Content uploaded by Nichole Egbert
Author content
All content in this area was uploaded by Nichole Egbert on May 29, 2014
Content may be subject to copyright.
Speaking the Language of Relational
Maintenance: A Validity Test of
Chapman’s (1992) Five Love Languages
Nichole Egbert & Denise Polk
This study examines the claims of Gary Chapman (1992) in his popular press book, The
Five Love Languages: How to Express Heartfelt Commitment to Your Mate. A confirma-
tory factor analysis showed that a five-factor fit of these data to Chapman’s proposed five
love languages was superior to unidimensional, three-factor, and four-factor solutions.
Results showed significant relationships between the love language factors and Stafford,
Dainton, and Haas’ (2000) relational maintenance typology, suggesting that Chapman’s
love languages may reflect behaviors performed to enact intended relational maintenance.
Keywords: Relational Maintenance; Scale Development; Love Languages; Confirmatory
Factors Analysis
In a book that lasted 50 weeks on Publisher’sWeekly’sbest-sellers list, The Five
Love Languages:How to Express Heartfelt Commitment to Your Mate,Dr.Gary
Chapman promotes a theory that has resonated with millions of people through his
books, videos, marriage conferences, or nationally syndicated radio program. Chapman’s
(1992) thesis is based on a metaphor of an emotional ‘‘love tank.’’ He argues that
romantic partners should fill one another’s love tanks by ‘‘speaking’’ their partner’s ‘‘love
languages’’ (LLs) to enhance relational quality. As popular books and academic research
have often been at cross-purposes, the goal of this project is to test the factor structure
of items derived from Chapman’s typology and then to test the construct validity of the
new measure by comparing it to established empirical measures of related constructs.
Incorporating and testing popular press ideas through research serves several valuable
Nichole Egbert (PhD, University of Georgia) is an Assistant Professor in the School of Communication Studies
at Kent State University. Denise Polk (PhD, Kent State University) is an Assistant Professor in the Department
of Communication Studies at West Chester University of Pennsylvania. Please direct all correspondence to
Nichole Egbert at: negbert@kent.edu
Communication Research Reports
Vol. 23, No. 1, January 2006, pp. 19–26
ISSN 0882-4096 (print)/ISSN 1746-4099 (online) #2006 Eastern Communication Association
DOI: 10.1080/17464090500535822
functions: (a) providing an empirical check of intuitive ideas that have found favor with
the public, (b) forcing researchers to keep current with mass-mediated messages that affect
how people view relationships, and (c) enhancing the validity of relational measurement
instruments.
Chapman’s Five Love Languages
According to Chapman (1992), people speak five primary LLs: words of affirmation
(encouraging messages), quality time (time spent engaged in shared activities), gifts
(tokens of affection), acts of service (help with necessary tasks), and physical touch
(hand holding to sexual intercourse). Furthermore, people tend to have one or
two favorite LLs and often show love to their partners using their own preferred
LL, but partners who enjoy higher quality relationships tend to express love accord-
ing to their partners’ preferred LLs.
Although not directly referring to relational maintenance, Chapman’s (1992)
explanations of LLs parallel descriptions of maintenance behaviors found in academic
research. Chapman refers to the term ‘‘love’’ instead, even though his purpose is not
to eliminate confusion about the meaning of the term. Throughout the book,
Chapman refers to the necessity of enacting certain behaviors to sustain one’s own
well-being, a spouse’s well being, and the well being or quality of the marital relation-
ship. Rather than to complain that love is gone or that a relationship is dead,
Chapman advises that people must make conscious efforts to speak a partner’s
desired LL. Such efforts are essential, but they do not come naturally. Maintaining
the love tank by keeping it full is as essential as maintaining the proper oil level in
an automobile (Chapman). All of these claims bear a striking resemblance to litera-
ture about relational maintenance.
Relational Maintenance Behaviors
Relational maintenance refers to behaviors enacted to preserve desired relational fea-
tures (Dindia & Canary, 1993). The use of these behaviors has been connected with
equity, love, satisfaction, and commitment (Canary & Stafford, 1992; Stafford &
Canary, 1991). Several typologies of maintenance behaviors exist (e.g., Dindia &
Baxter, 1987; Stafford & Canary, 1991), and many of these items parallel what
Chapman (1992) calls LLs.
One measure that assesses relational maintenance similar to Chapman’s (1992) LLs
is Canary and Stafford’s (1992) five-category typology for relational maintenance. The
five categories are assurances, openness, positivity, sharing tasks, and social networks.
In addition to the parallel of a five-category typology, descriptions of the items that rep-
resent the various categories also parallel one another. First, assurances provide partners
with information about their importance such as stressing relational commitment.
Similarly, Chapman claims that partners need to offer words that affirm one partner’s
importance to the other. Second, openness refers to discussing relational feelings
directly. Openness includes disclosing what a partner needs or wants from a relation-
ship. When discussing openness, Chapman refers to the need for couples to
20 N. Egbert & D. Polk
make requests of one another instead of demands. Third, Canary and Stafford define
positivity as efforts to keep interactions enjoyable like building up self-esteem by
complimenting a partner. Similarly, Chapman discusses the need to offer positive
comments. Fourth, sharing tasks involves performing instrumental activities with an
equitable approach. Chapman suggests that couples need to overcome role stereotypes
to share tasks. Fifth, social networks refers to sharing common friends to maintain the
relationship, like including other family members in activities. Likewise, Chapman
suggests that one way that couples may spend quality time together is by playing games
as a family. Stafford, Dainton, and Haas (2000) confirmed the validity of Canary and
Stafford’s five behaviors and identified advice (e.g., expressing opinions) and conflict
management (e.g., using strategies like apologizing) as two additional categories.
Relational maintenance behaviors and LLs diverge and converge in other ways as
well. For example, Dainton and Stafford (1993) suggest that maintenance involves
strategic behaviors (enacted intentionally with the relationship in mind) or routine
behaviors (enacted at a lower level of consciousness and not intended to help the
relationship). Chapman (1992) suggests that LLs are always strategic because people
must learn a partner’s LL and must make conscious efforts to speak that LL. This claim
does, however, parallel the hypothesis that a large maintenance repertoire does not
ensure the ability to select or enact an appropriate strategy (Dindia & Baxter, 1987).
In addition, Dainton’s (2000) results support Chapman’s claim about the relationship
between LL and satisfaction, in that the extent to which expectations about partner
maintenance behaviors were met related positively to relational satisfaction. Thus, fail-
ing to enact certain behaviors may cause partners to feel unloved. Dainton, however,
claims people expect their partners to perform behaviors that comprise all of the rela-
tional maintenance categories, whereas Chapman focuses on enacting behaviors in the
preferred LL categories instead of performing behaviors in all the categories.
These comparisons suggest the need for further investigation into the similarities
and distinctions between Chapman’s (1992) popular press book and academic
understandings of relational maintenance. Thus, the purpose of this study is to inves-
tigate the empirical validity of Chapman’s five LLs. Specifically, we pose the following
two research questions:
RQ1:Do Chapman’s Five LLs form five distinct and cohesive factors?
RQ2:What is the relationship between Chapaman’s five LLs and more established
measures of relational maintenance?
Method
Participants
Students (N¼110) involved in a current dating or marriage relationship of at least
two months and who were enrolled in a basic speech course received course credit
for participating. The majority of participants (n¼70;64:8%) was female and
Caucasian (n¼95;88:0%). Other ethnic groups included African American
(n¼8;7:4%), Asian American (n¼1;0:9%), and ‘‘Other’’ (n¼4;3:7%). Most
Love Languages 21
participants were between the ages of 18 and 22 (n¼97;88:0%), but some participants
were older, aged between 23–30 (n¼10;9:3%), 30–40 (n¼2;1:9%), and over 40
years (n¼1;0:9%). Relationship length ranged from 2–6 months (n¼17;15:7%),
6 months to 2 years (n¼51;47:2%), 2–5 years (n¼35;32:4%), and over 5 years
(n¼5;4:6%).
Procedures
This study employed an anonymous survey methodology to test the factor structure
and construct validity of a scale measuring Chapman’s (1992) LLs. Toward this goal,
a measure of the five LLs was: (a) developed from concepts found in Chapman, (b)
tested using confirmatory factor analysis (CFA), and (c) analyzed for its construct
validity by comparing the LL subscales to an established measure of relational main-
tenance behaviors.
Creating a measure of behaviors relating to Chapman’s (1992) five LLs (acts of ser-
vice, physical touch, words of affirmation, quality time, and gifts), involved generat-
ing a list of 20 behaviors discussed by Chapman in his book. Twenty behaviors (four
behaviors=items for each of the five LLs) followed the stem ‘‘I tend to express my
feelings to my partner by ...’’ along with Likert-type response options (1 ¼strongly
strongly disagree to 5 ¼strongly agree; see Table 1).
Participants completed this 20-item scale of the five LLs, along with a measure of
maintenance behaviors (Stafford et al., 2000) and several demographic items. To test
the factor structure of the LL scale, a CFA helped to determine how well the items
loaded on their respective subscales. Finally, we compared the final factor model with
Stafford et al.’s relational maintenance behavior to determine the construct validity of
the five LLs.
Results
A CFA using AMOS 5.0 (Arbuckle, 2003) helped to determine the internal consistency of
Chapman’s (1992) five LL dimensions. The v2test was significant (v2¼343:82;
df ¼165;p<:001). The ratio of the v2to the degrees of freedom was 2.08, and other
indices of fit suggested that the model did not provide an adequate fit to the data
(RFI ¼.71; CFI ¼.85; RMSEA ¼.10). To understand the factor structure better, these
fit indices were compared against several comparison models. The first comparison
model was a unidimensional model where all 20 items loaded onto the same ‘‘love lan-
guage’’ factor. In this case, the fit was worse than the five-factor model (v2¼471:14;
p<:001;df ¼170;RFI ¼:62;CFI ¼:75;RMSEA ¼:13). When examining the
factor correlations, the factors of ‘‘words’’ and ‘‘time’’ were highly intercorrelated, sug-
gesting that these two factors may be combined in a four-factor model with better fit.
However, the four-factor model did not provide better fit than the original five-factor
model tested (v2¼353:66;p<:001;df ¼166;RFI ¼:70;CFI ¼:84;RMSEA ¼:10).
Finally, a three-factor model was also tested where ‘‘gifts’’ and ‘‘touch’’ were combined
into one factor. As with the other comparison models, the five-factor fit was still slightly
better (v2¼350:16;p<:001;df ¼160;RFI ¼:70;CFI ¼:84;RMSEA ¼:10).
22 N. Egbert & D. Polk
Although the five-factor model improved the fit of the data as compared with the
unidimensional, four-factor, and three-factor models, we wanted to see how the fit
could be improved. An examination of the modification indices and item loadings
suggested that deleting one item from the ‘‘acts’’ factor would improve the model
significantly (‘‘Helping my partner out when he=she needs it’’). The resulting model
provided a better fit to the data, but was still marginal (v2¼294:63;p<:001;
df ¼147;RMSEA ¼:10;RFI ¼:73;CFI ¼:87). However, as the ratio of the v2
to the degrees of freedom was now 2.00, we argue that this model could be
considered to be a more acceptable, yet still marginal, fit to the data (see Bollen,
Table 1 Means, Standard Deviations, and Reliabilities of LL Subscales
LL subscales with corresponding items M SD
I tend to express my feelings to my partner by
Acts of Service (a¼:85) 17.75 2.49
Running errands for him=her
Finishing a chore for him=her when he=she doesn’t
have time to do it
Helping my partner helps out when he=she needs it
Helping my partner keep things cleaned up
Physical Touch (a¼:77) 17.97 2.48
Hugging my partner
Giving my partner a kiss
Holding my partner’s hand
Touching my partner
Words of Affirmation (a¼:81) 16.37 3.33
Complimenting my partner
Telling my partner that I appreciate him=her
Giving my partner credit for something good he=she did
Giving my partner positive comments
Quality Time (a¼:80) 16.59 3.07
Really listening to my partner
Spending time doing something we both like
Having a quality conversation
Spending my free time with him=her
Gifts (a¼:83Þ18.38 2.90
Giving him=her a thoughtful birthday gift
Picking up a greeting card for him=her
Giving him=her a surprise present when there’s
no special occasion
Giving him=her a small present when I come
home from a trip
Love Languages 23
1989; Carmines & McIver, 1981). The resulting five subscales all demonstrated
acceptable reliability, ranging from a¼:77 [touch] to a¼:85 [acts]).
To understand how these five factors compared with an established measure of
relational maintenance, they were correlated with Stafford et al.’s (2000) relational
maintenance typology. We selected Stafford et al.’s measure because it confirmed
the validity of Canary and Stafford’s (1992) widely used five-category relational
maintenance typology (assurances, social networks, openness, positivity, and shared
tasks) and also identified two other behaviors that reflect strategic and routine main-
tenance behaviors (advice and conflict management). First-order correlations
revealed significant correlations (ranging from .21 to .52) between the LL factors
and all but one of Stafford et al.’s items. Specifically, acts and positivity were not
significantly correlated (r¼:16;p¼:10; see Table 2).
Further testing helped determine the most highly related subscales. Five regressions
were computed, one for each of the five LLs, with Stafford et al.’s (2000) seven main-
tenance behaviors as predictors (entered simultaneously). None of the seven mainte-
nance behaviors significantly predicted the words factor when controlling the other
six behaviors. However, openness significantly predicted time (b¼:36;p¼:02),
social networks significantly predicted gifts (b¼:22;p¼:02), assurances and shared
tasks predicted touch (b¼:38;p¼:02;b¼:22;p¼:04), and shared tasks and
social networks predicted acts (b¼:38;p¼:01;b¼:22;p¼:03).
Table 2 Bivariate Correlations Between LL Subscales and Stafford et al.’s (2000)
Relational Maintenance Behaviors
Love languages
(Chapman, 1992)
Relational maintenance behaviors
(Stafford et al., 2000)
123456789101112
1. Words 1 .75 .56 .65 .52 .32 .49 .48 .39 .39 .30 .22
2. Time 1 .54 .65 .45 .22.44 .52 .45 .33 .38 .29
3. Gifts 1 .56 .65 .26 .51 .51 .33 .32 .28 .38
4. Touch 1 .52 .25 .46 .38 .21.37 .33 .33
5. Acts 1 .23.35 .27 .28 .44 .16 .29
6. Advice 1 .45 .47 .36 .51 .49 .23
7. Assurances 1 .83 .45 .44 .43 .38
8. Openness 1 .55 .45 .48 .35
9. Conflict
management
1 .45 .51 .28
10. Shared tasks 1 .45 .24
11. Positivity 1 .49
12. Social networks 1
Note.Denotes that correlation is significant at the p<.05 level. Denotes that correlation is significant at the
p<.01 level.
24 N. Egbert & D. Polk
Discussion
These exploratory analyses demonstrated that Chapman’s (1992) five-factor LL
model might have some psychometric validity, as compared with a unidimensional,
three-factor, and four-factor solution. Certainly, his ideas of five distinct love lan-
guages seem valid to the general public who made his book a best seller. Although
our efforts to produce a measurement instrument from his book did not conclude
with a fitting model, the data suggest that the model is close to fitting, and may
produce a valid, useable instrument when tested with a larger sample.
Furthermore, these five factors were related to an established instrument in appro-
priate ways, thereby demonstrating their construct validity. Specifically, these data
revealed significant relationships between Stafford et al.’s (2000) shared tasks with
time, social networks with gifts, assurances and shared tasks with touch, and two rela-
tional maintenance behaviors (shared tasks and social networks) with Chapman’s
(1992) acts. These subscale relationships suggest that perhaps the Stafford et al. scale
reflects the intentions of the communicator, whereas the LLs are the behaviors people
enact to carry those intentions to the recipient. For example, if partners wish to share
tasks, they would communicate this intention best by spending time with their part-
ners and performing acts of service. Similarly, to communicate that a partner has an
effective social network, giving a gift or token of affection is an appropriate behavior
to enact. Thus, according to these data, Chapman’s behaviors can be considered the
vehicle whereby people deliver to the recipient the relational maintenance items con-
structed by Stafford et al. This explanation makes the measurement of this process
slightly less abstract, a finding that could be meaningful to both empirical researchers
and the general public.
Limitations and Future Directions
Taken as a whole, the results of this study begin to paint a picture of how various
relational maintenance behaviors are desired and received by romantic partners.
One limitation to this study, however, is its homogenous college student sample.
Although traditional college students are highly invested in romantic relationships,
this sample limits our analysis because they have less relational experience, different
relational expectations, and their existing relationships tend to be less developed than
the general population’s. Future investigations should include a more diverse group
of participants so that the results are more generalizable.
Further refining the LL items used in this study could also improve the factor
structure, perhaps making the items represent behaviors that are even more concrete
and distinct, so that they form more cohesive factors. For example, an item like, ‘‘I
tend to express my feelings to my partner by giving my partner positive comments,’’
may be too conceptually broad. The factor of ‘‘words of affirmation’’ might be
defined more by adding common, affectionate phrases such as, ‘‘I love you,’’ or ‘‘I
missed you today.’’ These concrete phrases, if they demonstrate the same relationship
to the relational maintenance scale of Stafford et al. (2000) as these data did, would
Love Languages 25
further support our conclusion that LLs represent behaviors designed to communi-
cate goals related to relational maintenance.
We consider these studies important, as they help to bridge the current thinking
found in both popular and academic arenas. When issues resonate with the public
like Chapman’s (1992) LLs, communication researchers may be able to utilize these
concepts to provide new avenues for research and research dissemination instead of
working separately from the popular press. Similarly, the popular press and mental
health professionals receive valuable research findings for their own writing, which
then can be communicated to the general public in an effort to help them improve
their own lives and relationships. Thus, we hope this study serves as a springboard for
the intersection of popular and scholarly discourse on relationships.
References
Arbuckle, J. L. (2003). AMOS (Version 5.0) [Computer software]. Chicago: SmallWaters Corporation.
Bollen, K. A. (1989). Structural equations with latent variables. New York: John Wiley & Sons.
Canary, D. J. & Stafford, L. (1992). Relational maintenance strategies and equity in marriage.
Communication Monographs,59, 243–267.
Carmines, E. G. & McIver, J. P. (1981). Analyzing models with unobserved variables: Analysis of
covariance structures. In G. W. Bohrnstedt & E. F. Borgatta (Eds.), Social measurement:
Current issues (pp. 65–115). Beverly Hills, CA: Sage.
Chapman, G. (1992). The five love languages: How to express heartfelt commitment to your mate.
Chicago: Northfield Publishing.
Dainton, M. (2000). Maintenance behaviors, expectations for maintenance, and satisfaction: Link-
ing comparison levels to relational maintenance strategies. Journal of Social and Personal
Relationships,17, 827–842.
Dainton, M. & Stafford, L. (1993). Routine maintenance behaviors: A comparison of relationship
type, partner similarity and sex differences. Journal of Social and Personal Relationships,10,
255–271.
Dindia, K. & Baxter, L. (1987). Strategies for maintaining and repairing marital relationships. Jour-
nal of Social and Personal Relationships,4, 143–158.
Dindia, K. & Canary, D. J. (1993). Definitions and theoretical perspectives on maintaining relation-
ships. Journal of Social and Personal Relationships,10, 163–173.
Stafford, L. & Canary, D. J. (1991). Maintenance strategies and romantic relationship type, gender,
and relational characteristics. Journal of Social and Personal Relationships,8, 217–242.
Stafford, L., Dainton, M., & Haas, S. (2000). Measuring routine and strategic relational mainte-
nance: Scale revision, sex versus gender roles, and the prediction of relational characteristics.
Communication Monographs,3, 306–323.
26 N. Egbert & D. Polk
