Journal
of
Personality
and
Social
Psychology
1986,
W.
51,
No.
1,125-139
Copyright
1986
by
the
American Psychological Association, Inc.
0022-3514y86/$00.75
On the
Nature
of
Self-Monitoring: Matters
of
Assessment,
Matters
of
Validity
Mark Snyder
and
Steve
Gangestad
University
of
Minnesota
An
extensive network
of
empirical relations
has
been
identified
in
research
on the
psychological
construct
of
self-monitoring.
Nevertheless,
in
recent years some concerns have been expressed about
the
instrument used
for the
assessment
of
self-monitoring
propensities,
the
Self-Monitoring
Scale.
Both
the
extent
to
which
the
measure taps
an
interpretable
and
meaningful
causal variable
and the
extent
to
which
the
self-monitoring construct provides
an
appropriate theoretical understanding
of
this causal variable have been questioned.
An
examination
of
reanalyses
of
studies
of
self-monitor-
ing,
analyses
of the
internal structure
of the
Self-Monitoring
Scale,
and
further
relevant data suggest
that
the
measure does
tap a
meaningful
and
interpretable causal variable with pervasive
influences
on
social behavior,
a
variable
reflected
as a
general
self-monitoring
factor.
We
discuss
the
evaluation
and
furthering
of the
interpretation
of
this
latent causal variable,
offer
criteria
for
evaluating alterna-
tive
measures
of
self-monitoring,
and
present
a
new,
18-item
Self-Monitoring
Scale.
According
to
theoretical analyses
of
self-monitoring,
people
differ
in the
extent
to
which
they
can and do
observe
and
control
their expressive behavior
and
self-presentation (e.g., Snyder,
1979).
Individuals high
in
self-monitoring
are
thought
to
regu-
late
their expressive self-presentation
for the
sake
of
desired
public appearances,
and
thus
be
highly
responsive
to
social
and
interpersonal cues
of
situationally
appropriate
performances.
Individuals
low in
self-monitoring
are
thought
to
lack either
the
ability
or the
motivation
to so
regulate their expressive self-pre-
sentations. Their expressive behaviors, instead,
are
thought
to
functionally reflect
their
own
enduring
and
momentary inner
states,
including
their
attitudes,
traits,
and
feelings.
A
number
of
hypotheses—concerning,
among others,
the de-
terminants
of
specificity
and
consistency
in
social behavior,
the
origins
of
linkages between attitudes
and
action,
the
dynamics
of
social interaction,
and the
nature
and
consequences
of
con-
ceptions
of
self—have
followed from
these basic initial proposi-
tions. Research involving
a
measure
of
self-monitoring
propen-
sities
(the
Self-Monitoring
Scale;
for
information
on its
reliabil-
ity
and
validity,
see
Snyder, 1974)
has
provided empirical
support
for
these
and
many more hypotheses about
the
cogni-
tive,
behavioral,
and
interpersonal consequences
of
self-moni-
toring (e.g.,
Ajzen,
Timko,
&
White,
1982;
Becherer
&
Richard,
1978;
Caldwell&
O'Reilly,
1982;Danheiser&Graziano,
1982;
Ickes,
Layden,
&
Barnes, 1978;
Krauss,
Geller,
&
Olson, 1976;
Kulik&Taylor,
1981;Lippa, 1976,
1978;Lutsky,
Woodworth,
The
research
for and the
preparation
of
this article were
supported
in
part
by
National Science Foundation Grant
BNS
82-07632
to
Mark
Snyder,
in
part
by a
National Institute
of
Mental Health National
Re-
search
Council postdoctoral traineeship
to
Steve Gangestad,
and in
part
by
a
grant
from
the
University
of
Minnesota
Computer Center.
We
thank
Daryl
Bern
and
Jeffry
A.
Simpson
for
their
helpful
com-
ments
on the
article.
Correspondence concerning this article should
be
addressed
to
Mark
Snyder
or to
Steve Gangestad,
Department
of
Psychology,
University
of
Minnesota,
75
East River Road, Minneapolis, Minnesota 55455.
&
Clayton, 1980; McCann
&
Hancock, 1983; Paulhus, 1982;
Rarick,
Soldow,
&
Geizer, 1976; Ross, McFarland,
&
Fletcher,
1981; Shaffer,
Smith,
&
Tomarelli, 1982; Snyder, 1974; Snyder,
Berscheid,
&
Click,
1985;
Snyder
&
Cantor, 1980; Snyder
&
Gangestad,
1982;
Snyder, Gangestad,
&
Simpson,
1983;
Snyder
&Kendzierski,
1982a,
1982b;
Snyder
&Monson, 1975;Snyder
&
Simpson, 1984; Snyder
&
Swann,
1976;
Snyder
&
Tanke,
1976; Tunnell, 1980; Zanna, Olson,
&
Fazio, 1980).
More
recently,
elaboration
of the
construct
has led to its ap-
plication
in yet
other domains
of
social
behavior
and
interper-
sonal relationships. Some
of the
domains
in
which
the
Self-
Monitoring
Scale
has
proven
its
relevance
and
applicability
are
recent
and
ongoing explorations
of the
nature
of
friendships
(e.g., Snyder
&
Smith,
1986),
romantic relationships (e.g., Sny-
der
&
Simpson,
1986),
and
sexual involvements
(e.g.,
Snyder,
Simpson,
&
Gangestad, 1986),
as
well
as
applications
to the
psychology
of
advertising (e.g., Snyder
&
DeBono,
1985), per-
sonnel selection
(e.g.,
Snyder, Berscheid,
&
Matwychuk,
1984),
and
psychopathology
(e.g., Snyder
&
Smith,
1985).
Clearly
then, over
many
years
of
research,
a
large number
of
associations between
the
Self-Monitoring Scale
and a
wide
range
of
important behavioral
criteria
have been
documented;
for
reviews
of
these
associations,
see
Snyder
(1979,
in
press)
and
Shaw
and
Costanzo
(1982).
In a
phrase,
the
measure
empiri-
cally
works.
This
fact
notwithstanding,
in
recent years some
concerns
have
been expressed
about
why and how the
measure
works.
Thus, Briggs
and
Cheek (1986) have
stated,
The
Self-Monitoring Scale
is a
popular measure
of
personality
and
has
served
as the
centerpiece
for a
number
of
published articles.
The
scale
has
proved successful
in
predicting
a
variety
of
criteria
and
has
acted
to
stimulate
experimental
social
psychologists'
inter-
est in the
measurement
of
individual
differences.
The
problem
is
to
understand
why the
scale
works,
(p.
129)
Of
course,
one may
ask, isn't
the
obvious answer
that
the
Self-
Monitoring Scale works because
it
measures
the
differences
be-
tween
individuals explicated
by the
self-monitoring
construct?
Or,
if it
doesn't
measure
the
self-monitoring
construct,
doesn't
125
126
MARK
SNYDER
AND
STEVE
GANGESTAD
the
scale work
at
least because
it
measures
meaningful
and
sys-
tematic individual
differences
implicated
in
social behavior
and
interaction? Although these
may
appear
to be the
obvious
an-
swers, they are, according
to
Briggs
and
Cheek
as
well
as
others
(e.g.,
Lennox
&
Wolfe,
1984), wrong. For, according
to
these
authors,
the
Self-Monitoring Scale possesses psychometric
properties
that
pose threats
to its
ability
to
measure
meaningful
and
interpretable
individual
differences.
The
Self-Monitoring
Scale:
What
Does
It
Measure?
Specifically,
what
are the
concerns that have been expressed
about
the
measure
of
self-monitoring?
First,
it has
been claimed
that
the
Self-Monitoring
Scale,
whatever
it
measures, does
not
validly
measure
the
self-monitoring
construct.
The
tenor
of
this
claim
has
ranged
from
the
merely suggestive
to the
unshakably
resolute. Thus,
Briggs,
Cheek,
and
Buss
(1980) have suggested
that
there "may
be a gap
between
the
construct
of
self-monitor-
ing
and its
operationalization
in the
Self-Monitoring
Scale"
(p.
586).
And, Lennox
and
Wolfe
(1984)
have asserted
that
the
"measure
demonstrably
lacks
fidelity to the
construct
[of
self-
monitoring]"
(p.
1350).
Second,
it has
been claimed that
the
Self-Monitoring Scale
not
only
does
not
measure
the
construct
of
self-monitoring,
but
also
may not be a
good measure
of
anything
at
all—at
least
anything
that
is
interpretable
or
meaningful.
Thus,
Briggs
et
al.
(1980)
have suggested,
for
instance, that "subjects labeled high
self-monitors
in one
study might
be
different
from
those labeled
high
self-monitors
in
another
study"
(p.
684). And, Lennox
and
Wolfe
(1984)
have stated,
in no
uncertain terms, that "the total
score
on
[the] scale tends
to
defy
interpretation;
it is
impossible
to
determine what
the
scale
as a
whole might
be
measuring"
(p.
1350),
and
that
the
"measure
. . .
exhibits fundamental psy-
chometric
weaknesses"
(p.
1350).
It
may
seem that
any
measure that
not
only
fails
to tap the
latent construct
that
it is
intended
to
tap,
but
also
fails
to tap
anything
else, should
be a
rather worthless tool
for
purposes
of
doing research
to find
associations with important
aspects
of
social
behavior. But,
as we
have
noted,
a
very
large number
of
associations have been
found
to
exist between
the
Self-Monitor-
ing
Scale
and
behavioral criterion variables.
If the
measure
has
"fundamental
psychometric weaknesses,"
why and how
does
it
work
as
well
as it
does? Those
who
have raised concerns about
the
psychometric soundness
of the
measure have provided
one
answer
to
this
question. They claim that
the
measure
has
associ-
ations
in the
wide-ranging domains that
it
does because,
in
fact,
it
measures
not one
person variable but, rather, multiple person
variables
(e.g., Briggs
&
Cheek, 1986).
As a
result,
it is
claimed,
the
measure
has
associations
with
behavioral criterion variables
related
to
each
of the
multiple aspects
of
social behavior
that
the
Self-Monitoring Scale taps (none
of
which,
incidentally,
can
be
identified
as
self-monitoring). Thus, they claim,
the
number
and
range
of
effects
the
Self-Monitoring Scale achieves
are
achieved
illicitly.
What
are we to
make
of
these concerns expressed about
the
Self-Monitoring
Scale?
Does
the
measure have fundamental
psychometric weaknesses?
Does
the
measure achieve
its
effects
illicitly
by
tapping
not a
single entity
but
rather multiple aspects
of
individual
differences
simultaneously?
In the
present
article,
we
address these questions about
the
Self-Monitoring Scale.
Al-
though
we
explicitly
deal with
specific
issues about
the
Self-
Monitoring
Scale,
we
will
implicitly speak also
to
general con-
cerns
in the
development
and use of
measures
of
individual
differences
in
building psychological theory.
It is our
hope that
discussion
of
issues
surrounding
the
Self-Monitoring Scale
will
contribute
meaningfully
to
understanding fundamental issues
in
personality.
Factor
Analytic
Investigations
and
Their Implications
Lennox
and
Wolfe
(1984)
claimed that
if the
Self-Monitoring
Scale
measures something, this something
is
uninterpretable
(and
thus
presumably
unmeaningful)
due to
"fundamental psy-
chometric weaknesses"
(p.
1350). What
are
these purported
fundamental
psychometric weaknesses?
How
would they
render total
scores
on the
measure uninterpretable
and un-
meaningful?
The
major
claims
about
psychometric weaknesses
of
the
Self-Monitoring Scale have been based
on
interpretations
of
factor-analytic
investigations.
It is now
well established
that
factor
analysis
of the
Self-Monitoring Scale yields multiple
factors—generally
three
(e.g.,
Briggs
et
al.,
1980),
although
as
few
as two
(Sparacino
et
al.,
1983)
and as
many
as
four
(Ga-
brenya
&
Arkin,
1980)
have been
reported.
Our own
scree
test
(Cattell,
1966)
of the
eigenvalues
on a
large college sample
(N =
1918)
clearly suggested three reliable factors (Gangestad
&
Sny-
der,
1985b).
Rotated
factor
structures
of
three
factor solutions
(e.g.,
Briggs
et
al.,
1980)
have reliably identified three item-content
clusters.
A first
cluster (which
we
call
expressive
self-control)
concerns
the
ability
to
actively control expressive behavior,
for
example,
"I
would probably make
a
good
actor"
(keyed true)
and
"I can
look anyone
in the eye and
tell
a lie
with
a
straight
face
(if for a
right
end)"
(true).
A
second cluster
(social
stage
presence)
concerns
the
propensity
to
perform
in
social
situa-
tions
and
attract
social
attention
to
oneself,
for
example,
"In a
group
of
people
I am
rarely
the
center
of
attention"
(false)
and
"At a
party
I let
others keep
the
jokes
and
stories
going"
(false).
A
third cluster
(other-directed
self-presentation)
concerns
dis-
playing
what others expect
one to
display
in
social
situations,
for
example,
"I may
deceive people
by
being
friendly
when
I
really
dislike
them"
(true)
and "I
guess
I put on a
show
to im-
press
or
entertain
people"
(true).
How
might factor-analytic studies indicate psychometric
weaknesses
that
threaten
the
coherence
and
meaningfulness
of
the
Self-Monitoring Scale?
Briggs
and
Cheek (1986) have pro-
vided
the
clearest
statement
of the
concerns
that
follow
from
factor-analytic
studies (although similar concerns
are at
least
implicit
in
Briggs
et
al.,
1980;
Cheek
&
Briggs,
1981;
and
Len-
nox
&
Wolfe,
1984).
A
single
scale,
Briggs
and
Cheek
(1986)
asserted, should measure
a
single individual
difference
or
per-
son
variable. Quoting
McNemar
(1946):
"Measurement
im-
plies
that
one
characteristic
at a
time
is
being quantified"
(p.
268).
The
fact
that
factor-analytic
investigations
find
that
multi-
ple
person variables
account
for the
interrelations
between
self-
monitoring items, Briggs
and
Cheek claim, raises
the
strong
possibility
that
the
items
do not
measure
a
single person vari-
able.
What
are the
problems
that
follow
from
use
of
a
measure
that
does
not tap a
single person variable?
First,
the
total
scores
are
rendered
uninterpretable (Briggs
et
al.,
1980;
Lennox
&
Wolfe,
THE
NATURE
OF
SELF-MONITORING
127
1984).
If a
measure
taps
three
nearly orthogonal person vari-
ables, then total scores represent linear composites
of
three
nearly
orthogonal components. Thus,
the
same
total
score
may
represent
for one
person
a
very
different
pattern
of
factor
scores
than
it
represents
for
another person. Why, these authors ask,
should
these
two
persons' responses
be
represented
by the
same
score?
Second,
if
multiple person variables
are
tapped
by a
sin-
gle
measure,
the
relations that
the
measure possesses
with
exter-
nal
criterion variables
may be
attributable
to
different
person
variables; that
is,
they
may be
causally heterogeneous. Thus,
one
factor
may
account
for the
relations
found
in one set of
studies,
a
second
factor
for
another set,
and the
third factor
for
yet
a
third set. Clearly,
if the
relations
a
measure
possesses
are
causally
heterogeneous, they should
not be
systematized
or ex-
plained
in a
theoretical
framework
that
asserts
that
a
single
per-
son
variable accounts
for
them.
We
do not
take issue
with
the
premise that
a
single measure
should,
to the
extent possible, measure
a
single person variable
or
entity.
The
problems
that
ensue when
a
measure taps multi-
ple
person variables
to
approximately equal extent
and
when
the
variance
in a
measure
is not to
some substantial extent
at-
tributable
to a
single person variable,
we
agree,
are
genuine
problems.
We do,
however,
take issue with
the
claim that
factor-
analytic
investigations
have
yet
demonstrated that
the
variance
in
the
Self-Monitoring Scale
is
not,
to any
substantial extent,
attributable
to a
single person variable that
is
conceptually
meaningful.
Indeed,
we
have
pursued
a
line
of
research that
offers
an
alternative perspective
on the
factor-analytic
investiga-
tion
of the
Self-Monitoring
Scale,
one
that postulates
that
a
sin-
gle
person variable does underlie
a
substantial amount
of the
variance
in the
measure.
Interpreting
the
Implications
of
Factor
Analysis
Typically,
factor-analytic investigations have
focused
on
ro-
tated
factor
structures
(e.g.,
Briggs
et
al.,
1980;
Furnham
& Ca-
pon, 1983; Gabrenya
&
Arkin,
1980; Lennox
&
Wolfe,
1984;
Nowack
&
Kammer,
1984;
Riggio
&
Friedman, 1982; Spara-
cino
et
al.,
1983).
We,
however,
consider
it
most
informative
to
begin
with
the
unrelated
factor
structure. For,
if
some
major
portion
of the
variance
in the
full
Self-Monitoring Scale does
reflect
a
single latent person variable
or
entity,
that
entity
should
be
reflected
as a
general
factor,
a
factor
on
which
most
of the
items
positively
load. Given
the
manner
in
which
the
Self-Moni-
toring Scale
was
constructed (items chosen
to tap
coherently
a
hypothesized
latent variable), this general
factor,
if it
exists,
should
naturally
be
reflected
as the first
unrotated
factor
(Rum-
mel,
1970).
The first
unrotated
factor
emerging
from
a
factor
analysis
of
the
Self-Monitoring Scale
is
indeed
a
factor
on
which
most
of
the
items positively
load.
In our own
factor
analysis
of the re-
sponses
of
1,918
college students,
24 of the 25
items
had
positive
loadings
on
this
first
unrotated
factor;
18
of
these
24
items
had
loadings
of
+.15
or
greater (Gangestad
&
Snyder,
1985b).
Clearly,
this
first
unrotated factor must correlate
highly
with
the
total
Self-Monitoring Scale. And,
given
that
the
second
and
third unrotated
factors
are
bipolar (that
is,
some items load pos-
itively,
whereas others load negatively), these factors
do not
cor-
relate
highly
with
the
total Self-Monitoring Scale.
If
there
is
some
single
meaningful
person
variable
to
which
a
substantial
amount
of the
total score variance
is
attributable, then that
variable
should closely correspond
to the first
unrotated
factor.
Of
course,
we
cannot consider
the
fact
that
a
general
factor
exists
as, in and of
itself, justification that
this
general
factor
closely
corresponds
to a
single
meaningful
person variable.
In-
deed, others
who
have conducted factor-analytic studies
may al-
ready
have
dismissed,
even
if
implicitly,
the
general
factor's
pos-
sible
meaningfulness.
They
may
simply have
felt
it
justified,
on
a
priori grounds,
to
regard factors
rotated
to
simple structure
as the
interpretable,
meaningful,
and
"real"
sources
of
variation
underlying
the
Self-Monitoring Scale items. Accordingly, they
may
have dismissed
the
unrotated
factor
structure
as an
arbi-
trary,
and
thus psychologically meaningless, result
of the
real
coherences (i.e., rotated factors)
in the
self-monitoring items.
In
fact,
justifications
for
regarding rotated factors
as
real
and
unrotated ones
as
arbitrary
have been proposed.
It is
often
as-
sumed (e.g., Cattell,
1978)
that,
if a set of
overt variables
has
multiple
causative
influences,
then
any one
underlying
influ-
ence substantially
affects
only
a
portion
of
them.
It is
also
often
assumed
(e.g.,
Thurstone, 1947) that
any one
overt variable
is
likely
to be
influenced
substantially
by
only
one
real underlying
variable.
To the
extent that these assumptions
are
correct,
the
rotated
factor
structure
will
indeed reveal
the
real underlying
sources
of
variation.
Whether
or not one
accepts these assumptions
as
generally
appropriate,
one
must recognize
inferential
risks
associated
with
automatically interpreting
the
rotated factor structure
as
that containing
the
real,
nonarbitrary
underlying sources
of
variance.
In
fact,
leading
experts, including many proponents
of
simple structure rotation (e.g., Cattell,
1978;Eysenck,
1950;
Guilford
&
Zimmerman,
1963;
Overall,
1964),
have
discussed
these
risks.
In the
present context, CattelPs
(1978)
remarks
are
most
instructive.
He has
noted that
the
assumption
that
any
real
source
of
influence
substantially
affects
only
a
subset
of the
overt
variables
is not
likely
to
hold
if the
variables
one
factor
analyzes
are not
expected
to be
widely
heterogeneous. Thus,
simple
structure
requires
a
foresighted
choice
of the
sample
of
vari-
ables.
It is
obviously
absurd
to
expect
to get any
determinate
rota-
tion
of a
certain
factor
X if we
have
chosen
the
variables
in the
study
such
that
probably
all of
them
will
have
some
significant
loading
on X.
(Cattell,
1978,
p.
112)
This statement
has
obvious implications
for the
present
case.
The
Self-Monitoring
Scale, like
any
other rationally derived
and
empirically
refined
measure (e.g., Nunnally, 1967),
was
con-
structed
to
meet precisely these conditions
(Snyder,
1974).
Thus,
it is
true that
the
existence
of a
factor
with
general
posi-
tive
loadings does
not in
itself vindicate
the
meaningfulness
of
a
general
factor
underlying
responses
to the
Self-Monitoring
Scale.
However,
it is
also true that
the
lack
of a
general factor
after
rotation
to
simple
structure does
not
justify
the
conclusion
that
a
general
factor
is
unmeaningful.
To
further
address
the
possibility that
a
single
entity
can
account
for a
substantial por-
tion
of the
variance
in
self-monitoring scores,
we
next examine
relations
of the
Self-Monitoring Scale with external social
be-
havioral
variables.
External
Relations
of the
Measure
and the
Factors
A
major
concern about
the
possibility that
the
Self-Monitor-
ing
Scale measures multiple individual
differences
is the
ensu-
128
MARK
SNYDER
AND
STEVE
GANGESTAD
ing
possibility that
the
associations observed
in
self-monitoring
studies
are
causally heterogeneous.
That
is,
perhaps
some
asso-
ciations
are
attributable
to one
factor, others
to a
second
factor,
and
still others
to a
third
factor.
There
do
exist some indications
in
the
literature
of
differential
effects
of the
factors (e.g., Cheek,
1982;
Penner
&
Wymer,
1983;
Sypher
&
Sypher,
1983),
and we
will
discuss
these studies
later.
Let us
now, however, consider
a
set of
15
studies that
we
have reanalyzed
by
means
of the
factors,
as
well
as the
full
measure,
in
light
of
expressed concerns. These
studies cover
a
wide range
of
social phenomena,
and
include
studies
within
the
cognitive, behavioral,
and
interpersonal
do-
mains (see Table
1).
Was
it the
case
that
the
effects
in
these studies were
due to
one or
other factor? Clearly not.
In
general,
the
factors tended
to
be
associated similarly with
the
criterion variables. Moreover,
almost
as a
rule,
the
full
Self-Monitoring Scale outperformed
all of the
factors—in
spite
of the
fact
that
each
of the
factors
possesses internal consistency similar
to
that
of the
full
measure
(Briggs
et
al.,
1980).
In
only
4 out of 19
comparisons
in
these
15
studies
was any
factor more related
to the
criterion variable
than
the
full
scale.
1
Clearly
then,
these
studies,
at
least,
are
very
consistent
with
the
conjecture
that
the
general self-monitoring
factor
that
can be
found
in the
factor
space,
and
that
is
highly
related
to
full
scale scores,
accounts
for
many associations
the
scale possesses. Indeed,
the
most parsimonious explanation
of
these studies
is
that
it is the
general
factor
that
makes
the
scale
work.
Nevertheless, parsimony
aside,
an
alternative must
be
consid-
ered.
Perhaps
the
rotated
factors
are the
"real"
person variables
underlying
the
total measure,
and
each
of the
factors
has
inde-
pendent
and
additive
effects
on the
criterion
variables
in the
studies
we
examined.
This
may
seem
an
unlikely
possibility,
given
that
it
requires multiple parallel
effects
of
what
are
claimed
to be
very distinct
and
different
person variables;
but
it is a
possibility.
We now
turn
to
consider
further
evidence that
the
general factor
is a
meaningful
underlying causal variable.
Nature
of
the
General
Factor
Recently,
we
reported
extensive
taxometric
analyses
of the
internal structure
of the
self-monitoring items; these analyses
suggested that
a
major source
of
variation
in the
responses
to
the
Self-Monitoring
Scale
is a
discrete
or
quasi-discrete entity,
as
opposed
to a
normally distributed continuous variable
(Gangestad
&
Snyder,
1985b).
This postulated causal entity
ac-
counts
for
over
one
half
of the
reliable variance
in
full
Self-Mon-
itoring Scale
scores,
and in
fact
appears
to be
reflected
as the
first
unrotated factor
in a
factor
analysis
of the
items. Further-
more,
reanalyses
of a
sample
of
self-monitoring studies
in
light
of
these
findings
provides evidence
that
much
of the
criterion-
oriented
validity
of the
Self-Monitoring Scale
is
attributable
to
this
causal entity (Gangestad
&
Snyder, 1985b). Indeed,
a
mea-
sure
of the
causal entity performed,
on
average,
as
well
in
these
studies
as did the
full
Self-Monitoring Scale itself
and
better
than each
of the
rotated factors.
In
essence,
we
claimed
that
what
in
large
part
makes
the
Self-Monitoring
Scale
work
is the
existence
of
strong
effects
of
this latent causal entity that
is re-
flected in the
same region
of the
factor space
as the
general
self-
monitoring factor.
Of
course,
it is
natural
to ask
what this postulated discrete
causal entity might
be.
This question
has
been addressed
in
fur-
ther speculation
and
investigation. Indeed, evidence
is now
mounting
that
the
causal variable corresponding
to the
general
factor
has a
biological genetic basis
(Dworkin,
1977; Gangestad,
1984;
Gangestad
&
Snyder,
1985a).
Specifically,
it has
been esti-
mated that monozygotic (MZ) twins
are
nearly
always,
if not
always,
concordant
on
this latent factor, whereas dizygotic (DZ)
twins
are
concordant
at
better than
a
chance
rate,
but at a
rate
substantially less than
MZ
twins.
Presently,
we
wish
to
present
further
data
from
this twin
study relevant
to
assess whether
the
Self-Monitoring
Scale
taps
some
single person variable.
Let us
begin with
a
consideration
of
the
intraclass
correlations
between
MZ
twins (149 pairs)
on
each
of the
rotated factors. These correlations
reflect
the
opera-
tion
of any
influences (environmental, genetic,
or
combina-
tional)
that create greater similarity between
MZ
twins than
between
randomly paired individuals.
The MZ
correlations
of
.60, .56,
and .34 for the
factors
of
expressive self-control, social
stage
presence,
and
other-directed self-presentation, respec-
tively,
indicate that
in
fact,
for
each
of
these factors,
there
do
exist
these
influences.
But
now,
let us
further
question
the
nature
and
relation
of
these influences.
On the one
hand,
it is
possible
that
these
in-
fluences
underlying
the
three rotated factors
are
wholly
separate
and
distinct.
If
the
rotated
factors
do not
have
a
common under-
lying
substrate,
then
it
will
be the
case
that
these genetic
and
environmental
influences
are
distinct.
Alternatively, however,
it
is
possible
that
the
genetic
or
environmental influences
on
these
factors
are
overlapping
to a
substantial
degree.
That
is,
the ro-
tated factors
may
share, perhaps
even
to a
substantial extent,
a
common genetic
or
environmental influence.
In
fact,
we can
assess these possibilities.
If the
causal influences
on
rotated
fac-
tors
are
distinct,
then
the
similarity
of MZ
twins
on an
additive
composite
of the
three factors
will
be
less than
that
for the
fac-
tors themselves.
If, on the
other hand,
the
factors
share
some
causal
influence,
the
similarity
of MZ
twins
on a
composite
of
the
three
factors
may
exceed that
for
the
factors themselves.
Did
1
Of
course,
given
sampling
variability,
we
should
expect
on
occasion
one or
other
factor
to
outperform
the
full
measure,
even
if the
actual
effect
size
in the
population
was
always
greatest
for the
full
measure.
In
this
light,
it is
interesting
to
note
that
three
of the
four
studies
in
which
a
factor
outperformed
the
full
measure
were
studies
with
the
smallest
sample
sizes,
and
thus
studies
for
which
sampling
variability
should
be
greatest.
We may
also
note
that
for two of the
same
four
comparisons,
a
second
comparison
in the
same
study
demonstrated
greater
effect
size
for
the
full
measure
than
for the
factors.
We
also
may
note
that
we do not
expect
all of the
factors
to
perform
identically
well.
The
three
rotated
factors
all
correlate
positively
with
the
general
factor,
but not
equally:
expressive
self-control,
r = .6;
social
stage
presence,
r = .5;
other-directed
self-presentation,
r=
.3
(Gangestad
&
Snyder,
1985b).
The
relative
ability
of the
factors
to
predict
the
crite-
rion
variables
in the
studies
we
reanalyzed
do, in
fact,
correspond
to
their
relative
correlations
with
the
general
factor.
Although
samples
for
several
of the
studies
we
reanalyzed
were
se-
lected
for
extreme
scores
on the
Self-Monitoring
Scale,
a
procedure
that
generates
within
these
samples
increased
correlation
between
individ-
ual
factors,
samples
in six
studies
were
unselected
and in
another
two
studies
represented the
full
range
of
self-monitoring
scores.
For 9 out
of
10
comparisons
in
these
studies,
the
full
measure
outperformed
all
factors.
THE
NATURE
OF
SELF-MONITORING
129
Table
1
Assessment
of
the
General
Self-Monitoring
Factor
Comparison
of
the
Full
Self-Monitoring
Scale
With
the
Rotated
Factors
Investigation
Full
Self-
Monitoring
Scale
Social
stage
presence
factor
Other-directed
self-
presentation
factor
Expressive
self-
control
factor
Snyder&
Cantor
(1980)
Investigation
2
mF(4,
53)
^1,56)
Snyder
&
Gangestad
(1982)
Investigation
I:F([,
117)
Investigation
2
2.83
8.29
11.49
2.44
0.01
4.90
1.36
1.10
6.70
1.03
0.51
3.91
F(l,ao)
F([,
oo
)
Snyder
&Kendzierski(1982b)
Invesligation
l:P(l,
oo)
Snyder,
Gangestad,
&
Simpson
(1983)
Investigation
1:((28)
Investigation
2:
((58)
Snyder
&
Simpson
(1984)
Investigation
1:((30)
Investigation
2
<(28)
«28)
Investigation
3
«76)
((80)
Investigation
4:
((253)
Snyder,
Simpson,
&
Gangestad
(1986)
((143)
Snyder,
Berscheid,
&
Click
(1985)
Investigation
1:
((37)
Investigation
2:
((30)
Snyder
&
Smith
(1985)
Investigation
l:F(l,
198)
Investigation
2:
F{
1
,
70)
5.89
4.24
11.90
3.93
2.53
7.16
2.92
2.61
2.28
2.44
2.60
6.03
2.98
2.68
26.31
8.83
2.61
6.09
8.51
1.47
0.46
3.18
2.70
3.08
1.63
2.02
1.62
3.05
1.98
3.61
12.16
0.17
2.61
0.00
5.89
2.93
0.95
3.61
1.33
1.66
1.10
1.03
0.48
2.82
2.90
1.98
14.77
1.60
3.91
2.30
5.21
3.98
1.83
4.83
.73
.88
.63
.84
.63
5.13
1.38
0.72
10.28
1.23
Note.
All
effects
within
a
single
study
are in the
same
direction.
the
similarity
of
twins
on the
general
factor
(which
is a
weighted
composite
of
items
from
all
three rotated
factors)
exceed that
of
the
rotated
factors?
Indeed,
it
did.
The
intraclass
correlation
for
this general
factor
was
.64,
which
is
significantly
larger than
those
of the
rotated
factors
(all
/s
> 2, p <
.05; Gangestad,
1985).
2
What
do
these
analyses,
then,
show? They show
that
the
ro-
tated
factors
do
indeed appear
to
share some common genetic
influence.
They
are not
etiologically distinct. Moreover, this
common
genetic
factor
appears
to
correspond
closely
to a
gen-
eral factor
that
is the first
unrelated
factor.
Given
that
this
gen-
eral
factor
generally
outperformed
all
three
rotated
factors
in
our
reanalyses
of
15
self-monitoring studies,
it is our
view
that
the
causal
influence
contributing variance
to all
three factors
is
largely
responsible
for
the
relations
we
observed
in
these
studies.
We
believe that
the
evidence
not
only invites,
but
also encour-
ages,
one to
adopt this
view.
In any
case,
however,
surely
the
evidence
does
not now
permit
one to
dismiss this
view.
Links
Between
the
Factors
and
Other
Variables
We
have
claimed that more than
one
half
of the
reliable vari-
ance
in the
full
Self-Monitoring
Scale
is
attributable
to a
single
causal
influence
and
that
if our
reanalyses
are any
indication,
a
sizeable number
of the
associations
the
measure possesses
are
attributable
to
this
factor.
Still, although
we
believe
the
variance
attributable
to
this
single
latent variable
is
substantial, perhaps
40%
of the
reliable variance
in the
measure
is not
attributable
to
this variable (Gangestad
&
Snyder,
1985b).
What
are the im-
plications
of
this
fact?
Even
though
one may
conclude,
as we do,
that
all
three
ro-
tated
factors
possess
variance
attributable
to a
common
influ-
ence,
one
should also recognize that
the
rotated
factors
do
emerge,
and
that they emerge precisely because
the
items
on
the
Self-Monitoring Scale share systematic variance with each
other
that
is not
attributable
to
this
common
influence (see
2
A
factor
analysis
of the
monozygotic
(MZ)
cross-twin
correlations
(which
reflect
covariance
due
only
to
family
influences,
environmental
or
genetic)
yielded
two
factors.
The first
factor
closely
resembled
the
first
unrelated
factor
of the
large
college
sample
(factor
congruence
=
.91).
Furthermore,
varimax
rotation
to
simple
structure
did not
change
substantially
the
factor
structure.
Indeed,
the first
rotated
factor
also
closely
resembled
the first
unrelated
factor
of the
large
college
sample
(factor
congruence
=
.91).
This
result
shows
that
by
varying
the
influ-
ences
one
allows
to
contribute
to
covariation,
one
changes
the
factor
structure.
When
only
family
influences
are
allowed,
the
general
factor
emerges
in the
rotated
as
well
as
unrelated
factor
structures.
The
fact
that
the
dizygotic
(DZ)
twin
intraclass
correlation
on the
general
factor
was
only
.19
indicates
that
the
family
influence
is
substantially
genetic
in
nature.