purposeful sampling
DESCRIPTION
Designing Qualitative Studies, Patton, M. (1990).TRANSCRIPT
Patto
n, M
. (19
90).
Qua
litat
ive
eval
uatio
n an
d re
sear
ch m
etho
ds
(pp.
169
-186
). B
ever
ly H
ills,
CA
: Sag
e.
Des
ignin
g Q
ualit
ativ
e Stu
dies
16
9 PU
RPO
SEFU
L SA
MPL
ING
Pe
rhap
s no
thin
g be
tter
capt
ures
the
diff
eren
ce b
etw
een
quan
titat
ive
and
qual
itativ
e m
etho
ds
than
th
e di
ffer
ent
logi
cs
that
un
derg
ird
sam
plin
g ap
proa
ches
. Qua
litat
ive
inqu
iry t
ypic
ally
foc
uses
in d
epth
on
rela
tivel
y sm
all s
ampl
es, e
ven
sing
le c
ases
(n =
1),
sele
cted
pur
pose
fully
. Q
uant
itativ
e m
etho
ds t
ypic
ally
dep
end
on l
arge
r sa
mpl
es s
elec
ted
rand
omly
. Not
onl
y ar
e th
e te
chni
ques
for
sam
plin
g di
ffer
ent,
but
the
very
log
ic o
f ea
ch a
ppro
ach
is u
niqu
e be
caus
e th
e pu
rpos
e of
eac
h st
rate
gy is
diff
eren
t. Th
e lo
gic
and
pow
er o
f pr
obab
ility
sam
plin
g de
pend
s on
sel
ectin
g a
trul
y ra
ndom
and
sta
tistic
ally
rep
rese
ntat
ive
sam
ple
that
will
per
mit
conf
iden
t ge
nera
lizat
ion
from
the
sam
ple
to a
lar
ger
popu
latio
n. T
he
purp
ose
is g
ener
aliz
atio
n.
The
logi
c an
d po
wer
of
purp
osef
ul s
ampl
ing
lies
in s
elec
ting
in
form
atio
n-ric
h ca
ses f
or s
tudy
in d
epth
. Inf
orm
atio
n-ric
h ca
ses
are
thos
e fr
om w
hich
one
can
lear
n a
grea
t de
al a
bout
issu
es o
f ce
ntra
l im
por-
tanc
e to
the
purp
ose
of th
e re
sear
ch, t
hus
the
term
pur
posef
ul s
ampl
ing.
Fo
r ex
ampl
e, if
the
pur
pose
of
an e
valu
atio
n is
to
incr
ease
the
eff
ec-
tiven
ess
of a
pro
gram
in
reac
hing
low
er-s
ocio
econ
omic
gro
ups,
one
may
lear
n a
grea
t dea
l mor
e by
foc
usin
g in
dep
th o
n un
ders
tand
ing
the
need
s, in
tere
sts,
and
ince
ntiv
es o
f a
smal
l num
ber
of c
aref
ully
sel
ecte
d po
or f
amili
es th
an b
y ga
ther
ing
stan
dard
ized
info
rmat
ion
from
a la
rge,
st
atis
tical
ly r
epre
sent
ativ
e sa
mpl
e of
the
who
le p
rogr
am. T
he p
urpo
se
of p
urpo
sefu
l sam
plin
g is
to s
elec
t inf
orm
atio
n-ric
h ca
ses
who
se s
tudy
w
ill il
lum
inat
e th
e qu
estio
ns u
nder
stu
dy.
Ther
e ar
e se
vera
l di
ffer
ent
stra
tegi
es f
or p
urpo
sefu
lly s
elec
ting
info
rmat
ion-
rich
case
s. Th
e lo
gic
of e
ach
stra
tegy
ser
ves
a pa
rtic
ular
ev
alua
tion
purp
ose.
(1
) E
xtre
me
or d
evia
nt c
ase
sam
plin
g. Th
is a
ppro
ach
focu
ses
on c
ases
th
at a
re r
ich
in i
nfor
mat
ion
beca
use
they
are
unu
sual
or
spec
ial
in
som
e w
ay. U
nusu
al o
r sp
ecia
l cas
es m
ay b
e pa
rtic
ular
ly tr
oubl
esom
e or
es
peci
ally
en
light
enin
g,
such
as
ou
tsta
ndin
g su
cces
ses
or
nota
ble
failu
res.
If,
for
exam
ple,
the
eva
luat
ion
was
aim
ed a
t ga
ther
ing
data
he
lp a
nat
iona
l pro
gram
rea
ch m
ore
clie
nts,
one
mig
ht c
ompa
re a
few
pr
ojec
t si
tes
that
hav
e lo
ng w
aitin
g lis
ts w
ith t
hose
tha
t ha
ve s
hort
w
aitin
g lis
ts.
If s
taff
mor
ale
was
an
issu
e, o
ne m
ight
stu
dy a
nd
com
pare
hig
h-m
oral
e pr
ogra
ms
to lo
w-m
oral
e pr
ogra
ms.
170
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTI
ON
The
logi
c of
ext
rem
e ca
se s
ampl
ing
is th
at l
esso
ns m
ay b
e le
arne
d ab
out
unus
ual
cond
ition
s or
ext
rem
e ou
tcom
es t
hat
are
rele
vant
to
impr
ovin
g m
ore
typi
cal p
rogr
ams.
Let's
sup
pose
tha
t w
e ar
e in
tere
sted
in
stu
dyin
g a
natio
nal
prog
ram
with
hun
dred
s of
loca
l site
s. W
e kn
ow
that
man
y pr
ogra
ms
are
oper
atin
g re
ason
ably
wel
l, ev
en q
uite
wel
l, an
d th
at o
ther
pro
gram
s ve
rge
on b
eing
disa
ster
s. W
e al
so k
now
tha
t m
ost
prog
ram
s ar
e do
ing
"oka
y."
This
info
rmat
ion
com
es
from
kn
owle
dgea
ble
sour
ces
who
hav
e m
ade
site
visit
s to
eno
ugh
prog
ram
s to
ha
ve a
bas
ic id
ea a
bout
wha
t the
var
iatio
n is.
The
que
stio
n is
this:
How
sh
ould
pro
gram
s be
sam
pled
for
the
stu
dy?
If o
ne w
ante
d to
pre
cise
ly
docu
men
t th
e na
tura
l va
riatio
n am
ong
prog
ram
s, a
rand
om s
ampl
e w
ould
be
appr
opria
te, p
refe
rabl
y a
rand
om s
ampl
e of
suf
ficie
nt s
ize
to
be
trul
y re
pres
enta
tive
of
and
perm
it ge
nera
lizat
ions
to
th
e to
tal
popu
latio
n of
pro
gram
s. H
owev
er, s
ome
info
rmat
ion
is al
read
y av
aila
ble
on w
hat
prog
ram
var
iatio
n is
like.
The
que
stio
n of
mor
e im
med
iate
in
tere
st m
ay c
once
rn e
xtrem
e ca
ses.
With
lim
ited
reso
urce
s an
d lim
ited
time
an e
valu
ator
mig
ht le
arn
mor
e by
inte
nsiv
ely
stud
ying
one
or
mor
e ex
ampl
es o
f re
ally
poo
r pr
ogra
ms
and
one
or m
ore
exam
ples
of
real
ly
exce
llent
pro
gram
s. Th
e ev
alua
tion
focu
s, th
en, b
ecom
es a
que
stio
n of
un
ders
tand
ing
unde
r w
hat
cond
ition
s pr
ogra
ms
get
into
tro
uble
and
un
der
wha
t co
nditi
ons
prog
ram
s ex
empl
ify e
xcel
lenc
e. I
t is
not
even
ne
cess
ary
to r
ando
mly
sam
ple
poor
pro
gram
s or
exc
elle
nt p
rogr
ams.
The
rese
arch
ers
and
inte
nded
use
rs in
volv
ed in
the
stud
y th
ink
thro
ugh
what
cas
es th
ey c
ould
lea
rn t
he m
ost
from
and
tho
se a
re t
he c
ases
tha
t ar
e se
lect
ed fo
r stu
dy.
In a
sin
gle
prog
ram
the
sam
e st
rate
gy m
ay a
pply
. Ins
tead
of s
tudy
ing
som
e re
pres
enta
tive
sam
ple
of p
eopl
e in
the
set
ting,
the
eva
luat
or m
ay
focu
s on
stu
dyin
g an
d un
ders
tand
ing
sele
cted
cas
es o
f sp
ecia
l int
eres
t, fo
r ex
ampl
e, u
nexp
ecte
d dr
opou
ts o
r ou
tsta
ndin
g su
cces
ses.
In m
any
inst
ance
s m
ore
can
be l
earn
ed f
rom
int
ensiv
ely
stud
ying
ext
rem
e or
un
usua
l cas
es th
an c
an b
e le
arne
d fr
om s
tatis
tical
dep
ictio
ns o
f wha
t the
av
erag
e ca
se i
s lik
e. I
n ot
her
eval
uatio
ns d
etai
led
info
rmat
ion
abou
t sp
ecia
l cas
es c
an b
e us
ed to
sup
plem
ent s
tatis
tical
dat
a ab
out t
he n
orm
al
dist
ribut
ion
of p
artic
ipan
ts.
Eth
nom
etho
dolo
gist
s us
e a
form
of
extr
eme
case
sam
plin
g w
hen
they
do
thei
r fie
ld e
xper
imen
ts. E
thno
met
hodo
logi
sts
are
inte
rest
ed i
n ev
eryd
ay
expe
rienc
es
of
rout
ine
livin
g th
at
depe
nd
on
deep
ly
unde
rsto
od,
shar
ed u
nder
stan
ding
s am
ong
peop
le i
n a
setti
ng (
see
Cha
pter
3).
One
way
of
expo
sing
thes
e im
plic
it as
sum
ptio
ns a
nd n
orm
s on
whi
ch e
very
day
life
is ba
sed
is to
cre
ate
dist
urba
nces
that
Desi
gnin
g Qua
litat
ive S
tudi
es 17
1 de
viat
e fr
om th
e no
rm. O
bser
ving
the
reac
tions
to s
omeo
ne e
atin
g lik
e a
pig
in a
res
taur
ant
and
then
int
ervi
ewin
g pe
ople
abo
ut w
hat
they
saw
an
d ho
w th
ey f
elt w
ould
be
an e
xam
ple
of s
tudy
ing
a de
vian
t sam
ple
to
illum
inat
e th
e or
dina
ry.
The
Pete
rs a
nd W
ater
man
(19
82)
best
-sel
ling
stud
y of
"A
mer
ica's
be
st r
un c
ompa
nies
," I
n Se
arch
of
Exc
ellen
ce, e
xem
plifi
es t
he l
ogic
of
purp
osef
ul, e
xtre
me
grou
p sa
mpl
ing.
The
ir st
udy
was
bas
ed o
n a
sam
ple
of 6
2 co
mpa
nies
"ne
ver
inte
nded
to
be p
erfe
ctly
rep
rese
ntat
ive
of U
.S.
indu
stry
as
a w
hole
...
[but
] a
list
of c
ompa
nies
con
sider
ed t
o be
in
nova
tive
and
exce
llent
by
an i
nfor
med
gro
up o
f ob
serv
ers
of t
he
busin
ess
scen
e" (P
eter
s an
d W
ater
man
, 198
2: 1
9).
Ano
ther
exc
elle
nt e
xam
ple
of e
xtre
me
grou
p sa
mpl
ing
is A
ngel
a Br
owne
's (1
987)
stu
dy, W
hen
Batte
red W
omen
Kill
. She
con
duct
ed in
-dep
th
stud
ies
of t
he m
ost e
xtre
me
case
s of
dom
estic
vio
lenc
e to
elu
cida
te t
he
phen
omen
on o
f ba
tterin
g an
d ab
use.
The
ext
rem
e na
ture
of
the
case
s pr
esen
ted
are
wha
t re
nder
the
m s
o po
wer
ful.
Brow
ne's
book
is
an
exem
plar
of
qual
itativ
e in
quiry
usin
g pu
rpos
eful
sam
plin
g fo
r ap
plie
d re
sear
ch.
(2)
Inten
sity
sam
plin
g. In
tens
ity s
ampl
ing
invo
lves
the
sam
e lo
gic
as
extr
eme
case
sam
plin
g bu
t w
ith l
ess
emph
asis
on t
he e
xtre
mes
. A
n in
tens
ity s
ampl
e co
nsist
s of
inf
orm
atio
n-ric
h ca
ses
that
man
ifest
the
ph
enom
enon
of
inte
rest
int
ense
ly (
but
not
extr
emel
y).
Ext
rem
e or
de
vian
t ca
ses
may
be
so u
nusu
al a
s to
dist
ort
the
man
ifest
atio
n of
the
ph
enom
enon
of i
nter
est.
Usin
g th
e lo
gic
of in
tens
ity s
ampl
ing,
one
see
ks
exce
llent
or
rich
exam
ples
of
the
phen
omen
on o
f in
tere
st,
but
not
unus
ual c
ases
. H
euris
tic r
esea
rch
uses
inte
nsity
sam
plin
g. H
euris
tic r
esea
rch
draw
s ex
plic
itly
on t
he i
nten
se p
erso
nal
expe
rienc
es o
f th
e re
sear
cher
, fo
r ex
ampl
e, e
xper
ienc
es w
ith l
onel
ines
s or
jea
lous
y C
ores
earc
hers
who
ha
ve e
xper
ienc
ed t
hese
phe
nom
ena
inte
nsel
y al
so p
artic
ipat
e in
the
st
udy
(see
Cha
pter
3).
The
heur
istic
res
earc
her
is no
t ty
pica
lly s
eeki
ng
path
olog
ical
or
ex
trem
e m
anife
stat
ions
of
lo
nelin
ess,
jeal
ousy
, or
w
hate
ver
phen
omen
on is
of i
nter
est.
Such
ext
rem
e ca
ses
mig
ht n
ot le
nd
them
selv
es t
o th
e re
flect
ive
proc
ess
of h
euris
tic i
nqui
ry. O
n th
e ot
her
hand
, if
the
expe
rienc
e of
the
heu
ristic
res
earc
her
and
his
or h
er
core
sear
cher
s is
quite
mild
, th
ere
won
't be
muc
h to
stu
dy.
Thus
the
re
sear
cher
see
ks a
sam
ple
of s
uffic
ient
int
ensit
y to
elu
cida
te t
he p
he-
nom
enon
of i
nter
est.
The
sam
e lo
gic
appl
ies
in a
pro
gram
eva
luat
ion.
Ext
rem
e su
cces
ses
or u
nusu
al f
ailu
res
may
be
disc
redi
ted
as b
eing
too
extr
eme
or u
n-
172
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTI
ON
us
ual f
or g
aini
ng in
form
atio
n. T
here
fore
, the
eva
luat
or m
ay s
elec
t cas
es
that
man
ifest
suf
ficie
nt in
tens
ity t
o ill
umin
ate
the
natu
re o
f su
cces
s or
fa
ilure
, but
not
at t
he e
xtre
me.
In
tens
ity s
ampl
ing
invo
lves
som
e pr
ior
info
rmat
ion
and
cons
ider
able
ju
dgm
ent.
The
rese
arch
er m
ust d
o so
me
expl
orat
ory
wor
k to
det
erm
ine
the
natu
re o
f th
e va
riatio
n in
the
situ
atio
n un
der
stud
y O
ne c
an t
hen
sam
ple
inte
nse
exam
ples
of t
he p
heno
men
on o
f int
eres
t. (3
) M
axim
um V
aria
tion
sam
plin
g. T
his
stra
tegy
for
pur
pose
ful s
am-
plin
g ai
ms
at c
aptu
ring
and
desc
ribin
g th
e ce
ntra
l th
emes
or
prin
cipa
l ou
tcom
es
that
cu
t ac
ross
a
grea
t de
al
of
part
icip
ant
or
prog
ram
va
riatio
n. F
or s
mal
l sa
mpl
es a
gre
at d
eal
of h
eter
ogen
eity
can
be
a pr
oble
m b
ecau
se in
divi
dual
cas
es a
re s
o di
ffer
ent
from
eac
h ot
her.
The
max
imum
var
iatio
n sa
mpl
ing
stra
tegy
turn
s th
at a
ppar
ent w
eakn
ess
into
a
stre
ngth
by
appl
ying
the
fol
low
ing
logi
c: A
ny c
omm
on p
atte
rns
that
em
erge
fro
m g
reat
var
iatio
n ar
e of
par
ticul
ar i
nter
est
and
valu
e in
ca
ptur
ing
the
core
exp
erie
nces
and
cen
tral
, sha
red
aspe
cts
or im
pact
s of
a
prog
ram
. H
ow d
oes
one
max
imiz
e va
riatio
n in
a s
mal
l sam
ple?
One
beg
ins
by
iden
tifyi
ng d
iver
se c
hara
cter
istic
s or
crit
eria
for
cons
truc
ting
the
sam
ple.
Su
ppos
e a
stat
ewid
e pr
ogra
m h
as p
roje
ct s
ites
spre
ad a
roun
d th
e st
ate,
so
me
in r
ural
are
as, s
ome
in u
rban
are
as, a
nd s
ome
in s
ubur
ban
area
s. Th
e ev
alua
tion
lack
s su
ffic
ient
res
ourc
es t
o ra
ndom
ly s
elec
t en
ough
pr
ojec
t site
s to
gen
eral
ize
acro
ss th
e st
ate.
The
eva
luat
or c
an a
t lea
st b
e su
re t
hat
the
geog
raph
ical
var
iatio
n am
ong
sites
is
repr
esen
ted
in t
he
stud
y.
Whe
n se
lect
ing
a sm
all s
ampl
e of
gre
at d
iver
sity,
the
dat
a co
llect
ion
and
anal
ysis
will
yie
ld t
wo
kind
s of
fin
ding
s: (1
) hi
gh-q
ualit
y, d
etai
led
desc
riptio
ns o
f eac
h ca
se, w
hich
are
use
ful f
or d
ocum
entin
g un
ique
ness
, an
d (Z
) im
port
ant s
hare
d pa
ttern
s th
at c
ut a
cros
s ca
ses
and
deriv
e th
eir
signi
fican
ce fr
om h
avin
g em
erge
d ou
t of h
eter
ogen
eity
. Th
e sa
me
stra
tegy
can
be
used
with
in a
sin
gle
prog
ram
in s
elec
ting
indi
vidu
als
for
stud
y.
By
incl
udin
g in
th
e sa
mpl
e in
divi
dual
s th
e ev
alua
tor
dete
rmin
es h
ave
had
quite
diff
eren
t ex
perie
nces
, it i
s po
ssib
le
to
mor
e th
orou
ghly
de
scrib
e th
e va
riatio
n in
th
e gr
oup
and
to
unde
rsta
nd
varia
tions
in
ex
perie
nces
w
hile
al
so
inve
stig
atin
g co
re
elem
ents
an
d sh
ared
ou
tcom
es.
The
eval
uato
r us
ing
a m
axim
um
varia
tion
sam
plin
g st
rate
gy
wou
ld
not
be
atte
mpt
ing
to
gene
raliz
e fin
ding
s to
all
peop
le o
r all
grou
ps b
ut w
ould
be
look
ing
for i
nfor
mat
ion
that
elu
cida
tes
prog
ram
mat
ic v
aria
tion
and
signi
fican
t com
mon
pat
tern
s w
ithin
that
var
iatio
n.
Desi
gnin
g Qua
litat
ive S
tudi
es 17
3 (4
) Hom
ogen
eous
sam
ples
. In
dire
ct c
ontr
ast t
o m
axim
um v
aria
tion
sam
plin
g is
the
stra
tegy
of
pick
ing
a sm
all
hom
ogen
eous
sam
ple.
The
pu
rpos
e he
re i
s to
des
crib
e so
me
part
icul
ar s
ubgr
oup
in d
epth
. A
pr
ogra
m th
at h
as m
any
diff
eren
t kin
ds o
f par
ticip
ants
may
nee
d in
-dep
th
info
rmat
ion
abou
t a
part
icul
ar
subg
roup
. Fo
r ex
ampl
e,
a pa
rent
ed
ucat
ion
prog
ram
tha
t in
volv
es m
any
diff
eren
t ki
nds
of p
aren
ts m
ay
focu
s a
qual
itativ
e ev
alua
tion
on th
e ex
perie
nces
of
singl
e-pa
rent
fem
ale
head
s of
hou
seho
ld b
ecau
se th
at is
a p
artic
ular
ly d
iffic
ult g
roup
to re
ach
and
hold
in th
e pr
ogra
m.
Focu
s gr
oup
inte
rvie
ws
are
typi
cally
bas
ed o
n ho
mog
eneo
us g
roup
s. Fo
cus
grou
p in
terv
iew
s in
volv
e co
nduc
ting
open
-end
ed in
terv
iew
s w
ith
grou
ps o
f fiv
e to
eig
ht p
eopl
e on
spe
cial
ly t
arge
ted
or f
ocus
ed i
ssue
s. Th
e us
e of
focu
s gr
oups
in e
valu
atio
n w
ill b
e di
scus
sed
at g
reat
er le
ngth
in
the
chap
ter o
n in
terv
iew
ing.
The
poi
nt h
ere
is th
at s
ampl
ing
for f
ocus
gr
oups
ty
pica
lly
invo
lves
br
ingi
ng
toge
ther
pe
ople
of
sim
ilar
back
grou
nds
and
expe
rienc
es t
o pa
rtic
ipat
e in
a g
roup
inte
rvie
w a
bout
m
ajor
pro
gram
issu
es th
at a
ffec
t the
m.
(5) T
ypic
al c
ase
sam
plin
g. I
n de
scrib
ing
a pr
ogra
m o
r its
par
ticip
ants
to
peo
ple
not
fam
iliar
with
the
pro
gram
it c
an b
e he
lpfu
l to
prov
ide
a qu
alita
tive
prof
ile o
f on
e or
mor
e "t
ypic
al"
case
s. Th
ese
case
s ar
e se
lect
ed w
ith t
he c
oope
ratio
n of
key
info
rman
ts, s
uch
as p
rogr
am s
taff
or
kno
wle
dgea
ble
part
icip
ants
, who
can
hel
p id
entif
y w
hat i
s ty
pica
l. It
is
also
pos
sible
to
sele
ct t
ypic
al c
ases
fro
m s
urve
y da
ta,
a de
mog
raph
ic
anal
ysis
of a
vera
ges,
or o
ther
pro
gram
mat
ic d
ata
that
pro
vide
a n
orm
al
dist
ribut
ion
of
char
acte
ristic
s fr
om
whi
ch
to
iden
tify
"ave
rage
" ex
ampl
es. K
eep
in m
ind
that
the
purp
ose
of a
qua
litat
ive
prof
ile o
f on
e or
mor
e ty
pica
l cas
es is
to d
escr
ibe
and
illus
trat
e w
hat i
s ty
pica
l to
thos
e un
fam
iliar
with
the
prog
ram
—no
t to
mak
e ge
nera
lized
sta
tem
ents
abo
ut
the
expe
rienc
es
of
all
part
icip
ants
. Th
e sa
mpl
e is
illus
trat
ive
not
defin
itive
. W
hen
entir
e pr
ogra
ms
or c
omm
uniti
es a
re t
he u
nit o
f an
alys
is, i
t is
al
so p
ossi
ble
to s
ampl
e so
mew
hat
typi
cal
case
s. A
gain
, th
e st
udy
of
such
typ
ical
pro
gram
s do
es n
ot,
of c
ours
e, p
erm
it ge
nera
lizat
ions
in
any
rigor
ous
sens
e. I
t do
es,
how
ever
, m
ean
that
the
pro
cess
es a
nd
effe
cts
desc
ribed
for
the
typ
ical
pro
gram
nee
d no
t be
dis
mis
sed
as
pecu
liar
to "
poor
" si
tes
or "
exce
llent
" si
tes.
Whe
n th
e ty
pica
l si
te
sam
plin
g st
rate
gy is
use
d, t
he s
ite is
spe
cific
ally
sel
ecte
d be
caus
e it
is
not
in a
ny m
ajor
way
aty
pica
l, ex
trem
e, d
evia
nt, o
r in
tens
ely
unus
ual.
This
str
ateg
y is
ofte
n ap
prop
riate
in
sam
plin
g vi
llage
s fo
r co
mm
unity
de
velo
pmen
t stu
dies
in T
hird
Wor
ld c
ount
ries.
A s
tudy
of
a ty
pica
l
174
Q
UA
LITA
TIV
E D
ESI
GN
S A
ND
DA
TA C
OLL
EC
TIO
N
villa
ge il
lum
inat
es k
ey is
sues
that
mus
t be
cons
ider
ed in
any
dev
elop
-m
ent
proj
ect a
imed
at t
his
kind
of v
illag
e.
Dec
ision
mak
ers
may
hav
e m
ade
thei
r pe
ace
with
the
fac
t tha
t th
ere
will
alw
ays
be s
ome
poor
pro
gram
s an
d so
me
exce
llent
pro
gram
s, bu
t th
e pr
ogra
ms
they
rea
lly w
ant m
ore
info
rmat
ion
abou
t are
thos
e ru
n-of
-th
e-m
ill p
rogr
ams
that
are
"ha
rd t
o ge
t a
hand
le o
n."
It i
s im
port
ant,
whe
n us
ing
this
stra
tegy
, to
atte
mpt
to g
et b
road
con
sens
us a
bout
whi
ch
prog
ram
s ar
e "t
ypic
al."
If
a nu
mbe
r of
suc
h pr
ogra
ms
are
iden
tifie
d,
only
a f
ew c
an b
e st
udie
d, a
nd t
here
is
no o
ther
bas
is fo
r se
lect
ing
amon
g th
em p
urpo
sefu
lly,
then
it
is po
ssib
le t
o ra
ndom
ly s
elec
t fr
om
amon
g al
l "ty
pica
l" p
rogr
ams
iden
tifie
d to
sel
ect t
hose
few
typi
cal c
ases
th
at a
ctua
lly w
ill b
e in
clud
ed in
the
stud
y.
(6) S
tratif
ied p
urpo
sefu
l sam
plin
g. It
is a
lso c
lear
ly p
ossib
le to
com
bine
a
typi
cal c
ase
sam
plin
g st
rate
gy w
ith o
ther
s, es
sent
ially
taki
ng a
stra
tified
pu
rpos
eful
sam
ple
of a
bove
ave
rage
, av
erag
e, a
nd b
elow
ave
rage
cas
es.
This
is le
ss t
han
a fu
ll m
axim
um v
aria
tion
sam
ple.
The
pur
pose
of
a st
ratif
ied
purp
osef
ul s
ampl
e is
to c
aptu
re m
ajor
var
iatio
ns r
athe
r th
an to
id
entif
y a
com
mon
cor
e, a
lthou
gh t
he l
atte
r m
ay a
lso e
mer
ge i
n th
e an
alys
is. E
ach
of t
he s
trat
a w
ould
con
stitu
te a
fai
rly h
omog
eneo
us
sam
ple.
Thi
s st
rate
gy d
iffer
s fr
om s
trat
ified
ran
dom
sam
plin
g in
that
the
sam
ple
sizes
are
lik
ely
to b
e to
o sm
all
for
gene
raliz
atio
n or
sta
tistic
al
repr
esen
tativ
enes
s. (7
) C
ritica
l ca
se s
ampl
ing.
Ano
ther
str
ateg
y fo
r se
lect
ing
purp
osef
ul
sam
ples
is
to l
ook
for
criti
cal
case
s. C
ritic
al c
ases
are
tho
se t
hat
can
mak
e a
poin
t qu
ite d
ram
atic
ally
or
are,
for
som
e re
ason
, pa
rtic
ular
ly
impo
rtan
t in
the
sch
eme
of t
hing
s. A
clu
e to
the
exi
sten
ce o
f a
criti
cal
case
is a
sta
tem
ent t
o th
e ef
fect
that
"if
it ha
ppen
s th
ere,
it w
ill h
appe
n an
ywhe
re,"
or,
vice
ver
sa, "
if it
does
n't
happ
en t
here
, it
won
't ha
ppen
an
ywhe
re."
The
foc
us o
f th
e da
ta g
athe
ring
in t
his
inst
ance
is
on
unde
rsta
ndin
g w
hat
is ha
ppen
ing
in t
hat
criti
cal
case
. A
noth
er c
lue
to
the
exist
ence
of
a cr
itica
l ca
se i
s a
key
info
rman
t ob
serv
atio
n to
the
ef
fect
that
"if
that
gro
up is
hav
ing
prob
lem
s, th
en w
e ca
n be
sur
e al
l the
gr
oups
are
hav
ing
prob
lem
s."
Look
ing
for t
he c
ritic
al c
ase
is pa
rtic
ular
ly im
port
ant w
here
reso
urce
s m
ay li
mit
the
eval
uatio
n to
the
stu
dy o
f on
ly a
sin
gle
site.
Und
er s
uch
cond
ition
s it
mak
es s
trat
egic
sen
se t
o pi
ck t
he s
ite t
hat
wou
ld y
ield
the
m
ost
info
rmat
ion
and
have
the
gre
ates
t im
pact
on
the
deve
lopm
ent
of
know
ledg
e.
Whi
le
stud
ying
on
e or
a
few
cr
itica
l ca
ses
does
no
t te
chni
cally
per
mit
broa
d ge
nera
lizat
ions
to a
ll po
ssib
le c
ases
,
Des
igni
ng Q
ualit
ativ
e St
udie
s 17
5 lo
gical
gen
eral
izat
ions
can
oft
en b
e m
ade
from
the
wei
ght o
f ev
iden
ce
prod
uced
in s
tudy
ing
a si
ngle
, crit
ical
cas
e.
Phys
ics
prov
ides
a g
ood
exam
ple
of s
uch
a cr
itica
l ca
se.
In
Gal
ileo'
s st
udy
of g
ravi
ty h
e w
ante
d to
fin
d ou
t if
the
wei
ght
of a
n ob
ject
aff
ecte
d th
e ra
te o
f sp
eed
at w
hich
it
wou
ld f
all.
Rat
her
than
ra
ndom
ly s
ampl
ing
obje
cts
of d
iffer
ent w
eigh
ts in
ord
er to
gen
eral
ize
to a
ll ob
ject
s in
the
wor
ld, h
e se
lect
ed a
crit
ical
cas
e—th
e fe
athe
r. If
in
a v
acuu
m,
as h
e de
mon
stra
ted,
a f
eath
er f
ell
at t
he s
ame
rate
as
som
e he
avie
r ob
ject
(a
coin
), th
en h
e co
uld
logi
cally
gen
eral
ize
from
th
is o
ne c
ritic
al c
ase
to a
ll ob
ject
s. H
is f
indi
ngs
wer
e en
orm
ousl
y us
eful
and
cre
dibl
e.
The
re
are
man
y co
mpa
rabl
e cr
itica
l ca
ses
in
soci
al
scie
nce
rese
arch
—if
one
is
crea
tive
in
look
ing
for
them
. Fo
r ex
ampl
e,
supp
ose
natio
nal
polic
ymak
ers
wan
t to
ge
t lo
cal
com
mun
ities
in
volv
ed i
n m
akin
g de
cisi
ons
abou
t ho
w t
heir
loca
l pr
ogra
m w
ill b
e ru
n, b
ut t
hey
aren
't su
re t
hat
the
com
mun
ities
will
und
erst
and
the
com
plex
reg
ulat
ions
gov
erni
ng t
heir
invo
lvem
ent.
The
firs
t cr
itica
l ca
se i
s to
eva
luat
e th
e re
gula
tions
in
a co
mm
unity
of
wel
l-edu
cate
d ci
tizen
s; i
f th
ey c
an't
unde
rsta
nd t
he r
egul
atio
ns,
then
les
s-ed
ucat
ed
folk
s ar
e su
re
to
find
the
regu
latio
ns
inco
mpr
ehen
sibl
e.
Or
conv
erse
ly,
one
mig
ht c
onsi
der
the
criti
cal
case
to
be a
com
mun
ity
cons
istin
g of
peo
ple
with
qui
te lo
w le
vels
of
educ
atio
n: "
If t
hey
can
unde
rsta
nd th
e re
gula
tions
, any
one
can.
" Id
entif
icat
ion
of c
ritic
al c
ases
dep
ends
on
reco
gniti
on o
f th
e ke
y di
men
sion
s th
at m
ake
for
a cr
itica
l ca
se.
A c
ritic
al c
ase
mig
ht b
e in
dica
ted
by t
he f
inan
cial
sta
te o
f a
prog
ram
; a p
rogr
am w
ith p
artic
-ul
arly
hig
h or
par
ticul
arly
low
cos
t-pe
r-cl
ient
rat
ios
mig
ht s
ugge
st a
cr
itica
l ca
se.
A c
ritic
al c
ase
mig
ht c
ome
from
a p
artic
ular
ly d
iffic
ult
prog
ram
loca
tion.
If
the
fund
ers
of a
new
pro
gram
are
wor
ried
abou
t re
crui
ting
clie
nts
or p
artic
ipan
ts in
to a
pro
gram
, it m
ay m
ake
sens
e to
st
udy
the
site
whe
re r
esis
tanc
e to
the
pro
gram
is
expe
cted
to
be
grea
test
to
prov
ide
the
mos
t rig
orou
s te
st o
f th
e po
ssib
ility
of
prog
ram
rec
ruitm
ent.
If t
he p
rogr
am w
orks
in
that
site
, "I
t co
uld
wor
k an
ywhe
re."
W
orld
-ren
owne
d m
edic
al h
ypno
tist
Milt
on H
. Eric
kson
bec
ame
a cr
itica
l cas
e in
the
fie
ld o
f hy
pnos
is. E
ricks
on w
as s
o sk
illfu
l tha
t he
be
cam
e w
idel
y kn
own
for
"his
abi
lity
to s
ucce
ed w
ith 'i
mpo
ssib
les'—
peop
le w
ho h
ave
exha
uste
d th
e tr
aditi
onal
med
ical
, de
ntal
, ps
ycho
-th
erap
eutic
, hyp
notic
and
rel
igio
us a
venu
es f
or a
ssis
ting
them
in th
eir
176
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTI
ON
ne
ed,
and
have
not
bee
n ab
le t
o m
ake
the
chan
ges
they
des
ire"
(Grin
der
et a
l., 1
977:
109
). If
Milt
on E
ricks
on c
ould
n't
help
, no
one
co
uld
help
. He
was
abl
e to
dem
onst
rate
tha
t an
yone
cou
ld b
e hy
pno-
tized
. (S
) Sn
owba
ll or
cha
in s
ampl
ing.
This
is
an a
ppro
ach
for
loca
ting
info
rmat
ion-
rich
key
info
rman
ts o
r cr
itica
l ca
ses.
The
proc
ess
begi
ns
by a
skin
g w
ell-s
ituat
ed p
eopl
e: "
Who
kno
ws
a lo
t ab
out
___
_
?
Who
sho
uld
I ta
lk to
?" B
y as
king
a n
umbe
r of
peo
ple
who
els
e to
talk
w
ith,
the
snow
ball
gets
big
ger
and
bigg
er a
s yo
u ac
cum
ulat
e ne
w
info
rmat
ion-
rich
case
s. In
mos
t pro
gram
s or
sys
tem
s, a
few
key
nam
es
or in
cide
nts
are
men
tione
d re
peat
edly
. Tho
se p
eopl
e or
eve
nts
reco
m-
men
ded
as v
alua
ble
by a
num
ber
of d
iffer
ent
info
rman
ts t
ake
on
spec
ial
impo
rtan
ce.
The
chai
n of
rec
omm
ende
d in
form
ants
will
typ
i-ca
lly d
iver
ge in
itial
ly a
s m
any
poss
ible
sou
rces
are
rec
omm
ende
d, t
hen
conv
erge
as
a fe
w k
ey n
ames
get
men
tione
d ov
er a
nd o
ver.
The
Pete
rs a
nd W
ater
man
(19
82)
stud
y In
Sea
rch o
f Exc
ellen
ce be
gan
with
sno
wba
ll sa
mpl
ing,
ask
ing
a br
oad
grou
p of
kno
wle
dgea
ble
peop
le
to i
dent
ify w
ell-r
un c
ompa
nies
. A
noth
er e
xcel
lent
and
wel
l-kno
wn
exam
ple
was
Ros
abet
h M
oss
Kan
ter's
(19
83)
stud
y of
inn
ovat
ion
repo
rted
in
The
Cha
nge
Mas
ters.
Her
boo
k fo
cuse
d on
ten
cur
e ca
se
stud
ies.
She
bega
n he
r se
arch
for
the
"be
st"
or "
mos
t in
nova
tive"
co
mpa
nies
by
getti
ng t
he v
iew
s of
cor
pora
te e
xper
ts i
n hu
man
re-
sour
ce f
ield
s. N
omin
atio
ns f
or c
ases
to
stud
y sn
owba
lled
from
the
re
and
then
con
verg
ed in
to a
sm
all n
umbe
r of
cor
e ca
ses
nom
inat
ed b
y a
num
ber o
f diff
eren
t inf
orm
ants
. (9
) C
riter
ion s
ampl
ing.
The
logi
c of
crit
erio
n sa
mpl
ing
is t
o re
view
an
d st
udy
all c
ases
tha
t m
eet
som
e pr
edet
erm
ined
crit
erio
n of
impo
r-ta
nce.
Thi
s ap
proa
ch i
s co
mm
on i
n qu
ality
ass
uran
ce e
ffor
ts.
For
exam
ple,
the
exp
ecte
d ra
nge
of p
artic
ipat
ion
in a
men
tal
heal
th o
ut-
patie
nt p
rogr
am m
ight
be
4 to
26
wee
ks.
All
case
s th
at e
xcee
d 28
w
eeks
are
rev
iew
ed a
nd s
tudi
ed t
o fin
d ou
t w
hat
is h
appe
ning
and
to
mak
e su
re th
e ca
se is
bei
ng a
ppro
pria
tely
han
dled
. C
ritic
al in
cide
nts
can
be a
sou
rce
of c
riter
ion
sam
plin
g. F
or e
xam
-pl
e, a
ll in
cide
nts
of c
lient
abu
se i
n a
prog
ram
may
be
obje
cts
of i
n-de
pth
eval
uatio
n in
a q
ualit
y as
sura
nce
effo
rt. A
ll fa
rmer
men
tal h
ealth
cl
ient
s w
ho
com
mit
suic
ide
with
in
thre
e m
onth
s of
re
leas
e m
ay
cons
titut
e a
sam
ple
for
in-d
epth
, qua
litat
ive
stud
y. I
n a
scho
ol s
ettin
g,
all s
tude
nts
who
are
abs
ent
mor
e th
an h
alf
the
time
may
mer
it th
e in
-de
pth
atte
ntio
n of
a q
ualit
ativ
e ca
se s
tudy
. Th
e po
int
of c
riter
ion
sam
plin
g is
to
be s
ure
to u
nder
stan
d ca
ses
that
are
like
ly t
o be
Des
ignin
g Qua
litat
ive S
tudi
es
177
info
rmat
ion-
rich
beca
use
they
may
rev
eal
maj
or s
yste
m w
eakn
esse
s th
at b
ecom
e ta
rget
s of
opp
ortu
nity
for
pro
gram
or
syst
em im
prov
e-m
ent. Crit
erio
n sa
mpl
ing
can
add
an im
port
ant q
ualit
ativ
e co
mpo
nent
to
a m
anag
emen
t in
form
atio
n sy
stem
or
an o
ngoi
ng p
rogr
am m
onito
r-in
g sy
stem
. All
case
s in
the
dat
a sy
stem
tha
t ex
hibi
t ce
rtai
n pr
edet
er-
min
ed c
riter
ion
char
acte
ristic
s ar
e ro
utin
ely
iden
tifie
d fo
r in
-dep
th,
qual
itativ
e an
alys
is. C
riter
ion
sam
plin
g al
so c
an b
e ap
plie
d to
iden
tify
case
s fr
om q
uant
itativ
e qu
estio
nnai
res
or t
ests
for
in-
dept
h fo
llow
-up
. (10)
The
ory-
base
d or
ope
ratio
nal c
onstr
uct s
ampl
ing.
A m
ore
form
al b
asic
re
sear
ch v
ersi
on o
f cr
iterio
n sa
mpl
ing
is t
heor
y-ba
sed
sam
plin
g. T
he
rese
arch
er s
ampl
es in
cide
nts,
slic
es o
f lif
e, t
ime
perio
ds, o
r pe
ople
on
the
basi
s of
th
eir
pote
ntia
l m
anife
stat
ion
or
repr
esen
tatio
n of
im
port
ant
theo
retic
al c
onst
ruct
s. Th
e sa
mpl
e be
com
es,
by d
efin
ition
, re
pres
enta
tive
of t
he p
heno
men
on o
f in
tere
st. A
n ec
olog
ical
psy
chol
-og
ist
(see
Cha
pter
3)
is in
tere
sted
, for
exa
mpl
e, in
stu
dyin
g th
e in
ter-
actio
n be
twee
n a
pers
on a
nd t
he e
nviro
nmen
t. In
stan
ces
of s
uch
inte
ract
ion
mus
t be
def
ined
bas
ed o
n th
eore
tical
pre
mis
es in
ord
er t
o st
udy
exam
ples
that
repr
esen
t the
phe
nom
enon
of i
nter
est.
Thi
s di
ffer
s fr
om t
he m
ore
prac
tical
sam
plin
g in
pro
gram
eva
lua-
tion.
The
eva
luat
or d
oesn
't ne
ed a
the
ory-
base
d de
finiti
on o
f "p
ro-
gram
" be
caus
e th
e en
tity
to b
e st
udie
d is
usu
ally
lega
lly o
r fin
anci
ally
de
fined
. How
ever
, to
sam
ple
soci
al s
cien
ce p
heno
men
a th
at r
epre
sent
th
eore
tical
con
stru
cts
of in
tere
st, o
ne m
ust d
efin
e th
e co
nstr
uct t
o be
sa
mpl
ed,
such
as
pers
on-e
nviro
nmen
tal
inte
ract
ions
or
inst
ance
s of
so
cial
dev
ianc
e, id
entit
y cr
isis
, cre
ativ
ity, o
r po
wer
inte
ract
ions
in a
n or
gani
zatio
n.
Whe
n on
e is
stud
ying
peo
ple,
pro
gram
s, or
gani
zatio
ns,
or c
omm
u-ni
ties,
the
popu
latio
n of
int
eres
t ca
n be
fai
rly r
eadi
ly d
eter
min
ed.
Con
stru
cts
do n
ot h
ave
as c
lear
a fr
ame
of re
fere
nce;
nei
ther
doe
s tim
e.
The
prob
lem
with
tim
e sa
mpl
ing
is th
at t
here
are
no
conc
rete
pop
ula-
tions
of i
nter
est,
and
we
are
anyw
ay u
sual
ly re
stric
ted
to th
e lim
ited
time
span
ove
r w
hich
a s
tudy
is c
ondu
cted
or
to th
e on
ly s
light
ly lo
nger
tim
e sp
an,
hist
oric
ally
spe
akin
g ov
er w
hich
the
lite
ratu
re o
n a
topi
c ha
s ac
cum
ulat
ed. F
or s
ampl
ing
oper
atio
nal i
nsta
nces
of
cons
truc
ts, t
here
is
also
no
conc
rete
tar
get
popu
latio
n....
Mos
tly, t
here
fore
, we
are
forc
ed
to s
elec
t on
a pu
rpos
ive
basi
s th
ose
part
icul
ar in
stan
ces
of a
con
stru
ct
that
pas
t va
lidity
stu
dies
, con
vent
iona
l pr
actic
e, i
ndiv
idua
l in
tuiti
on,
or c
onsu
ltatio
n w
ith c
ritic
ally
min
ded
pers
ons
sugg
est o
ffer
the
clos
est
178
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTI
ON
corr
espo
nden
ce t
o th
e co
nstru
ct o
f in
tere
st. A
ltern
ativ
ely,
we
can
use
the
sam
e pr
oced
ures
to
selec
t m
ultip
le op
erat
iona
l re
pres
enta
tions
of
each
co
nstru
ct,
chos
en
beca
use
they
ov
erlap
in
re
pres
entin
g th
e cr
itica
l th
eore
tical
com
pone
nts
of th
e co
nstru
ct a
nd b
ecau
se th
ey d
iffer
from
eac
h ot
her
on i
rrele
vant
dim
ensio
ns.
This
seco
nd f
orm
of
sam
plin
g is
calle
d m
ultip
le op
erat
iona
lism
, an
d it
depe
nds
mor
e he
avily
on
in
divi
dual
judg
men
t th
an
does
th
e ra
ndom
sa
mpl
ing
of
pers
ons
from
a
well
-de
signa
ted,
targ
et p
opul
atio
n. Y
et su
ch ju
dgm
ents
, whi
le in
evita
ble,
are
less
w
ell u
nder
stoo
d th
an f
orm
al sa
mpl
ing
met
hods
and
are
larg
ely
igno
red
by
sam
plin
g ex
perts
. (Co
ok e
t al.,
198
5: 1
63-6
4)
"Ope
ratio
nal
cons
truc
t" s
ampl
ing
sim
ply
mea
ns t
hat
one
sam
ples
fo
r st
udy
real
-wor
ld e
xam
ples
(i.e
., op
erat
ional
exa
mpl
es)
of t
he c
on-
stru
cts
in w
hich
one
is in
tere
sted
. Stu
dyin
g a
num
ber
of s
uch
exam
ples
is
cal
led
"mul
tiple
ope
ratio
nalis
m"
(Web
b et
al.,
196
6).
(11)
Con
firm
ing
and
disco
nfirm
ing
case
s. In
the
ear
ly p
art
of q
ualit
ativ
e fie
ldw
ork
the
eval
uato
r is
exp
lorin
g—ga
ther
ing
data
and
beg
inni
ng t
o al
low
pat
tern
s to
em
erge
. Ove
r tim
e th
e ex
plor
ator
y pr
oces
s gi
ves
way
to
con
firm
ator
y fie
ldw
ork.
Thi
s in
volv
es t
estin
g id
eas,
conf
irmin
g th
e im
port
ance
and
mea
ning
of
poss
ible
pat
tern
s, an
d ch
ecki
ng o
ut t
he
viab
ility
of
emer
gent
fin
ding
s w
ith n
ew d
ata
and
addi
tiona
l cas
es. T
his
stag
e of
fie
ldw
ork
requ
ires
cons
ider
able
rig
or a
nd in
tegr
ity o
n th
e pa
rt
of t
he e
valu
ator
in
look
ing
for
and
sam
plin
g co
nfirm
ing
as w
ell a
s di
sconf
irmin
g cas
es.
Con
firm
ator
y ca
ses
are
addi
tiona
l ex
ampl
es
that
fit
al
read
y em
erge
nt p
atte
rns;
thes
e ca
ses
conf
irm a
nd e
labo
rate
the
fin
ding
s, ad
ding
ric
hnes
s, de
pth,
and
cre
dibi
lity.
Dis
conf
irmin
g ca
ses
are
no le
ss
impo
rtan
t at t
his
poin
t. Th
ese
are
the
exam
ples
that
don
't fit
. The
y ar
e a
sour
ce o
f riv
al in
terp
reta
tions
as
wel
l as
a w
ay o
f pl
acin
g bo
unda
ries
arou
nd c
onfir
med
fin
ding
s. Th
ey m
ay b
e "e
xcep
tions
tha
t pr
ove
the
rule
" or
exc
eptio
ns t
hat
disc
onfir
m a
nd a
lter
wha
t ap
pear
ed t
o be
pr
imar
y pa
ttern
s. Th
e so
urce
of
ques
tions
or
idea
s to
be
conf
irmed
or
disc
onfir
med
m
ay b
e fr
om s
take
hold
ers
or p
revi
ous
scho
larly
lite
ratu
re r
athe
r th
an
the
eval
uato
r's f
ield
wor
k. A
n ev
alua
tion
may
in p
art s
erve
the
purp
ose
of
conf
irmin
g or
di
scon
firm
ing
stak
ehol
der's
or
sc
hola
rs'
pre-
conc
eptio
ns,
thes
e ha
ving
bee
n id
entif
ied
durin
g ea
rly,
conc
eptu
al
eval
uato
r-st
akeh
olde
r des
ign
disc
ussi
ons
or li
tera
ture
revi
ews.
Thin
king
abo
ut th
e ch
alle
nge
of fi
ndin
g co
nfirm
ing
and
disc
on-
firm
ing
case
s em
phas
izes
the
rela
tions
hip
betw
een
sam
plin
g an
d
Des
ignin
g Qua
litat
ive S
tudi
es
179
rese
arch
con
clus
ions
. Th
e sa
mpl
e de
term
ines
wha
t th
e ev
alua
tor
will
ha
ve s
omet
hing
to
say
abou
t—th
us t
he im
port
ance
of
sam
plin
g ca
re-
fully
and
thou
ghtf
ully
. (1
2) O
ppor
tuni
stic
sam
plin
g. Fi
eldw
ork
ofte
n in
volv
es o
n-th
e-sp
ot
deci
sion
s ab
out
sam
plin
g to
tak
e ad
vant
age
of n
ew o
ppor
tuni
ties
durin
g ac
tual
dat
a co
llect
ion.
Unl
ike
expe
rimen
tal
desi
gns,
qual
itativ
e in
quiry
des
igns
can
incl
ude
new
sam
plin
g st
rate
gies
to
take
adv
anta
ge
of u
nfor
esee
n op
port
uniti
es a
fter
field
wor
k ha
s be
gun.
Bei
ng o
pen
to
follo
win
g w
here
ver
the
data
lea
d is
a p
rimar
y st
reng
th o
f qu
alita
tive
stra
tegi
es i
n re
sear
ch.
This
per
mits
the
sam
ple
to e
mer
ge d
urin
g fie
ldw
ork.
W
hen
obse
rvin
g, i
t is
not
pos
sibl
e to
cap
ture
eve
ryth
ing.
It
is,
ther
efor
e, n
eces
sary
to
mak
e de
cisi
ons
abou
t w
hich
act
iviti
es t
o ob
-se
rve,
whi
ch p
eopl
e to
obs
erve
and
int
ervi
ew, a
nd w
hat
time
perio
ds
will
be
sele
cted
to
colle
ct d
ata.
The
se d
ecis
ions
can
not
all b
e m
ade
in
adva
nce.
The
pur
pose
ful
sam
plin
g st
rate
gies
dis
cuss
ed a
bove
oft
en
depe
nd o
n so
me
know
ledg
e of
the
setti
ng b
eing
stu
died
. Opp
ortu
nist
ic
sam
plin
g ta
kes
adva
ntag
e of
wha
teve
r unf
olds
as
it un
fold
s. (1
3) P
urpo
sefu
l ran
dom
sam
plin
g. Th
e fa
ct th
at a
sm
all s
ampl
e si
ze w
ill
be c
hose
n fo
r in
-dep
th q
ualit
ativ
e st
udy
does
not
aut
omat
ical
ly m
ean
that
the
sam
plin
g st
rate
gy s
houl
d no
t be
rand
om. F
or m
any
audi
ence
s,
rand
om s
ampl
ing,
eve
n of
sm
all s
ampl
es, w
ill s
ubst
antia
lly in
crea
se th
e cr
edib
ility
of
the
resu
lts.
I re
cent
ly w
orke
d w
ith a
pro
gram
tha
t an
nual
ly a
ppea
rs b
efor
e th
e st
ate
legi
slat
ure
and
tells
"w
ar s
torie
s"
abou
t cl
ient
suc
cess
es,
som
etim
es e
ven
incl
udin
g a
few
sto
ries
abou
t fa
ilure
s to
pr
ovid
e ba
lanc
e.
They
de
cide
d th
ey
wan
ted
to
begi
n co
llect
ing
eval
uatio
n in
form
atio
n.
Beca
use
they
w
ere
striv
ing
for
indi
vidu
aliz
ed
outc
omes
th
ey
reje
cted
th
e no
tion
of
basi
ng
the
eval
uatio
n en
tirel
y on
a s
tand
ardi
zed
pre-
post
inst
rum
ent.
They
wan
ted
to c
olle
ct c
ase
hist
orie
s an
d do
in-
dept
h ca
se s
tudi
es o
f cl
ient
s, bu
t th
ey h
ad v
ery
limite
d re
sour
ces
and
time
to d
evot
e to
suc
h da
ta
colle
ctio
n. I
n ef
fect
, sta
ff a
t ea
ch p
rogr
am s
ite, m
any
of w
hom
ser
ve
200
to 3
00 f
amili
es a
yea
r, fe
lt th
at t
hey
coul
d on
ly d
o 10
or
15
deta
iled,
in-d
epth
clin
ical
cas
e hi
stor
ies
each
yea
r. W
e sy
stem
atiz
ed th
e ki
nd o
f in
form
atio
n th
at w
ould
be
goin
g in
to th
e ca
se h
isto
ries
at e
ach
prog
ram
site
and
the
n se
t up
a r
ando
m p
roce
dure
for
sel
ectin
g th
ose
clie
nts
who
se c
ase
hist
orie
s w
ould
be
reco
rded
in
dept
h. E
ssen
tially
, th
is p
rogr
am t
here
by s
yste
mat
ized
and
ran
dom
ized
the
ir co
llect
ion
of
"w
ar s
tori
es."
Whi
le t
hey
cann
ot g
ener
aliz
e to
the
ent
ire
clie
nt
180
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTI
ON
po
pula
tion
on t
he b
asis
of
10 c
ases
fro
m e
ach
prog
ram
site
, the
y w
ill
be a
ble
to t
ell
legi
slat
ors
that
the
sto
ries
they
are
rep
ortin
g w
ere
rand
omly
sel
ecte
d in
adv
ance
of kn
owled
ge o
f how
the
out
com
es wo
uld
appe
ar
and
that
the
info
rmat
ion
colle
cted
was
com
preh
ensi
ve. T
he c
redi
bilit
y of
sys
tem
atic
and
ran
dom
ly s
elec
ted
case
exa
mpl
es i
s co
nsid
erab
ly
grea
ter
than
the
per
sona
l, ad
hoc
sel
ectio
n of
cas
es t
o re
port
afte
r th
e fa
ct—
that
is, a
fter
out
com
es a
re k
now
n.
It is
crit
ical
to
unde
rsta
nd, h
owev
er, t
hat
this
is a
pur
pose
ful r
ando
m
sam
ple,
not a
rep
rese
ntat
ive
rand
om s
ampl
e. T
he p
urpo
se o
f a sm
all r
ando
m
sam
ple
is cre
dibi
lity,
not
repr
esen
tativ
enes
s. A
sm
all,
purp
osef
ul r
ando
m
sam
ple
aim
s to
red
uce
susp
icio
n ab
out w
hy c
erta
in c
ases
wer
e se
lect
ed
for
stud
y,
but
such
a
sam
ple
still
do
es
not
perm
it st
atis
tical
ge
nera
lizat
ions
. (1
4) S
ampl
ing
polit
ically
im
porta
nt c
ases
. E
valu
atio
n is
inh
eren
tly a
nd
inev
itabl
y po
litic
al t
o so
me
exte
nt (
see
Palu
mbo
, 19
87;
Patto
n, 1
986,
19
87b;
Tur
pin,
198
9). A
var
iatio
n of
the
criti
cal c
ase
sam
plin
g st
rate
gy
invo
lves
sel
ectin
g (o
r so
met
imes
avo
idin
g) a
pol
itica
lly s
ensi
tive
site
or
unit
of a
naly
sis.
For
exam
ple,
a s
tate
wid
e pr
ogra
m m
ay h
ave
a lo
cal
site
in th
e di
stric
t of
a st
ate
legi
slat
or w
ho is
par
ticul
arly
influ
entia
l. By
st
udyi
ng c
aref
ully
the
pro
gram
in t
hat
dist
rict,
eval
uatio
n da
ta m
ay b
e m
ore
likel
y to
attr
act
atte
ntio
n an
d ge
t us
ed. T
his
does
not
mea
n th
at
the
eval
uato
r th
en u
nder
take
s to
mak
e th
at s
ite l
ook
eith
er g
ood
or
bad,
dep
endi
ng o
n th
e po
litic
s of
the
mom
ent.
This
is
sim
ply
an
addi
tiona
l sa
mpl
ing
stra
tegy
for
try
ing
to i
ncre
ase
the
usef
ulne
ss a
nd
utili
zatio
n of
info
rmat
ion
whe
re r
esou
rces
per
mit
the
stud
y of
onl
y a
limite
d nu
mbe
r of c
ases
. Th
e sa
me
(bro
adly
spe
akin
g) p
oliti
cal p
ersp
ectiv
e m
ay in
form
cas
e sa
mpl
ing
in a
pplie
d or
eve
n ba
sic
rese
arch
stu
dies
. A p
oliti
cal s
cien
tist
or h
isto
rian
mig
ht s
elec
t th
e W
ater
gate
or
Iran
-Con
tra
scan
dals
for
st
udy
not
only
bec
ause
of
the
insi
ghts
the
y pr
ovid
e ab
out
the
Am
er-
ican
sys
tem
of
gove
rnm
ent
but
beca
use
of t
he li
kely
atte
ntio
n su
ch a
st
udy
wou
ld a
ttrac
t. A
soc
iolo
gist
's st
udy
of a
rio
t or
a p
sych
olog
ist's
st
udy
of a
fam
ous
suic
ide
wou
ld li
kely
invo
lve
som
e at
tent
ion
durin
g sa
mpl
ing
to th
e po
litic
al im
port
ance
of t
he c
ase.
(1
5) C
onve
nien
ce s
ampl
ing.
Fina
lly, t
here
is
the
stra
tegy
of
sam
plin
g by
con
veni
ence
: do
ing
wha
t's f
ast
and
conv
enie
nt.
Thi
s is
pro
babl
y th
e m
ost
com
mon
sam
plin
g st
rate
gy—
and
the
leas
t de
sira
ble.
Too
of
ten
eval
uato
rs u
sing
qua
litat
ive
met
hods
thi
nk t
hat,
beca
use
the
sam
ple
size
the
y ca
n st
udy
is t
oo s
mal
l to
per
mit
gene
raliz
atio
ns,
it do
esn'
t mat
ter
how
cas
es a
re p
icke
d, s
o th
ey m
ight
as
wel
l pic
k on
es
Des
ignin
g Qua
litat
ive S
tudi
es
181
that
are
eas
y to
acc
ess
and
inex
pens
ive
to s
tudy
. Whi
le co
nven
ience
and
co
st ar
e re
al c
onsid
erat
ions
, th
ey s
houl
d be
the
las
t fa
ctors
to
be t
aken
int
o ac
coun
t af
ter
stra
tegi
cally
del
iber
atin
g on
how
to
get
the
mos
t in
for-
mat
ion
of g
reat
est
utili
ty f
rom
the
lim
ited
num
ber
of c
ases
to
be
sam
pled
. Pur
pose
ful,
stra
tegi
c sa
mpl
ing
can
yiel
d cr
ucia
l inf
orm
atio
n ab
out
criti
cal
case
s. C
onve
nien
ce s
ampl
ing
is ne
ither
pur
pose
ful
nor
stra-
tegic.
Infor
mat
ion-R
ich C
ases
Tabl
e 5.
5 su
mm
ariz
es t
he 1
5 pu
rpos
eful
sam
plin
g st
rate
gies
dis
-cu
ssed
abo
ve,
plus
a 1
6th
appr
oach
—co
mbi
natio
n or
mix
ed p
ur-
pose
ful
sam
plin
g.
For
exam
ple,
an
ex
trem
e gr
oup
or
max
imum
he
tero
gene
ity a
ppro
ach
may
yie
ld a
n in
itial
pot
entia
l sam
ple
size
that
is
still
larg
er th
an th
e st
udy
can
hand
le. T
he f
inal
sel
ectio
n, t
hen,
may
be
mad
e ra
ndom
ly—
a co
mbi
natio
n ap
proa
ch. T
hus
thes
e ap
proa
ches
are
no
t m
utua
lly e
xclu
sive
. E
ach
appr
oach
ser
ves
a so
mew
hat
diff
eren
t pu
rpos
e.
Beca
use
rese
arch
an
d ev
alua
tions
of
ten
serv
e m
ultip
le
purp
oses
, m
ore
than
one
qua
litat
ive
sam
plin
g st
rate
gy m
ay b
e ne
ces-
sary
. In
long
-ter
m f
ield
wor
k al
l of
thes
e st
rate
gies
may
be u
sed
at s
ome
poin
t. Thes
e ar
e no
t th
e on
ly w
ays
of s
ampl
ing
qual
itativ
ely.
The
und
er-
lyin
g pr
inci
ple
that
is
com
mon
to
all
thes
e st
rate
gies
is
sele
ctin
g in
form
atio
n-ric
h ca
ses.
Thes
e ar
e ca
ses
from
whi
ch o
ne c
an l
earn
a
grea
t de
al a
bout
mat
ters
of
impo
rtan
ce. T
hey
are
case
s w
orth
y of
in-
dept
h st
udy.
In
the
pro
cess
of
deve
lopi
ng t
he r
esea
rch
desi
gn, t
he e
valu
ator
or
rese
arch
er is
try
ing
to c
onsi
der
and
antic
ipat
e th
e ki
nds
of a
rgum
ents
th
at w
ill le
nd c
redi
bilit
y to
the
stud
y as
wel
l as
the
kind
s of
arg
umen
ts
that
mig
ht b
e us
ed to
atta
ck th
e fin
ding
s. R
easo
ns fo
r si
te s
elec
tions
or
indi
vidu
al c
ase
sam
plin
g ne
ed t
o be
car
eful
ly a
rtic
ulat
ed a
nd m
ade
expl
icit.
Mor
eove
r, it
is i
mpo
rtan
t to
be
open
and
cle
ar a
bout
the
st
udy'
s lim
itatio
ns,
incl
udin
g ho
w a
ny p
artic
ular
pur
pose
ful
sam
plin
g st
rate
gy m
ay l
ead
to d
isto
rtio
n in
the
fin
ding
s—th
at i
s, to
ant
icip
ate
criti
cism
s th
at w
ill b
e m
ade
of a
par
ticul
ar s
ampl
ing
stra
tegy
. H
avin
g w
eigh
ed
the
evid
ence
an
d co
nsid
ered
th
e al
tern
ativ
es,
eval
uato
rs a
nd p
rimar
y st
akeh
olde
rs m
ake
thei
r sa
mpl
ing
deci
sion
s, so
met
imes
pai
nful
ly, b
ut a
lway
s w
ith t
he r
ecog
nitio
n th
at t
here
are
no
perf
ect
desi
gns.
The
sam
plin
g st
rate
gy m
ust
be s
elec
ted
to f
it th
e pu
rpos
e of
the
stu
dy, t
he r
esou
rces
ava
ilabl
e, t
he q
uest
ions
bei
ng
Desi
gnin
g Qua
litat
ive S
tudi
es
183
182
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTIO
N
Tab
le 5
.5 S
ampl
ing
Stra
tegi
es
Tab
le 5
.5 (c
ontin
ued)
Ty
pe
Purp
ose
Type
Pu
rpos
e
9. c
riter
ion
sam
plin
g Pi
ckin
g all
cas
es th
at m
eet s
ome
crite
rion,
su
ch a
s all
child
ren
abus
ed in
a tr
eatm
ent
facil
ity. Q
ualit
y as
sura
nce.
A. R
ando
m p
roba
bilit
y sa
mpl
ing
Repr
esen
tativ
enes
s: Sa
mpl
e siz
e a
func
tion
of p
opul
atio
n siz
e an
d de
sired
co
nfid
ence
leve
l. 10
. the
ory-
base
d or
ope
ratio
nal
Find
ing
man
ifest
atio
ns o
f a th
eore
tical
l. sim
ple
rand
om sa
mpl
e Pe
rmits
gen
eral
izat
ion
from
sam
ple
to
the
popu
latio
n it
repr
esen
ts.
cons
truct
sam
plin
g
cons
truct
of i
nter
est s
o as
to e
labor
ate
an
d ex
amin
e th
e co
nstru
ct.
2. st
ratif
ied
rand
om a
nd c
lust
er
Incr
ease
s con
fiden
ce in
mak
ing
gene
r-
11. c
onfir
min
g an
d di
scon
firm
ing
E
labor
atin
g an
d de
epen
ing
initi
al
sam
ples
al
izat
ions
to p
artic
ular
subg
roup
s or
area
s.
ca
ses
analy
sis, s
eeki
ng e
xcep
tions
, tes
ting
va
riatio
n.
B. P
urpo
sefu
l sam
plin
g Se
lect
s inf
orm
atio
n-ric
h ca
ses f
or in
-de
pth
stud
y. S
ize
and
spec
ific
case
s de
pend
on
stud
y pu
rpos
e.
12. o
ppor
tuni
stic
sam
plin
g
Follo
win
g ne
w le
ads d
urin
g fie
ldw
ork,
ta
king
adv
anta
ge o
f the
une
xpec
ted,
fle
xibi
lity.
l. ex
trem
e or
dev
iant
cas
e sa
mpl
ing
Lear
ning
from
hig
hly
unus
ual m
ani-
fest
atio
ns o
f the
phe
nom
enon
of i
nter
est,
such
as o
utst
andi
ng su
cces
ses/
not
able
failu
res,
top
of th
e cla
ss/
drop
outs
, exo
tic
even
ts, c
rises
.
13. r
ando
m p
urpo
sefu
l sam
plin
g
Add
s cre
dibi
lity
to sa
mpl
e w
hen
pote
n-
(stil
l sm
all sa
mpl
e siz
e)
tial p
urpo
sefu
l sam
ple
is lar
ger t
han
one
can
hand
le. R
educ
es ju
dgm
ent w
ithin
a
purp
osef
ul c
ateg
ory.
(Not
for g
ener
ali-
zatio
ns o
r rep
rese
ntat
iven
ess.)
2.
inte
nsity
sam
plin
g In
form
atio
n-ric
h ca
ses t
hat m
anife
st th
e ph
enom
enon
inte
nsel
y, b
ut n
ot
extre
mel
y, su
ch a
s goo
d st
uden
ts/
poor
st
uden
ts, a
bove
ave
rage
/bel
ow a
vera
ge.
14. s
ampl
ing
polit
ically
impo
rtant
A
ttrac
ts a
ttent
ion
to th
e st
udy
(or a
void
s
ca
ses
attra
ctin
g un
desir
ed a
ttent
ion
by
purp
osef
ully
elim
inat
ing
from
the
sam
ple
polit
ically
sens
itive
cas
es).
3. m
axim
um v
aria
tion
sam
plin
g-
Doc
umen
ts u
niqu
e or
div
erse
var
iatio
ns
15. c
onve
nien
ce sa
mpl
ing
Save
s tim
e, m
oney
, and
effo
rt. P
oore
st
ratio
nale;
low
est c
redi
bilit
y. Y
ields
in
form
atio
n-po
or c
ases
. pu
rpos
eful
ly p
icki
ng a
wid
e th
at h
ave
emer
ged
in a
dapt
ing
to d
iffer
- ra
nge
of v
aria
tion
on d
imen
sions
en
t con
ditio
ns. I
dent
ifies
impo
rtan
t of
inte
rest
co
mm
on p
atte
rns t
hat c
ut a
cros
s va
riatio
ns.
16. c
ombi
natio
n or
mix
ed p
urpo
sefu
l Tr
iangu
latio
n, fl
exib
ility
, mee
ts m
ultip
le
sam
plin
g in
tere
sts a
nd n
eeds
. 4.
hom
ogen
eous
sam
plin
g Fo
cuse
s, re
duce
s var
iatio
n, si
mpl
ifies
an
alys
is, fa
cilit
ates
gro
up in
terv
iew
ing.
5. ty
pica
l cas
e sa
mpl
ing
Illus
trate
s or h
ighl
ight
s wha
t is t
ypic
al,
norm
al, a
vera
ge.
aske
d, a
nd th
e co
nstra
ints
bein
g fa
ced.
Thi
s hol
ds tr
ue fo
r sam
plin
g st
rate
gy
as w
ell a
s sam
ple
size.
6. st
ratif
ied
purp
osef
ul sa
mpl
ing
Illus
trat
es
char
acte
ristic
s of
pa
rtic
ular
su
bgro
ups
of
inte
rest
; fa
cilit
ates
co
m-
paris
ons.
7.
crit
ical
cas
e sa
mpl
ing
Perm
its lo
gica
l gen
eral
izat
ion
and
max
i-m
um a
pplic
atio
n of
info
rmat
ion
to o
ther
ca
ses b
ecau
se if
it's
true
of th
is on
e ca
se
SAM
PLE
SIZ
E
Q
ualit
ativ
e in
quiry
is
rife
with
am
bigu
ities
. Th
ere
are
purp
osef
ul
stra
tegi
es
inst
ead
of
met
hodo
logi
cal
rule
s. Th
ere
are
inqu
iry
ap-
proa
ches
ins
tead
of
stat
istic
al f
orm
ulas
. Q
ualit
ativ
e in
quiry
see
ms
to
wor
k be
st f
or p
eopl
e w
ith a
hig
h to
lera
nce
for
ambi
guity
. (A
nd w
e're
st
ill o
nly
disc
ussi
ng d
esig
n. I
t get
s w
orse
whe
n w
e ge
t to
anal
ysis
.)
it'
s lik
ely
to b
e tru
e of
all
othe
r cas
es.
8. sn
owba
ll or
cha
in sa
mpl
ing
Iden
tifie
s cas
es o
f int
eres
t fro
m p
eopl
e w
ho k
now
peo
ple
who
kno
w p
eopl
e w
ho
know
wha
t cas
es a
re in
form
atio
n ric
h,
that
is, g
ood
exam
ples
for s
tudy
, goo
d in
terv
iew
subj
ects
.
184
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTI
ON
Now
here
is th
is a
mbi
guity
cle
arer
than
in th
e m
atte
r of s
ampl
e si
ze.
I ge
t let
ters
. I g
et c
alls
. "Is
10
a la
rge
enou
gh s
ampl
e to
ach
ieve
m
axim
um v
aria
tion?
" "I
sta
rted
out
to in
terv
iew
20
peop
le fo
r 2 h
ours
eac
h, b
ut I
've
lost
2
peop
le. I
s 18
larg
e en
ough
, or d
o I
have
to fi
nd 2
mor
e?"
"I w
ant t
o st
udy
just
one
org
aniz
atio
n, b
ut in
terv
iew
20
peop
le in
th
e or
gani
zatio
n. I
s m
y sa
mpl
e si
ze 1
or 2
0 or
bot
h?"
My
univ
ersa
l, ce
rtai
n, a
nd c
onfid
ent r
eply
to th
ese
ques
tions
is th
is:
"it d
epen
ds."
Th
ere
are
no r
ules
for
sam
ple
size
in q
ualit
ative
inq
uiry
. Sa
mpl
e si
ze
depe
nds
on w
hat y
ou w
ant t
o kn
ow, t
he p
urpo
se o
f th
e in
quiry
, wha
t's
at s
take
, wha
t will
be
usef
ul, w
hat w
ill h
ave
cred
ibili
ty, a
nd w
hat c
an b
e do
ne w
ith a
vaila
ble
time
and
reso
urce
s. E
arlie
r in
thi
s ch
apte
r, I
disc
usse
d th
e tr
ade-
off
betw
een
brea
dth
and
dept
h. W
ith t
he s
ame
fixed
res
ourc
es a
nd l
imite
d tim
e, a
re-
sear
cher
cou
ld s
tudy
a s
peci
fic s
et o
f ex
perie
nces
for
a la
rger
num
ber
of p
eopl
e (s
eeki
ng b
read
th) o
r a
mor
e op
en r
ange
of
expe
rienc
es f
or a
sm
alle
r nu
mbe
r of
peo
ple
(see
king
dep
th).
In-d
epth
info
rmat
ion
from
a
smal
l num
ber
of p
eopl
e ca
n be
ver
y va
luab
le, e
spec
ially
if t
he c
ases
ar
e in
form
atio
n-ric
h. L
ess
dept
h fr
om a
larg
er n
umbe
r of
peo
ple
can
be
espe
cial
ly
help
ful
in
expl
orin
g a
phen
omen
on
and
tryi
ng
to
docu
men
t di
vers
ity o
r un
ders
tand
var
iatio
n. I
rep
eat,
the
size
of
the
sam
ple
depe
nds
on w
hat y
ou w
ant t
o fin
d ou
t, w
hy y
ou w
ant t
o fin
d it
out,
how
the
findi
ngs
will
be
used
, and
wha
t res
ourc
es (i
nclu
ding
tim
e)
you
have
for t
he s
tudy
. To
und
erst
and
the
prob
lem
of
smal
l sam
ples
in q
ualit
ativ
e in
quiry
, it'
s ne
cess
ary
to p
lace
thes
e sm
all s
ampl
es in
the
cont
ext o
f pr
obab
ility
sa
mpl
ing.
A q
ualit
ativ
e in
quiry
sam
ple
only
seem
s sm
all
in c
ompa
rison
w
ith t
he s
ampl
e si
ze n
eede
d fo
r re
pres
enta
tiven
ess
whe
n th
e pu
rpos
e is
gen
eral
izin
g fr
om a
sam
ple
to t
he p
opul
atio
n of
whi
ch it
is a
par
t. Su
ppos
e th
ere
are
100
peop
le in
a p
rogr
am t
o be
eva
luat
ed. I
t w
ould
be
nec
essa
ry t
o ra
ndom
ly s
ampl
e 80
of
thos
e pe
ople
(80
%)
to m
ake
a ge
nera
lizat
ion
at t
he 9
5% c
onfid
ence
leve
l. If
the
re a
re 5
00 p
eopl
e in
th
e pr
ogra
m, 2
17 p
eopl
e m
ust b
e sa
mpl
ed (4
3%) f
or th
e sa
me
leve
l of
conf
iden
ce.
If t
here
are
1,0
00 p
eopl
e, 2
78 p
eopl
e m
ust
be s
ampl
ed
(28%
); an
d if
ther
e ar
e 5,
000
peop
le in
the
pop
ulat
ion
of in
tere
st, 3
57
mus
t be
sam
pled
(7%
) to
ach
ieve
a 9
5% c
onfid
ence
lev
el i
n th
e ge
nera
lizat
ion
of f
indi
ngs.
At
the
othe
r ex
trem
e, i
f th
ere
are
only
50
peop
le in
the
prog
ram
, 44
mus
t be
rand
omly
sam
pled
(88%
) to
achi
eve
Des
ignin
g Qua
litat
ive S
tudi
es
185
a 95
% le
vel o
f con
fiden
ce. (
See
Fitz
gibb
on a
nd M
orris
, 198
7: 1
63, f
or
a ta
ble
on d
eter
min
ing
sam
ple
size
from
a g
iven
pop
ulat
ion.
) Th
e lo
gic
of p
urpo
sefu
l sa
mpl
ing
is q
uite
diff
eren
t fr
om t
he l
ogic
of
pro
babi
lity
sam
plin
g. T
he p
robl
em is
, how
ever
, tha
t th
e ut
ility
and
cr
edib
ility
of s
mal
l pur
pose
ful s
ampl
es a
re o
ften
judg
ed o
n th
e ba
sis
of
the
logi
c, p
urpo
se,
and
reco
mm
ende
d sa
mpl
e si
zes
of p
roba
bilit
y sa
mpl
ing.
Wha
t sh
ould
hap
pen
is t
hat
purp
osef
ul s
ampl
es b
e ju
dged
on
the
basi
s of
the
pur
pose
and
rat
iona
le o
f ea
ch s
tudy
and
the
sam
plin
g st
rate
gy u
sed
to a
chie
ve th
e st
udy'
s pu
rpos
e. T
he s
ampl
e, li
ke a
ll ot
her
aspe
cts
of q
ualit
ativ
e in
quiry
, m
ust
be j
udge
d in
con
text
—th
e sa
me
prin
cipl
e th
at u
nder
gird
s an
alys
is a
nd p
rese
ntat
ion
of q
ualit
ativ
e da
ta.
Ran
dom
pr
obab
ility
sa
mpl
es
cann
ot
acco
mpl
ish
wha
t in
-dep
th,
purp
osef
ul s
ampl
es a
ccom
plis
h, a
nd v
ice
vers
a.
Piag
et c
ontr
ibut
ed a
maj
or b
reak
thro
ugh
to o
ur u
nder
stan
ding
of
how
chi
ldre
n th
ink
by o
bser
ving
his
ow
n tw
o ch
ildre
n at
leng
th a
nd in
gr
eat
dept
h. F
reud
est
ablis
hed
the
field
of
psyc
hoan
alys
is b
ased
on
few
er t
han
ten
clie
nt c
ases
. Ba
ndle
r an
d G
rinde
r (1
975a
, 19
75b)
fo
unde
d ne
urol
ingu
istic
pro
gram
min
g (N
LP)
by s
tudy
ing
thre
e re
-no
wne
d an
d hi
ghly
eff
ectiv
e th
erap
ists
: M
ilton
Eric
kson
, Fr
itz P
erls
, an
d V
irgin
ia S
atin
Pet
ers
and
Wat
erm
an (1
982)
form
ulat
ed th
eir
wid
ely
follo
wed
eig
ht p
rinci
ples
for
org
aniz
atio
nal e
xcel
lenc
e by
stu
dyin
g 62
co
mpa
nies
, a
very
sm
all
sam
ple
of t
he t
hous
ands
of
com
pani
es o
ne
mig
ht s
tudy
. Th
e va
lidity
, m
eani
ngfu
lnes
s, an
d in
sight
s ge
nera
ted
from
qua
litat
ive
inqu
iry
have
mor
e to
do
with
the
inf
orm
atio
n-ri
chne
ss o
f th
e ca
ses
selec
ted
and
the
obse
rvat
iona
l/an
alyt
ical
capa
bilit
ies o
f th
e re
sear
cher
tha
n wi
th
sam
ple
size.
This
issu
e of
sam
ple
size
is a
lot l
ike
the
prob
lem
stu
dent
s ha
ve
whe
n th
ey a
re a
ssig
ned
an e
ssay
to w
rite.
St
uden
t: "H
ow lo
ng d
oes
the
pape
r hav
e to
be?
"
Inst
ruct
or: "
Long
eno
ugh
to c
over
the
assi
gnm
ent."
St
uden
t: "B
ut h
ow m
any
page
s?"
Inst
ruct
or: "
Eno
ugh
page
s to
do
just
ice
to th
e su
bjec
t—no
mor
e, n
o le
ss."
Linc
oln
and
Gub
a (1
985:
202
) rec
omm
end
sam
ple
sele
ctio
n
to t
he p
oint
of
redu
ndan
cy ..
. In
purp
osef
ul s
ampl
ing
the
size
of
the
sam
ple
is d
eter
min
ed b
y in
form
atio
nal c
onsi
dera
tions
. If t
he p
urpo
se
186
QU
ALI
TATI
VE
DE
SIG
NS
AN
D D
ATA
CO
LLE
CTI
ON
is t
o m
axim
ize
info
rmat
ion,
the
sam
plin
g is
ter
min
ated
whe
n no
new
in
form
atio
n is
for
thco
min
g fr
om n
ew s
ampl
ed u
nits
; thu
s re
dund
ancy
is
the
prim
ary
crite
rion.
(em
phas
is in
the
orig
inal
) Th
is s
trat
egy
leav
es th
e qu
estio
n of
sam
ple
size
ope
n.
Ther
e re
mai
ns, h
owev
er, t
he p
ract
ical
pro
blem
s of
how
to n
egot
iate
an
eva
luat
ion
budg
et o
r ho
w t
o ge
t a
diss
erta
tion
com
mitt
ee t
o ap
prov
e a
desi
gn if
you
don
't ha
ve s
ome
idea
of
sam
ple
size
. Sam
plin
g to
the
poi
nt o
f re
dund
ancy
is a
n id
eal,
one
that
wor
ks b
est
for
basi
c re
sear
ch, u
nlim
ited
time
lines
, and
unc
onst
rain
ed re
sour
ces.
The
solu
tion
is j
udgm
ent
and
nego
tiatio
n. I
rec
omm
ende
d th
at
qual
itativ
e sa
mpl
ing
desi
gns
spec
ify m
inim
um s
ampl
es b
ased
on
ex-
pect
ed r
easo
nabl
e co
vera
ge o
f th
e ph
enom
enon
giv
en t
he p
urpo
se o
f th
e st
udy
and
stak
ehol
der
inte
rest
s. O
ne m
ay a
dd t
o th
e sa
mpl
e as
fie
ldw
ork
unfo
lds.
One
may
cha
nge
the
sam
ple
if in
form
atio
n em
erge
s th
at in
dica
tes
the
valu
e of
a c
hang
e. T
he d
esig
n sh
ould
be
unde
rsto
od
to b
e fle
xibl
e an
d em
erge
nt.
Yet
, at
the
beg
inni
ng,
for
plan
ning
and
bu
dget
ary
purp
oses
, on
e sp
ecifi
es a
min
imum
exp
ecte
d sa
mpl
e si
ze
and
build
s a
ratio
nale
for
tha
t m
inim
um, a
s w
ell a
s cr
iteria
tha
t w
ould
al
ert
the
rese
arch
er t
o in
adeq
uaci
es in
the
orig
inal
sam
plin
g ap
proa
ch
and/
or s
ize.
In
the
end,
sam
ple
size
ade
quac
y, li
ke a
ll as
pect
s of
res
earc
h, is
sub
ject
to
pee
r re
view
, con
sens
ual v
alid
atio
n, a
nd ju
dgm
ent.
Wha
t is
cru
cial
is
that
th
e sa
mpl
ing
proc
edur
es
and
deci
sion
s be
fu
lly
desc
ribed
, ex
plai
ned,
and
jus
tifie
d so
tha
t in
form
atio
n us
ers
and
peer
rev
iew
ers
have
the
appr
opria
te c
onte
xt f
or ju
dgin
g th
e sa
mpl
e. T
he r
esea
rche
r or
ev
alua
tor
is a
bsol
utel
y ob
ligat
ed t
o di
scus
s ho
w t
he s
ampl
e af
fect
ed
the
findi
ngs,
the
stre
ngth
s an
d w
eakn
esse
s of
the
sam
plin
g pr
oced
ures
, an
d an
y ot
her
desi
gn d
ecis
ions
tha
t ar
e re
leva
nt f
or i
nter
pret
ing
and
unde
rsta
ndin
g th
e re
port
ed
resu
lts.
Exe
rcis
ing
care
no
t to
ov
erge
nera
lize
from
pur
pose
ful s
ampl
es, w
hile
max
imiz
ing
to t
he f
ull
the
adva
ntag
es o
f in
-dep
th,
purp
osef
ul s
ampl
ing,
will
do
muc
h to
al
levi
ate
conc
erns
abo
ut s
mal
l sam
ple
size
.