ABSTRACT
Aim
Loneliness is common in later life and may be shaped by aging stereotypes. Drawing on Stereotype Embodiment Theory, this study examined whether the components of aging‐stereotype internalization predict loneliness independently of sociodemographic factors and whether effects differ by dimension.
Methods
In this cross‐sectional study, 485 community‐dwelling adults aged 65 or older in Turkey completed the Elderly Internalized Stigma Scale (EISS), indexing awareness, agreement, and self‐application, and the Social and Emotional Loneliness Scale for Adults (SELSA‐S). Spearman's correlations and multiple linear regression were used, with separate models for total, emotional, and social loneliness.
Results
Self‐application showed the strongest bivariate association with loneliness (ρ = 0.25, p < 0.001), whereas agreement was unrelated. It was independently associated with total loneliness (β = 0.23) and both dimensions (β = 0.18 emotional, β = 0.21 social; all p < 0.001), whereas awareness was not. Agreement showed a negative coefficient consistent with suppression. Being unmarried was the strongest correlate of emotional loneliness (β = 0.55) but was unrelated to social loneliness (β = 0.04), for which self‐application had the largest standardized coefficient among the included predictors. Separating the components explained more variance than the composite score.
Conclusion
Loneliness was associated with applying aging stereotypes to oneself rather than with awareness or general agreement, and this held for both dimensions. Being married was associated with lower emotional loneliness but not with social loneliness. Findings for social loneliness should be interpreted cautiously given the low internal consistency of that subscale. Interventions targeting self‐directed stereotypes merit evaluation.
Keywords: aged, ageism, loneliness, self concept, stereotyping
In 485 community‐dwelling older adults, loneliness was associated with applying aging stereotypes to oneself, not with awareness of them or general agreement. Self‐application was independently associated with both emotional and social loneliness, suggesting that interventions should target the self‐relevant component of internalization rather than awareness alone.
1. Introduction
Globally, one in six people experiences loneliness, which contributes to roughly 871 000 deaths each year, and the burden falls disproportionately on older adults [1]. Loneliness is not simply the absence of social contact; it reflects a subjective gap between the relationships a person has and those they desire [2], and social loneliness, an unmet need for a broader network, need not move together with emotional loneliness, the absence of close bonds [2, 3]. In later life it is linked to sex, employment, education, and network size [4, 5, 6], and marital status is among its most robust correlates, with loss of a spouse exerting a particularly pronounced effect [7, 8]. In a Turkish community sample, ageism and loneliness were both negatively associated with successful aging [9]. Most of these risk factors are events that may be difficult to reverse once they have occurred; beliefs about one's own aging differ in this respect, being acquired over the life course, open to psychosocial modification, and specific to old age.
Ageism comprises the stereotypes, prejudice, and discrimination directed at people on the basis of age [10], and since the term was introduced [11] a central insight has been that such attitudes may be turned inward, toward oneself. Stereotype Embodiment Theory specifies how this inward form develops: negative age stereotypes are acquired across the life course before they become personally relevant, can operate without awareness, and are embodied in old age as self‐perceptions of aging that are held to affect health through psychological, behavioral, and physiological pathways [12]. Applied to loneliness, the theory implies a behavioral route in which older adults who apply aging stereotypes to themselves expect rejection, take part in fewer activities, and withdraw from contact, each response widening the gap between the relationships they have and those they want [12, 13].
Testing this requires attention to measurement. The instrument used here distinguishes three sequential components of age‐stereotype processing [14]: awareness, agreement, and self‐application, of which self‐application most directly represents the stage at which aging stereotypes become personally relevant; endorsing a stereotype while excluding oneself from the group it describes is a documented alternative response [15]. Summing the three therefore combines distinct stages and can obscure their separate associations with an outcome. Empirical work is consistent with a self‐relevant route: age‐based devaluation appears to reach loneliness partly through the individual's own aging self‐perceptions [13, 16], and internalized aging stereotypes have been linked to poorer mental health with loneliness as a mediator [17]. Which part of the internalization process carries the association has not been established, and because international loneliness policies have focused mainly on high‐income countries [18], the gap is especially significant in middle‐income settings.
1.1. Study Aims
This study aimed to examine the association between the internalization of aging stereotypes and loneliness in community‐dwelling older adults and to determine whether it predicts loneliness independently of sociodemographic factors. Because the loneliness measure distinguishes emotional from social loneliness, each dimension was also modeled separately, and the three stereotype components were examined both separately and as a composite.
2. Methods
2.1. Study Design and Sampling
This cross‐sectional, correlational study was conducted among adults aged 65 and over living in the city center of Kütahya, Turkey, drawn from the 80 370 residents of Kütahya in that age group recorded by the Turkish Statistical Institute in 2021 [19]. An a priori G*Power analysis for a nine‐predictor multiple regression, assuming α = 0.05, power = 0.80, and a small incremental effect (f 2 = 0.02) for two tested predictors, indicated a required sample of 485 [20]; a sensitivity analysis for the achieved sample is reported in the Supporting Information. Participants were recruited with equal allocation (about 162 per stratum) from three neighborhoods classified by the Kütahya Provincial Governorate as low, middle, or high in socioeconomic level. Volunteers meeting the eligibility criteria were enrolled through communal areas identified by local headmen until each quota was filled; selection within strata was not random, so this is a nonprobability sample. Neighborhood socioeconomic level structured recruitment only; participants' own perceived income adequacy was recorded separately.
2.2. Participants
The study included community‐dwelling volunteers aged 65 and over who resided in the city center of Kütahya and were literate in Turkish. Individuals who might have difficulty responding to scale items due to a diagnosis of dementia, legal guardianship, or sensory impairment were excluded.
2.3. Data Collection Instruments
Data were collected through face‐to‐face interviews between June 20 and September 30, 2025, with each interview lasting approximately 25–30 min. Three instruments were used:
The Sociodemographic Characteristics Form comprised 13 items covering age, sex, education, marital status, perceived income adequacy (income below, equal to, or above expenses; referred to below as low, middle, and high income), children and grandchildren, cohabitation, employment, experience of age‐based discrimination, chronic illness, frequency of healthcare use, and functional dependence, the last recorded with a single self‐report item on whether help from another person was needed for daily activities.
The Elderly Internalized Stigma Scale (EISS), developed by Karasoy and Baydur [14] from a progressive model of self‐stigma [21], assesses three sequential components of age‐stereotype processing on a 3‐point format: stereotype awareness (5 items), agreement (5 items), and self‐application (7 items). Following the original scoring, items were reverse‐coded and each subscale linearly transformed to a 0–100 metric, with the total taken as the mean of the three subscales; higher scores indicate greater internalization. In the present sample α was 0.86 for the total scale, 0.81 for awareness, 0.77 for agreement, and 0.74 for self‐application, against 0.89, 0.82, 0.67, and 0.88 in the development study [14]; the self‐application subscale was thus less consistent here than in the original. Because α is downwardly biased for three‐category items, McDonald's ω was also computed: 0.92 for the total scale, 0.88 for awareness, 0.78 for agreement, and 0.81 for self‐application.
The Social and Emotional Loneliness Scale for Adults (SELSA‐S), developed by DiTommaso and Spinner [2], follows Weiss's distinction between emotional and social isolation [3]. The 15‐item short form yields a social and an emotional subscale, the latter combining the family and romantic items, which together form the total. Items are rated on a 7‐point scale, with totals ranging from 15 to 105 and higher scores indicating greater loneliness. In the Turkish adaptation, α was 0.83 for the total scale, 0.82 for the social subscale, and 0.76 and 0.85 for the family and romantic components of the emotional subscale [22]; in the present sample it was 0.79 for the total scale and 0.80 for the emotional dimension but only 0.49 for the five‐item social dimension, and findings for that dimension are interpreted with the caution set out in the Section 4.4.
2.4. Data Analysis
Data were analyzed in IBM SPSS Statistics 26.0, and the factor analyses reported as Supporting Information in JASP 0.19. Bivariate associations were examined with Spearman's rank‐order correlations, and group differences with independent‐samples t‐tests (Welch‐corrected where appropriate) and one‐way ANOVA with Games–Howell post hoc tests, reporting Cohen's d and η 2. Internal consistency was estimated with Cronbach's α and McDonald's ω.
Because the EISS is used here at the subscale level, its structure was examined in the same sample by ordinal confirmatory factor analysis with WLSMV and by exploratory factor analysis with minimum residual extraction and oblique rotation; full results are reported in the Supporting Information and their implications in Section 4.5.
Loneliness was modeled by multiple linear regression for total, emotional, and social loneliness. For each outcome the primary model entered the three EISS components separately, because, per Stereotype Embodiment Theory [12], self‐application most directly reflects internalization; a secondary model entered the composite score alone, for comparability with prior total‐score research, and the two were compared on adjusted R 2. Covariates were fixed in advance from established correlates of later‐life loneliness, not chosen on bivariate significance: age, sex, marital status, income, employment, and functional dependence [4, 5, 6, 7, 8]. Living arrangement was omitted because it overlaps with marital status; educational attainment, a plausible confounder of both, was entered in a sensitivity analysis rather than the primary model to limit the number of predictors. Categorical predictors were dummy‐coded; because the low‐ and high‐income groups did not differ on SELSA‐S scores (p = 0.536), they were merged and used as the reference. Linearity, multicollinearity, influential observations, independence of residuals, normality of residuals, and homoscedasticity were examined; the first four raised no concern, whereas residual variance was uneven and residuals were positively skewed, as noted in Section 4.5. Significance was set at p < 0.05.
3. Results
3.1. Descriptive Statistics
Sociodemographic characteristics are given in Table 1 and descriptive statistics for the EISS and SELSA‐S scores in Table 2. Skewness and kurtosis fell within ±1 [23], but because normality tests flagged minor departures and the items are ordinal, Spearman's coefficients were used.
TABLE 1.
Comparison of Social and Emotional Loneliness and Elderly Internalized Stigma Scale Mean Scores by sociodemographic characteristics of participants (N = 485).
| Variable | n (%) | SELSA‐S (M ± SD) | SELSA‐S test | EISS (M ± SD) | EISS test |
|---|---|---|---|---|---|
| Age group | |||||
| 65–74 | 366 (75.5) | 37.42 ± 15.45 |
t = −0.21 p = 0.830 d = 0.02 |
63.76 ± 24.12 |
t = −0.68 p = 0.490 d = 0.07 |
| ≥ 75 | 119 (24.5) | 37.82 ± 18.27 | 65.43 ± 22.78 | ||
| Sex | |||||
| Male | 287 (59.2) | 37.17 ± 16.68 |
t = −0.566 p = 0.572 d = 0.052 |
63.59 ± 24.90 |
t = −0.651 p = 0.515 d = 0.059 |
| Female | 198 (40.8) | 38.02 ± 15.43 | 65.00 ± 22.10 | ||
| Educational attainment | |||||
| Primary school | 258 (53.2) | 39.78 ± 16.66 |
F(2, 482) = 5.573 p = 0.004** η 2 = 0.023 |
71.66 ± 20.88 |
F(2, 482) = 40.522 p < 0.001*** η 2 = 0.144 |
| Secondary school | 145 (29.9) | 35.30 ± 16.36 | 60.30 ± 23.47 | ||
| University | 82 (16.9) | 34.30 ± 12.96 | 47.44 ± 22.88 | ||
| Games–Howell post hoc, SELSA‐S: primary > secondary (p = 0.010), primary > university (p = 0.002). EISS: primary > secondary (p < 0.001), primary > university (p < 0.001), secondary > university (p < 0.001). | |||||
| Income level | |||||
| Low | 162 (33.4) | 35.41 ± 15.22 |
F(2, 482) = 4.796 p = 0.009** η 2 = 0.020 |
63.45 ± 21.80 |
F(2, 482) = 0.392 p = 0.676 η 2 = 0.002 |
| Middle | 161 (33.2) | 40.66 ± 16.62 | 63.53 ± 25.66 | ||
| High | 162 (33.4) | 36.49 ± 16.22 | 65.52 ± 23.76 | ||
| Games–Howell post hoc, SELSA‐S: middle > low (p = 0.003), middle > high (p = 0.024). | |||||
| Marital status | |||||
| Married | 360 (74.2) | 32.98 ± 14.18 |
t = −11.880 p < 0.001*** d = 1.233 |
63.13 ± 23.16 |
t = −1.632 p = 0.103 d = 0.169 |
| Unmarried | 125 (25.8) | 50.58 ± 14.42 | 67.16 ± 25.34 | ||
| Having children | |||||
| Yes | 427 (88.0) | 37.13 ± 15.93 |
t = −1.402 p = 0.161 d = 0.196 |
64.45 ± 23.86 |
t = 0.706 p = 0.481 d = 0.099 |
| No | 58 (12.0) | 40.31 ± 17.71 | 62.09 ± 23.31 | ||
| Having grandchildren | |||||
| Yes | 373 (76.9) | 36.69 ± 15.84 |
t = −2.045 p = 0.041* d = 0.220 |
65.37 ± 23.19 |
t = 2.041 p = 0.042* d = 0.220 |
| No | 112 (23.1) | 40.25 ± 17.01 | 60.15 ± 25.34 | ||
| Living arrangement | |||||
| With spouse | 264 (54.4) | 32.86 ± 14.21 |
F(2, 482) = 57.607 p < 0.001*** η 2 = 0.193 |
63.53 ± 23.55 |
F(2, 482) = 1.875 p = 0.154 η 2 = 0.008 |
| Living alone | 130 (26.8) | 36.92 ± 14.19 | 62.48 ± 23.04 | ||
| Other | 91 (18.8) | 51.87 ± 15.93 | 68.42 ± 25.10 | ||
| Games–Howell post hoc, SELSA‐S: other > with spouse (p < 0.001), other > living alone (p < 0.001), living alone > with spouse (p = 0.023). EISS: no significant pairwise differences. | |||||
| Employment status | |||||
| Retired/not employed | 453 (93.4) | 37.02 ± 15.99 |
t = −2.561 p = 0.011* d = 0.468 |
64.05 ± 23.82 |
t = −0.398 p = 0.691 d = 0.073 |
| Employed | 32 (6.6) | 44.56 ± 17.26 | 65.79 ± 23.51 | ||
| Experience of age‐based discrimination | |||||
| Yes | 118 (24.3) | 37.84 ± 18.90 |
t = 0.224 p = 0.823 d = 0.027 |
71.89 ± 20.75 |
t = 4.442 p < 0.001*** d = 0.435 |
| No | 367 (75.7) | 37.41 ± 15.21 | 61.68 ± 24.19 | ||
| Functional dependence | |||||
| Yes | 121 (24.9) | 41.93 ± 18.82 |
t = 3.114 p = 0.002** d = 0.367 |
74.44 ± 18.93 |
t = 6.372 p < 0.001*** d = 0.592 |
| No | 364 (75.1) | 36.05 ± 14.93 | 60.75 ± 24.28 | ||
| Chronic illness | |||||
| Yes | 287 (59.2) | 37.09 ± 16.02 |
t = −0.691 p = 0.490 d = 0.064 |
67.38 ± 23.23 |
t = 3.614 p < 0.001*** d = 0.334 |
| No | 198 (40.8) | 38.13 ± 16.40 | 59.52 ± 23.86 | ||
| Frequency of healthcare use | |||||
| Several times/week | 51 (10.5) | 35.47 ± 15.57 |
F(2, 482) = 1.243 p = 0.290 η 2 = 0.005 |
72.72 ± 17.82 |
F(2, 482) = 5.005 p = 0.007** η 2 = 0.020 |
| Several times/month | 365 (75.3) | 37.33 ± 15.88 | 63.95 ± 24.21 | ||
| Never | 69 (14.2) | 40.00 ± 17.86 | 58.99 ± 23.80 | ||
| Games–Howell post hoc, EISS: several times/week > several times/month (p = 0.003), several times/week > never (p < 0.001). | |||||
Note: N = 485. Values are M ± SD. Two‐group comparisons used independent‐samples t‐tests (Welch‐corrected where variances were unequal) with Cohen's d; three‐group comparisons used one‐way ANOVA with η 2 and Games–Howell post hoc tests.
Abbreviations: EISS = Elderly Internalized Stigma Scale; SELSA‐S = Social and Emotional Loneliness Scale for Adults.
p < 0.05.
p < 0.01.
p < 0.001.
TABLE 2.
Spearman's rank‐order correlations among EISS and SELSA‐S total and subscale scores (N = 485).
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|---|
| EISS—Stereotype awareness | 58.35 | 33.51 | 1.00 | ||||||
| EISS—Stereotype agreement | 83.13 | 25.07 | 0.41** | 1.00 | |||||
| EISS—Stereotype self‐application | 51.02 | 28.95 | 0.64** | 0.35** | 1.00 | ||||
| EISS—Total | 64.17 | 23.83 | 0.89** | 0.61** | 0.86** | 1.00 | |||
| SELSA‐S—Social loneliness | 12.99 | 6.22 | 0.19** | −0.01 | 0.24** | 0.19** | 1.00 | ||
| SELSA‐S—Emotional loneliness | 24.52 | 12.83 | 0.13* | −0.02 | 0.21** | 0.16** | 0.28** | 1.00 | |
| SELSA‐S—Total | 37.51 | 16.20 | 0.18** | −0.03 | 0.25** | 0.19** | 0.62** | 0.90** | 1.00 |
Note: N = 485. Values are Spearman's rank‐order correlation coefficients (ρ). The Elderly Internalized Stigma Scale (EISS) comprises three subscales (stereotype awareness, stereotype agreement, and stereotype self‐application) and a total score (0–100). The SELSA‐S yields social and emotional loneliness subscales and a total score. Correlations between each total and its own subscales are part–whole associations and are reported for completeness.
p < 0.01.
p < 0.001.
3.2. Comparison of Loneliness and Internalized Aging Stereotypes by Sociodemographic Characteristics
Mean SELSA‐S and EISS scores by sociodemographic characteristics are presented in Table 1; loneliness varied most with marital status, unmarried participants scoring substantially higher than married ones (d = 1.23, the largest effect in the study), and with living arrangement (η 2 = 0.19), whereas internalized stereotypes differed by neither and varied most with education (η 2 = 0.14). Age was unrelated to both outcomes, whether treated as continuous (ρ = −0.05 and 0.06) or as two groups (65–74, n = 366, versus 75 or older, n = 119; all p > 0.05).
3.3. Relationships Among the EISS and SELSA‐S Dimensions
The three stereotype dimensions related to loneliness in clearly different ways (Table 2). Self‐application showed the strongest associations, with the SELSA‐S total (ρ = 0.25) and its social (ρ = 0.24) and emotional (ρ = 0.21) subscales, all p < 0.001; awareness was more weakly related (ρ = 0.18, p < 0.001), and agreement was unrelated bivariately (ρ = −0.03).
3.4. Subscale‐Level Predictors of Total, Emotional, and Social Loneliness: Multiple Linear Regression
Regression results are given in Tables 3, 4, 5. All three models were significant, F(9, 475) = 28.32, 37.50, and 6.32, respectively, all p < 0.001, but differed considerably in explanatory power, accounting for 33.7%, 40.4%, and 9.0% of the variance (adjusted R 2). For each outcome the corresponding model using the EISS total score explained less (adjusted R 2 = 0.287, 0.369, and 0.056), so separating the composite improved prediction throughout. Adding educational attainment as a covariate changed nothing of substance: no education term was significant and self‐application remained a significant predictor of all three outcomes.
TABLE 3.
Multiple linear regression predicting total loneliness from sociodemographic covariates, age, and EISS subscales (N = 485).
| Variable | B | SE | β | t | p | 95% CI |
|---|---|---|---|---|---|---|
| Age | −0.152 | 0.107 | −0.054 | −1.416 | 0.157 | [−0.363, 0.059] |
| Sex (ref: female) | 3.518 | 1.291 | 0.107 | 2.725 | 0.007** | [0.981, 6.055] |
| Income level (ref: low or high income) | 4.148 | 1.334 | 0.121 | 3.109 | 0.002** | [1.527, 6.769] |
| Marital status (ref: married) | 16.797 | 1.426 | 0.454 | 11.775 | < 0.001*** | [13.994, 19.599] |
| Employment status (ref: not employed) | 6.541 | 2.438 | 0.100 | 2.683 | 0.008** | [1.750, 11.332] |
| Functional dependence (ref: no) | 3.806 | 1.462 | 0.102 | 2.603 | 0.010* | [0.933, 6.679] |
| EISS—Stereotype awareness | 0.043 | 0.024 | 0.089 | 1.778 | 0.076 | [−0.005, 0.091] |
| EISS—Stereotype agreement | −0.119 | 0.027 | −0.184 | −4.419 | < 0.001*** | [−0.172, −0.066] |
| EISS—Stereotype self‐application | 0.126 | 0.028 | 0.226 | 4.511 | < 0.001*** | [0.071, 0.181] |
| R 2 = 0.349, adjusted R 2 = 0.337, F(9, 475) = 28.32, p < 0.001 | ||||||
Note: N = 485. Dependent variable: total loneliness (SELSA‐S). VIF values ranged from 1.02 to 1.84, indicating no multicollinearity, and no observation exceeded a Cook's distance of 1. For comparison, the composite model (same covariates + EISS total score instead of subscales) yielded Adjusted R 2 = 0.287 with EISS total β = 0.13 (p = 0.001); decomposing the EISS into subscales increased explained variance and revealed divergent subscale effects (self‐application positive, agreement negative, awareness nonsignificant).
Abbreviations: B: unstandardized coefficient; CI: confidence interval; SE: standard error; β: standardized coefficient.
p < 0.05.
p < 0.01.
p < 0.001.
TABLE 4.
Multiple linear regression predicting emotional loneliness from sociodemographic covariates, age, and EISS subscales (N = 485).
| Variable | B | SE | β | t | p | 95% CI |
|---|---|---|---|---|---|---|
| Age | −0.119 | 0.081 | −0.053 | −1.481 | 0.139 | [−0.278, 0.039] |
| Sex (ref: female) | 2.293 | 0.969 | 0.088 | 2.366 | 0.018* | [0.389, 4.198] |
| Income level (ref: low or high income) | 3.484 | 1.001 | 0.128 | 3.479 | 0.001** | [1.516, 5.452] |
| Marital status (ref: married) | 16.246 | 1.071 | 0.554 | 15.170 | < 0.001*** | [14.141, 18.350] |
| Employment status (ref: not employed) | 3.426 | 1.831 | 0.066 | 1.872 | 0.062 | [−0.170, 7.023] |
| Functional dependence (ref: no) | 2.626 | 1.098 | 0.089 | 2.392 | 0.017* | [0.469, 4.783] |
| EISS—Stereotype awareness | 0.026 | 0.018 | 0.069 | 1.443 | 0.150 | [−0.010, 0.062] |
| EISS—Stereotype agreement | −0.084 | 0.020 | −0.165 | −4.164 | < 0.001*** | [−0.124, −0.045] |
| EISS—Stereotype self‐application | 0.081 | 0.021 | 0.183 | 3.864 | < 0.001*** | [0.040, 0.122] |
| R 2 = 0.415, Adjusted R 2 = 0.404, F(9, 475) = 37.50, p < 0.001 | ||||||
Note: N = 485. Dependent variable: emotional loneliness (SELSA‐S). The same predictors were entered as in Table 3. VIF values ranged from 1.02 to 1.84, indicating no multicollinearity, and no observation exceeded a Cook's distance of 1. For comparison, the composite model (same covariates + EISS total score instead of subscales) yielded Adjusted R 2 = 0.369 with EISS total β = 0.09 (p = 0.014).
Abbreviations: B: unstandardized coefficient; CI: confidence interval; SE: standard error; β: standardized coefficient.
p < 0.05.
p < 0.01.
p < 0.001.
TABLE 5.
Multiple linear regression predicting social loneliness from sociodemographic covariates, age, and EISS subscales (N = 485).
| Variable | B | SE | β | t | p | 95% CI |
|---|---|---|---|---|---|---|
| Age | −0.033 | 0.048 | −0.030 | −0.677 | 0.499 | [−0.127, 0.062] |
| Sex (ref: female) | 1.225 | 0.580 | 0.097 | 2.111 | 0.035* | [0.085, 2.365] |
| Income level (ref: low or high income) | 0.663 | 0.599 | 0.050 | 1.107 | 0.269 | [−0.514, 1.841] |
| Marital status (ref: married) | 0.551 | 0.641 | 0.039 | 0.859 | 0.391 | [−0.709, 1.810] |
| Employment status (ref: not employed) | 3.114 | 1.096 | 0.125 | 2.843 | 0.005** | [0.962, 5.267] |
| Functional dependence (ref: no) | 1.180 | 0.657 | 0.082 | 1.796 | 0.073 | [−0.111, 2.471] |
| EISS—Stereotype awareness | 0.017 | 0.011 | 0.091 | 1.547 | 0.123 | [−0.005, 0.038] |
| EISS—Stereotype agreement | −0.035 | 0.012 | −0.141 | −2.876 | 0.004** | [−0.059, −0.011] |
| EISS—Stereotype self‐application | 0.045 | 0.013 | 0.210 | 3.583 | < 0.001*** | [0.020, 0.070] |
| R 2 = 0.107, adjusted R 2 = 0.090, F(9, 475) = 6.32, p < 0.001 | ||||||
Note: N = 485. Dependent variable: social loneliness (SELSA‐S). The same predictors were entered as in Table 3. VIF values ranged from 1.02 to 1.84, indicating no multicollinearity, and no observation exceeded a Cook's distance of 1. For comparison, the composite model (same covariates + EISS total score instead of subscales) yielded adjusted R 2 = 0.056 with EISS total β = 0.15 (p = 0.001).
Abbreviations: B: unstandardized coefficient; CI: confidence interval; SE: standard error; β: standardized coefficient.
p < 0.05.
p < 0.01.
p < 0.001.
The EISS effects were consistent across outcomes. Self‐application independently predicted total loneliness (β = 0.23) and both dimensions (β = 0.18 emotional, β = 0.21 social), all p < 0.001, whereas awareness predicted none (p = 0.076, 0.150, and 0.123). Agreement was a significant negative predictor in all three models (β = −0.18, −0.17, and −0.14) once the other subscales were controlled; because it was unrelated bivariately, this is best read as statistical suppression, discussed below.
The sociodemographic predictors behaved differently across the two dimensions. Marital status was the strongest predictor of total loneliness (unmarried β = 0.45) and dominated the emotional dimension (β = 0.55, both p < 0.001) but was unrelated to social loneliness (β = 0.04, p = 0.391), whereas employment showed the reverse pattern, predicting social loneliness (β = 0.13, p = 0.005) but not the emotional dimension (β = 0.07, p = 0.062). Male sex predicted all three outcomes, and income and functional dependence the total and emotional models; age predicted none. Within the social loneliness model, self‐application was the strongest predictor of any variable entered.
4. Discussion
This study examined how internalized aging stereotypes relate to loneliness in community‐dwelling older adults. When the construct was separated into its components, the association was carried by self‐application alone, and modeling the two loneliness dimensions separately showed that this holds for both, whereas the sociodemographic predictors are dimension‐specific.
4.1. Stereotype Self‐Application as the Active Internalization Mechanism
Self‐application predicted loneliness independently of sociodemographic factors and age, whereas awareness did not and agreement did not at the bivariate level. This is what Stereotype Embodiment Theory predicts: an older adult may recognize aging stereotypes, and even regard them as broadly true, without reporting greater loneliness; higher loneliness was observed primarily when they were applied to the self [12, 17]. The pattern also explains why the composite score performed less well in every outcome, since summing the three components dilutes the contribution of self‐application with two that do not share its relationship to loneliness.
The finding specifies the route described by existing work on ageism and loneliness [13, 16]. What accounts for the association is not exposure to stereotypes, nor general endorsement of them, but their application to oneself.
The negative coefficient for agreement warrants caution. Agreement was unrelated to loneliness on its own and turned negative only when self‐application and awareness were controlled, which points to statistical suppression rather than a protective effect. One possible reading is that older adults who regard aging stereotypes as broadly true but decline to apply them to themselves are distancing themselves from the stereotyped group rather than internalizing it [15], and the coefficient would be consistent with such dissociation. The recurrence of the coefficient across the loneliness outcomes indicates that it is not confined to a single scoring specification; because those outcomes are correlated and partly overlapping, however, this does not constitute independent replication. Dissociation was not measured in this study, so the reading remains a hypothesis requiring direct assessment.
4.2. Differentiated Effects on Emotional and Social Loneliness
Self‐application was associated with both dimensions and agreement retained its negative coefficient in both, but the models differed elsewhere. Marital status dominated emotional loneliness, of which the model explained 40%, whereas it did not predict social loneliness at all; there the model explained 9% and self‐application was the strongest predictor entered. The contrast partly follows from what the dimensions measure: emotional loneliness reflects the absence of a close attachment figure [3] and marital status lies close to that definition. It may also reflect item‐content overlap, since the emotional score combines family and romantic items and the married–unmarried difference was far larger on the romantic component, though it remained moderate on the family component (Table S7). Social loneliness, by contrast, was related to self‐application and to continued employment rather than to partnership, consistent with the behavioral route specified by Stereotype Embodiment Theory [12, 13]. Because the social model explained far less variance and its outcome had limited internal consistency (see Section 4.5), these results support the conclusion that self‐application was associated with both dimensions, not that it was more strongly associated with one.
4.3. Sociodemographic Correlates
Marital status was the strongest sociodemographic correlate of loneliness, consistent with systematic‐review [4] and cross‐national evidence [24], and the dimension‐specific analysis places that association in emotional loneliness; living arrangement followed the same logic, indicating that the presence of others does not compensate for the absence of an intimate bond [2, 25]. Male sex was associated with higher loneliness on both dimensions in the adjusted models, consistent with norms discouraging male disclosure [26, 27], education was inversely related to both outcomes and most strongly to stereotypes [12, 28, 29], and functional dependence predicted both [30, 31]. The remaining associations, involving employment, income [32], chronic illness [33], and having grandchildren, were outcome‐specific, rested in some cases on small groups, and were unexpected in direction; they were not central to the study hypotheses and require replication.
4.4. Clinical and Policy Implications
The findings suggest that interventions should address the application of aging stereotypes to the self rather than awareness alone, an approach on which programs based on Stereotype Embodiment Theory already draw [34, 35]. Because some older adults already decline to apply stereotypes they otherwise regard as true, such programs may build on a response that arises spontaneously rather than having to create it [15]. Such work may complement programs targeting the social determinants of loneliness. Because the study was cross‐sectional, intervention effects require evaluation in longitudinal or experimental research.
4.5. Limitations
The cross‐sectional design precludes causal inference, so the direction of the relationship between self‐application and loneliness requires longitudinal study [36]. Measurement warrants particular caution. This is the first large‐scale independent application of the EISS, and although the confirmatory model showed acceptable fit, the exploratory analysis indicated instability and cross‐loadings within the self‐application subscale; the two analyses were run in the same sample, so the confirmatory findings are an internal examination of the proposed structure rather than an independent validation. The three components were strongly intercorrelated and should be read as related stages of one process rather than as independent constructs, and measurement invariance across older‐adult subgroups remains to be established. Imprecision of this kind would generally attenuate rather than inflate the associations reported. The social subscale of the SELSA‐S also showed low internal consistency, which may partly account for the smaller variance explained in that model, so those findings need replication with a more reliable measure. The negative agreement coefficient, interpreted here as suppression, requires independent replication with a direct measure of stereotype dissociation [15]. Functional dependence was indexed by a single self‐report item and reflects perceived need for assistance rather than measured ability. Residuals were positively skewed and heteroscedastic, so the standard errors and confidence intervals in Tables 3, 4, 5 should be read as approximate. Sampling was not probability‐based: the sample was limited to literate, urban older adults and quotas were filled by volunteers approached in communal areas, so those who seldom leave home are probably underrepresented. Finally, the models left substantial variance unexplained.
5. Conclusion
In this sample of community‐dwelling older adults, self‐application was, among the three components of aging‐stereotype processing, the one most consistently associated with loneliness: it was independently associated with emotional and social loneliness alike after adjustment for sociodemographic factors and age, whereas neither awareness nor general agreement was. Marital status was the strongest correlate of emotional loneliness but was unrelated to the social dimension, for which self‐application had the largest standardized coefficient among the included predictors. Loneliness in later life therefore appears to depend less on exposure to aging stereotypes than on the degree to which older adults apply them to themselves. Psychosocial interventions addressing self‐application, alongside the social determinants of loneliness, may be worth developing and evaluating, and longitudinal, psychometrically rigorous studies are needed to confirm these effects and establish the measurement model.
Author Contributions
Rahime Aslan: conceptualization, methodology, data collection, formal analysis, investigation, writing – original draft, writing – review and editing, supervision, project administration. Mahsun Toplu: data collection, methodology, investigation, writing – review and editing.
Funding
This research was supported by the Scientific and Technological Research Council of Türkiye (TÜBİTAK) under the University Students Research Projects Support Program (2209‐A; Grant No: 1919B012464103).
Ethics Statement
This study was approved by the Non‐Interventional Clinical Research Ethics Committee of Kütahya Health Sciences University (Decision No: 2025/06‐47; Date: May 6, 2025). Institutional permission was additionally obtained from the Kütahya Provincial Governorship (Date: June 2, 2025). All procedures were conducted in accordance with the ethical standards of the Declaration of Helsinki.
Consent
Written informed consent was obtained from all individual participants included in the study prior to data collection.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1: Standardized factor loadings for the three‐factor EISS model (WLSMV, N = 485).
Table S2: Correlations among the three EISS factors (WLSMV, N = 485).
Table S3: Fit of the three‐factor EISS model (WLSMV, N = 485).
Table S4: Reliability of the EISS subscales (N = 485).
Table S5: Pattern matrix for the three‐factor EISS solution (minimum residual extraction, promax rotation, N = 485).
Table S6: Correlations among the three factors of the exploratory solution (N = 485).
Table S7: SELSA‐S component scores by marital status (N = 485).
Acknowledgments
The authors thank all participants who volunteered to take part in this study.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
References
- 1. World Health Organization , From Loneliness to Social Connection: Charting a Path to Healthier Societies. Report of the WHO Commission on Social Connection (WHO, 2025). [Google Scholar]
- 2. DiTommaso E. and Spinner B., “The Development and Initial Validation of the Social and Emotional Loneliness Scale for Adults (SELSA),” Personality and Individual Differences 14, no. 1 (1993): 127–134, 10.1016/0191-8869(93)90182-3. [DOI] [Google Scholar]
- 3. Weiss R. S., Loneliness: The Experience of Emotional and Social Isolation (MIT Press, 1973). [Google Scholar]
- 4. Dahlberg L., McKee K. J., Frank A., and Naseer M., “A Systematic Review of Longitudinal Risk Factors for Loneliness in Older Adults,” Aging & Mental Health 26, no. 2 (2022): 225–249, 10.1080/13607863.2021.1876638. [DOI] [PubMed] [Google Scholar]
- 5. Maes M., Qualter P., Vanhalst J., Van den Noortgate W., and Goossens L., “Gender Differences in Loneliness Across the Lifespan: A Meta‐Analysis,” European Journal of Personality 33, no. 6 (2019): 642–654, 10.1002/per.2220. [DOI] [Google Scholar]
- 6. Jutengren G. and Ståhl F., “Determinants of Social Loneliness Among Older Adults in Job Retirement and the Role of Emotional Expressivity,” Aging & Mental Health 28, no. 10 (2024): 1153–1161, 10.1080/13607863.2024.2338205. [DOI] [PubMed] [Google Scholar]
- 7. Belfiori M., Salis F., Puxeddu B., and Mandas A., “The Cognitive and Mood‐Related Costs of Loneliness: Why Marital Status Matters in Old Age,” Geriatrics 10, no. 5 (2025): 117, 10.3390/geriatrics10050117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Yang H. J., Kim G. Y., Park S. J., Choi C., and Bum C. H., “An Analysis of Differences in Successful Aging, Loneliness, and Depression According to Marital Status Among Older Golf Participants,” Behavioral Science 16, no. 2 (2026): 266, 10.3390/bs16020266. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Aslan R., “Associations of Ageism and Social‐Emotional Loneliness With Aging Outcomes Among Elderly Individuals Living in the Community: A Cross‐Sectional Study,” Current Psychology 45 (2026): 277, 10.1007/s12144-025-08812-w. [DOI] [Google Scholar]
- 10. World Health Organization , Global Report on Ageism (World Health Organization, 2021), https://www.who.int/teams/social‐determinants‐of‐health/demographic‐change‐and‐healthy‐ageing/combatting‐ageism/global‐report‐on‐ageism. [Google Scholar]
- 11. Butler R. N., “Age‐Ism: Another Form of Bigotry,” Gerontologist 9, no. 4 Part 1 (1969): 243–246, 10.1093/geront/9.4_Part_1.243. [DOI] [PubMed] [Google Scholar]
- 12. Levy B. R., “Stereotype Embodiment Theory,” in Encyclopedia of Gerontology and Population Aging, ed. Gu D. and Dupre M. E. (Springer, 2021), 4763–4767, 10.1007/978-3-030-22009-9_604. [DOI] [Google Scholar]
- 13. Shiovitz‐Ezra S., Shemesh J., and McDonnell/Naughton M., “Pathways From Ageism to Loneliness,” in Contemporary Perspectives on Ageism, ed. Ayalon L. and Tesch‐Römer C. (Springer, 2018), 131–147, 10.1007/978-3-319-73820-8_9. [DOI] [Google Scholar]
- 14. Karasoy G. and Baydur H., “Development, Validity and Reliability of the Elderly Internalized Stigma Scale,” Balikesir Saglik Bilimleri Dergisi 14, no. 2 (2025): 251–263. [Google Scholar]
- 15. Weiss D. and Kornadt A. E., “Age‐Stereotype Internalization and Dissociation: Contradictory Processes or Two Sides of the Same Coin,” Current Directions in Psychological Science 27, no. 6 (2018): 477–483, 10.1177/0963721418777743. [DOI] [Google Scholar]
- 16. Terán‐Mendoza O. and Cancino V., “Ageism and Loneliness in People Over 50: Understanding the Role of Self‐Perception of Aging and Social Isolation in a Chilean Sample,” Journal of Applied Gerontology 44, no. 10 (2025): 1681–1690, 10.1177/07334648251314283. [DOI] [PubMed] [Google Scholar]
- 17. Pedroso‐Chaparro M. S., Antón‐López J. C., Cabrera I., Márquez‐González M., Martínez‐Huertas J. Á., and Losada‐Baltar A., ““I Feel Old and Have Aging Stereotypes”: Internalized Aging Stereotypes and Older Adults' Mental Health: The Mediational Role of Loneliness,” Aging & Mental Health 27, no. 8 (2023): 1619–1626, 10.1080/13607863.2022.2163376. [DOI] [PubMed] [Google Scholar]
- 18. Goldman N., Alemdar M., Megges H., et al., “National Policy Responses to Address Loneliness: A Global Scoping Review of 194 WHO Member States,” Health Policy 165 (2026): 105553, 10.1016/j.healthpol.2025.105553. [DOI] [PubMed] [Google Scholar]
- 19. Turkish Statistical Institute , Elderly Statistics, 2021 (TurkStat, 2022), https://www.tuik.gov.tr/media/announcements/istatistiklerle_yaslilar_2021.pdf. [Google Scholar]
- 20. Faul F., Erdfelder E., Lang A. G., and Buchner A., “G*Power 3: A Flexible Statistical Power Analysis Program for the Social, Behavioral, and Biomedical Sciences,” Behavior Research Methods 39, no. 2 (2007): 175–191, 10.3758/BF03193146. [DOI] [PubMed] [Google Scholar]
- 21. Corrigan P. W., Larson J. E., and Rüsch N., “Self‐Stigma and the “Why Try” Effect: Impact on Life Goals and Evidence‐Based Practices,” World Psychiatry 8, no. 2 (2009): 75–81, 10.1002/j.2051-5545.2009.tb00218.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Akgül H., “Yetişkinler için Sosyal ve Duygusal Yalnızlık Ölçeği'nin (SELSA‐S) Türk kültürüne uyarlaması: geçerlilik ve güvenirlik çalışması [Adaptation of the Social and Emotional Loneliness Scale for Adults (SELSA‐S) to Turkish Culture: Validity and Reliability Study],” Yalova Sosyal Bilimler Dergisi 10, no. 21 (2020): 54–69, https://izlik.org/JA66HP95LS. [Google Scholar]
- 23. Tabachnick B. G. and Fidell L. S., Using Multivariate Statistics, 6th ed. (Pearson, 2013). [Google Scholar]
- 24. Taniguchi H. and Kaufman G., “Family, Collectivism, and Loneliness From a Cross‐Country Perspective,” Applied Research in Quality of Life 17, no. 3 (2022): 1555–1581, 10.1007/s11482-021-09978-8. [DOI] [Google Scholar]
- 25. Stephens C. and Phillips H., “Older People's Neighborhood Perceptions Are Related to Social and Emotional Loneliness and Mediated by Social Network Type,” Gerontologist 62, no. 9 (2022): 1336–1346, 10.1093/geront/gnac087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Takagi E., Saito Y., and Chan A., “Gender Differences in the Association Between Social Relationships and Loneliness Among Older Adults in Singapore,” Journal of Population Research 37 (2020): 243–263, 10.1007/s12546-020-09242-0. [DOI] [Google Scholar]
- 27. Ratcliffe J., Galdas P., and Kanaan M., “Older Men and Loneliness: A Cross‐Sectional Study of Sex Differences in the English Longitudinal Study of Ageing,” BMC Public Health 24 (2024): 354, 10.1186/s12889-024-17892-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Allen J. O. and Sikora N., “Aging Stigma and the Health of US Adults Over 65: What Do We Know?,” Clinical Interventions in Aging 18 (2023): 2093–2116, 10.2147/CIA.S396833. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Balki E., Hayes N., and Holland C., “The Indirect Impact of Educational Attainment as a Distal Resource for Older Adults on Loneliness, Social Isolation, Psychological Resilience, and Technology Use During the COVID‐19 Pandemic: Cross‐Sectional Quantitative Study,” JMIR Aging 6 (2023): e47729, 10.2196/47729. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Tomida K., Shimoda T., Nakajima C., Kawakami A., and Shimada H., “Social Isolation/Loneliness and Mobility Disability Among Older Adults,” Current Geriatrics Reports 13 (2024): 86–92, 10.1007/s13670-024-00414-x. [DOI] [Google Scholar]
- 31. Tully‐Wilson C., Bojack R., Millear P. M., Stallman H. M., Allen A., and Mason J., “Self‐Perceptions of Aging: A Systematic Review of Longitudinal Studies,” Psychology and Aging 36, no. 7 (2021): 773–789, 10.1037/pag0000638. [DOI] [PubMed] [Google Scholar]
- 32. Tapia‐Muñoz T., Staudinger U. M., Allel K., et al., “Income Inequality and Its Relationship With Loneliness Prevalence: A Cross‐Sectional Study Among Older Adults in the US and 16 European Countries,” PLoS One 17, no. 12 (2022): e0274518, 10.1371/journal.pone.0274518. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Chrisler J. C., Barney A., and Palatino B., “Ageism Can Be Hazardous to Women's Health: Ageism, Sexism, and Stereotypes of Older Women in the Healthcare System,” Journal of Social Issues 72, no. 1 (2016): 86–104, 10.1111/josi.12157. [DOI] [Google Scholar]
- 34. Shimizu Y., Hashimoto T., and Karasawa K., “Decreasing Anti‐Elderly Discriminatory Attitudes: Conducting a “Stereotype Embodiment Theory”‐Based Intervention,” European Journal of Social Psychology 52, no. 1 (2022): 174–190, 10.1002/ejsp.2823. [DOI] [Google Scholar]
- 35. Murphy D. J., Jacob J. E., Mackenzie C. S., and Porter M. M., “Changes in Narratives About Aging Among Older Adults in an Internalized Ageism Intervention,” Aging & Mental Health 29 (2025): 2058–2066, 10.1080/13607863.2025.2521367. [DOI] [PubMed] [Google Scholar]
- 36. Kornadt A. E., Wettstein M., Lepinteur A., Vögele C., and D'Ambrosio C., “Associations of Subjective Age Trajectories With Loneliness and Stress Across Adulthood,” PLoS One 20, no. 4 (2025): e0320673, 10.1371/journal.pone.0320673. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Standardized factor loadings for the three‐factor EISS model (WLSMV, N = 485).
Table S2: Correlations among the three EISS factors (WLSMV, N = 485).
Table S3: Fit of the three‐factor EISS model (WLSMV, N = 485).
Table S4: Reliability of the EISS subscales (N = 485).
Table S5: Pattern matrix for the three‐factor EISS solution (minimum residual extraction, promax rotation, N = 485).
Table S6: Correlations among the three factors of the exploratory solution (N = 485).
Table S7: SELSA‐S component scores by marital status (N = 485).
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.