Two major
considerationsaffectthe value of the 70 adultRorschach
protocols, insofaras they aretobe used asdata for the elucidation of
group
personality characteristics: sampling procedures;and admin-
istration
and
scoring procedures.According
to thewriter's census, in July 1948 the totalnumber
of adult (16 yearsand
older) sociological Tuscarora^^was
352,ofwhom
179
were
malesand
173 were females.The
figure 352 the writer considersto be accurate towithin about 20 persons.The
populationwas
further classified according to age. Since ageswere
inmost
casesascribed
by
one informant(Mina
Smith),there certainlywould
be a large probability ofan error ofafew
yearsmore
orlessinregard toany
one individual. If these errors wererandomly
distributed, however,no
consistent biaswould
affect the proportionsby
ageand
sex inany
one age-sexgroup
; if the errors were consistently plus or minus, the errorswould
be approximately evenly distributed in the age-sex categories, again not materially affecting the distribution.The
only serious likelihood of errorwould
be inmaking
consistently largemistakes with regardtoone orseveral of the age-sex categories:
forinstance,considerablyunderestimatingtheages offemalesbetween 40
and
50, while considerably overestimating the ages of females be-tween
20and
30,and
thereby (perhaps)adding
noticeably to thenum-
ber of females in the 30 to 40 age-group.
The numbers
involved in the total adult population are so small, however,and
the absolutenumbers
in the sample itself somuch
smaller still (so small that a difference of 20 percentfrom
the present count of females between 30and
40, forinstance,would mean
the addition or rejectionof only^*"Sociological Tuscarora" are Indians who spent the first 6 or 7 years of their lives
onthe reservation, orwho were born and/orraisedoff the reserve by aTuscarora mother andlaterreturned to spenda major portion of theiradult liveson the reservation. Only thoseIndianslivingonornearthe reservewereavailable forstudy.
Wallacb]
modal personality
OFTU&CARORA
INDIANS41
one Rorschach protocol) that the possibility of such bias does not seemtothreaten thevalidityofthesampling procedureused.The
distribution of the populationamong
the age-sex categories having been established, the writer then calculated thenumber
of records necessaryto preserve thesame
proportions in the sample ashad
been observed in the population at large. These figureswere
calculatedon
the assumptionthat the samplewould
total 100 recordsand
that itwould
include proportionatenumbers
in all age grades,from
infancy to old age.As
thework
proceeded, however, itwas
decidedtoconcentrate
on
collectingaslightlylargernumber
of adults thanwas
needed in the original sampling scheme,and
the full pro- portion of preschool recordswas
not filled.A
total of 103 records weresecured. Seventy-oneof thesecame from
adults (age 16and
up).
One
ofthe adult recordswas
recordedonwire,withasomewhat
differ- ent technique of administrationthan thatwhich was
used throughout the series; this recordwas
excludedon
the grounds that theadmin-
istrativeprocedure
would make
itnotstrictlycomparabletotheothers.Thus
thesampleof adult records available forstudyis70. This sam- ple does not strictly follow theform
ofthe population distribution, but the deviations are not large enough,inthe writer'sview,seriously to bias the results. Table 1 (p. 42) presents the essential data.The
adult records thus represent a proportional sample inwhich
the frequencies ofcases inthespecifiedsex-age categories are carefully controlled soas tocorrespond,in theirmutual
proportions,tothe pro- portions of those categories in the whole population.The
selection of individuals for inclusion in the samplewas
at first left almost entirely to one of the writer's informants, a personwho had
(along with the rest of his family) taken a Rorschach,and was
willing to introduce the ethnographerand
his ink-blots to possible candidates.He would
betoldthattherecordswere neededtofillin certainsex-age categories.He would
take theproblem
under advisement, suggest certain persons of the propercategory he thoughtwould
"look at the cards,"go
withthe writerto theirhomes,introducehim
ifhe was
notknown
to them,and
sometimes personally persuade subjects to co- operate,explaining thathehad
takenthe testand
giving otherneces- sary particulars. This techniquewas
phenomenally successful; dur- ing thetwo summers
ofwork
onlytwo
or three persons approached in thismanner
refused totake Rorschachs.Most
of the recordswere
secured in this fashion.As
the writer got toknow more
people, however, hewas
able inmany
cases to suggest the Rorschach and,when
hewas
not refused, to give the teston
hisown
presentation.During
the secondsummer,
infact,several peoplesolicitedhim
toletthem
look atthe cardsand
itbecame
almost aminor
fad, in certain circles, tohave "ProfessorWallace" dropby
withhis ink-blots.906418—51 4
42 BUREAU
OFAMERICAN ETHNOLOGY
[Bull. 150Table 1.
—
The sampling of the Tuscarora population byage and sexAgegroup
Wallace]
MODAL PERSONALITY
OFTUSCARORA
INDIANS43
cent). This difference
was
found not to be significant at the 0.01 probability levelwhen
testedby
the large-sample critical ratio method. This suggests that those subjectswho
were selectedby
the informant constituted a psychological subpopulationwhich was
in- distinguishablefrom
the total sample,and
hence that his selection ofcases did notbias the formulation of themodal
personality struc- ture.On
the other hand, however, this technique of evaluatingbias does not fully take account of the possibility that the whole samplewas
biased, becausemost
(but notby any means
all) of the writer's contactswere
in that area of the reservation nearest to the inform- ant'shouse.From
his personalknowledge
of the homogeneityof the reservation, the writer does not believe thatany
significant biaswas
introducedby
thismeans
either, but he cannot demonstrate its im- probabilitystatistically.A
cross-validation study, inwhich
the sub- jects were chosenrandomly by
mechanicalmeans and
were propor- tionately distributed geographically aswell asby
ageand
sex,would
bedesirable.The
sizeof thesampleitselfisamatter ofsome moment. A
largersample would, ofcourse,have beenbetterthan this; this
was
the best procurable for the timeand
money.A
sample of 100 records is de-manded
instandard Eorschachpractice before a statement ofwhat
the popularand
original responsesare.The
writerwillhavetogoahead to definewhat
heregards as the popular responses on the basisof 70 records.A
sampleof70isconsidered "large" forstatisticalpurposesand
does not require the use of small-sample theory.From
such a small population, of352,asample of70 isreally afairlygood
size: itincludes 20 percent of the total population,