Академический Документы
Профессиональный Документы
Культура Документы
In this article, I argue that the relationships among mental health, disorder, and
distress are a key source of conflict in the sociology of mental health and that
the features of the conflict have the potential to call into question much of the
accumulated scientific knowledge on mental health. To address this issue, I at-
tempt to empirically assess three competing frameworks regarding these rela-
tionships: (1) the “modal perspective,” (2) the “Mirowsky and Ross perspec-
tive,” and (3) the “positive psychology perspective.” Results, however, support a
“discontinuous perspective:” no underlying continuum among any of the three
concepts. These results suggest that researchers need to avoid the common prac-
tice of “lumping together” distress, disorder, and mental health and study each
in their own right. Subsequent tests attempt to further specify the relationships
among these concepts. Results indicate a strong positive directional association
from distress to disorder, a strong negative directional association from distress
to mental health, and no significant relationship between mental health and dis-
order. These results are used to generate a number of directions for future re-
search.
In a 2002 special issue of the Journal of nature of what it is that they are studying. It dri-
Health and Social Behavior entitled “Selecting ves to the very heart of the enterprise for how
Outcomes for the Sociology of Mental Health: we can make claims to understand the sociolo-
Issues of Measurement and Dimensionality” a gy of mental health and to be accumulating
number of researchers wrestled with the rela- knowledge regarding mental health if we oper-
tionships among distress, disorder, and mental ationalize our dependent variable on the basis
health. Aneshensel (2002), in a commentary on of multiple, opaque concepts that contain con-
the articles, referred to these relationships as tradictory assumptions about the nature of one
“several recurrent and unresolved epistemo- another.
logical issues that are especially consequential In this issue, Mirowsky and Ross (2002) re-
for sociologists” (p. 236). Indeed, without a iterated their long-standing framework that
clear conceptualization of the relationships links psychological distress with mental health
among these concepts researchers in the soci- and that views disorder as a conceptually dis-
ology of mental health must question the very
tinct and socially constructed phenomenon.
Additionally, Keyes (2002) offered evidence to
* I would like to express my most profound appreci- support the positive psychology movement,
ation to Peggy Thoits, Ken Bollen, and Andy Perrin. which seeks to confront the tendency toward
I also wish to thank the Editor and the anonymous treating mental health and mental disorder as a
reviewers at JHSB. All shortcomings are of my own continuous phenomenon. Aneshensel’s review
doing. Please address correspondence to Andrew
Payton, University of North Carolina at Chapel Hill, of the articles in this special issue suggested
Department of Sociology, CB#3210, Hamilton Hall, that the important insights of Mirowsky and
Chapel Hill, NC 27599-3210 (e-mail: arpayton Ross and Keyes offered compelling arguments
@unc.edu). in favor of pulling apart mental disorder and
Downloaded from hsb.sagepub.com at NORTHERN KENTUCKY UNIV on June 19, 2015
213
214 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
mental health, but in no way provided a defin- ized only by the relative presence or absence of
itive assessment of the relationships among disorder, as defined by the Diagnostic and
distress, disorder, and mental health. This de- Statistical Manual of Mental Disorders (DSM)
bate resurfaced in a recent issue (late 2007) of of the American Psychiatric Association. This
Health: An Interdisciplinary Journal for the perspective is perhaps best signified by the fact
Social Study of Health, Illness, and Medicine, that the American Sociological Association
with contributions from some of the same, and section entitled the “Sociology of Mental
other, researchers. While a large body of re- Health” has very little research explicitly fo-
search has accumulated on mental health, dis- cused on mental health. While labeled “mental
order, and distress, the relationships among health” research, the actual substance is almost
these phenomena remain contested and, be- exclusively focused on disorders: depression,
cause they carry assumptions about one anoth- anxiety, alcohol and drug abuse, comorbidity,
er, implicitly challenge the accumulated re- etc. (Keyes 2002). To be fair, the terms mental
search. health and mental illness naturally lead to the
The goal of this article is to move beyond the assumption that they exist on a single continu-
present impasse surrounding the dependent um. Within the modal perspective distress is ei-
variable(s) of interest to sociologists of health ther ignored or relegated to some subsidiary
and illness. It seeks to do so by turning the key status.
features of the debate into a series of testable The positive psychology perspective is an
hypotheses subject to empirical analysis. First, explicit alternative to the modal perspective. It
I review the literature on distress, disorder, and is both not new and emergent (Ryff 1989a). It
mental health and distill alternative under- argues that the absence of disorder is a neces-
standings of the relationships among these sary but insufficient condition for mental
concepts into three competing perspectives. I health and thus seeks to study mental health in
label these perspectives (1) the “modal per- its own right, as a separate yet related continu-
spective,” which treats mental health and men- um to disorder. As such, it maintains that re-
tal disorder as a single continuum and either search stemming from the modal perspective
ignores distress or relegates it to some sub- tells us very little about mental health. One of
sidiary status; (2) the “Mirowsky and Ross per- the leading approaches within the positive psy-
spective,” which treats distress and mental chology perspective conceptualizes mental
health as a single continuum, and disorder as health as consisting of six key dimensions:
artificial thresholds imposed along the contin- self-acceptance, positive relations with others,
uum; and (3) the “positive psychology per- personal growth, purpose in life, environmen-
spective,” which argues explicitly against the tal mastery, and autonomy (Ryff 1989a,
modal perspective, instead viewing disorder 1989b). These dimensions represent liking
and mental health as distinct phenomena. The oneself, having warm and trusting relation-
positive psychology perspective almost wholly ships with others, having goals of personal de-
ignores distress. I test these three perspectives velopment, having direction in life, having the
using confirmatory factor analysis and then ability to control and satisfy one’s needs, and
turn to structural equation modeling (SEM) in having a sense of self-determination (Ryff
order to further assess the relationships among 1989b).
distress, disorder, and mental health. The fol- Marie Jahoda (1958) was perhaps the first
lowing section offers a brief sketch of the de- scholar to articulate this perspective when she
bate organized around these competing per- argued that the idea of “normality” is “unspe-
spectives. cific and bare of psychological content”
(1958:22). I trace the empirical origins of the
LITERATURE REVIEW modern perspective to the classic finding that
The modal perspective is less a well-articu- positive affect (e.g., good spirits) and negative
lated position supported by a substantial body affect (e.g., hopelessness) are weakly correlat-
of empirical research, but more a prevailing ed (Bradburn 1969). If we assume that positive
norm representing the long-standing tendency affect indicates mental health and negative af-
for researchers to conflate mental health and fect indicates disorder then this finding repre-
mental illness. From this perspective, disorder sents early evidence in favor of separating
(or mental illness) is the “master” or “umbrel- mental health and disorder into two distinct
la” concept. Mental health is a state character- continua.
2000) a respondent is not diagnosed as having reliability for the latent construct disorder is
experienced a major depressive episode unless .84. Internal reliability is measured using the
they experienced either (1) depressed mood or equation
(2) loss of interest or pleasure most of the day
and nearly every day during a two-week peri- (⌺i)2 Var () ,
od, which mirrors the screening mechanism
employed in the MIDUS. In short, there may (⌺i)2 Var () + ⌺ Var (⑀i) + 2⌺⌺Cov (εiεj)
i<j
be a “true” zeros problem, but the instrument
does not undercount “true” disorders following following Bollen (1980), where i,j = 1, 2, .|.|. ,
the DSM-IV assessment criteria. This holds total number of indicators. I depart from the
true for all measures of disorder. standard use of Cronbach’s alpha because it
Respondents screening positively for GAD makes two assumptions that do not hold for my
were administered a series of 10 questions. analyses. First, Cronbach’s alpha assumes that
These items are restlessness, nervousness, irri- there are no correlated errors, but CFA models
tability, trouble falling asleep, trouble staying correlated error terms. In the above equation
asleep, inability to concentrate, trouble re- the last term of the denominator controls for
membering things, low energy, tiring easily, this assumption. Second, Cronbach’s alpha as-
and sore or aching muscles. Two measures— sumes that all of the indicators of the underly-
“trouble remembering things” and “sore or ing construct equally contribute to it. While I
aching muscles” due to worry—do not fit eas- have no clear theoretical basis to suspect that
ily into the DSM criteria. Additionally, the this assumption is violated, in practice it is
DSM definition of GAD has a time frame of common. The factor loadings in the present
six months, while these questions cover the analysis demonstrate that this assumption is vi-
past 12 months. While these two points are olated (see Figure 1).
noteworthy, the overall measurement of GAD Distress. Distress is constructed differently
corresponds quite closely to the DSM-IV as- from the three measures of disorder. The ques-
sessment criteria. Generalized anxiety disorder tions for distress were framed within the past
was then constructed as a continuous variable 30 days and asked how much of the time re-
summing the yes responses for the 10 ques- spondents felt “so sad nothing could cheer you
tions and coded so 10 represents the most se- up,” “hopeless,” “worthless,” “nervous,” “rest-
vere symptoms. The mean value of GAD is less or fidgety,” and “that everything was an ef-
quite low as well (see Table 1). fort.” Responses range from “all of the time”
Respondents passing the screen for panic at- (coded 1) to “none of the time” (5). Variables
tacks were administered six items concerning were reverse-coded so that 5 represents nega-
feelings and sensations of heart pounding; tive symptoms “all of the time” and 1 “none of
tightness, pain, or discomfort in the chest or the time.” Based on Mirowsky and Ross’s view
stomach; sweating; trembling or shaking; hot that distress contains two types of symptoms,
flashes or chills; and whether things around mood and malaise, I intercorrelated the first
them seem unreal. “Yes” responses were three items as questions concerning mood, and
summed, and the descriptive statistics are the second three items as questions concerning
shown in Table 1. malaise to model hypothesized covariation.
Conceptually, disorder is the underlying The final sample size is 3,675 cases, with a
construct to test. In order to do so, MDE, GAD, mean score of 1.57 (see Table 1). The smaller
and panic attacks are used to create a single sample size reflects the fact that questions con-
second-order factor for disorder. The internal cerning disorder were asked in the initial tele-
Method Factor, AU = Autonomy, EM = Environmental Mastery, PG = Personal Growth, PR = Positive Relations, PU = Purpose in Life, and SA = Self-Acceptance.
220 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR
phone survey, while the questions concerning worded items, which was the approach of the
distress and mental health were asked in the previous research.
follow-up mail questionnaire. Although the items for a given dimension
It is important to note that these questions are generally distributed throughout the psy-
are quite similar to many of the questions used chological well-being instrument, there are
for disorder, which may raise concerns of op- some inconsistencies. Thus, the second adjust-
erational confounding. However, this concern ment is for a question ordering effect, which is
is attenuated by the fact that the questions are represented by the straight arrow from one
not identical, that they involve very different item to the next in the 18 items used to con-
time frames, and that they do not overlap en- struct psychological well-being. The previous
tirely with any one dimension of disorder. In approach (Springer and Hauser 2006) correlat-
the results I report on a test that created a high- ed a given item with the next item in the ques-
er-order factor composed of distress and disor- tionnaire; however, I believe my approach to be
der in order to explicitly model this possibility. more intuitively appealing because it can be in-
The results provide little evidence to support terpreted, quite literally, as saying a response to
such a model. a given question has an effect on the response
Mental health. Mental health consists of six to the subsequent question (Bollen and
latent variables corresponding to Ryff’s six di- Medrano 1998). Measures of overall model fit
mensions of psychological well-being, and one suggested that the inclusion of these adjust-
second-order factor that synthesizes these six ments improved model fit (data available upon
dimensions. Each of the six dimensions is mea- request).
sured with three items giving a total of 18 in- One final note is in order before I proceed.
dicators (see Appendix A for the wording of all The various latent measures are based on ob-
18 items). Response categories range from served indicators that employ different time
“strongly disagree” (coded 1) to “strongly frames. The measures of disorder are based on
agree” (7). Each dimension ranges from 3 to 21 a 12-month time frame, the measures of dis-
(the sum of three items that range from 1 to 7 tress are based on a 30-day time frame, and the
each). These six dimensions are then used to measures of mental health are based on a sub-
create a single second-order factor of mental jective time frame. This is an artifact of the pre-
health (see Figure 1). vailing measurement norms for each of the
In the descriptive statistics presented below three concepts. At this time, no theoretical or
(Table 2), all negatively worded items have empirical basis leads me to expect bias in a par-
been reverse-coded so that 1 corresponds to ticular direction due to this measurement arti-
poor mental health while 7 corresponds to ex- fact. If there are time-based measurement ef-
cellent mental health. Positively worded items fects, then it would be most suitably addressed
receive higher scores, all with a mean above using longitudinal analyses, which is beyond
5.0, and most above 6.0, while negatively the scope of the present work. On the positive
worded items are slightly lower, with the low- side, these differential time frames may help
est mean (“demands get me down”) equal to further avoid issues of operational confound-
4.23. The means for all six dimensions are ing.
slightly above 16.0, except for personal I use the data analysis program Mplus
growth, which has a mean of almost 18.0 (see (Muthén and Muthén 1998) to estimate maxi-
Table 2). The internal reliability for mental mum likelihood for confirmatory factor analy-
health is .895. ses. I rely on CFA for a number of reasons.
First, I intend to test specific hypotheses re-
Analysis garding the relationships among mental health,
Based on previous research (Springer and disorder, and distress, which lends itself to
Hauser 2006), I introduce two effects for the CFA rather than a “brute force” exploratory
questionnaire design; however, I model the ef- factor-analytic approach. Further, these con-
fects in a slightly different fashion. First, I cepts are measured with a number of different
model a method factor to control for the lower observed variables and modeled as latent fac-
correlations among negatively-worded items tors. Using CFA it is possible to construct sev-
(labeled “MF” in the path diagram, Figure 1). eral latent factors and to test the relationships
This approach is much more common in the among multiple factors simultaneously. Last,
CFA tradition than intercorrelating negatively- CFA controls for measurement error by corre-
lating error terms, modeling the question or- suggests some people have more severe disor-
dering effect, and modeling the question word- ders or more severe symptoms of a disorder
ing effect. This is perhaps the most important than others. Expressions such as “borderline
reason to use CFA because measurement error depression” suggest there is a continuum with
is a key aspect of the evidence used to support a threshold. Finally, if there is bias in treating
the Mirowsky and Ross perspective and to re- disorder as continuous then it is in favor of the
fute the positive psychology perspective. Mirowsky and Ross perspective, which oppos-
A related yet distinct argument concerns es treating disorder as a discrete state.
whether disorder should be treated as continu- Figure 1 presents the path diagram for the
ous or discrete in nature (Mirowsky and Ross test of the relationships among distress, disor-
[1989] 2003; Horwitz 2002a; Horwitz 2002b). der, and mental health, as well as the factor
The one downside of CFA and structural equa- loadings for the model and the correlations
tion approaches is that they treat all latent vari-
among the three key constructs. Each of the
ables as continuous, thus making an assump-
squares represents an observed variable, which
tion about the nature of disorder. From the
SEM perspective, this is a softer assumption is then used to construct the latent variables un-
than it may first appear because even the most der question. Latent variables are signified by
categorical measures often obscure important circles. The present analysis concentrates on
underlying variation. Gender is one example. the three circles labeled “MH,” “Disorder,” and
Quantitative researchers almost uniformly treat “Distress,” which correspond to the concepts
gender as a categorical measure, which ignores under question. The three curved two-headed
the underlying heterogeneity that is one of the arrows between each of these circles represent
important insights of Gender and Queer the three hypotheses presented above. For ex-
Theories stemming from the work of Foucault ample, the curved two-headed arrow between
(1990), Judith Butler (2006), and many others. MH and Disorder represents the hypothesis
In relation to disorder, treating it as continuous that mental health and disorder exist on a sin-
may be a rather soft assumption that simply gle continuum.
associated with changes in distress, while the rectional association of –.63 from distress to
alternative model hypothesizes the opposite. mental health. This suggests changes in dis-
The two models on the second row switch tress tend to be negatively associated with
the direction of the arrows between distress changes in mental health. Finally, and perhaps
and mental health and disorder and mental most interestingly, the parameter from disorder
health. These models suggest that distress and to mental health is quite small, –.03, and this
disorder tend to be associated with changes in effect is not significant (p = .537). This sug-
mental health, but the changes need not vary gests disorder has no clear directional associa-
systematically. These models also test the di- tion with mental health. Counter to the positive
rection of the association between distress and psychology perspective, which views the ab-
disorder. sence of disorder as a necessary but insuffi-
The BIC for each of these models suggests cient condition for mental health, this result in-
good overall fit; however, using the BIC allows dicates that the absence of disorder may be nei-
for adjudication between competing models in ther a necessary nor sufficient condition for
a straightforward manner: the lower the BIC, mental health.
the better the model (Raftery 1995). Of the
four models tested, the model in the second DISCUSSION
row and second column performs the best. The Using the MIDUS, these analyses focused
BIC difference between it and the next best on a perennial debate in the sociology of men-
model is 53.37. A difference greater than 10 is tal health: the relationships among mental
very strong evidence in favor of a particular health, illness, and distress. The goal was to
model (Raftery 1995). Other goodness-of-fit distill the debate into testable hypotheses and
measures give a similar portrait (data available assess these with new empirical evidence in or-
upon request). der to move beyond several impasses in the ex-
The parameter estimates for the best-fitting isting literature. The results yielded several im-
model (see Figure 2) reveal interesting rela- portant findings.
tionships among mental health, disorder, and First, results of the confirmatory factor
distress that are more specific than those found analysis supported a “discontinuous perspec-
in the CFA-based analysis above. There is a tive” that views each concept as distinct rather
strong directional association of .67 from dis- than treating any two or all of them as points
tress to disorder. This suggests that changes in on a single, underlying continuum. These re-
distress tend to be positively associated with sults demonstrate the necessity of studying
changes in disorder. There is also a strong di- each of these phenomena in their own right