Professional Documents
Culture Documents
The Differentiation of Self Inventory, Development and Initial Validation PDF
The Differentiation of Self Inventory, Development and Initial Validation PDF
0022-0167/98/S3.00
236
tions. Thus, we undertook development of the Differentiation of Self Inventory (DSI) to create a self-report instrument for adults, age 25+, capable of (a) testing theoretical
assumptions, (b) assessing individual differences in adult
functioning, and (c) evaluating psychotherapeutic outcomes
from a systemic perspective. By defining adulthood with a
lower limit of 25 years of age, we sought to ensure that the
samples obtained consisted of those individuals who, from a
family life cycle perspective (Carter & McGoldrick, 1988),
could be considered adults (i.e., postcollege or working,
living apart from the parental home, and largely financially
independent).
To adequately measure differentiation, we included both
the intrapsychic and interpersonal components, that is, the
thinking-feeling and separateness-togetherness dimensions.
Historically, transgenerational theorists (e.g., BoszormenyiNagy & Ulrich, 1981; Framo, 1992) have described individual and family functioning solely in terms of interpersonal and intergenerational family processes. Self-report
instruments developed within this tradition include Kear's
(1978) Differentiation of Self Scale, the Emotional Cutoff
Scale (McCollum, 1991), the Family-of-Origin Scale (Hovestadt, Anderson, Piercy, Cochran, & Fine, 1985), and the
Personal Authority in the Family System Questionnaire
(Bray, Williamson, & Malone, 1984). Although each represents an important contribution to the field, none attempts to
operationalize the range of interpersonal components of
differentiation (i.e., fusion to emotional cutoff), and none
focuses on the intrapsychic aspects of differentiation (see
Bowen, 1978; Kerr & Bowen, 1988).
For example, Kear's (1978) Differentiation of Self Scale
consists of three factors: Separation of Thinking and Feeling, Emotional Maturity, and Emotional Autonomy; yet
items reflect only interpersonal components of differentiation and ignore quality of relations with spouse or partner.
The Differentiation of Self Scale suffers also from significant methodological limitations. For instance, a factor
analysis used to create its subscales was conducted on 72
initial items using only 50 participants (see Nunnally, 1978).
McCollum's (1986, 1991) Emotional Cutoff Scale is an
excellent measure of the degree to which respondents
manage their emotional attachment to each parent through
cutoff. Yet its limited focus on relations with parents ignores
the presence of emotional cutoff in current significant
relationships as well as other aspects of differentiation. To
respond to the Family of Origin Scale (Hovestadt et al.,
1985), adults provide retrospective perceptions of their
family of origin relations, whereas adolescents are asked to
give their current perceptions of relations with family (e.g.,
Niedermeier, Handal, Brown, Searight, & Manley, 1992).
The retrospective ratings emphasize the past and ignore the
respondent's current relations with family members. And
although the Personal Authority in the Family System
Questionnaire (Bray et al., 1984) includes items about
current relationships, it neglects the concept of emotional
cutoff as well as the intrapsychic aspects of Bowen's (1976,
1978) concept of differentiation.
There also exist several self-report measures of separationindividuation based on object relations theory (e.g., Hoff-
DIFFERENTIATION OF SELF
man, 1984; Levine, Green, & Millon, 1986; Olver, Aries, &
Batgos, 1990). These separation-individuation measures
were designed for use with late adolescents rather than
adults, and none contain items that deal with marital
relations or that reflect problems in achieving a balance
between intimacy and autonomy. The concept of differentiation, as defined by Bowen (1976, 1978), is often misinterpreted in the family therapy literature and equated with
individuation or autonomy. Although similar in some respects, separation-individuation is not equivalent to differentiation of self. Individuation, from an object relations
perspective (e.g., Bios, 1975; Mahler, Pine, & Bergman,
1975), involves the achievement of independence and a
unique sense of identity. Differentiation of self is the
capacity to maintain autonomous thinking and achieve a
clear, coherent sense of self in the context of emotional
relationships with important others.
To create the DSI, a series of studies was undertaken
based on three different samples. The purpose of these
studies was to develop and validate the DSI using a construct
approach to test construction (e.g., Jackson, 1970; Jackson
& Messick, 1958; Loevinger, 1957; NunnalLy, 1978). Jackson's recommendations for personality scale development
were used to construct items that would adequately reflect
the domain (i.e., differentiation of self), be clear and
unambiguous, be relatively free of social desirability bias
and other content biases, have high discriminatory power,
and, as a set, sufficiently represent the underlying construct
of differentiation (Jackson, 1970).
Study 1
The purpose of this study was to create the DSI. First,
definitions, descriptions, and examples from Bowen (1976,
1978; Anonymous, 1972) and his successors (Kerr, 1985;
Kerr & Bowen, 1988; Nichols, 1984; Nichols & Schwartz,
1998; Papero, 1990) were used to generate a pool of items
that exemplify differentiation of self. Items (N = 96) generated by our research team reflected the ability to distinguish
and balance (a) thinking and feeling and (b) the capacity for
intimacy with and autonomy from others in current important relationships as well as with parents and siblings.
Differentiation was operationalized in a multidimensional
fashion, given that Bowen (1976, 1978) described many
components of differentiation in his writings. Further, Gurman (1978) argued that differentiation, like any complex
psychological construct, is inherently multidimensional. We
used a principal-components analysis to identify the DSFs
dimensionality and determine final item selection. Theoretical relations between differentiation and chronic anxiety
were tested to assess the initial construct validity of the DSI
(i.e., Bowen's proposition that poorly differentiated individuals also experience more chronic anxiety).
Method
Participants. Participants were adults (A^ =313) living in New
York, Ohio, and California, including (a) randomly selected faculty
and staff at a large state university, (b) parents of children on
237
238
Method
Participants, Adults (n = 169, 111 women and 58 men), age
25+, who were employed at a large northeastern state agency, took
part in the research. Participants averaged 42.34 years of age
(SD = 8.59). The majority were married (70.2%; M - 15.04
years), 13.7% were single, 6.3% were unmarried and living with a
partner, and 9.5% were separated or divorced. In terms of ethnicity,
90.4% were White, 5.4% African American, 0.6% Asian American,
0.6% Latino-Latina, 0.6% Native American, and 2.4% other.
Approximately 15% of participants were currentLy in therapy; 45%
had sought therapy in the past.
Instruments. The DSI used in Study 2 contained 78 items
constituting four subscales: Emotional Reactivity, I Position,
Emotional Cutoff, and Fusion With Others. To rate each item,
respondents used a 6-point Likert-type scale, ranging from not at
all true of me (1) to very true of me (6). Crowne and Marlowe's
(1964) Social Desirability Scale (SDS), a 33-item true-false
self-report measure, was used to estimate the tendency to describe
oneself in favorable terms. Internal consistency reliability has been
estimated at .88, with test-retest correlations at .88 and .89
(Crowne & Marlowe, 1960,1964; Robinson & Shaver, 1973).
Procedure. Adults, age 25+, who were employed at a large
northeastern state agency, took part in the research. Clustersampling procedures were used to randomly select 2 departments
from a large northeastern state agency, out of 32 total from which to
solicit participants. Three hundred potential participants were
contacted by interagency mail. Each packet included a cover letter
explaining the voluntary and anonymous nature of the research, the
two counterbalanced questionnaires, and a demographic sheet. The
study was described as "focusing on adults' interpersonal relationships, relationships with family members, and (their) general
attitudes." Participants returned completed packets by mail in
sealed envelopes. One hundred sixty-nine participants returned
completed questionnaires, for a 56% return rate.
Results
Item analyses. Statistical analyses were performed at
the item level to discern the DSI's inherent factor and to
ensure that each subscale was homogeneous and distinct
from the other three subscales (Campbell & Fiske, 1959;
Jackson, 1970). All items met a priori criteria for response
distribution (i.e., items with skewness and kurtosis values
between 1.5 and 1.5 and SDs >: 1 were retained). Thirtyfive items were eliminated due to low item-scale correlations (i.e., items with item-subscale correlations <.45 were
eliminated). Seven items were rekeyed because they loaded
highly on another subscale and demonstrated good discrimination between subscales and because the Bowen experts we
consulted suggested that those items corresponded more
highly to that respective subscale.
No additional items were eliminated on the basis of
criteria for evaluating social desirability bias. Correlations
between DSI items and social desirability scores ranged
from -.15 to .49. None of the remaining 43 items were
found to lower the internal consistency reliability of their
239
DIFFERENTIATION OF SELF
respective subscales; thus, no additional items were eliminated. (Tables illustrating item response distributions, skewness and kurtosis values, item correlations with the SDS,
item-subscale correlations, and item reliability analyses are
available from Elizabeth A. Skowron.)
Description of the DSL The resulting 43-item DSI (see
Appendix) contains four subscales: Emotional Reactivity, I
Position, Emotional Cutoff, and Fusion With Others. The
11-item Emotional Reactivity subscale reflects the degree to
which a person responds to environmental stimuli with
emotional flooding, emotional lability, or hypersensitivity.
The I Position subscale contains 11 items that reflect a
clearly defined sense of self and the ability to thoughtfully
adhere to one's convictions when pressured to do otherwise.
The 12-item Emotional Cutoff subscale reflects feeling
threatened by intimacy and feeling excessive vulnerability
in relations with others. Items reflect fears of engulfment and
behavioral defenses like overfunctioning, distancing, or
denial. Finally, the 9-item Fusion With Others subscale
reflects emotional overinvolvement with others, including
triangulation and overidentification with parents.
To compute the DSI full-scale score, raw scores on all
items in the Emotional Reactivity, Emotional Cutoff, and
Fusion With Others subscales and on one item in the I
Position subscale (#35) are reversed, so that higher scores
signify greater differentiation. Scores on all items are then
summed and divided by the total number of items, so that the
full-scale score ranges from 1 (low differentiation) to 6 {high
differentiation). To facilitate comparison of full-scale and
subscale scores, each subscale is also computed by reversing
respective items, summing item scores, and then dividing by
the number of items in the subscale (Emotional Reactivity = 11, I Position = 11, Emotional Cutoff = 12, Fusion
With Others = 9). Scores on each subscale thus range from
1 to 6, with higher scores reflecting greater differentiation.
Descriptive statistics. All scores were normally distributed; subscale means ranged from 2.07 to 4.34 (full-scale
M = 3.73, SD = 0.58; see Table 1). Subscale-full-scale
correlations were moderate to high, ranging from .43
(Fusion With Others) to .80 (Emotional Reactivity), all ps <
.001. Intercorrelations among the subscales were low to
moderate, ranging from .08 (Fusion With Others and I
Position) to .53 (Fusion With Others and Emotional Reactivity; see Table 2). Correlations between DSI subscales and
SDS scores were negligible to moderate {r .42 for Emotional Reactivity, r = .49 for I Position, r = .34 for Emotional Cutoff, and r = - .02 for Fusion With Others).
Cronbach's alpha was used to estimate internal consistency
reliabilities for the DSI full scale and each of the four
subscales (DSI a = .88, Emotional Reactivity a = .84; I
Position a = .83, Emotional Cutoff a = .82; Fusion With
Others a = .74).
Study 3
After the DSI subscale revisions in Study 2, a third sample
was obtained to evaluate the DSI's factor structure using
confirmatory factor analyses and to test theoretically predicted relations between differentiation of self, psychological symptoms, and marital satisfaction. It was hypothesized
that (a) significant inverse relationships between symptomatology and the DSI subscales would support Bowen's
(1976,1978) assumption that highly differentiated individuals are more free of symptoms and generally better adjusted
and (b) significant positive relationships between marital
satisfaction and the DSI subscales would support Bowen's
(1976,1978; Kerr & Bowen, 1988) proposition that individuals with higher levels of differentiation establish more
satisfying marriages.
Method
Participants. A total of 127 adults (118 employees and 9
spouses) participated, with only 91 married adults completing the
Dyadic Adjustment Scale (DAS). Participants were 53 men and 73
women (1 gender unspecified), with an average age of 42.23 years
(SD 10.22, range 25-72 years). Married respondents constituted
59.5% of the sample (M = 12.72 years married, SD = 9.74). Of the
remainder, 15.9% were single, 9.4% were unmarried and living
with a partner, 13.5% were separated or divorced, and 1,6% were
widowed. More than half of the participants (61.8%) were parents
(M = 2.13 children, SD = 0.94). In terms of ethnicity, 90.5% of the
sample were White, 4.0% African American, 2.4% Asian American, 1.6% Latino-Latina, and 0.8% Native American. Ten percent
Table 1
Total
Men
Women
Scale
SD
SD
SD
ER
IP
EC
FO
DSI
3.35
4.01
4.34
2.97
3.73
0.90
0.83
0.87
0.88
0.58
3.37
4.08
4.53
2.92
3.74
0.94
0.85
0.79
0.85
0.60
3.69a
4.24
4.44
3.05
3.87
0.88
0.90
0.77
0.89
0.55
SD
3.18b
3.97
4.61
2.82
3.64
0.92
0.81
0.81
0.82
0.61
Note. For Study 2, n = 169; for Study 3, n = 127 (53 men, 73 women, 1 unspecified). DSI =
Differentiation of Self Inventory; ER = Emotional Reactivity; IP = I Position; EC = Emotional
Cutoff; FO = Fusion With Others. Scores range from 1 to 6. Higher scores on all scales reflect greater
differentiation of self. Means in the same row that do not share subscripts differ at p < .01.
240
Table 2
Intercorrelations Among Subscales
and Subscale-Full-Scale Correlations
Scale
1
Study 2
1.
2.
3.
4.
5.
DSI
ER
IP
EC
FO
.80**
.61**
.64**
.43**
.46**
.27*
.53**
.31*
.08
-.12
.28*
.12
-.04
Study 3
1.
2.
3.
4.
5.
DSI
ER
IP
EC
FO
.84**
.69**
.58**
.52**
.53**
.25
.48**
of participants were currently in therapy; 52.70% had past experience in psychotherapy, predominantly (59.40%) in individual
treatment.
Instruments. The four subscales in the 43-item DSI were used
as predictor variables. In the Study 3 sample, internal consistency
reliabilities were moderate to high and similar to those obtained in
previous studies (DSI a = .88, Emotional Reactivity a = .88; 1
Position a = .85; Emotional Cutoff a = .79; Fusion With Others
a = .70).
The Hopkins Symptom Checklist (Derogatis, Lipman, Rickels,
Uhlenhuth, & Covi, 1974) is a well-known self-report measure
assessing psychological symptomatology on five dimensions: Somatization, Obsessive-Compulsive, Interpersonal Sensitivity, Depression, and Anxiety. Items are rated on a 4-point Likert-type
scale, ranging from 1 (not at all) to 4 (extremely), reflecting the
degree of distress experienced within the past 7 days. Its General
Severity Index (GSI) reflects intensity of distress independent of
the number of symptoms endorsed (Derogatis, Yevzeroff, &
Wittelsberger, 1975). The GSI is sensitive to symptom changes
over the course of psychotherapy (Rickels et al., 1971) and is used
most often to provide a summary measure of symptomatology
(Derogatis et al., 1974). The GSI is computed by summing the five
raw symptom subscales and dividing by 58; scores range from 1 to
4, with higher scores indicating greater symptomatology. Internal
consistency reliabilities range from .84 to .87, and test-retest
coefficients range from .75 to .84 (Derogatis et al., 1974). GSI
scores in the present sample ranged from 1.0 to 2.6 (M = 1.5,
SD = 0.33).
Spanier's (1976) DAS assesses relationship discord and overall
marital satisfaction. The DAS yields a total score ranging from 0 to
151, with higher scores reflecting better marital adjustment.
Internal consistency reliability of the DAS full-scale score has been
reported at .96 (Spanier, 1976). Construct validity is supported by
significant correlations with other well-known measures of marital
adjustment and by results showing that divorced couples score
significantly lower than married couples (Spanier, 1976, 1988).
Scores in the present sample ranged from 50 to 150 (M = 104,
SD = 17.9). On the basis of the accepted DAS cutoff score of 98
(Eddy, Heyman, & Weiss, 1991; Jacobson et al., 1984), 30 married
participants (30.90%) in Study 3 were classified as maritally
distressed.
Results
Preliminary analyses. Means and standard deviations
on the four DSI subscale and full-scale scores were highly
similar to those obtained in Study 2 (see Table 1). As in
Study 2, scores were normally distributed, subscale intercorrelations were moderate (Emotional Reactivity and I Position r .53; Emotional Reactivity and Fusion With Others
r = .48) to negligible (e.g., Fusion With Others and Emotional Cutoff r= - .04; see Table 2).
Factor analyses. We conducted a confirmatory factor
analysis, using Lisrel 7, to evaluate the four-factor structure
of the DSI. Given the limitations inherent in using Lisrel
procedures when fitting models with large numbers of
single-item indicators such as the DSI, an item-clustering
procedure was used to increase the stability of the indicators
(c.f. Anderson & Gerbing, 1988; Joreskog & Sdrbom, 1986;
MacCallum, 1986). For each of the 4 DSI subscales, single
items were randomly summed into meta-items, comprising 3
or 4 items each, resulting in 3 indicators per subscale, or 12
indicators in all. Table 3 contains the correlation matrix
among the 12 confirmatory factor analysis indicator variables.
Fit indices used to evaluate the model included the
goodness-of-fit index (GFI), adjusted goodness-of-fit index
(adjusted GFI), a chi-square to degrees of freedom ratio
(X2'4O, and the root mean squared of the residuals (RMS).
GFI values greater than .90, adjusted GFI values greater than
.80, a x W r a t i o less than 2.0, and RMS values less than . 10
indicate a well-fitting model (cf, Cole, 1987; Joreskog &
Sorbom, 1986; Marsh & Hocevar, 1985).
The four-factor model of differentiation, corresponding to
the four DSI subscales, was tested. Indicators expected to
load on each factor were stated in advance (Nunnally, 1978).
Each of the indicators was permitted to load freely on its
respective factor and was constrained to 0 on the other
factors. Each latent variable was scaled to the first indicator
by fixing its value to 1.00. A maximum-likelihood solution
was used to fit the model to the data. The fit of this
four-factor model was good, x2(48, N = 137) = 89.35,
p < .0001, GFI = .91, adjusted GFI = .85, ^idf = 1.86,
RMS = .07.
Figure 1 illustrates the four latent variables (Emotional
Reactivity, I Position, Emotional Cutoff, and Fusion With
Others), the individual indicator loadings on the latent
variables, and the residual error terms. Correlations among
the four factors were negligible to moderate, ranging from
241
DIFFERENTIATION OF SELF
Table 3
Correlation Matrix for Confirmatory Factor Analyses
7
2
3
4
5
6
Variable
1
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
ER,
ER2
ER3
IP!
IP2
IP3
ECi
EC2
EC3
FO!
FO2
FO 3
.74
.76
.35
.49
.45
.63
.24
.22
.42
.40
.26
.16
.06
.34
.37
.32
.27
.13
.39
.33
.33
.35
.46
.53
.26
.24
.18
.30
.31
.32
.70
.47
.20
.28
.06
.01
.13
-.05
.64
.23
.18
.01
.02
.21
.05
10
11
12
.62
.02
-.06
-.05
-.01
-.12
-.07
.49
.48
.50
.32
.39
.15
.00
.14
.09
.51
.50
.07
.04
-.08
General Discussion
Our aim, to construct a reliable, valid self-report measure
of differentiation of self for adults using a construct approach to test development, was realized in these investigations. While a portion of the items were created or reworded
during the Study 2 revision of the DSI, its multidimensional
structure was retained. Confirmatory factor analyses demonstrated support for the DSI subscales, Emotional Reactivity,
I Position, Emotional Cutoff, and Fusion with Others, as
identifiable, empirically distinct dimensions of a single
construct, differentiation of self. Subsequent analyses with
the DSI subscales supported the internal consistency reliability and initial construct validity of the measure. Tests of
Bowen theory supported the hypothesized relations between
self-reported differentiation, symptomatology, and marital
satisfaction, providing important psychometric support for
the DSI.
Differentiation of self, estimated by the DSI subscales,
correlated significantly with amount and intensity of symptomatic distress. The unique predictors of global maladjustment were Emotional Reactivity and Emotional Cutoff, the
242
.17
r
ER1
.36
1'
32
1r
ER3
ER2
.83
.15
if
^f
IP1
IP2
.51
1r
IP3
.92
34
.58
EC1
er,
.44
EC2
.82
EC3
.52
.51
.50
FO1
FO2
FO3
.70
Figure 1. Study 3: four-factor model confirmatory factor loadings, error variances, and correlations
among the latent variables (ER = Emotional Reactivity; IP = I Position; EC = Emotional Cutoff;
FO = Fusion With Others).
DIFFERENTIATION OF SELF
243
244
References
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC:
Author.
Anderson, J. C , & Gerbing, D. W. (1988). Structural equation
modeling in practice: A review and recommended two-step
approach. Psychological Bulletin, 103, 411423.
Anonymous. (1972). Toward the differentiation of a self in one's
own family. In J. Framo (Ed.), Family interaction (pp. 111-173).
New York: Springer.
Bios, P. (1975). The second individuation process of adolescence.
New York: International Universities Press.
Boszormenyi-Nagy, I., & Ulrich, D. N. (1981). Contextual family
therapy. In A. S. Gurman & D. P. Kniskera (Eds.), Handbook of
family therapy (pp. 159-186). New York: Brunner/Mazel.
Bowen, M. (1976). Theory in the practice of psychotherapy. In P. J.
Guerin, Jr. (Ed.), Family therapy: Theory and practice (pp.
42-90). New York: Garner Press.
Bowen, M. (1978). Family therapy in clinical practice. New York:
Jason Aronson.
Bray, J. H., Harvey, D. M., & Williamson, D. S. (1987). Intergenerational family relationships: An evaluation of theory and measurement. Psychotherapy, 24, 516-528.
Bray, J. H., Williamson, D. S., & Malone, P. E. (1984). Personal
authority in the family system: Development of a questionnaire
to measure personal authority in intergenerational family processes. Journal of Marital and Family Therapy, 10, 167-178.
Campbell, D. X, & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, 81-105.
Carter, B., & McGoldrick, M. (1988). The changing family life
cycle: A framework for family therapy (2nd ed.). New York:
Garner Press.
Chodorow, N. (1978). The reproduction of mothering: Psychoanalysis and the sociology of gender. Berkeley: University of California Press.
Cole, D. A. (1987). Utility of confirmatory factor analysis in test
validation research. Journal of Consulting and Clinical Psychology, 55, 584-594.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in
psychological tests. Psychological Bulletin, 52, 281-302.
Crowne, D., & Marlowe, D. (1960). A new scale of social
desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349-354.
Crowne, D., & Marlowe, D. (1964). The approval motive. In D. P.
Crowne (Ed.), The experimental study of personality (pp.
153-183). New York: Wiley.
52, 497-504.
DIFFERENTIATION OF SELF
245
{Appendix follows)
246
Appendix
Differentiation of Self Inventory
These are questions concerning your thoughts and feelings about yourself and relationships with others. Please read each statement carefully and decide
how much the statement is generally true of you on a 1 (nor at all) to 6 (very) scale. If you believe that an item does not pertain to you (e.g., you are not
currently married or in a committed relationship, or one or both of your parents are deceased), please answer the item according to your best guess about
what your thoughts and feelings would be in that situation. Be sure to answer every item and try to be as honest and accurate as possible in your responses.
Not at all
Very true
true of me
of me
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
1
1
I
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
1
1
1
2
2
2
3
3
3
4
4
4
5
5
5
6
6
6
1
1
1
1
1
1
1
1
I
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
2
2
2
2