SWLS Properties

The SWLS was developed using classical test construction approaches to both select an initial item pool and subsequently narrow the instrument to five items that inquire about respondents' overall assessment of their lives. Generally the cognitive appraisal involved in measuring life satisfaction is believed to require a comparison by individuals between the current status of their lives and self-defined expectations regarding what they would like their lives to be. This comparison may be in relation to a self-defined ideal, to other people, or to one's own past. The SWLS does not measure satisfaction with specific domains of life (e.g., family, employment, and income). Combining domain-specific ratings to attain a global indication of life satisfaction has been criticized for assuming that individuals place equal value on all life areas. Summing or other weightings imposed by the instrument lose the subjective valence of the domains.

Content Validity
The initial item selection for the SWLS included 48 items that all had face validity as indicators of one's appraisal of life. Initial factor analysis revealed a three-factor structure, with 10 items loading highly (greater than .60) on a factor reflecting cognitive-judgmental evaluative processes. Five of these items were considered redundant in wording or word choice, thus leading to the current five items of the SWLS.

Reliability
Initial and subsequent studies have examined the internal consistency of the SWLS and alpha coefficients have repeatedly exceeded .80 (see Pavot & Diener, 1993). For instance, Diener and colleagues' original study found an alpha coefficient equaling .87 for 176 undergraduates at the University of Illinois.

Similarly, test-retest reliabilities have been generally acceptable (see Pavot & Diener, 1993), though lower than desirable scores result when the time span between test and re-test is longer and situational influences affect responses. In the original validation study Diener and colleagues found two-month test-retest correlation coefficients equal to .82 for 76 students re-tested from their original sample. Alfonso and Allison (c.f., Pavot & Diener, 1993) reported a test-retest correlation coefficient of .89 for subjects retested after 2 weeks.

In the initial validation study Diener and colleagues factor analyzed the five-item scale and, based on a scree test of eigen values, concluded that it represented a single factor. That factor accounted for 66% of the variance in the instrument. Again, subsequent studies have replicated these factor analytic findings. Item to total score correlations have ranged from .57 to .66. When comparing individual item factor loadings, as well as item to total score correlations, the first item is consistently the most associated, while the fifth item is frequently the least well associated. However, factor loadings in the subsequent analyses all exceeded the minimum criterion of .60 used by Diener and colleagues in the original development.

Allen Heinemann and colleagues at the Rehabilitation Institute of Chicago (personal communication) computed a Rasch analysis on 777 respondents to the TBI State Demonstration Grant needs assessment conducted in Illinois. As with previous factor analyses, Item #5 had the poorest fit and the validity of its inclusion may need to be examined for persons with TBI.

Criterion-related Validity
The original validation studies correlated the SWLS with ten other measures of subjective well being. Most measures correlated at an r = .50 or higher for each of the two samples from the original work. Subsequent studies have found comparable or higher correlations with other populations when interviewer ratings, informant reports, or other objective measures (e.g. the Andrews/Withey scale, Fordyce global scale) are used.

Construct Validity
Evidence for the construct validity of the SWLS can be drawn from a number of investigations. For instance, test-retest stability has been found to decline as time between testing increases, suggesting that the instrument is sensitive to changes that occur with life and not just a direct effect of stable personality traits. Additionally, the SWLS has shown consistent differences between populations that would be expected to have different qualities of life (e.g., psychiatric patients, or male prison inmates). The SWLS has also been found to change in the expected directions in response to major life events, such as elderly caregivers who had a spouse diagnosed with primary degenerative dementia (Vitaliano et al, 1991), and patients receiving psychotherapy (Friedman reported in Pavot & Diener, 1993).

The SWLS has been administered in conjunction with measures of positive and negative affective appraisal. The SWLS tends to correlate with scales measuring both constructs even though the two are unrelated. The absolute values of these correlations range from .26 to .47, indicating that the SWLS taps a dimension of subjective well being different from either positive or negative affectivity. Also consistent with theoretical postulations about subjective well being, SWLS scores have been found to be positively correlated with extroversion and inversely correlated with neuroticism (Diener et al, 1985; Pavot & Diener, 1993, Pavot, Diener, Colvin & Sandvik, 1991).

Normative Data
Corrigan, Smith-Knapp & Granger (1998) reported a mean of 19.0 and standard deviation of 7.6 for a cross-sectional sample of 98 patients with TBI six months to 5 years post-discharge from acute rehabilitation. Time post-injury was significantly associated with higher SWLS total score.

Corrigan et al (1999) reported a mean and standard deviation of 19.58 (8.29) 1 year post-injury and 20.98 (9.39) 2 years post-injury for 179 patients admitted to a specialized brain injury rehabilitation unit. Higher life satisfaction was associated with employment, less depression, the absence of a prior history of substance abuse, greater social integration, and lower Glasgow Coma Scale score at emergency department admission. Changes in SWLS scores from Year 1 to Year 2 were associated with depression and marital status.

Sokol et al (1999) reported a mean of 16.2 for a sample of 875 persons with TBI living in the community a median of 7 years post-injury. The sample was identified from the state vocational rehabilitation roles, mailing lists of the Brain Injury Association of Illinois, and brain injury rehabilitation providers in the state. A linear transformation of the Total score using Rasch analysis revealed associations between higher life satisfaction and lower levels of unmet needs, having a spouse or partner, the perception that the TBI had a minimal impact on one's life, having more than a high school education, the perception that one's friendships, living situation and employment had not changed for the worse, better emotional status since TBI onset, and greater time post-injury.

The TBI Model Systems database as of April, 2000 included a limited number of cases with SWLS total scores. Descriptive statistics are provided in the following table. (The SWLS was added to the TBI Model Systems database April 4, 1998.)

	1 year post-injury	2 years post-injury	3 years post-injury	4 years post-injury	5 years post-injury
N	211	107	84	63	85
Mean	20.14	20.38	20.90	20.70	21.06
SD.	8.47	8.32	8.67	8.51	8.19
Kurtosis	-1.12	-1.22	-1.05	-.875	-1.24
Minimum	5	5	5	5	6
Maximum	35	35	35	35	35
25th percentile	13.00	13.00	12.25	15.00	14.00
50th percentile	21.00	21.00	23.00	21.00	21.00
75th percentile	27.00	28.00	27.75	28.00	29.00

See Pavot and Diener (1993) for data on a variety of subpopulations, including students with disabilities.

See Consortium for Spinal Cord Medicine (1999) for data on persons with traumatic spinal cord injury.