Advances in Consumer Research Volume 15, 1988
MEASURING THE MEANING OF CONSUMPTION OBJECTS:
AN EMPIRICAL INVESTIGATION
Robert E. Kleine, III, University of Cincinnati
Jerome B. Kernan, University of Cincinnati
Since Holbrook and Hirschman's (1982) elucidation of the experiential
perspective, the "meaning" of consumption objects
has been accorded renewed prominence in consumer research. Surprisingly
few studies explicate meaning as a scientific construct, however.
Using ordinary consumption objects, this paper proposes a definition
of meaning and a measure MOCOM -- derived from that definition.
An initial test of MOCOM (Measure Of Consumption Object Meaning)
suggests that it elicits comparatively rich qualitative data
and that it also has desirable psychometric properties.
The proposition that consumers respond according to the meanings
hey ascribe to marketplace stimuli was, until Holbrook and Hirschman's
(1982) elucidation of the experiential perspective, accepted
virtually without question. As a consequence, only a few consumer
researchers (e.g., Csikszentimihali and Rochberg-Halton 1981;
Douglas and Isherwood 1979; Friedman 1986; Hirschman 1980; Kernan
and Sommers 1967; Levy 1959; 1981; McCracken 1986; Mick 1986)
have even discussed, much less studied meaning. This suggests
that (1) there is a gap in appropriate theory and methodology
and/or (2) that the influence of meaning is thought to be so
obvious that little is to be gained from its study.
To the extent that we accept the notion that objects have subjective
meanings, we must simultaneously reject the account that meaning's
influence on consumers is too obvious to study. As though by
default, then, it appears that consumer researchers have not
explicated meaning as a scientific construct because they have
lacKed the requisite theory and methodology. Before meaning
can assume its place 'at the nucleus of consumer behavior"
(Mick 1986), it must be formalized as a construct that reflects
the key characteristics of consumption phenomena, i.e., one
that incorporates people's subjective reactions to consumption
experiences, those stimulated by objects and/or behaviors. This
requires that we overcome the limitations of past efforts, which
typically have yielded either rich qualitative content (e.g.,
anthropological and psychoanalytic approaches) or well-behaved
quantitative representations (e.g., semantic differentiation)
and develop a procedure that simultaneously provides both these
This paper is an initial attempt to formalize meaning as a
scientific construct. We propose a definition of consumption-object
meaning and a measure based on that definition. We also offer
the results of an empirical study as initial evidence of our
conceptualization s validity.
TOWARD A DEFINITION OF OBJECT MEANING
Kleine and Kernan (1987) have outlined the multidisciplinary
traditions through which the meaning construct has been studied.
Although they identified two dimensions -- contextual sensitivity
of meaning and level of analysis -- that distinguish the research
traditions reviewed, they fail to specify which research tradition
(or combination) should serve as the foundation for developing
the meaning construct in a consumption context. Ogden and Richards'
(1953) "triangle of reference", however, explicates
a key assumption common to most researchers, regardless of tradition:
the relation between a referent, the perceived stimulus (e.g.,
an automobile), and a symbol, the result of the referent's interpretation
(e.g., "that's a Porsche 924S"), is an indirect one.
Referent and symbol are related through thought (or reference3.
Meaning is not inherent in the referent; it is ascribed to it
by the interpreter.
Yet Ogden and Richards' triangle raises two questions important
in the development of a definition of meaning. First, who is
doing the thought (or reference); i.e., what is the level of
analysis? Kleine and Kernan suggest that most researchers determine
this by their trainingCthose of a psychological and/or social-psychological
tradition tend to prefer a micro (i.e., individual) level of
analysis, while those trained in the anthropological tradition
prefer a macro (e.g., cultural system) level of analysis. Although
this paper adopts the individual level of analysis, it recognizes
the embeddedness of the individual within a social world. The
second question raised by the triangle is what type of referent
does the researcher wish to examine; i.e., what is the unit
of analysis? Although objects and behaviors are paramount in
the experiential perspective, the present research emphasizes
the consumption object as the analytic unit, for reasons of
tractability. This combination of individual levels of analysis
and consumption-object stimuli influences our definition of
meaning and, consequently, the type of measure proposed. However,
it presents us with an opportunity to fill an important gap
in the literature. Most meaning research uses words as the focal
stimuli. The meaning individuals ascribe to objects is perforce
an under-researched phenomenon.
Our definition of object meaning incorporates three essential
characteristics: (1) polysemy; (2) contextual sensitivity; and
(3) consensus. Polysemy refers to the fact that a given object
can mean many things -- baking soda, for example, can be a refrigerator
deodorizer, a dentifrice or an antacid. Contextual sensitivity
suggests that the meaning of a turkey on a Thanksgiving Day
dinner table probably differs from that of a turkey placed on
a dinner table during mid-May. And consensus refers to the fact
that, even though each person holds idiosyncratic information
about an object, some minimal amount of object information (meaning)
must be shared by people in order for them to communicate about
the object. These characteristics, plus the forgoing discussion,
locate our conceptualization of meaning in what Kleine and Kernan
call a social-psychological research tradition. Apropos of that
tradition is Szalay and Deese's (1978) characterization of meaning
as an individual's subjective reaction to a stimulus, composed
of certain salient elements. Although they propose no a priori
structure regarding these elements (preferring instead to rely
on post hoc "semantic clustering"), Kernan and Sommers
(1967) advance two orthogonal dimensions of an object's meaning
-- its attributes and performance potential. Attribute refers
to the physical dimension of meaning, to one's perception of
an object's palpable characteristics (baking soda is white and
powdery). Performance refers to the functional dimension of
meaning, to one's perception of the object's action potential
or what the object can do or be used for (a dentifrice or an
antacid, in the case of baking soda). The Szalay and Deese and
Kernan and Sommers definitions of meaning are complementary.
Both define meaning as an aggregate perception. Szalay and Deese
suggest that this perception consists of a constellation of
"reactions" that vary in salience, while Kernan and
Sommers introduce the possibility that these "reactions"
are manifestations of attribute and performance. Thus we propose
that the meaning to an individual of a consumption object is
that person's aggregate perception of the object. One's perception,
in turn, consists of two dimensions: an interpretation of the
object's physical attributes and of its action potential. These
perceptual dimensions vary in salience among objects and individuals
and according to the context in which the object is perceived.
A MEASURE OF CONSUMPTION OBJECT MEANING
Based on the preceding discussion, we can identify several
properties a measure of consumption-object meaning should possess.
It should: (1) adduce the attribute and performance dimensions
ascribed to the object; (2) determine the salience of those
dimensions; and (3) distinguish meaning that is shared from
that which is idiosyncratic. Additionally, the measure should
be amenable to large samples.
The method of continued associations (Szalay and Deese 1978)
provides a foundation from which to build such a measure. The
method requires subjects to produce all the one-word stimulus-bound
responses they can muster within a 60-second interval. These
responses are the salient elements of the object's meaning.
Since a subject's first responses are assumed to be more dominant
(i.e., salient), each response is assigned a dominance score
(DS) that is a measure of its relative salience. In our adaptation
of the procedure -- which we designate MOCOM (for Measure Of
Consumption Object Meaning) -- subjects' responses are then
categorized as attribute or performance, the two proposed dimensions
of meaning. Three totals, useful for both inter- and intra-subject
and/or object analysis, can be computed: a total DS for the
attribute dimension, a total DS for the performance dimension,
and a total DS for all meaning elements. Thus, the procedure
can be used to assess the idiosyncratic meaning perceived by
an individual or that shared by a group of individuals.
AN EMPIRICAL TEST
A study was conducted to assess the proposed MOCOM and the
conceptualization upon which it rests. We admonish the reader
that, in this initial test, our focus was not on the actual
meanings subjects ascribed to the focal objects (although qualitative
results are available to adduce these). Following scaling convention,
our quantitative results portray meaning in the form of inter-object
similarities, rather than in terms of absolute object perceptions.
Two measures that purport to measure the same construct should
be highly correlated. To explore this psychometric property
of MOCOM, the semantic differential (Osgood, Suci, and Tannenbaum
1957), a measure of meaning with which most consumer researchers
are familiar, was chosen as a criterion measure of meaning.
H1: MOCOM and the semantic differential will be significantly
Kernan and Sommers (1967) propose that one's attitude toward
an object ("value" in their terminology) is a function
of an object's perceived meaningCi.e., of attribute and performance.
Thus, the meaning construct should be discriminable from the
attitude construct. Szalay and Bryson (1974) present an analysis
in which this distinction is supported. Since Osgood et al.
make no assertions regarding the functional relationships among
the three dimensions of the semantic differential, the second
hypothesis is advanced:
H2: The correlation between MOCOM and a reduced form of the
semantic differential, composed of the activity and potency
dimensions, will be significantly greater than the correlation
between MOCOM and the full, three-dimensional, semantic differential
Our definition of meaning postulates two dimensions of perception
that underlie meaning: attribute and performance. Hence a goal
of this study is to explore the psychological reality of these
dimensions. If consumers' perceptions of consumption objects
are characterized by these two dimensions it follows that the
responses produced by subjects on the continued-association
task should be characterizable as either attribute or performance.
H3: The meaning dimensions revealed via the method of continued
associations can be reliably characterized as either attribute
Common sense indicates that any measure of meaning should reveal
every object to have a unique meaning and some objects to be
more similar in meaning than others. It follows that consumption
objects determined a priori to be similar should be indicated
by the meaning measure to be more similar than objects deemed
dissimilar, a priori. Two natural categories of consumption
objects were identified -- food and clothing -- to assess whether
the proposed MOCOM possesses this characteristic. In order to
test the efficacy of our measurement procedure (and to avoid
any affect-laden contamination that might result from subjects'
familiarity with branded stimuli), three generic objects were
selected to represent each category. A black fedora, a white
cotton pullover sweater with a button-up v-neck, and a "dirty
buck" shoe were selected to represent the clothing category.
The food category was represented by one-quarter pound of uncooked
#19 linguine, a ripe banana, and a slice of whole wheat bread.
Thus, Hypothesis 4:
H4: Objects within a natural category should be perceived as
more similar in meaning to other objects within that category
than to objects from other categories (i.e., a given clothing
item should be more similar in meaning to other clothing items
than it is to any food items).
SCORES FOR EACH OBJECT ON EACH SEMANTIC DIFFERENTIAL DIMENSION
The notion of inter-object meaning similarity raises the interesting
question of which dimension, attribute or performance, is meaning
similarity largely a function? We make no prediction regarding
this intriguing question.
Ninety-six students at a Midwestern university were subjects
for- this study. They were processed in four groups.
Each subject received a packet containing: (1) instructions
for the continued-association task; (2) seven continued-association
response forms; (3) instructions for the semantic differential
task; and (4) seven semantic- differential response forms. The
three dimensions of the semantic differential were operationalized
with 34 bi-polar adjective scales selected because of their
frequent use in studies reported by Osgood et al. (1957).
Subjects first responded to a trial object, an 18" inflatable
lobster, to familiarize them with the continued-association
task. They were then exposed to the six experimental objects,
one at a time, in one of two random orders, for 60 seconds.
Subjects recorded their associations during this time. Stimuli
were referred to only by an identifying number. Subjects were
then instructed how to use the semantic differential. They then
responded to the warm-up object and then to the six experimental
objects. No time constraint was imposed on this task. A new
stimulus object was presented when all subjects were ready.
Data from the 34 semantic-differential scales were reverse
coded where necessary and a confirmatory factor analysis (SAS
procedure) was performed, with a three-factor solution specified.
Both eigenvalues and a scree test clearly indicated the appropriateness
of the three-factor solution. Following Osgood et al., the five
scales with the highest loading on each of the three semantic-differential
factors were retained for subsequent data analyses (see Table
Because an object's meaning is its location within the three-dimensional
semantic space, the mean factor scores for each stimulus object
on each dimension of the semantic differential were computed
(Table 1). Inter-object meaning similarity is also of interest
so Dis, or inter-object Euclidian distances within the three-dimensional
space, were calculated from the mean factor scores (Table 2).
Note that smaller values of Di indicate greater inter-object
DISTANCES WITHIN THE THREE-DIMENSIONAL SEMANTIC SPACE
Data produced by the continued-association task require extensive
preparation prior to analysis. The desired result is a list,
for each stimulus object, of associations, and their dominance
scores. These scores were assigned according to Szalay and Deese's
recommendation: 6 to the first response produced by a subject,
5 to the second response, 4 to the third response, 3 to the
fourth through seventh responses, 2 to the eighth and ninth
responses, and 1 to each subsequent response. Dominance scores
for common responses were summed across subjects. Affinity refers
to the degree to which persons see relations of any sort between
any two stimuli and is operationalized as the amount of overlap
between two response lists (i.e., the number of meaning elements
two objects have in common). Affinity is thus analogous to meaning
similarity and is assessed by means of an index. Calculation
of the inter-object affinity index involves summing dominance
scores across the overlapping elements and across stimuli. This
total is then divided by the sum of the total dominance scores
of the objects being compared. The resulting index value is
the proportion of the combined total dominance scores accounted
for by the affinial relations. The index has a theoretic range
of zero to one and increases in value as inter-object affinity
increases. The affinity index for each pair of stimulus consumption
objects is given in Table 3.
AFFINITY INDEX VALUES
With Hypothesis 1 we predicted that the correlation between
MOCOM and the semantic differential would be significantly greater
than zero. Because the affinity index (representing MOCOM) increases
and Di (representing the semantic differential) decreases as
inter-object meaning similarity increases, a negative correlation
is expected. To test this hypothesis, the off-diagonal elements
of the affinity-index matrix (Table 3) were correlated with
the corresponding elements in the matrix of inter-object Dis
in the three-dimensional semantic space (Table 2). Since the
obtained Pearson correlation of -0.52 is significantly greater
than zero (p < .05), the data are consistent with Hypothesis
1: There is some evidence that MOCOM has convergent validity.
With Hypothesis 2 we advance a stronger assertion than in Hypothesis
1, namely that the correlation between MOCOM and a two-dimensional
version of the semantic differential, composed of the activity
and potency dimensions, will be significantly greater than the
correlation explored in Hypothesis 1. (This follows from the
fact that MOCOM associations -which are categorized into attribute
or performance (or "other") dimensions -- do not measure
affect directly.) To test this hypothesis the off-diagonal elements
of the matrix of inter-object Dis in the two-dimensional semantic
space (Table 4) were correlated with the off-diagonal elements
of the matrix of inter-object affinity indices (Table 3). Inter-measure
correlations (all significant at p = .05) are summarized in
Table 5. The difference between the two theoretically important
correlations, -0.52 and -0.73, is statistically significant
(Fisher's z = 2.41, p < .01)Ca result consistent with Kernan
and Sommers' assertion (and our H2) that the meaning construct
is discernible from the attitude construct. This theoretic consistency
increases our confidence that MOCOM is indeed a measure of meaning,
hence subsequent analyses employ the reduced, two-dimensional
DISTANCES WITHIN THE REDUCED (2-D) SEMANTIC SPACE
BETWEEN MEANING MEASURES
Hypothesis 3 seeks to assess the psychological reality of attribute
and performance as dimensions of meaning. To test H3, standard
measure-development procedures were followed: operational definitions
of attribute and performance were developed (see Table 6) and
two trained judges applied the coding categories to subjects'
responses to each of the six focal objects. Inter-coder agreement
averaged 85%, a rate acceptably high for an exploratory study.
Unresolved disputes, refereed by one of the authors, were categorized
as "other." The distribution of attribute and performance,
expressed as a percent of the total dominance score for each
stimulus object, is presented in Figure 1. Overall, the attribute
and performance dimensions capture 87% of the total dominance
score and 96% of the responses, a result that strongly supports
H3 -- a consumption object's psychological meaning is recoverable
with its attribute and performance dimensions.
Hypothesis 4 contends that MOCOM should recover the two natural
categories of stimulus objects; that objects within a category
will be judged more similar to other items within that category
than to items in the other category. For example, a shoe should
be indicated as more similar to other clothing items than to
any food item.
Using the reduced-space semantic differential measures, Figure
2 plots the locations of the stimuli within the semantic space.
(Identical inter-object similarity can be gleaned from the distance
measures in Table 4.) Figure 2 reveals the food category to
be recovered nicely -- its three items are clustered. In contrast,
the clothing category was not completely recovered -- the sweater
and the shoe are quite similar but the hat has a very dissimilar
meaning. A reading of subjects' association data suggests the
explanation that, compared to the shoe and sweater, the hat
is an extremely expressive consumption object. Accordingly,
a two-group (food and clothing) MANOVA was performed (SAS's
GLM procedure). The significant MANOVA (F(2 577) = 102.59, p
< .01) indicates that the group mean vectors on activity
and potency factor scores differ. Univariate ANOVAs reveal a
significant effect for potency (F(1, 578) = 204.08, p < .01)
but not for activity (F(1, 578) = .50, p > .01) -- within
category variance exceeded between category variance on this
latter dimension, reinforcing our previous observation that
the clothing category was not recovered well by the semantic
differential. Thus, a three-group MANOVA (splitting out the
hat) was performed and, as expected, significant univariate
effects were observed on both the activity (F(2, m) = 60.23,
p < .01) and potency (F(2, 573) = 113.53, p < .01) dimensions.
Thus, the semantic differential did a good job of recovering
our food category but it did not completely recover our clothing
DEFINITIONS OF ATTRIBUTE AND PERFORMANCE
TOTAL DOMINANCE SCORE ACCOUNTED FOR BY THE MEANING DIMENSIONS
To assess the efficacy of MOCOM relative to Hypothesis 4, we
redirect the reader's attention to Table 3, the matrix of inter-object
affinity indices. Although casual observation of that matrix
reveals the clothing items to be similar, the food items to
be similar, and little meaning similarity between objects in
different categories, a statistical test is necessary to establish
the existence of our natural categories. Accordingly, a Mann-Whitney
U test was used, with the null hypothesis that affinity indices
of pairs of objects within a category are from the same distribution
as affinity indices of pairs of objects from different categories.
Comparing either the three within-clothing cells to the nine
between-category cells or the three within-food cells to the
nine between-category cells, the null hypothesis must be rejected
(U = 0, p < .01). This result, supportive of Hypothesis 4,
indicates that MOCOM recovered the two natural categories nicely.
Even the hat is indicated to be more similar to other clothing
items than it is to any of the food items. In this respect,
our proposed measure has out-performed the semantic differential.
Yet the nature of inter-object similarity remains to be established.
The semantic differential reveals, for example, that the sweater
and the shoe (the two objects most similar in meaning) are moderately
high on activity and low on potency. We could extend this investigation
by examining the respective factor scores for these objects,
but this is a cumbersome procedure that guarantees little by
way of interpretable information. In contrast, we can use the
lists of common responses from which MOCOM's affinity indices
are derived and discover a richness the semantic differential
cannot approach. The sweater and the shoe, for example, share
these elements: casual, clean, comfortable, conservative, fall,
L. L. Bean, man's, nice, plain, preppy, soft, stylish, ugly,
warm, and worn. Thus, these two objects share warmth, conservativeness,
and preppiness, and are worn by men in the autumn. Whereas the
semantic differential requires researcher specification of scales,
the MOCOM "provides" the researcher with scales --
that are texture-laden with qualitative interpretability in
But MOCOM does not restrict us to qualitative analysis. By
separating subjects' associations into their attribute? performance,
and "other" dimensions we can compute affinial relations
(shared dimensions) among the objects. Table 7, which displays
such relations, reveals that meaning similarity within a category
is dominated by performance while such similarity as exists
among objects from different categories is largely a function
of attribute. In other words, "natural" categories
seem to consist of objects that do the same thing.
A series of x2 tests supports this interpretation. The distribution
of attribute and performance does not differ across the three
objects representing each natural category (food category: X2(2)
= 6.75, p > .05; clothing category: X2(2)= 2.34, p > .05)
nor do the distributions of attribute and performance differ
between the two natural categories (x2(l)= 1.70, p > .05).
This indicates that the six cells representing the two categories
can be pooled. A x2 test to explore homogeneity of the distributions
of attribute and performance in the nine between-category cells
could not be performed because all cells have expected values
less than five. Thus, those nine cells were collapsed into a
single between-category group. The dramatically significant,
but hardly surprising difference (X2(1) = 262.99, p < .01)
between the distributions of attribute and performance in the
pooled between-category cells and the pooled within-category
cells is illustrated in Figure 3.
OF INTER-OBJECT AFFINIAL RELATIONS INTO THEIR DIMENSIONS ATTRIBUTE,
PERFORMANCE, AND OTHER
Thus, Hypothesis 4 seems supported. MOCOM not only recovered
our a priori defined categories, it did a better job of reproducing
them than did the semantic differential. We also explored the
insight that each measure can yield into the nature of inter-object
meaning similarity. In this regard we found MOCOM to provide
richer information than the semantic differential. Exploring
the inter-object similarities revealed by our proposed measure
further, we assert tentatively that similarities between objects
within a natural category lie along the performance dimension
of meaning. In contrast, meaning similarity, if any, between
objects from different categories derives from the attribute
dimension of meaningCwhat physical attributes the objects appear
Although this exploratory effort encourages us that our definition
and measure of consumption-object meaning is right-headed, we
recognize that much remains to be done. Indeed, the reader might
muse that we have raised more questions than we have answered.
And we would agree. For example, we have not tested branded
products -- some of which are extraordinarily rich in symbolic
meaningCor services, and we have not addressed the complex issue
of how object meaning (however generated) relates to affectCand
other reactions deeper in the consumer's response hierarchy.
And we have not dealt with the tenuous relationship between
object meaning and that which attends consumption experiences
(particularly those accompanied by aesthetic responses) where
no physical entity can be regarded as focal. But theseCand many
other critical issues -- must await more mundane developmental
OF AFFINAL RELATIONS INTO PERFORMANCE, ATTRIBUTE, AND OTHER
Our current program of research is focusing on two problems.
First, experiments are being conducted to explicate the effect
of contextual variation on an object's meaning. We believe this
issue relates to both meaning creation and meaning change. Second
(for all the obvious reasons), we are examining the process
through which information is transformed into meaning. If we
are lucky and learn something about these two problems, we shall
feel confident (or foolhardy) enough to press on to the more
Csikszentmihalyi, Mihaly and Eugene Rochberg-Halton (1981),
The Meaning of Things: Domestic Symbols and the Self, New York:
Cambridge University Press.
Douglas, Mary and Baron Isherwood (1979), The World of Goods,
New York: W. W. Norton.
Friedman, Roberto (1986), "Psychological Meaning of Products:
Identification and Marketing Applications," Psychology
and Marketing, 3 (Spring), 1-15.
Hirschman, Elizabeth (1980), "Attributes of Attributes
and Layers of Meaning," in Advances in Consumer Research,
Vol. ?, ed. Jerry C. Olson, Ann Arbor, MI: Association for Consumer
Holbrook, Morris B. and Elizabeth C. Hirschman (1982), "The
Experiential Aspects of Consumption: Consumer Fantasies, Feelings,
and Fun," Journal of Consumer Research, 9 (September),
Kernan, Jerome B. and Montrose S. Sommers (1967), "Meaning,
Value, and the Theory of Promotion," Journal of Communication,
2 (June), 109-135.
Kleine, Robert E., III, and Jerome B. Kernan (1987), "Toward
an Epistemology of Symbolism Some Preliminary Considerations,"
in Advances in Consumer Research, Vol. 14, eds. Paul F. Anderson
and Melanie Wallendorf, Provo, UT: Association for Consumer
Levy, Sidney J. (1959), "Symbols for Sale," Harvard
Business Review, 37 (July-August), 117-124.
Levy, Sidney J. (1981), "Interpreting Consumer Mythology:
A Structural Approach to Consumer Behavior," Journal of
Marketing, 45 (Summer), 49-61.
McCracken, Grant (1986), "Culture and Consumption: A Theoretical
Account of the Structure and Movement of the Cultural Meaning
of Consumer Goods," Journal of Consumer Research, 13 (June),
Mick, David Glen (1986), "Consumer Research and Semiotics:
Exploring the Morphology of Signs, Symbols, and Significance,"
Journal of Consumer Research, 13 (September), 196-213.
Ogden, C. K. and I. A. Richards (1953), The Meaning of Meaning,
8th ed., New York: Harcourt, Brace.
Osgood, Charles E., George J. Suci, and Percy H. Tannenbaum
(1957), The Measurement of Meaning, Urbana, IL: University of
Szalay Lorand B. and Jean A. Bryson (1974), "Psychological
Meaning: Comparative Analyses and Theoretical Implications,"
Journal of Personality and Social Psychology, 30 (December),
Szalay, Lorand B. and James Deese (1978), Subjective Meaning
and Culture: An Assessment Through Word Associations, Hillsdale,
NJ: Lawrence Erlbaum Associates.