LINGUIST List 5.917

Thu 25 Aug 1994

Sum: Factors in personality traits

Editor for this issue: <>

Directory

1. "Rianne Doeleman", summary "factors in personality traits"

Here is my summary based on the answers to my question, posted in
july, which concerned factor analysis of personality trait
descriptions, especially on differences and similarities between:
1. the "Big Five" i.e., five factors called: I surgency (also called
extraversion); II agreeableness; III conscientiousness; IV emotional
stability; and V culture (or intellect) (cf. Peabody & Goldberg
1989);
2. "Osgood's three" I evaluation; II activity; and III potency
(Osgood, Succi & Tannenbaum 1957);
3. "Sociolinguistic factors" I status (or competence) and II
solidarity (or benevolence); with solidarity sometimes divided in II
social attractiveness and III personal integrity (Ryan 1973, Giles &
Robinson 1990).

I want to thank all of the people that have responded:
Achim Grabowski: grabclarity.Princeton.EDU
Alice Horning: HORNINGargo.acs.oakland.edu
Carol W. Slater: CSLATERalma.edu
Danny Rouckhout: 100277.1755compuserve.com
Ines Shaw: ISHAWVM1.NoDak.EDU
Jacques Steyn: STEYNJalpha.unisa.ac.za
Jean-Marc Dewaele: jmdewaelvnet3.vub.ac.be

The first few responses were negative and suggested that these three
types of research have nothing in common except for the fact that
they all use factor analysis. A phone call to M. Verkuyten, social
psychologist at the University of Utrecht, resulted in the same
viewpoint: the big five factors are underlying personality
characteristics and have nothing in common with Osgood's three. The
latter are underlying evaluative attitude toward whatever object.
These responses are best summarized in what Achim Grabowski wrote:
"The big five are personality traits, i.e. constructs that are
designed to describe latent behavioral sources in an individual. But
Osgood's three factors are factors that describe the best solution
for a similarity matrix of words! And the third-mentioned analyses
concern factors of interpersonal relationships. Thus, factor analysis
is a good tool to simplify the complex information that is included
in correlation or similarity matrices, but it does not include
theoretical aspects of personality. ... Thus, it seems to me
perfectly clear that the lines of research you mentioned in your
query are not compatible. And you should probable not try to make
them compatible."
But, I also received some enthusiastic, positive responses. Although
nobody could name an article on the comparison of the three lines of
research, those responses did give some useful references and
insights into the matter.
In Carol Slater's point of view the shift of focus (that is, the
objects under study) results into the differences in factors. "The
Osgood dimensions are what emerge when all kind of objects are
considered: people, rocks, colors, the lot. They provide an
allpurpose space in which to place all and anypotential objects of
attitudes and were, indeed, appropriated early on by people who were
interested specifically in attitude formation and change. ... So,
given that what you get out of factor analysis depends on what
you put in, it is not surprising that when we constrain the domain to
descriptions of people and nothing else, things look somewhat
different, we get the Big Five (plus or minus two), instead. When we
shift focus again, and look primarily at characteristics of people's
social interaction, in particular, we get yet another set of
dimensions, the status/solidarity configurations."

I was not quite satisfied yet with Slater's viewpoint and I continued
by thoroughly reading three studies representing the three different
types of research I wanted to relate: (1) Peabody & Goldberg 1989:
research on the Big Five, factors underlying personality ratings as
found in implicit personality theories on the structure of personality
trait descriptions; (2) Osgood, Succi & Tannenbaum 1957: the
invariant universal factor structure of the explicit personality
theory of Osgood e.a; and (3) Ryan 1973: the sociolinguistic
dimensions found to be important in language attitude studies.

Most helpful in understanding and comparing all this was the book
suggested by Danny Rouckhout: Susan Hampson's introduction on the
construction of personality (Hampson 1988). Hampson's book is
organized around three different perspectives on personality, which
have generated three kinds of personality theory: (1) explicit
theories of personality from the personality psychologists'
perspective; (2) implicit theories of personality from the lay
person's perspective; and (3) the perspective from the standpoint of
the self which is concerned with the theories people have about their
own personalities. I was particularly interested in the second part of
this book which is concerned with the lay's perspective.
Hampson relates Osgood and the Big Five and concludes: "In implicit
personality theory, as with explicit personality theory, the number
and nature of the underlying dimensions is dependent upon the level
of detail favoured by the theorist." (p. 118). Metaphorically she
describes it as: "...when the microscope is low-powered, then
evaluation stands out as the most salient dimension. However, when the
power is turned up, then a more complex structure is revealed." (p.
119). Hampson's conclusion is confirmed by Van der Pligt & De Vries,
who describe the dimensions found by Osgood as an abstraction at a
higher dimension of the Big Five (Van der Pligt & De Vries 1991;160-
161).
In comparing implicit and explicit personality theories Hampson (p.
156-157) concludes that most studies in explicit personality theory
come up with five underlying factors of personality ratings. Osgood's
three factors were sometimes found in personality structures derived
from trait ratings viewed as a measure of people's implicit
personality theories. The first dimension (evaluation) was clearly
present, but the second and third (potency and activity) often merged.
However, Hampson argues that evaluation alone is not enough to
describe adequately the structure of implicit personality theory,
since this ignores all the variation in the descriptive content of
the "trait language" (i.e., words used to describe personality traits
R.D.). When the structure of the descriptive content is explored the
same five-factor structure as that found for explicit personality
theory is commonly found.

If both the first and second line of research actually deal with the
same underlying dimensions, why do different studies find different
factors or give the dimensions different names?
Hampson mentions three determinants of variation in personality
structures in different studies: (1) the original item selection
(e.g., Factor IV - emotional stability, has assumed greater
significance than it otherwise would because psychologists have a
professional interest in this area of personality and have tended to
add more scales to measure it than the natural language of personality
description would warrant); (2) the selection of the targets for
personality description (e.g., if only the self and liked targets are
used, then the resulting factors tend to be smaller, i.e. account for
less variance); (3) the effect of raters' categorization style (e.g.,
broad categorizers' ratings yield a few broad factors, like those
found in semantic-similarity studies, narrow categorizers are more
discriminating and their ratings produced more complex structures,
characteristic of the rating studies of real people). I think there is
at least one other important determinant: (4) the habit or tactic of
researchers to base their work on that of others. Most researchers
chose the adjectives on the rating scales based on the results of
earlier studies and/or theories, e.g., when a researcher chooses to
follow a personality theory s/he will use (and therefore find in the
analysis) something like the Big Five, but if s/he bases the study on
social psychological attitude research something like Osgoods Three
will come up.
Hampson concludes (p. 159-160): "Comparisons of personality structures
obtained from different data sets show that the five factors
identified by Norman (1963) remain the most useful framework (Goldberg
1981b). The Big Five (Factor I - surgency, Factor II - agreeableness,
Factor III - conscientiousness, Factor IV - emotional stability, and
Factor V - culture) are the typical number of factors, not all of
which are found in every data set. The Osgood factors of potency and
activity combine to form factor I - surgency. Factors I, II and III
are by far the largest and most robust of the Big Five, whereas
Factors IV and V are more controversial."

How to relate all this to speaker evaluation?
First of all it has become clear from both Osgood, Succi and
Tannenbaum (1957) and Hampson (1988) that Osgood's Three are NOT
likely to be found in speaker evaluation. In Osgood e.a. (1957) we
find factor analysis of ratings of all kind of different objects (from
concrete inanimate objects as "knife" and "snow" to abstract objects
like "sin" and "symphony", as well as animate objects like "me",
"Adlai Stevenson" - a politician well known at the time, or "mother").
Osgood e.a. try to find a way to measure the meaning of objects. They
describe a universal semantic space in which the meaning of all
objects can be placed. In this semantic space they discover three
major dimensions which are able to explain most of the variance for
most objects. Initially Osgood e.a. distinguish five factors: I
evaluation, II potency, III activity, IV stability, V receptivity. The
fourth and fifth factor are left out in a later stage, because they
add too little to the explanation of variance. If we do not average
the results for all objects, but separate the human objects from the
rest we find (extracted from Osgood e.a. 1957;table 30 p. 181-185, V=
percentage of explained variance):
"Foreigner": FI V=.102, FII V=.052 FIII n.m. FIV V=.038 FV V=.051
"Mother": FI V=.185, FII V=.034 FIII V=.039 FIV V=.038 FV V=.039
"Me": FI V=.075, FII V=.053 FIII V=.086 FIV V=.044 FV V=.072
"Adlai Stevenson": FI V=.253, FII V=.044 FIII n.m. FIV n.m. FV n.m.

In this part of the picture Factor I still is the most explaining
factor (except for the object "me") but the explained variance
registered for the other factors does not support the selection of the
first three factors as defining semantic space and leaving out the
others. So even though Osgood e.a. are probably right in defining the
semantic space for meaning in general to be based on the three
dimensions, their analysis is not sufficient in the measurement of
personality ratings of human objects. The first factor still stands
out. However just evaluation is not sufficient to adequately describe
the lay's perspective on personality (see above in reviewing Hampson
1988).
Personally, I think Hampson's argumentation on the importance of the
Big Five and the usefulness of these factors as a framework for the
analysis of lay person's ratings of personality is rather convincing.
It is certainly more convincing than Osgood e.a.'s argumentation on
the selection of their three factors, especially in the field of
personality ratings.

Ryan (1973) based her idea of two factors in attitude measurement on
Brown's discussion of forms of address (Brown 1965). According to Ryan
(1973, p. 68) the social psychologist Brown "demonstrated the roles
played by status (relative degree to which a person possesses the
characteristics valued by the society as a whole) and solidarity
(degree to which a person is similar to the perceiver, in terms of
group membership, shared experiences, age, socioeconomic class etc.)."
In other words, people judge the personality characteristics of others
by group values and personal similarities. I did not find any
literature on the relation between these sociolinguistic factors and
the Big Five or Osgood. However, personally I think that a person's
Social Economic Status (SES) is very important in the ideas that
society has about her/him. This SES is related to adjectives loading
on the Big Five factors I Surgency, III Conscientiousness and V
Culture (or intellect). The factor II Agreeableness and IV emotional
stability include adjectives that seem to refer to more personal ideas
and feelings, which would place them with the sociolinguistioc factor
II Solidarity. Osgood's Evaluation factor seems to me to be comparable
to the sociolinguistic factor I Status, the group evaluation. Most
studies in which personality ratings based on speech were factor
analysed came up with two or three factors. Not all five of the Big
Five were found; there are (at least) three possible explanations.
(1) As both Hampson (1988) and Peabody & Goldberg (1989) indicated,
the first three factors of the Big Five are the most robust ones, the
fourth and fifth are more controversal, therefore the sociolinguistic
factors may be comparable to the "Big Three". (2) The big Five were
found in lay person's ratings of personality in all kinds of different
settings, e.g., in rating personality of stereotypes, of persons from
memory, of persons the raters had only recently met, of persons whom
the raters had known for ages, etc. It does not seem illogical that
the three robust factors also emerge in lay person's ratings of
personality of speakers on the basis of speech. (3) In most studies
the personality characteristics that are to be rated are chosen
because of good results in earlier research. That is, the dimensions
which are thought to be important because they emerged in earlier
studies are represented by a couple of scales. If factor analysis of
the result comes up with those same dimensions that does not support
the idea of those being the important underlying dimensions. It proves
only that the adjectives used to represent the personality
characteristics on the rating scales are good representations of the
dimensions.

Finally, I will present the measurement instrument which I want to
use to investigate the personality ratings of speakers based on
speech, also called speaker evaluations. I would like to get comments
from those of you who took the trouble of reading this summary, on the
ideas presented in the summary and especially on the resulting scaling
instrument.
Peabody and Goldberg (1989) described the adjectives which turned out
to represent the Big Five factors in their study. Ideally one would
like to put a large sample of these adjectives in the scaling
instrument which will be used to collect the speaker evaluations. At
the same time the three factors which are often used in language
attitude research must be properly represented. However, all this
would result in a rating instrument with very many adjective scales,
which would take to much time and concentration of the raters. A first
step in minimizing the number of scales is to represent only the
robust factors of the Big Five, which were: I Surgency, II
Agreeableness, and III Conscientiousness. Secondly, for the same
reason, only the one robust factor of Osgood is included: I
Evaluation. All three factors from sociolinguistics, after all that is
my field of interest, are also present: I Status, II Social
attractiveness, III Personal integrity.

In the rating experiments all speakers will be rated for personality
on bipolar seven point scales with the following adjectives at both
ends (I wrote down both the Dutch and the English terms because my
little English vocabulary could not give a perfect translation of the
Dutch terms):

hoog opgeleid - laag opgeleid (high educated - low educated)
rijk - arm (rich - poor)
wilskrachtig - slap (forceful - submissive/weak)
hard werkend - lui (hard wordking - lazy)
vriendelijk - onvriendelijk (friendly - unfriendly)
goedlachs - zuur (good natured - irritable)
aantrekkelijk - onaantrekkelijk (attractive - unattractive)
rechtvaardig - onrechtvaardig (fair - unfair)
betrouwbaar - onbetrouwbaar (trustworthy - untrustworthy)
zachtaardig - agressief (gentle - aggressive)

The three robust factors from "Big Five" are represented:
I Surgency or Extraversion: forcefull - submissive, and gentle -
agressive; II Agreeableness: friendly - unfriendly, and good natured -
 irritable; III Conscientiousness: trustworthy - untrustworthy, and
hard working - lazy.

The robust factor Evaluation of "Osgood's three" is represented by:
friendly - unfriendly, rich - poor, and trustworthy - untrustworthy.

The factors of sociolinguistic language attitude and speaker
evaluation research are represented:
I Status: rich - poor, high education - low education, hard working -
lazy; II Social attractiveness: friendly - unfriendly, goodnatured vs
irritable, attractive - unattractive; III Personal integrity: fair -
unfair, trustworthy - untrustworthy, gentle - aggressive.

Please send any comments, remarks and critisism to:
RIANNE.DOELMANKUB.NL
I hope reading from you soon.

Rianne Doeleman
Research group on Language and Minorities - Tilburg University
P.O. box 90153 - 5000 LE Tilburg - The Netherlands
telephone: +3113-663123

Brown, R. (1965) Social Psychology. New York: Free Press.
Giles, H. & Robinson, W.P. (Eds.) (1990). Handbook of language and
social psychology. Chichester: John Wiley & Sons.
Hampson, S.E. (1988). The construction of personality: An
introduction. London: Routledge.
Osgood, C.E., Suci, G.J. and Tannenbaum, P.H. (1957). The measurement
of meaning. Urbana, IL: University of Illinois Press.
Peabody, D. & Goldberg, L.R. (1989) Some determinants of factor
structures from personality-trait descriptors. Journal of
Personality and Social Psychology, 57, 3, 552-567.
Pligt, J. van der & Vries, N.K. de (1991). Cognitieve sociale
psychologie. Meppel: Boom.
Ryan, E.B. (1973). Subjective reactions toward accented speech. In:
R.W. Fasold & R.W. Shuy (Eds.), Analyzing variation in language.
p. 60-73. Washington: Georgetown University Press.