934

Period 2 3 4 5 7 8

1788-91.. 8 7.3 16.8 3.0 21.5 14.01792-96 . .8 6.2 19.3 8.5 13.8 15.01797-1801 .7 2.8 11.2 4.8 14.6 4.81802-06 .... 1.5 2.6 11.5 11.7 5.5 10.11807-11 .... .6 5.2 6.1 8.8 4.0 5.5

1812-16 .... 2.1 5.4 7.3 9.6 3.4 7.21817-21 .... 2.2 10.8 14.2 3.5 7.3 10.51822-26 .... 1.8 4.5 9.6 6.8 8.1 9.91827-31 .... 1.4 4.8 2.0 6.9 6.8 6.41832-36 .... 1.8 1.5 4.0 2.4 1.6 9.7

1837-41 .2 1.8 1.7 4.2 2.4 2.81842-46 .3 .9 1.8 1.9 2.5 4.11847-51 .3 1.9 4.1 3.6 1.7 .81852-56 .1 1.8 4.1 2.0 3.3 2.01857-61 .3 1.2 6.3 5.1 1.7 .5

1862-66 .4 1.5 1.6 2.9 4.1 2.11867-71 .6 1.4 6.3 4.7 3.9 5.31872-76. 2 3.5 3.5 4.9 5.7 7.11877-81 .2 2.2 5.1 6.6 2.0 4.51882-86 .4 1.9 5.2 2.9 3.8 4.6

1887-91 .4 .5 3.0 5.1 5.0 2.81892-96 .3 2.6 1.4 5.0 3.6 4.21897-1901 .3 .7 4.4 9.2 4.3 9.11902-06 .1 2.8 7.3 4.7 2.9 12.61907-11. 4 3.2 6.1 8.2 5.6 3.9

1912-16 .... 2.1 2.2 5.8 12.4 3.2 9.91917-21 .... 1.5 2.5 4.0 25.1 1.7 5.11922-26 .... 3.6 7.6 6.0 10.1 3.8 3.11927-31 .... 5.9 5.1 7.9 5.9 6.9 4.91932-34 .... 1.1 2.3 8.7 7.6 14.6 9.8

are under 1.0. Waist length rises to a maximumof 10.8, and only thrice falls below 1.0. Decol-letage depth, on the other hand, rises as highas 19.3, and never goes below 1.4. The trans-verse diameters, which of course average lowerthan the longitudinal, run about like decolle-tage depth.

Next, it is clear that while 1835-1910 is atime of small deviations or high year-to-yearstability for all six traits, these traits varyconsiderably among themselves as to whethertheir greatest instability falls in the periodbefore or after the long quiet span, and whetherearly or late in 1788-1835. Thus, skirt length

and width (Nos. 2 and 5) attain their greatestvariability in post-World War years; the waistand decolletage dimensions vary most before 1821--in fact, all except waist length (No. 3), be-fore 1803. These dozen early years 1788-1799,in the moving-average record, are only fairlyconspicuous for number of maxima or minima at-tained, in spite of their high deviations. 1802-16 show more wave crests, but the year-to-yearvariability averages lower in four of the sixdimensions (tables 20, 21, 22).

TABLE 22Ten-Year Averages of Annual Deviations

from Trend, 1788-1934~ ~ ~ ~ ~

Period 2 3 4 5 7 8

1788-96 ..8 6.8 18.1 5.8 17.6 14.51797-1806 .. 1.1 2.7 11.4 8.3 10.1 7.51807-16 .... 1.4 5.3 6.7 9.2 3.7 6.41817-26 .... 2.0 7.7 11.9 5.2 7.7 10.21827-36 .. 1.6 3.2 3.0 4.7 4.2 8.11837-46 ... .3 1.4 1.8 3.1 2.5 3.51847-56 ... .2 1.9 4.1 2.8 2.5 1.41857-66 ... .4 1.4 4.0 4.0 2.9 1.31867-76 ... .4 2.5 4.9 4.8 4.8 6.21877-86 ..3 2.1 5.2 4.8 2.9 4.61887-96 ..4 1.6 2.2 5.1 4.3 3.51897-1906 .2 1.8 5.9 7.0 3.6 10.91907-16 1.3 2.7 6.0 10.3 4.4 6.91917-26 2.6 5.1 5.0 17.7 2.8 4.11927-34 .... 3.5 1 3.7 8.3 6.8 10.8 7.4

Figure 7 shows graphically aali deviations ofthe year from the trend, above a certain magni-tude. This magnitude has been chosen so thatthe number of large deviations represented wouldbe about the same for each of the six dimensions.Convenient values are 3 per cent for No. 2, skirtlength; 6 per cent for No. 3, waist length; and12 per cent for the others. These are desig-nated in the figure as "fluctuation units." Forinstance, for trait No. 4, decolletage depth, theyear-from-trend percentage deviations beginningin 1788 (table 20) are 3.5, 12.0, 19.4, 32.2,8.9, 47.0. In terms of 12-per-cent units, theseequal 0, 1, 1, 2, 0, 3; and they are entered byas many crosses on the vertical line denotingdimension No. 4 in the figure.

The number of crosses in this figure is ap-proximately the same for the six traits. Thus:

Dimension 2Dimension 3Dimension 4Dimension 5Dimension 7Dimension 8

Number offluctua-tion units

162824292123

Number of yearsin which these

occur132319221822

139

10NTIROPOLOGICAL RECORDS

That there is a relation between large year-to-year fluctuations and wave crests or troughsis clear from figure 7, as compared with figures3-6, 8-11. It is also clear that the relationis by no means simple or complete. Sometimesthe fluctuations pile up in the years surround-ing a peak; thus, dimension 2, 1926-27.21 Inother cases, the fluctuations are most extremesome years before or after: dimension 3, 1807;4, 1804; 5, 1926; 7, 1811, 1923. Several timesthe fluctuations cluster continuously between anear-by crest and trough: dimension 4, 1788-1803;5, 1912-96; 8, 1795-1832. On the other hand,there are crests without any accompanying markedannual fluctuations: dimension 3, 1860, 1903; 4,1892, 1902/03; 7, 1852; and smaller clusters offluctuations remote from any peak: dimension 7,1824-31; 8, 1897-1903.

Essentially, each larger fluctuation repre-sents a one-year reversal of the current five-year trend. Periods of many accentuated fluc-tuations therefore are periods in which style isas it were two-minded or under strain; eventhough it may be moving rapidly in a certain di-rection, the movement is meeting with resistance.Periods of only minor fluctuation, on the con-trary, may be construed as times in which styleis progressing harmoniously and whole-mindedly,whether the change be rapid or slow. It isclear that 1840-1900 was such a period of har-mony, although it attained maxima in fullness ofskirt and slenderness of waist and near-maximain length of skirt and both long and high waist-edness. Table 23 summarizes these differencesby both five- and fifteen-year intervals.

It will be seen that the pre-1836 period ofunsettlement is really double. The fluctuationsare most marked and most numerous before 1800,then diminish, to resume again after 1815 andstraggle along until about 1835. In historicalterms, the Revolution-Directoire epoch was highlyunstable, the Empire fairly settled, the twentyyears after Moscow and Waterloo unsettled again.By 1830 quiet was impending, and 1848 was wellwithin a long calm.

Unsettlement began again, in one feature,about 1900; became acute in another in 1911; instill others about 1920, 1923, 1930. By 1933it had definitely diminished, except possibly inone feature: waist width. It is evident thatthe beginning of the era is pre-World War, itspeak post-War. Only in one trait, skirt diame-ter, do the greatest fluctuations occur duringthe War itself. The specific cause of this ex-ception seems to be a sharp reversal about 1915in a narrowing trend which had come to a pre-liminary peak in 1912, but did not reach its ex-treme until 1926. This extreme was reached andpassed with much less wobbling.

DIMENSION

FLUCT. UNITS

1790-

23

36

4+

4+

1800-

10-

20-

412

4+4.

4.

+

30-

52

712

+ +4

+ 44

44. r

4.

812

4

+t

$

40-

50-

60-

70-

4.

+.

80-+.

90-

1900-4.

10-

4.

20-

30- t.

+4-

+

4

4.

+

Fig. 7. Frequency of deviations fromfive-year moving-average trend, by fluc-tuation units per year, 1788-1934.

21 This is probably at least in part a functionof the moving average rounding a sharp crest.

-140

+E

F


If we compare the two eras of frequent annualreversals, it is apparent that the earlier,1788-1837, is more accentuated and may prove tohave been longer; at any rate if the quietingdown since 1932 continues after 1936. Fluctua-tions in all waist and decolletage dimensionsare definitely more marked during the earlierunsettled era; in skirt proportions, during thelater. This difference is of interest because

TABLE 23

Frequency of Fluctuation Units, 1788-1934(As per fig. 7)

1788-91 .......

1792-96.1797-1801 .....

1802-06 .......

1807-11 .......

1812-16 .......

1817-21 .......

1822-26 .......

1827-31 .......

1832-36 .......

1837-41 .......

1842-46 .......

1847-51 .......

1852-56 .......

1857-61 .......

1862-66 .......

1867-71.1872-76 .......

1877-81 .......

1882-86 .......

1887-91 .......

1892-96 .......

1897-1901.1902-06 .......

1907-11 .......

1912-16 .......

1917-21 .......

1922-26 .......

1927-31 .......

1932-34 .......

*At rate of 40

tAt rate of 33

In 5-yearperiods

141768)6513)66J300O0O1JO0O2JO101 125J4]48 j

12 1125J

in 15 years.in 15 years.

In 15-yearperiods

37 (14 years)*

19

25

3

1

2

8

16

29 (13 years)t

the Napoleonic period also attained sharp cli-maxes in shortness and narrowness of skirt; butrather peacefully, so to speak, as compared withthe 1926-27 climaxes. It would seem as if 1811-14 manipulated the skirt so far as it couldwithout basically questioning its nature, where-as 1926-27 was calling its very existence intoquestion; temporarily trying to rupture thebasic pattern of skirt, so to speak. The earlierera was somewhat similarly, though on the wholeless acutely, disturbed about waist and decol-letage proportions. In brief, its revolution-izing attempts concerned the bust; the recentones, the legs.

In connection with the somewhat greater fre-quency of early fluctuations, a statistical

caution must be noted. Before 1834, the averagenumber of observations per year is well underten; since 1920, above ten. The annual meansare therefore less well founded for the earlyera. Where observations number only five, three,two, or one for a year, fluctuations from thetrend may be due to smallness of the randomsample used; in other words, they may be appar-ent rather than real.

However, the long nineteenth-century or Vic-torian calm of small fluctuations is clearly be-yond possibility of doubt.

VARIABILITY WITHIN THE YEAR

This is the standard deviation or sigma ofthe individual measures around their mean forthe year. For uniformity among the six dimen-sions, these sigmas are expressed in percentagesof their means; that is, they have been convertedinto Coefficients of Variability, V - 100 a/M.

The full list of V's is given in table 24;their five- and ten-year averages in tables 25and 26.

TABLE 24

Percentage Sigmas of Annual Means, 1787-1936(V = 100 a/M)

Year 2 3 4 5 7 8

1787 .. 1.0 3.4 18.3 14.31788 .. 3.3 2.3 38.2 11.91789 .. 3.3 7.4 34.7 10.6 25.9 56.0

1790 .. 0.9 16.7 65.7 10.6 16.81791 .. 0.7 5.8 11.2 7.1 8.3 23.31792 1.8 3.81793 .. 0.3 6.2 59.2 4.1 16.31794 .. 0.0 6.4 7.1 10.2 14.6 21.11795 .. 3.3 7.6 24.6 15.8 6.9 10.71796 .. 0.5 8.7 44.7 3.81797 .. 0.9 3.6 0.0 16.2 1.71798 .. 0.8 5.2 12.9 7.2 11.1 22.21799 .. 1.1 8.1 21.0 14.5 15.3 26.3

1800 .. 0.7 12.4 22.1 14.4 16.1 12.91801 .. 1.4 11.8 9.5 29.2 9.4 5.91802 .. 0.0 6.7 3.51803 .. 1.3 11.1 17.5 39.1 7.6 16.61804 .. 3.6 9.3 36.6 27.9 5.9 18.61805 .. 1.1 8.0 8.3 24.7 12.7 12.31806 .. 1.1 7.4 18.5 33.6 11.4 21.61807 .. 1.9 10.0 18.0 28.8 10.6 10.81808 .. 2.0 11.2 11.8 41.7 10.4 28.91809 .. 2.5 11.9 16.5 38.0 10.7 19.3

1810 .. 3.1 12.6 13.0 37.6 11.7 18.21811 .. 4.0 14.9 11.3 41.1 6.6 12.61812 .. 1.7 13.1 7.0 49.6 11.9 12.01813 .. 9.0 18.6 17.7 42.2 9.5 16.71814 .. 5.5 8.5 4.2 14.3 8.6 15.81815 .. 2.9 9.6 7.4 33.9 9.7 17.61816 .. 2.8 9.4 31.7 17.1 4.1 12.5

141

142 ~~~~~~~~ANTHROPOLOGICAL RECOR?DS

TABLE 24 (Continued) TABLE 24 (Concluded)Year ] 2 I 3 1 4 I 8 Year I 2 I I 4~ I 5 I 7 I 8

1817 . 2.4 5.0 30.3 14.1 19.8 1877 ..1.0 4.2 8.3 13.5 10.8 16.11818 ..0.5 24.7 1.8 9.0 4.9 6.1 1878 ..1.6 5.9 11.8 15.1 6.5 17.81819 .1879 ..1.1 3.6 10.0 13.0 4.9 9.4

1820 ..5.7 7.5 1.9 4.6 5.8 2.3 1880 ..1.1 3.3 8.9 25.1 4.1 7.41821 ..2.3 7.9 6.9 3.8 4.0 10.0 1881 ..1.7 4.9 10.3 30.9 5.9 8.51822 ..1882 ..2.4 3.8 11.1 39.4 4.7 12.91823 ..1.7 7.2 1.4 1.9 4.4 2.7 1883 ..0.9 6.4 7.1 34.3 6.1 8.11824. 6.5 12.6 18.1 33.0 8.9 2.2 1884 ..1.8 4.5 6.8 25.9 9.9 7.11825 ..1885 ..1.9 5.3 8.5 33.5 9.6 13.41826 ..2.1 23.7 0.0 6.2 3.5 12.6 1886 ..1.4 7.5 12.0 27.0 16.5 21.11827 ..3.9 9.9 9.5 7.4 17.1 5.2 1887 ..1.7 4.3 7.4 23.6 9.1 9.91828 ..2.9 7.0 8.0 2.8 8.1 10.1 1888 ..1.0 5.9 12.3 32.4 12.0 14.11829 ..2.5 7.9 11.6 6.7 5.4 1889 ..2.0 5.6 8.1 26.7 13.8 17.1

1830 ..2.3 8.2 10.4 6.2 14.9 2.0 1890 ..1.1 5.2 9.6 21.6 7.9 13.81831 ..2.5 13.9 3.4 9.6 7.1 1891 ..1.2 6.9 10.4 27.2 7.0 8.11832 ..1.2 4.5 2.2 12.6 5.5 5.0 1892 ..0.9 6.4 11.8 22.2 7.7 20.81833 ..1893 ..1.2 7.8 15.2 27.5 14.2 16.31834 ..4.5 4.2 13.8 6.8 5.7 12.0 1894 ..1.0 6.9 7.5 15.3 8.2 13.51835 ..1.4 5.7 19.7 12.0 1.1 37.2 1895 ..1.2 8.5 10.6 15.6 8.5 19.21836 ..1.7 7.5 14.2 4.6 4.1 12.5 1896 ..1.2 5.4 12.5 7.1 13.5 12.71837 ..1.4 6.9 16.2 16.1 8.5 22.2 1897 ..0.3 3.3 15.6 9.2 5.8 15.41838 ..0.9 5.8 12.7 5.5 7.6 12.5 1898 ..0.6 4.8 24.7 9.0 10.1 20.71839 ..0.5 4.4 12.4 11.2 15.8 6.5 1899 ..0.0 3.7 8.1 11.0 12.7 12.5

1840 ..2.0 6.5 14.2 10.3 16.7 15.4 1900 ..0.8 5.0 12.8 8.2 7.3 14.21841 ..0.7 4.1 11.2 8.5 8.4 15.9 1901 ..0.6 7.1 14.0 6.0 9.4 31.81842 ..2.9 5.8 12.4 13.6 16.8 21.4 1902 ..0.0 9.1 15.5 14.2 7.9 22.31843 ..1.4 5.1 12.4 9.0 8.8 10.7 1903 ..0.0 8.3 14.0 18.5 7.1 22.91844 ..1.2 4.2 11.1 6.2 6.8 5.8 1904 ..0.0 5.0 17.8 10.3 6.8 14.31845 ..0.5 3.7 9.1 4.7 5.8 4.2 1905 ..0.0 3.0 9.3 7.7 3.6 12.41846 ..1.1 3.5 8.6 1.6 5.7 5.0 1906 ..0.9 8.3 13.7 5.0 10.5 20.61847 ..0.8 5.8 10.4 1.7 4.2 6.9 1907 ..0.8 6.2 13.2 20.2 16.5 17.21848 ..0.0 4.5 10.6 3.9 5.9 7.6 1908 ..1.0 6.2 23.1 11.3 13.8 19.21849 ..1.0 2.2 8.7 6.7 8.2 6.5 1909 ..0.6 7.4 25.1 15.7 13.5 14.9

1850 ..0.4 4.2 8.0 5.7 8.1 6.1 1910 ..1.8 6.8 17.9 24.7 7.8 13.01851 ..0.9 3.8 10.9 8.5 6.0 5.2 1911 ..1.1 4.3 15.9 24.4 7.8 12.01852 ..1.3 3.7 9-.6 7.4 5.5 10.4 1912 ..2.1 8.4 8.7 19.4 9.3 9.31853 ..0.7 2.8 10.1 5.4 5.6 11.9 1913 ..4.4 7.3 20.5 37.1 15.0 16.21854 ..0.3 2.9 9.3 12.7 8.2 7.6 1914 ..2.2 6.1 18.8 46.0 14.5 11.81855 ..1.0 5.9 9.0 9.6 7.9 8.0 1915 ..3.4 13.1 12.0 51.5 10.8 10.41856 ..1.3 6.1 15.1 3.5 5.7 5.8 1916 ..4.3 8.0 21.7 27.3 12.4 21.81857 ..0.8 6.5 13.9 3.7 14.8 13.7 1917 ..5.3 5.7 5.3 36.0 19.1 10.71858 ..0.8 2.6 15.5 7.2 8.1 7. 1918 ..1.9 7.6 14.9 40.8 7.6 11.51859 ..0.0 3.2 14.9 4.1 10.1 19.2 1919 ..6.6 10.6 13.9 64.5 8.6 12.5

1860 ..0.6 5.9 2.7 2.7 6.4 7.6 1920 ..3.8 6.2 16.4 44.3 9.3 17.41861.. 0.0 4.2 11.5 7.0 12.5 4.2 1921 ..5.1 14.8 27.5 35.8 13.5 26.61862 ..0.8 6.3 14.6 6.2 11.4 7.6 1922 ..3.8 10.7 31.9 54.5 10.2 12.61863 ..1.4 1.4 7.6 4.7 0.0 9.9 1923 ..3.0 17.6 30.6 52.6 6.9 21.51864 ..1.0 4.7 3.7 3.9 7.4 9.6 1924 ..4.6 13.3 19.7 44.7 11.4 9.61865 ..0.4 5.0 12.7 4.7 5.8 8.6 1925 ..3.3 19.8 28.0 30.4 11.3 15.91866 ..0.6 4.9 13.3 1.5 9.5 7.8 1926 ..3.1 19.2 18.5 17.6 8.3 15.51867 ..1.5 4.6 15.9 7.2 7.9 12.0 1927 ..1.9 13.2 17.3 19.0 13.8 15.01868 ..1.5 6.8 8.1 15.1 12.5 12.9 1928 ..5.4 7.7 18.3 52.3 11.3 8.01869 ..0.0 2.8 10.4 10.7 0.0 9.4 1929 ..4.4 10.8 12.7 23.0 24.2 23.1

1870 .~1.3 6.3 10.9 16.3 17.2 17.0 1930 ..6.4 12.1 10.4 19.2 31.0 10.11871 ..0.9 7.7 12.5 21.8 0.0 17.5 1931 ..3.8 11.3 21.7 19.7 17.4 28.51872 ..1.1 8.7 16.1 12.6 7.5 7.4 1932 ..2.6 12.4 22.3 29.6 17.3 24.21873 ..1.3 5.8 9.6 9.5 8.2 22.1 1933 ..3.0 8.8 28.6 27.3 26.2 21.91874 ..1.0 3.9 4.5 12.4 5.3 14.1 1934 ..1.4 8.1 44.7 23.8 16.4 35.31875 ..0.0 7.3 8.0 10.2 4.7 3.7 1935 ..2.1 14.1 33.8 29.2 23.5 29.11876 1.21.2 7.9 10.4 8.9 1 1.8 _25.9 1936 ..2.1 110.0 32.8 139.3 117.3 136.6

142


TABLE 25Five-Year Averages of Percentage Sigmas

1787-1936(V = 100 T/N)

Period 2 3 4 5 7 8

1787-91 .... 1.8 7.1 33.6 10.9 10.2 19.81792-96 .... 0.8 5.8 27.1 6.8 8.3 15.91797-1801 .. 1.0 8.2 13.1 16.3 10.7 13.51802-06 ... 1.4 8.5 16.2 25.1 8.2 13.81807-11 .... 2.7 12.1 14.1 37.4 10.0 18.01812-16 .... 4.4 11.8 13.6 31.4 8.8 14.91817-21 .... 2.2 9.0 8.2 6.3 3.0 9.61822-26 .... 2.6 10.5 4.9 10.3 4.2 4.41827-31 .... 2.8 9.4 6.3 7.5 10.8 4.51832-36 .... 2.2 5.5 12.5 9.0 4.1 16.7

1837-41 .... 1.1 5.5 13.3 10.3 11.4 14.51842-46 .... 1.4 4.5 10.7 7.0 8.8 9.41847-51 .... 0.6 4.1 9.7 5.3 6.5 6.51852-56 .... 0.9 4.3 10.6 7.7 6.6 8.71857-61 .... 0.4 4.5 11.7 4.9 10.4 10.41862-66 .... 0.8 4.5 10.4 4.2 6.8 8.71867-71 .... 1.0 5.6 11.6 14.2 7.3 13.81872-76 .... 0.9 6.7 9.7 10.7 7.5 14.61877-81 .... 1.3 4.4 9.9 19.5 6.4 11.81882-86 .... 1.7 5.5 9.1 32.0 9.4 12.5

1887-91 .... 1.4 5.6 9.6 26.3 10.0 12.61892-96 .... 1.1 7.0 11.5 17.5 10.4 16.51897-1901 .. 0.5 4.8 15.0 8.7 9.1 18.91902-06 .... 0.2 6.7 14.1 11.1 7.2 18.51907-11 .... 1.3 6.2 19.0 19.3 11.9 15.31912-16 .... 3.5 8.6 16.3 36.3 12.4 13.91917-21 .... 4.5 9.0 15.6 44.3 11.6 15.71922-26 .... 3.6 16.1 25.7 40.0 9.6 15.01927-31 .... 4.4 11.0 16.1 26.6 19.5 16.91932-36 .... 2.2 10.7 32.4 29.8 20.1 29.4

TABLE 26Ten-Year Averages of Percentage Sigmas, 1787-1936

(V = 100 a/N)

Period 2 3 4 5 7 8

1787-96 .... 1.3 7.0 *30.4 8.9 9.3 t17.91797-1806 .. 1.2 8.4 14.7 20.7 9.5 13.71807-16 .... .t3.6 t12.0 13.9 f34.4 9.4 16.51817-26 .... 2.4 10.0 6.6 8.3 3.6 6.01827-36 .... 2.5 7.5 9.4 8.3 7.5 10.61837-46 .... 1.3 5.0 12.0 8.7 t10.1 12.01847-56 .... 0.8 4.2 10.2 6.5 6.5 7.61857-66 .... 0.6 4.5 11.1 4.6 8.6 9.61867-76 .... 0.9 6.2 10.7 12.5 7.4 14.21877-86 .... 1.5 5.0 9.5 25.8 7.9 12.21887-96 .... 1.1 6.3 10.6 21.9 10.2 14.61897-1906 .. 0.3 5.8 14.6 9.9 8.2 18,71907-16 .... 2.5 7.4 17.7 27.8 12.2 14.61917-26 .... *4.1 *12.6 20.7 *42.2 10.6 15.41927-36 .... 3.3 10.9 124.3 28.2 *19.8 *23.2

*Highest value in column.ISecond highest (other than in adjacent

decennia).

It is at once evident that the variability ismarkedly different among the six dimensions.Dress length, dimension 2, again shows much thelowest variability, and waist length next least.The four other dimensions run about alike; thoughthe two decolletage measures show a strong pre-ponderance of V's between 10 and 20. The littlesubjoined table (no. 27, based on table 25) showsthe distribution of size of five-year averagedV's.

TABLE 27Distribution of Size of Variability Coefficients

(Five-Year Averages) Among the Six Measures

Variability 2 3 4 7coefficients 2 3 4 5 7 8

1.9 or less .. 192.0-4.9 ...... 11 7 1 2 3 25.0-9.9 ...... - 17 7 8 14 510.0-19.9 .... -- 6 18 10 12 2220.0-50.0 .... __-. 4 10 1 1

It can be concluded from this that dresslength, and next to it waist length, can bevaried least from the ideal norm of a given mo-ment if a dress is to be within fashion. Withrespect to decolletage and all transverse dimen-sions, the style is much less strict, and muchmore variability is exercised, within the yearand within a five-year period. What our aesthetictaste assumes as primary in the style norm, andinhibits too great departures therefrom, is thelength of the dress as a whole; next, the positionof the waist constriction. Skirt fullness, waistdiameter, and length and breadth of decolletageare allowed much more individual variation fromdress to dress.22

The first thing that is evident from tables 25and 26 is that there are once more an early periodof high variability, a middle one of low, and arecent one that is high again. A table could beconstructed that would be similar to table 23.Instead, in table 28 we give the maxima of V infive-year means.

22These different behaviors of the six dimen-sions are perhaps partly a function of theirabsolute size: No. 2 is of course by far thelargest measurement, and Nos. 4, 7, and 8 thesmallest. With the small dimensions, the proba-bility of error of caliper measurement isgreater, presumably tending to increase thevariability. However, the differences are notwholly a function of size, because skirt width(No. 5) consistently runs larger than waistlength (No. 3), yet has a variability like thatof the small dimensions. At least a consider-able part of the variability difference betweenthe dimensions therefore represents stylisticquality.

cz - -- X q

143


TABLE 28Maxima of Five-Year Averages of Coefficients

of VariabilityPeriod 2 3 4 5 7 8

1787-91 34 401792-96.1797-1801 111802-06.1807-11 .12 371812-16 ......41817-21.1822-26 ..........1827-31 .......... 111832-36 ..........1837-41 111842-46 ..........1847-51 ..........1852-56 ..........1857-61 ..........1862-66 ..........1867-71.1872-76.1877-81.1882-86......1887-91......1892-96 ...........1897-1901 .......1902-06......1907-11 .......1912-16 .....1917-21 ...........5 441922-26 .161927-31.1932-36 ....32_..3220 29

Much the same appears from the stars anddouble daggers in the ten-year table 26.

As before, high variability tends to be as-sociated with extreme of dimension, but not con-sistently so. The reason for the inconsistencyis in this case clear, and will be the nextpoint discussed.

It occurred to us to plot together the di-mension means and their variability coefficientson scales calculated to bring out such simi-larity of course as they might or might notpossess. Five-year averages were used to plotskirt and waist,.ten-year for decolletage.Figures 8 to 10 show the results.

It is clear that in four cases out of six,and mainly in a fifth, there is a definite andsurprising relation between large dimension andlow variability; conversely, when the dimensionshrinks, the variability goes up. This is veryconspicuous for both skirt and both decolletagediameters (Nos. 2, 5, 4, 8; figs. 8, 10). Itholds also fairly well for waist length (No. 3;fig. 9), except before 1821 and after 1921 .whenit.revbrses.3 Waist width (No. 7; fig. 93 must

23In No. 3 of fig. 9, V above 10 and M below20 or above 32 have been indicated by extrablackness of line, to emphasize that before 1821and after 1921 the variability reacts to extrememeans in opposite manner.

be read reversed (low variability accompanyinglow mean values) throughout, to achieve the bestfit.

Now what is the meaning of this relation ofdimension magnitude and variability? Evidentlythat when fashion brings a given trait to a

17f1 1101 III I1 311 4II I I 71 | 11 1ITI III 21 3II 41MV

82 5

M 4

20 3

92

202

5

nD_ O

M Vso 45

30 40

9Q 35

.W30

W 25

D 20

eD 15

2Q 10

IQ: 5

llQ O

2

5

./

I

Fig. 8. Relation of variability (V) and ampli-tude of dimension (M) in skirt length (2) andwidth (5), 1787-1936.

certain magnitude, the style is harmonious andwell-knit on that point, and individual produc-tions, or designs, are in close concord. Con-versely, when this magnitude is departed from,the style is under strain as regards that fea-ture, and efforts are made simultaneously to re-cede from the magnitude attained and to advancebeyond it. In other words, from the angle ofunderlying pattern of style, there seems to bean optimum magnitude or proportion for each fea-ture, when variability is low, and the style isconcurred in because it is felt to be satisfying.

There appears no reason why this explanationshould not be applicable to the minority of casesin which low variability accompanies low meanvalues. That is to say, in most of our traitsthe basic style is felt as satisfying, and re-

144

RICHARDSON AND KROEBER: TH3REE CENTURIES OF WOMEN'S DRESS FASHIONS

mains stable, when the silhouette dimension isample; but in other traits, when it is small ormedium.

On this interpretation we can construct abasic or ideal pattern of Occidental women'sevening or formal dress during the past 150years. It has a long skirt, ample at the bot-tom; an expanse of bare breast and shoulders,as deep and wide as possible, although for me-chanical reasons only one diameter can well beat maximum at the same time; as slender a waistas possible; and a middle or natural waist-lineposition, between 22 and 30 by our scale; whenthe waist line gets beyond these limits, andcrowds either the breasts or the hips, the basicpattern is violated, resistance and extravaganceare developed, and the variability rises.

To put it differently, a confining corsetmay be uncomfortable to the wearer, but it isfelt as aesthetically satisfying by Europeansof the last century and a half, even if it con-stricts unnaturally, provided it comes at ornear the natural waist. Skirts on the otherhand cannot be too full or too long, and breast

I7 I III I1 311 11 5 II 7,1 8 1 1T 111 1 311 1MVIa 16

20 13

22 0C247

2 4424

20

32

a0

MV& 20

1fi 18

14 16

12 12

uL 10

la 8

6

.& 4

7 2

and shoulder exposure too ample in eveningdresses, to satisfy the ideal of the style.

However, we have not only this basic patternor ideal style, which is aesthetic with a ting-ing of the erotic, but also a concept of tempo-rary mode or fashion as such, which demands change,and, when it has exhausted the possibilities ofmaterial, color, and accessories, goes on to al-ter the fundamental proportions, in other wordsthe basic aesthetic pattern. With such altera-tion there comes strain, simultaneous pullingforward and back; violent jumps in opposite di-rections within one or two or three years, andheightened statistical variability.

The several proportions are successfully at-tacked and distorted by fashion at somewhat dif-ferent times, and hence the picture is compli-cated. Nevertheless, there emerge periods of ageneration or so when fashion is particularlyactive in its attempts to break up or pervertthe basic pattern. Such are the decades 1785 to1835, and 1910 to the present. Between them,there lies a longer period of essential agree-ment and stability and low variability, in which

1796I16T3f AA 5r 66 7 af gt 1906If 26 3

3- MV- 31

- a227

- LH23

IL 19

- 1L 15

- 14 11

- jL 7

7M V

-_ L 29r _ c 26

\1-4 20

\ - 16 7

IL1 1412.23

-2 11

- 22 8

'v/ v A I

' I

iV

4

8

I I

Fig. 9. Relation of variability (V) and ampli-tude of dimension (M) in waist length (3) andwidth (7), 1787-1936.

Fig. 10. Relation of variability (V) and ampli-tude of dimension (M) in decolletage depth (4)and width (8), 1787-1936.

145


fashion accepted, or fulfilled, the patternwhile modifying it in superficial detail.

We have too few data to compute variabili-ties before 1787. This is unfortunate becausemost of the eighteenth century evidently re-sembled the middle and late nineteenth in hold-ing fairly close to what we have determined asthe basic pattern: the skirt full and ratherlong, at least not markedly short; the waist,if not narrow, at least accentuated, and inmedian position; decolletage considerable. Ifour hypothesis holds, the bulk of the eighteenthcentury should accordingly.prove to be a periodof low variability, on assembly of sufficientdata.24

However, we can make the trial assumptionthat the specific associations of variabilitieswith crests which we have found to hold since1787 also held before that date, and see how theresults plot out. That is to say, while we haveno reliably computable variabilities for most ofthe seventeenth and eighteenth centuries, we dohave fair approximations to the points in timeat which the maxima and minima of dimensionsfall; and by plotting the maxima and minima forthe whole three-hundred-odd years, we may hopeto discover whether the pre-1787 period shows atendency toward clustering of crests comparableto that after 1787. The result of the experi-ment is shown in figure 11. Solid circles showthose dimension crests, whether maxima or minima,which since 1787 have been associated with lowvariabilities and pattern stability; hollowcircles, crests with the opposite association.

The diagram makes it evident that there wasa clustering of crests between 1630 and 1680.Seven of the ten crests in this period fall be-tween 1645 and 1665.

However, the ten crests are nearly evenlydivided between those hypothetically associatedwith high and with low variability, and the twokinds are interdistributed scatteringly. Whilewe may accordingly infer that the mid-seven-teenth century was a period of attainment ofstyle dimension extremes, and rapid alternationbetween extremes in at least some features,there is nothing to prove that the particularpost-1787 associations of one of a pair of ex-

241t would also be desirable to try to definethe basic pattern of dress by inclusion of morefeatures than the six so far dealt with. Thetreatment of the arms, bust, and hips, in thebasic pattern, have not been considered at all.There are important traits here: sleeves; promin-ence and position of the bust; proportion of thehips to shoulders, bust, and base of skirt--compare for instance the Grecian bend and bustlefashions with the recent one of hips larger thanbase of the skirt. But the difficulties areconsiderable in dealing with these features overlonger ranges of times: some disappear and re-appear, others require profile views for fullmeasurement. Nevertheless something could nodoubt be ascertained by further analysis.

1620-

30-

40-

50-

60-

70-

2 5 3 7 4 8

00

0 00

0

80- 0

90->

1700- NO

10J

20-

30-

40-

50-

60-

70-

80- 0

90- *

000-

10-0

20-

30-

40-

50

60- -0

70-

DATA

0

-i0

I-

.~~~~~~~~~~~~p

I-

.0 :O ~D

0~~

O (~~~~~~~~LUo 0.

0 t0

*00Q~~~~~~~~~~'.

* 0~~

*, 0

0

Fig. 11.Clustering ofextremes as re-lated to varia-bility.

Solid circles:Maxima in di-mensions 2, 5,4, 8; minima in7; moderatecrests in 3;associated withlow variabili-ties and con-formity to pat-tern since 1787(no earlierdata availableon variability).

Hollow circles:Minima in di-mensions 2, 5,4, 8; maximain 7; extremecrests in 3;associated withhigh variabili-ties and pat-tern strainsince 1787 (noearlier dataavailable onvariability).

80-

90

1900-

10

20-

30- 0

0

0

0 0

0

0

0 0

146


tremes with heightened variability already heldin the seventeenth century.

Now follow nearly a hundred years, from 1680to 1777, with but two peaks, in dimensions 5 and7. Even on allowance for there being no datafor 1690-1710, there remains a three-quarter-century span with but these two crests. Undersuch pervading stability, variability may bepresumed to have been low.

The next forty years, 1778 to 1817, showeleven crests. Four of these are of the typedetermined as of low variability, seven of high.Moreover, the low-variability ones fall mostlywithin the first half of the period (mean date1788), the high in the latter half (mean date1802).

The following eighty-odd years, to 1900,possess nine crests, seven of them of low-variability type. The period seems to consistof two spans. First, some two decades, 1825-42, of quieting down from the preceding turbu-lence. Then a long calm, only slightly ruffledby four low-variability crests (and one high)in the 1860-69 decade, and none at all forthree decades after. This Victorian era wascertainly placid in fashion.

From 1900 to date, there are ten crests, thefour earliest of low variability, the last sixof high. Or we might say that 1903-17 was atime in which variability was increasing butonly low-variability maxima were reached; itwas a final phase of the preceding stable period:strain was already manifest but reaffirmationsof the dying pattern were being made. The period1923-34, by contrary, shows in every featureexamined a crest which is in extreme oppositionto what prevailed during the long Victorian calmand is in each instance accompanied by very highvariability. Table 29 summarizes this.

TABLE 29Extremes of Dimension and Variability by Period

Period Dimension extremes(rounded) Years Low High Total Per___________ var. var. decade

1630-1680 ... 50 (6) (4) 10 2.01681-1777 ... 97 (1) (1) 2 0.21778-1817 ... 40 4 7 11* 2.71818-1902 ... 105 7 2 9 0.91903-1934 ... 32 4 6 10 3.1

*The variability type of three crests before1787 is not known.

It is clear from this table as well as figure11 that there occur in European women's dressalternating longer periods in which a basic pat-tern of style is rather stably adhered to, rela-tively few extremes of proportion or dimensionare sought, and those all in a direction accom-panied by only low variability from year to year

and dress to dress; and shorter periods in whichbasic pattern is disrupted or transformed, ex-tremes of proportion are numerous, and high vari-ability prevails.

This differentiation of periods is positivein all respects for the era since 1787. It holdsas regards stability of basic pattern and infre-quency of extremes for the hundred and eightyyears preceding. Whether it also holds for theassociation of variability with those extremeswhich conflict with enduring stable patterns,we have not the evidence to prove or disprove;but at least there is nothing in the imperfectpre-1787 picture to argue against such variabil-ity association.

CAUSALITY OF CHANGE

We are now in position better to weigh theseveral possible causes of changes in variability.

The primary factor would seem to be adherenceto or departure from an ideal though unconsciouspattern for formal clothing of women. The con-sistent conformity of variability to certain mag-nitudes of proportion--mostly a conformity of lowvariabilities to high magnitudes--leaves littleroom for any other conclusion.

A second possible explanation, that high vari-ability is a function of extremes of proportion,falls as such. It is true for a full waist anda narrow or short skirt, untrue for slender waistor full or long skirt. The explanation holdsonly so far as it is subsumed in that of thebasic pattern.

A third possible explanation, that generic ornonstylistic factors unsettle fashion at certaintimes, is not eliminated, but is pushed into thebackground of further investigation. After all,such a cause would be an ultimate, not an immedi-ate one. It may well be that unsettled timesmake for unsettled styles. Revolution, Napole-onic and World wars, struggles over the rightsof man, Communism and Fascism, the motor and jazz,may contribute to fashion's trying to stretch anddisrupt its fundamental stylistic pattern. Butwhile such an influence is easily conjectured,it is difficult to prove. In any event, thereseems no clear reason for the specific fashionextremes which such a set of causes might bethought to produce. Social and political un-settlement as such might produce stylistic un-settlement and variability as such; but there isnothing to show that it would per se producethick waists, ultra-high or low ones, short andtight skirts. If there is a connection here, itseems that it must be through alteration of thebasic semi-unconscious pattern, through an urgeto unsettle or disrupt this; and that when in-creased fashion variability occurs, it is as adirect function of pattern stress, and only in-directly, and less certainly, of sociopoliticalinstability. In short, generic historic causes

147

ANTHROPOLOGICAL .RECORDS

tending toward social and cultural instabilitymay produce instability in dress styles also;but their effect on style is expressed in stressupon the existent long-range basic pattern ofdress, and the changes effected have meaningonly in terms of the pattern.

Concretely, it would be absurd to say thatthe Napoleonic wars, or the complex set ofhistoric forces underlying them, specificallyproduced high-waisted dresses, and the World Warlow-waisted ones. They both probably did pro-duce an unsettlement of style, which, however,resulted in extremity of high and low waisted-ness respectively.25

Herewith arises another question: whether thecrests and troughs of waves of fashion, its peri-odicities discussed in section IV, are perhapsalso to be sought not in anything inherent infashion, but rather in more general historiccauses. In favor of such a view is the heavierclustering of trait extremes in Revolutionary-Napoleonic, World-War, and immediately subse-quent decades. But again there are crests alsoin the intervening period. What is specifically

characteristic of the agitated periods is notso much extremes of dimension or proportion, asextremes of high variability; and these in turncorrelate with certain minima and maxima of pro-portion, but not with their opposites. The sig-nificalnt fact remains that high variability isnot associated with any dimensional crest, butalways26 with only one of a pair of opposing ex-tremes. This throws us back on the basic pat-tern as something that must be recognized.

Now, one can indeed accept this basic pat-tern, but accept it as something intrinsicallytending to remain more or less static over along period, or the whole of a civilization; andthen attribute the more marked variations fromit to broader historic disturbing causes, ratherthan to anything stylistically inherent andtending from within toward swings away from andback toward the pattern. On this view the cen-tury-long cycle which we have found to hold for.most of our fashion traits would not be a prop-erty of style per se, but a by-product of thefact that Europe happened to be generically dis-turbed in the decades around 1800 and 1920.

VIII. CONCLUSIONS

Our first finding is that the basic dimen-sions of modern European feminine dress alter-nate with fair regularity between maxima andminima which in most cases average about fiftyyears apart, so that the full-wave length oftheir periodicity is around a century.

By comparison, annual changes, and even thoseof moderately long periods of moderate length,generally are markedly less in degree or ampli-tude. This conclusion applies to the major pro-portions of the total silhouette. Superstruc-tural features have not been examined quanti-tatively, but appear to develop and pass awaycompletely in briefer cycles. The present studyis concerned with the variations in persistentfeatures.

There appear accordingly to be two componentsin dress fashions. One is mode in the propersense: that factor which makes this year'sclothes different from last year's or from thoseof five years ago. The other is a much morestable and slowly changing factor, which eachyear's mode takes for granted and builds upon.It cannot be pretended that these two factorsare definably distinguishable throughout. Be-havioristically, however, they can mostly beseparated by the length and regularity of thechanges due to the more underlying component.

It is evident that the basic features ofstyle as distinct from more rapidly fluctuatingmode, being taken for Tranted at any given mo-ment, are largely unconscious in the sense thatthey are felt as axiomatic and derivations are

made from them, but they are not tampered with,except again unconsciously.

This in turn seems to imply that the r6le ofparticular individuals in molding basic dressstyle is slight. The influence of creative orimportant individuals is probably largely exertedon the accessories of transient mode. How greatit is there, has never been objectively examined,and would be difficult to investigate. Histori-ans of fashion may be partly right or mainlyfictitious in the influence they assign to MarieAntoinette, Recamier, Eug6nie, and the variousPrinces of Wales. The reverse is much morelikely, that individuals conform to the stylewhich they find in existence, operate in minorways within its configuration, and at times ofcoincidence receive false credit for "causing"one or more of its features.

25The Empire mode was consciously Greek or Neo-Classic. It professed to take over from anti-quity a full and rather high waist and a fallinginstead of flaring skirt. It obviously did nottake over from antiquity its own short skirt andwide decolletage, nor its ultra-high waist, nora tight, undraped skirt, nor short puff-sleeves.In brief, Empire dress style fell in with thecatchwords of its day, and in consonance withthe social currents and political currents ofits time, which aimed toward the Classic, ac-cepted just as much of Classic dress style assuited its own trends, and for the rest followedthese trends while calmly ignoring or violatingall the remaining features of its supposed model.

26Excepting dimension 3, waist length, where lowvariability is associated with medium magnitude.

148


The long swings of proportion which we havedetermined seem comparable to what economistscall secular trends, which also carry oscilla-tions or lesser cyclic movements on their sur-face. No one attributes either these largereconomic trends or the fluctuations to individualinitiative. It is of course conceivable thateconomic determinants are social in their natureand stylistic ones individual. In fact this isoften assumed. However, such an assumption isnaive in the sense of being critically untested.It is rather more likely that what holds in onedomain of human culture holds also in another.At any rate, the burden of proof must rest onthe contrary view. And this burden has cer-tainly doubled since we have shown that dress-style changes behave historically somewhat likeeconomic ones; in the stateliness of their march,or trend, for instance, and in their superim-posed cycles or oscillations.

The evidence to date shows that when a pro-portion has swung one way to its extreme andgone halfway the other, it may oscillate for adecade or two part way back to the first ex-treme, but normally it resumes its swing towardthe opposite. But this is a behavioristic find-ing, and a priori may just as well be due tocultural as to personal causes. So far as in-dividuals are concerned, the total situationseems overwhelmingly to indicate that their ac-tions are determined by the style far more thanthey can determine it.

No generic significance can be claimed forthe value of a century found for the averageperiodicity or wave length of dress proportions.It is only a mean, though it is rather closelyadhered to in three of our six features. Obvi-ously, other features, or styles other thanmodern European ones, may possess quite differ-ent periodicities. In fact, there is no reasonwhy style in general, or even dress style,should necessarily swing rhythmically back andforth. Our findings apply only to the materialanalyzed.

Definitely significant is the fact that thereare periods of high and of low variability ofstyle. These come out much alike whether it isa matter of variations of yearly averages fromthe five-year moving mean, or of variations ofindividual dresses from the year's mean. Withinthe last century and a half, 1787-1835 (espe-cially before 1820) and 1910-36 are periods ofhigh variability. The intervening seventy-fiveor more years show low variability. The avail-able measures scarcely allow of variability com-putations for most of the eighteenth century,but the general pattern apparently underwent novery marked alterations in that century untilafter 1775.

The two high-variability periods also con-tain more crests or extremes of proportion thanthe intervening seventy-five years or than thestable bulk of the eighteenth century. There is

therefore a relation between extremes and vari-ability.

However, this relation is one-sided. Forfour of the six proportions examined, variabilityrises as the proportion or diameter shrinks, be-comes low as this reaches ampleness. For a fifthproportion, waist width, the relation is the op-posite. For the sixth, variability becomes acutewhen the measure is either very high or very low.

High variability thus is more completelylimited to certain periods than are extremes ofthe proportions or diameters themselves. Thoseof the diameter extremes which are accompaniedby low variability fall in some cases into thelong stable interval.

The best explanation that we are able to sug-gest for these phenomena is that of a basic pat-tern of women's dress style, toward which Europeanculture of recent centuries has been tending asan ideal. This pattern comprises amplitude inmost dimensions, scantness or medium value inothers. As these proportions are achieved, thereare equilibrium, relative stability, and low vari-ability. The pattern may be said to be saturated.At other times, most or all of the proportionsare at the opposite extreme, which may be con-strued as one of strain, and variability riseshigh. This basic or ideal pattern, for Europeof the last two or three centuries, requires askirt that is both full and long, a waist thatis abnormally constricted but in nearly properanatomical position, and decolletage that is am-ple both vertically and horizontally.

The periods of computed high variability andtherefore of "strain" or perversion of patterncoincide fairly closely with the Revolutionary-Napoleonic and World War-post-War eras. Genericcultural or historic influences can thereforeprobably be assumed to affect dress-style changes.Sociocultural stress and unsettlement seem toproduce fashion strain and instability. However,they exert their influence upon an existing sty-listic pattern, which they dislocate or invert.Without reference to this pattern, their effectwould not be understood.

While we have no reliable variability meas-ures before 1787, it is clear that in the dec-ades surrounding 1650-60 there was an accumula-tion of proportion extremes similar to those of1787-1835 and 1910-36. The mid-seventeenth cen-tury may thus have been a third period of pat-tern strain, rapid change, and variability.

The explanation propounded is not that revo-lution, war, and sociocultural unsettlement inthemselves produce scant skirts and thick andhigh or low waists, but that they disrupt theestablished dress style and tend to its over-throw or inversion. The directions taken inthis process depend on the style pattern: theyare subversive or centrifugal to it. By con-trary, in "normal" periods dress is relativelystable in basic proportions and features: itsvariations tend to be slight and transient--

149


fluctuations of mode rather than changes ofstyle. In another civilization, with a differ-ent basic pattern of dress style, generic socio-cultural unsettlement might also produce un-settlement of dress style but with quite dif-ferent specific expressions--slender waists andflaring skirts, for instance, or the introduc-tion or abolition of decolletage.

It is conceivable that the method pursued inthis study may be of utility as a generic measureof sociocultural unsettlement. Also, it pro-vides an objective description of one of thebasic patterns characteristic of a given civili-zation for several centuries, and may serve as aprecedent for the more exact definition of otherstylistic patterns in the same or other civili-zations.

It also seems possible that the correlationwith general conditions explains the near-regu-larity in the periodicities of dress. If theselargely express pattern disturbances due todisturbances more general in the culture, there

is no need to fall back on assumptions of anunknown factor inherent in dress itself and mak-ing for rhythmic change.

We have deliberately avoided explanation ofour phenomena in terms of psychological factorssuch as imitation, emulation, or competition,which are a stock explanation: the leaders wantto surpass the mass, so they keep going onestep farther, until a physical limit is reached,when they turn about and head the processionback. We do not deny that such psychologicalmotivations may be operative. We do believethat as explanations they are conjectural, andscientifically useless, because, to date atleast, they depend on factors which are unmeas-urable and undefinable. On the contrary, wethink we have shown that through behavioristicand inductive procedures operating wholly withinthe sociocultural level, functional correla-tions can be established for such supposedlyrefractory cultural manifestations as style andfashion changes.

150

APPENDIXBy A. L. KROEBER

Allport and Hartman27 have analyzed themethod of my paper of 1919, along with Chapin'sbook on Cultural Change. Some of their criti-cisms will presumably be raised, by them orothers, against the present study. It seemstherefore worth while to consider their argu-ments.

First of all, it is significant that Allportand Hartman call their analysis "The Predictionof Cultural Change."28 This was certainly notmy main purpose. I find only two predictionsin my paper; and neither of these is material.29Both are affirmations of emphasis of my convic-tion that there is "order" or "regularity" instyle changes. If it will clarify the issue,I shall be glad to withdraw both predictionsas having been unnecessary.

Why the emphasis of the analysis is so strongon prediction, is hard to see, especially asChapin's Cultural Change also is concerned pri-marily with change and not with prophecy. Theprobable explanation is that the analysts, incommon with most sociologists and economists,are themselves interested in the future, in thepractical consequences or applications of study,in short are so committed to what they call the"telic" approach, that they assume it and thehistoric approach to be the only ones possiblein the field of social data.

It seems to me that my approach in the formerpaper was a "natural science" one in the sensethat it was empirical, inductive, objective,and free of any motivation of applicability or

27Floyd H. Allport and Dale A. Hartman, ThePrediction of Cultural Change: A Problem illus-trated in studies by F. Stuart Chapin and A. L.Kroeber, pp. 307-350 of Stuart A. Rice, editor,Methods in Social Science: A Case Book, 1931.

28 That this is not a mere matter of an editor'Etitle is shown by the analysts' statement onpage 316: "The central question of ProfessorKroeber's research concerns the possibility ofprediction from this linear stylistic approach";and by their devoting nine pages to an analysisof the natural-science, genetic, telic, andstylistic approaches with reference to theirrespeotive predictive value.29pp. 249-250: "By 1912 the tide has once

more turned--no doubt to continue now for an-other two or three score years unless the peri-odicity of the rhythm is accelerated. . .This forecast has been proved premature by theevents. 1911-12 was indeed a low point forskirt width, and the amplitude increased until1917, but the narrowing resumed until a stilllower point was reached in 1926, which pre-sumably marks the real trough of the wave.P. 258, paraphrased: skirts will be longer about1928-40 than in 1919. This is correct since1930. For anyone setting out to be a prophet,one verification out of two would be a sorryrecord.

social control. It is also a "genetic" approachin the sense that it deals with unique histori-cal phenomena not subject to experimental veri-fication. How far the approach is "stylistic"in the Allport-Hartman sense, other than thatthe material dealt with concerns style, I cannotsay, because I do not understand their definitionof this approach. It seems rather a subdivisionof the "genetic," in so far as styles are his-torical phenomena. In fact, is not the primaryconcern of history and cultural anthropology, sofar as they transcend the writing of individualbiographies, precisely a dealing with "styles"of human behavior? The one thing that my paperwas free from, as I see it, was telic or practi-cal approach. I may be wrong here: one is a poorjudge of his own motivations: but it is also pos-sible that Allport and Hartman have started outwith preconceptions which have caused them tomisunderstand my objectives, underlying and pat-ent. At any rate, so long as theirs is the onlyformal criticism which my article has evoked, Ishould like to go on record.

Where I have presumably given provocation isin the loose use of deterministic terms such aslaw (usually in quotation marks), principle,cause, order, and regularity. If it will clearthe atmosphere toward understanding, I willcheerfully retract any of these that are confus-ing. There is no use quarreling over metaphors.That there is a certain "order" or "regularity"in the phenomena, enough to prevent their beingconstrued as due to the caprices of individualhuman wills, I continue to believe, and thinkDr. Richardson and I have overwhelmingly provedin the present paper. "Determinism" in thissense I adhere to: that the actions of individ-ual persons are determined much more by stylesand other sociocultural influences than they de-termine them. As to "laws," it seems perfectlyclear from the context even where there is noexplicit qualification, that I was not claimingto have found laws comparable to those of phys-ics. However, let us consider all statementson this score as withdrawn.

The argument about the sense in which winter"determines" spring seems another verbalisticone. We happen to know the astronomical caus-ality of our seasons. If we did not, we wouldstill know empirically the regular order of theswing of the seasons; and if any medicine-man,after spring had followed winter, asserted orwas credited with bringing on summer by magicsubservient to his personal will, a very limitednatural science experience would protect usagainst such superstition, even if it led us totalk a bit loosely about a "law of the seasons."In the field of cultural happenings, it is patentthat we do not know at all what the real causality

[151]


is. ''e also know very little about such ordera'md reRularity as there may be. I was tryingto make a contribution, at one concrete pointwhere the data looked promising, toward ascer-taining what order could be found in culturalphenomena--as did Chapin in his Cultural Change.The alternative to rejecting such endeavors isto hold that history is a series of accidents.This in turn, since most human minds will notremain content with utter negativism, usuallyresults in a reversion to the popular assump-tion that cultural events are produced by thefree volition of personalities. This assump-tion, again, in twentieth-century scholars,seems nearly as naive as the belief that di-rected magic can produce summer. .

If it is mere philosophy or mysticism to be-lieve that culture determines the actions ofpersonalities, at any rate determines them farmore than their uncontrolled volitions determineculture, then I am a mystic. However, the long,persistent swinLgs of style seem an empiricalfact which does directly support this belief.How do Allport and Hartman meet the very realand fundamental issue that there is a clear-cutfinding here which rests on evidence?

They fall back, tentatively, on a psycho-logical explanation: emulation, which leads eachindividual to try to outdo the rest until aphysical or other limit is reached, when theyall race back competitively. Interestinglyenough, this very explanation of emulation wasadduced nineteen hundred years ago by VelleiusPaterculus when he wrestled with the problem ofwhy most high cultural activities seem to comein cyclic bursts. Now, that a view is old doesnot prove it unsound; but if twentieth-centurysociologists can do no better than to speculatelike a soldier-historian contemporary of Augus-tus--well, they also have not traveled muchfarther than the mystics. Of course there ispsychology involved. Every cultural event in-volves psychological happenings. What my an-alysts fail to realize is that they do not, justas Paterculus did not, connect their psychologi-cal explanation with the cultural phenomena byany evidence. They;have made a guess, an un-verifiable conjecture. This is fine for after-dinner conversation before the fire, where toomuch evidence destroys social affability. Imay be mystic or telic, but I have at leasttried to relate observed phenomena with observedphenomena. I may add--as my conviction which Icannot prove--that my guess is the same: I alsobelieve that emulation or imitation is involvedin style changes. But I have deliberately leftit out of my interpretation. Paterculus saw avery real and still unsolved problem in a larreand important group of phenoniiena. But if wecannot attack the problem any better than by hisnonevidential method, I for one will call itquits and play another game.

The other tentative explanation of the ana-

lysts is through such factors as changes in theeconomic and social position of women, sports,mode of dancing, cost of materials, and such.These are cultural factors, such as I believeshould be correlated with stylistic changes be-fore the last ditch of psychological explana-tion is retreated to. However, they are spe-cific, immediate, and temporary factors whoseprimary effectiveness seems to be largely pre-cluded by the long range of the swings. Againthere is a suspicion of preference for the un-ordered interpretation: style may not be capri-cious, but it is accidental; the concept ofsuperindividual pattern is avoided and resisted.

Now why this resistance? Why all this quar-reling with my language, the imputation of mys-ticism to a quantitative investigation such aseconomists conduct by the thousand, the longand partly irrelevant wind-up about scientificand telic approaches, the near-pedantic pickingon the statistical reliability of my data inthe last paragraph?30

I suspect that the resistance goes back tothe common and deeply implanted assumption thatour wills are free. As this assumption has hadto yield ground elsewhere, it has taken refugein the collective, social, and historical sphere.Since the chemists, physiologists, and psycholo-gists have unlimbered their artillery, the per-sonal freedom of the will is thankless terrainto maintain. Culture they have not yet attacked;so that becomes a refuge. Whatever the degreeto which we have ceased to assert being freeagents as individuals, in the social realm wecan still claim to shape our destinies. Thetheologian is piping pretty small, but the so-cial reformer very loud. We are renouncing thekingdom of heaven, but going to establish anear-millennium on earth. Our personal willsmay be determined, but by collectivizing themwe can still have social freedom.

Of course this is not conscious motivation.But I see no other motivation for the resistanceto yielding an inch to any form or degree ofcultural determinism.

The amount of determinism shown is reallyvery little, in the original essay: no raorethan that there are stylistic trends of an am-plitude, effectiveness, and duration indicatingthat they are governed by factors which are un-

30The tenor of the original article makes itplain that there is no insistence on the reli-ability of any one figure or group of figures,but only on the duration of swings and persis-tence of trends. The absence of sigmas andprobable errors is cited as "a vital defect."As a matter of fact the changes found are sogreat that if we had data on only a single dressfor each year from 1844 to 1919, instead of fourto ten, the fact of notable swing in severaldimensions would be convincing to common sense;and, I suppose, provable as statistically sig-nificant if one were minded to show his tech-nical virtuosity.

152


known but which must be superindividual: a ran-dom series of free wills could not pull togetherin one direction so long and decisively. Apartfrom its presentation and analysis of the data,the 1919 paper is essentially a statement ofthis thesis--no doubt reiteratively and at timeswith some looseness; but it is difficult to beconcise and exact when one is compelled to swimagainst the intellectual current.

In the present monograph I believe that Dr.Richardson and I have gone a little farther. Wehave strengthened the case for authenticity ofthe swings by considerably enlarging the range ofobservation. We have also examined the vari-ability of the phenomena and found this to cor-relate definitely both with certain periods ofhistory and with certain statuses of the swings.From this double correlation we have inferred asprobable, for the last two centuries of Europe,a relation between periods of general socio-political and cultural unsettlement and periodsof extremity of fashion, through the medium ofan expressible and approximately definable pat-tern of dress style. We have not fallen backon emulation as an explanation of specificchanges; not because we deny its influence, butbecause we do not know any specific way of cor-relating emulation with particular historical

phenomena. We also do not deny that sports orthe invention of rayon may have an effect ondress styles. We have left them out of thereckoning, for the time being, because theirinfluence is presumably special and limited,and therefore secondary to the major swings ofdimension and variability. We are aware that"general historic unsettlement" involves psy-chological attitudes as much as emulation in-volves them; but they are attitudes which areat least partially measurable and definable interms of wars, changes of boundary or form ofgovernment, abolition of old institutions, newcodes, intensity of class struggles, aestheticinnovations, and the like; and historians areagreed in recognizing these phenomena as havingoccurred. The stone we have tried to lay onthe wall of the understanding of the history ofhuman civilization may be quite small; but wefeel that it is at least tangible and weighableas evidence.

I realize that any interpretation which di-minishes the range of free personality and en-larges the effectiveness of superpersonal cul-tural influences is likely to be unpalatable. Itwill irritate many and it will elicit rejections.But I am compelled to adhere to it--no doubt bythe strand of culture of which I am part.

15pu3

three centuries of women´s dress

Documents

california press berkeley

actual year means

full wave length

year moving average

anthropological records

wave lengths

marie antoinette

hollow circles