Article Text

Download PDFPDF

Sense of coherence and psychological well-being: improvement with age
  1. K W Nilsson1,
  2. J Leppert1,
  3. B Simonsson2,
  4. B Starrin3,4
  1. 1Centre for Clinical Research, Uppsala University, Central Hospital Västerås, Sweden
  2. 2Division of International Health (IHCAR), Karolinska Institutet, Stockholm, Sweden
  3. 3Department of Social Studies, Karlstad University, Karlstad, Sweden
  4. 4Faculty of Health and Social Sciences, Lillehammer University College, Lillehammer, Norway
  1. Correspondence to Kent W Nilsson, Centre for Clinical Research, Uppsala University, Central Hospital, 721 89 Västerås, Sweden; kent.nilsson{at}


Background Psychological well-being is important for individuals, communities and health services throughout the world because of the costs associated with psychological ill-health and the loss of quality of life for those affected and their relatives. Following a salutogenic approach, there is a link between health-promoting resources, such as generalised resistance resources and a positive state of health. Generalised resistance resources have been proposed to relate to an individual's sense of coherence (SOC). The objectives of the present study were (i) to investigate SOC in relation to age and sex, (ii) to investigate psychological well-being in relation to age and sex, and (iii) to investigate the relationship between generalised resistance resources and psychological well-being.

Methods A random sample of 43 598 respondents (54% female) aged 18–85 years participated in the present study via a postal survey questionnaire. SOC was measured by the SOC-13 and well-being by the General Health Questionnaire-12 questionnaire.

Results Males had both stronger SOC and well-being compared to females. There was a relationship between SOC and age, with stronger SOC in the older age groups. There was a larger proportion of individuals who experienced well-being as a function of age. In addition, an increase in SOC was related to a decrease in psychological well-being, that is, a stronger SOC corresponded to higher well-being.

Conclusion Males showed a stronger SOC and more well-being than females. Moreover, SOC and well-being increased with age in both sexes. Our findings suggest that SOC may develop over a entire lifetime.

  • Mental health
  • psychological adaptation
  • sex difference
  • public health
  • age factor
  • population characteristics
  • ageing RB
  • mental health DI
  • population surveys
  • psychological stress
  • public health FQ
View Full Text

Statistics from

The theory of salutogenesis, in which health is seen as a continuum ranging from “total ill health” to “total health”, was introduced by the American-Israeli medical sociologist Antonovsky. He also suggested “sense of coherence” (SOC) to be closely linked to health.1 SOC is a global orientation to view the world and interact with the environment in a comprehensive, manageable and meaningful way.1 According to Antonovsky, a person with a strong SOC, when coping with a stressor, has the ability to find appropriate solutions to a specific situation and to successfully resolve conflicts through adaptability.2 A person with a strong SOC is less likely to perceive stressful situations as threatening and anxiety provoking than a person with a weak SOC.3

Antonovsky believes that SOC is built up during childhood and adolescence and stabilises around the age of 30.1 After 30, an individual's SOC becomes fairly resistant to change. The assumption of stability in SOC has been the subject of discussion,4 and it has been stated that empirical evidences are weak. A recent study discovered that subjects >30 years old have a stronger and more stable SOC than those <30 years old.5 Moreover, in another recent study with a 13-year follow-up, it was found, in line with Antonovsky's hypothesis,1 that individuals with an initial strong SOC had more stable SOC than individuals with an initial weak SOC.6 However, in a recent 5-year follow-up study of negative life events and SOC, it was found that the more recent the life event, the lower the SOC. Initially strong SOC was not more stable than initially weak SOC, and no specific sex differences were discovered.7

Results from studies on the relation between age and SOC are somewhat ambiguous and contradictory. Follow-up studies show small differences in mean values over time.8–12 However, these studies have not usually comprised changes over a long time span.

Cross-sectional studies tend to point in different directions. In a Swedish study, it was middle-aged individuals who had the strongest SOC. Those >65 years had the weakest SOC.13 In another Swedish, as well as in a Danish study,14 15 however, the SOC increased with advancing age.

Although Antonovsky believes that a SOC should be considered as being relatively stable during adulthood, he does not exclude the possibility that it can change if radical and lasting changes in the life situation of the individual occur.1 This idea is in line with his notion of general resistance resources, which give rise to and reinforce a strong SOC.

In a systematic review of the literature, SOC was strongly related to health, especially mental health.16 However, if SOC is considered to be stable after the age of 30, it is not self-evident how the strong correlation between it and mental health should be understood. Does the strong relation between the two presuppose stability in mental health through the years after 30? The reasons why the results from cross-sectional studies investigating the relation between age and SOC point in different directions might be that the populations studied have been too small to consider the dynamic impact of age on SOC.

The objectives of the present study were (i) to investigate SOC in relation to age and sex, (ii) to investigate psychological well-being in relation to age and sex, and (iii) to investigate the relationship between SOC and psychological well-being, and, in the case of such a relationship, to describe it with regard to males and females and different age groups in a large population sample.



In the five Swedish counties of Uppsala, Sörmland, Västmanland, Värmland and Örebro, together comprising approximately 1 400 000 inhabitants, a random sample of 68 000 individuals aged from 18 to 85 years was asked to participate in the present study via a postal survey questionnaire. In all, there were 43 598 respondents (64% response rate), of which 20 060 (46%) were males and 23 529 (54%) were females. The mean age was 52.8 years (SD 18.3). The educational level among participants was compulsory school 63%, college or university ≤2 years 4.4%, and college or university >2 years 19.6%. Self-reported educational information was lacking in 12.7% of the participants. Between the counties of Uppsala, Sörmland, Västmanland, Värmland and Örebro, there was a moderate difference in response rates. More females responded to the questionnaire, with the response rate varying from 56% to 61% among males and 67% to 71% among females. The response rate also varied with respect to age, education and ethnicity. With a rise in age, the response rate increased: 50–56% among those 18 to 34 years old, 58–61% among those 35–49 years old, 66–72% among those 50 to 64 years old, 74–79% among those 65 to 79 years old and 61–69% among those 80 to 84 years old. A higher educational level resulted in a higher proportion of respondents: 57–64% of participants with compulsory school educational level responded, 59–65% with upper secondary school, 64–67% with college or university ≤2 years and 70–75% with college or university >2 years. The response rate was 62–66% among Swedish-born individuals, 60–66% among participants from other Nordic countries and 46–54% among participants from other non-Nordic countries.


Random samples stratified by sex, age and city (and parts of the city for larger cities) were drawn from the total population by Statistics Sweden. The data collection was performed from September to November 2004. The questionnaire was a postal form supplied with prepaid return envelopes. After 10 days, a reminder was sent to the participants who had not responded. Three weeks after the first reminder, a new questionnaire was sent together with a second reminder and a third reminder including a third questionnaire was sent after 10 additional days. The questionnaires were scanned and transformed into a data file with no personal identification of the participants. Therefore, no follow-up study is possible.


The study followed the Swedish guidelines for studies of social science and humanities, according to the Declaration of Helsinki. According to Swedish regulations, this type of study no longer need apply for ethical approval by the medical faculty.

Sense of coherence

The short version of Antonovsky's Sense of Coherence Questionnaire (SOC-13) with the seven grade scale of response alternatives was used.1 After answering the items, each individual was given a SOC score ranging from 13 to 91.

Psychological well-being

As a measure of psychological well-being, the short version of the General Health Questionnaire (GHQ-12) was used. In this study, Goldberg's GHQ scoring method was applied. Here, the responses can score 0–0–1–1, which gives a total sum ranging from 0 to 12. A result of ≥3 was categorised as reduced psychological well-being.17 18 In addition to the GHQ scoring method, the Likert method was also used. In the Likert method, the responses can score 0–1–2–3, which after summation results in a GHQ score ranging from 0 to 36.19 20

Statistical analysis

Differences between males and females in SOC and GHQ Likert scores were analysed using t test. Analysis of variance was applied to study the differences between the individual age groups in SOC and GHQ Likert scores. Differences in the proportions of individuals with ill-health or well-being (GHQ scoring) between the sexes and between the different age groups were investigated by χ2 test. To study the relationship between the variation in SOC and the variation in GHQ, regression analysis was performed.


Sense of coherence

There was a high internal consistency of the SOC-13 items (Cronbach's α=0.878). In the total sample, we found that there was a significant trend for stronger SOC as a function of age, with a difference of approximately 10 points between the youngest age group (18–24 years) compared with the oldest age group (80–84 years). The same pattern was found among males and females when analysed separately. However, the difference between mean SOC scores was somewhat lower among males (table 1). Furthermore, there was a significant difference between the sexes in the total sample, with a stronger SOC among males than females, and this difference between the sexes persisted in most of the analysed age groups (table 1).

Table 1

Sense of coherence mean scores (SD) in the total sample, among males and females, in the different age groups and differences between the sexes

Psychological well-being

The internal consistency of the GHQ-12 items was high (Cronbach's α=0.901). The mean GHQ scores applying the Likert method showed significant differences between the age groups, with a constantly decreasing mean score with increasing age until the group of 65- to 69-year olds, then rising continuously until the oldest age group. A comparison between males and females showed that the GHQ age trend was more noticeable among females. Moreover, females had significantly higher GHQ Likert scores in all age groups (p<0.02 to p<0.001), except among 60- to 64-year olds (not significant).

The mean GHQ score according to the Goldberg method was 1.22 in the total sample, and it was found that the proportion of individuals with reduced psychological well-being differed between the age groups. The highest proportion of reduced well-being was found in the youngest age groups, decreasing up to the group of 65- to 69-year olds before increasing again. The proportion was, however, still lower in the oldest age groups than in the younger groups comprising individuals from 18 to 54 years old (table 2). The percentage of participants with reduced well-being was significantly higher among females than males (χ2=404.035, df=1, p<0.001). This finding persisted when comparing the sexes within each different age group (table 2).

Table 2

The number and proportions of individuals reporting reduced psychological well-being measured by GHQ scores, means (SD) of GHQ Likert method in the total sample, among males and females, and differences in reduced psychological well-being between males and females in the different age groups

Relationship between SOC and GHQ

There was a negative correlation between SOC and GHQ in the total sample (r=−568, adjusted R2=0.322, p<0.001) and among both males and females (r=−555, adjusted R2=0.308, p<0.001 and r=−572, adjusted R2=0.327, p<0.001, respectively) (not shown in Table). Because both the SOC (table 1) and the GHQ (table 2) variables have a strong age-dependent function, the model was reanalysed with age as a covariable. As seen in table 3, age was significantly related to GHQ both among males and females, but did not, however, affect the relation between SOC and GHQ.

Table 3

The relation between SOC and GHQ in the total sample, among males and females, and adjusted for age, education and ethnicity

The model was reanalysed adjusting for demographic covariates. Both education and ethnicity had a significant relationship with GHQ (p<0.01) in separate analyses. Therefore, adjustments were made for the level of education, that is, (i) elementary school, 9 years; (ii) secondary school, 11–12 years; (iii) post-secondary school, 13–14 years; and (iv) university for ≥3 years. Ethnicity, however, coded as three dummy variables ((i) Sweden, (ii) other Nordic countries, (iii) other non-Nordic countries), was not significant when entering the final model. As shown in table 3, in the final model excluding ethnicity, both age and education had a significant relation to GHQ. There were, however, minor effects on the relationship between SOC and GHQ among both males and females.

In all age groups, and among both sexes, there was a significant relationship between SOC and GHQ (p<0.001; table 4). As discernable from table 4, males tended to have a somewhat weaker relationship between SOC and GHQ in the older age groups, whereas females showed a somewhat more complex pattern.

Table 4

The relation between SOC and GHQ in the different age groups among males and females

In figure 1, the relationship between age, well-being and SOC is illustrated. In the group of individuals who experience well-being, SOC increased with increasing age in a linear fashion among both males and females. The same effect could be discerned among both males and females in the group with reduced psychological well-being, although not in the same strictly linear fashion, possibly as a result of the smaller group size. There was, however, a notable difference between mean SOC scores with a level consistently approximately 20 points stronger among those that experienced well-being compared to those with reduced well-being.

Figure 1

Mean sense of coherence values among males and females, according to age and psychological well-being.


In this study of a large population sample, we found that SOC scores and the proportion of individuals experiencing well-being increased with age. Moreover, we found stronger SOC scores and a lower proportion of individuals with reduced psychological well-being among males compared with females in almost all age groups. In addition to the age-dependent function of both SOC and GHQ, we found an inverse relationship between SOC and GHQ, that is, psychological well-being increased with a stronger SOC. This was still true after adjustment for GHQ-related variables such as age, education and ethnicity.

The present study has several limitations. First, it has a cross-sectional design that restricts the conclusions that can be drawn, particularly those regarding cause and effect must be considered tentative. A prospective study spanning several decades would have been the ideal choice for the evaluation of SOC and GHQ stability in different age groups and the direction of their effects. It is also possible that the present results merely describe different cohort effects because of the different life experiences of the older and younger participants.

Second, an optimal balance between the sensitivity and specificity of the GHQ-12 measurement would have required a separate validity study. The threshold scores we used were chosen to obtain as high a specificity as possible while still ensuring a high sensitivity. We should not, therefore, overestimate the reduced psychological well-being in the study population.

Third, the overall response rate was not optimal (64%). We also found differences in the response rates among different groups—for example, those of males, younger participants, the less educated and individuals born outside of Sweden were all lower. This might have had an impact on our results as each of these variables affected both SOC and GHQ to some extent.

Regarding non-respondents, even if they had been included in the study, this would probably not have weakened our analysis because it is reasonable to hypothesise that they probably have weaker SOC and higher GHQ scores, and this would have strengthened the results. Unfortunately, because of the study design, it was not possible to do a thorough non-response analysis to confirm this.

It is also worth stating that the weaker SOC found among females could to some extent be explained by the higher non-response rate among males, if the non-responding males have a significantly weaker SOC than those that did respond.

Fourth, there was an internal non-response rate of 4% in the GHQ questions index and 8% in the SOC. We decided not to artificially modify these missing data with mean or median values because there were enough individuals in each group to perform the present analysis. Moreover, a transformation of this kind is more questionable than the lower power because of non-responses.

Finally, both SOC and GHQ might be related to several demographic or psychosocial variables that were not investigated in the present study. Such variables could have confounding effects on our model and might explain other relationships between SOC and GHQ.

The limitations of the present study might be balanced by its statistical power. Overall, there were 43 598 respondents, and the subgroup sizes ranged from 978 to 2361 among males and from 1378 to 2206 among females. These subgroup sizes equal total population samples of many other studies and overrule the risk for errors of random findings in small sample designs.

Initially, Antonovsky's theory came from interviewing Israeli women about their adaptation to menopause. He studied a group of women who had remained healthy despite their experiences in the holocaust concentration camps of the Second World War.1 He suggested this to be a result of the view they had of their lives and of the essence of their existence. Three components emerged: comprehensibility, manageability and meaningfulness. It was suggested that an individual's SOC was developed and their location on the SOC continuum more or less fixed by early adulthood, with only minor and temporary changes coming through later life experiences.1 In another recent cross-sectional study of our group, boys had a significantly stronger SOC than girls in a sample of 4305 16- and 19-year-old adolescents. Both sexes had lower SOC scores than the youngest age group (18–24 years old) in the present study.21 Thus, the results from adolescent individuals are in accordance with the timeline of SOC development in the present study. Other cohort studies have investigated SOC stability over time and its relation to health, disease and psychosocial changes10 22. Contrary to the present results, a decrease in SOC scores was found in the oldest age group (45–74 years), and during a 5-year follow-up, individuals with the lowest initial scores experienced the largest decrease in SOC. These findings are in line with the prediction (hypothesis?) that individuals who have the weakest SOC, that is, the lowest self-esteem, are thereby disadvantaged and have an affected ability to make balanced judgements and cope with situations successfully. In the long run, this affects their SOC negatively.4 It has, however, also been shown that a strong SOC is in fact no more stable than a weak one.7 Although the present study is not based on individual prospective data, it suggests that, on a population level, SOC continuously develops until the mid-70s.

There are several studies that have investigated GHQ, and different threshold levels have been proposed. For the GHQ-12, the suggestions range from as low as 1/2 up to 6/7.18 In a cross-national study of primary care settings in 15 cities, it was observed that for a mean GHQ <1.85, a threshold of 1/2 seemed to work best; for a GHQ of 1.85–2.7 and >2.7, these thresholds were 2/3 and 3/4, respectively.17 In the present study, the mean value was 1.22 by the GHQ scoring method. In the total population, a score of 0–2 was regarded as a marker for psychological well-being and the cut-off value for reduced well-being was a score of ≥3. This choice of threshold certainly reduces sensitivity; on the other hand, a higher threshold increases specificity. Therefore, our results concerning the proportion of psychological ill-health should not be an overestimation.

SOC has been related to GHQ in other studies. The SOC of severe multiple trauma patients was weakly related to GHQ at admission, highly related at discharge and moderately to highly related at follow-up 1 to 3 years after the trauma.23 In the present study, we found SOC and GHQ to be moderately to highly related in the total population (r=−568), with an explained variance of 32%. Variables that are associated with both SOC and GHQ, such as age, education and ethnicity, did not affect the relationship between SOC and GHQ. Moreover, the relationship between SOC and GHQ was stable in all age groups, both among males and females. Theoretically, SOC should have an impact on psychological well-being (GHQ). However, it is likely that psychological well-being also has an impact on SOC24 and that the two concepts are intermingled.

In a systematic review of the literature, SOC was strongly related to health, especially mental health. SOC has a main role in the explanation of health; however, there are other contributing factors such as age, social support and education.25 Our results therefore agree with the review of Eriksson and Lindström25 in that there is a possible link between health promoting resources, by which SOC is strengthened and a positive subjective state of health is developed.

This study has demonstrated a strong association between age and SOC up to the age of 80 years. In one way, this opposes Antonovsky's own hypotheses. On the other hand, such an association seems explicable when considering his notion of general resistance resources. It is assumed that general resistance resources such as income, social network, education and culture give rise to and reinforce a strong SOC. One possible explanation for the strong link between age and SOC in the present study might be that individuals continue to develop and refine their pool of generalised resistance resources up to old age.26 Furthermore, there was a strong negative relation (approximately −0.5 to −0.6) between SOC and GHQ, although not autocorrelated. Although both SOC and GHQ were related to age, we interpret the present results as showing a relation between SOC and age, independently of GHQ; that is, although some individuals had reduced psychological well being, and subsequently some 20 points lower SOC scores, the age–SOC relation remained reasonably constant.

The present study identifies an interesting paradox. Sweden has a good standard of living and one of the world's highest life expectancies and levels of health, all of which tend to increase with each generation. However, in spite of a high standard of living and better physical health, the youngest age groups have the lowest psychological well-being. Compared to many other countries, the Swedish attempts to achieve sex equality have shown far-reaching effects. Our results, though, indicate significant sex differences in SOC and well-being. A simple explanation for this could be a sex and age bias in the measurements of SOC and GHQ. Another more complex explanation could be that Swedish society, despite a high standard of living, good health care and high sex equality, has failed to generate health promoting resources. Seemingly, there is still much to do, especially among the young and among females.

To summarise, this study investigated SOC and GHQ among males and females, in different age groups. Males had a stronger SOC and higher well-being compared to females. SOC showed a trend to increase with the passing of years, and perceived psychological well-being followed the same general pattern. Furthermore, SOC explained a reasonable proportion of the variation in psychological well-being among 18- to 85-year olds.

What is already known on this topic

  • A SOC is developed until early adulthood.

  • A relation between SOC and health, especially mental health, has been proposed.

  • Differences between males and females in their SOC and psychological well-being have been suggested.

What this study adds

  • Population-based normative data for SOC and GHQ-12 in different age groups from a large study sample.

  • Support for a continuously increasing SOC and psychological well-being throughout adult life.

  • The support for a substantial relationship between SOC and psychological well-being even after adjusting for demographic variables.

  • A confirmation of differences between males and females in self-reports of both SOC and psychological well-being.


We wish to thank the departments of community medicine of the county councils of Uppsala, Sörmland, Västmanland, Värmland and Örebro for giving us the opportunity to use these data to perform the present study.


View Abstract


  • Funding Grants and funds from the county councils of Västmanland and Värmland are acknowledged. The sponsors of the study had no role in data analysis, data interpretation or writing of the report.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.