A third of the studies (N = 17, 33%) used a version of the UCLA-LS, a well-established measure of loneliness (see Russell, 1996, for portrayal of the questionnaire). The scale has shown dependably high internal consistency, with a coefficient alpha of >0.90 in college student populations and good retest reliability after 12 months (r = 0.73) (Cramer and Barry, 1999). Nearly all of the studies using a version of this scale were conducted in the U.S. (N = 15, 29%), the country of origin of the scale developers. The UCLA-LS was based on an early conceptualization of loneliness (Russell et al., 1978). The emphasis on uni-dimensionality is noteworthy (Russell, 1996, p. 30), contrasting with the emotional and social loneliness dimensionality of the DJG-LS (see below). Russell et al. (1980), see also Russell et al. (1984) further described a cognitive model of loneliness in which the perception is that “social relationships are “too few,” and people feel “lonely” (Russell et al., 1980, p. 472). Thus, consistent with what is usually understood as the nature of the experience, these researchers placed emphasis on the person's own subjective perception of loneliness. It is worth noting that intensity as well as frequency were incorporated (as reflected in the scale).

The second most-frequently used scale in bereavement studies is the De Jong Gierveld Loneliness Scale (N = 8, 16%) (DJG-LS; De Jong Gierveld and Van Tilburg, 1999, 2010). Internal consistency is reported to be good, particularly among older adult samples, with Cronbach's alpha coefficients ranging from 0.80 to 0.90 (De Jong Gierveld and Van Tilburg, 1999; for comparison of the UCLA-LS with the DJG-LS see Penning et al., 2014). Overall, authors have concluded the scale demonstrates reliability and validity (De Jong-Gierveld and Kamphuis, 1985; Pinquart and Sörensen, 2001a,b; Gierveld and Tilburg, 2006). Nearly all of the studies using this scale were conducted in The Netherlands (N = 6, 75%), again the country of origin of the scale developers. The DJG-LS followed a definition of loneliness as “the manner in which the person perceives, experiences, and evaluates his or her isolation and lack of communication with other people” (De Jong-Gierveld, 1987, p. 120). The DJG-LS therefore emphasizes some rather different components of loneliness (e.g., communication) compared with the UCLA-LS. However, quite similarly to the DJG-LS, loneliness is here considered to be a subjective negative feeling originating from perceived deficits in social relationships, indicating a lack of intimacy or support in relationships, one that reflects a person's social participation and isolation. The scale also adopted the developers' so-called “cognitive theoretical approach” (De Jong Gierveld and Van Tilburg, 1999), which emphasizes the discrepancy between what the person desires in terms of interpersonal affection and intimacy and what one perceives they actually have. Again, this is similar to the cognitive approach reflected in the UCLA-LS. Furthermore—and uniquely at the time the scale was developed—De Jong Gierveld and colleagues built on Weiss's (Weiss, 1973) constructs of social and emotional loneliness, as compared with the unidimensional measurement approach reflected in the UCLA-LS.

Turning to the actual use of these loneliness scales in bereavement research: Within the empirical studies examined, different versions of the UCLA-LS–including various shortened versions–have been utilized (see Table 1; adapted from Vedder et al., 2021). Instructions ask respondents to indicate how often each of the statements applies to them. Items are typically rated on a 4-point scale, from “never feeling this way” to “often feeling this way.” Inspection of these scales shows that the items do not include the actual word loneliness (we consider this strategy in Discussion), not even in the instructions, in order to reduce response bias. Instead, they cover aspects such as: lacking companionship, having no one to turn to, feeling isolated, feeling unhappy being so withdrawn. As such, they reach beyond the above-cited definition of loneliness (Valtorta and Hanratty, 2012), to include social isolation as well as related concepts that could function as moderators, mediators, or outcomes (e.g., the included item about unhappiness). Such extensions in the scope also represent a systematic difference compared to the single- or few-item measures of loneliness (discussed below). As Cramer and Barry (1999, p. 493) note, the UCLA-LS does not specify a time frame for respondents, so it remains “unclear whether a state or trait measure has been designed.” There appears to be no norm or reference information concerning UCLA-LS scores, limiting the extent to which scores from a particular study can be compared with those found for other populations or relevant sub-groups. Nor do there appear to be agreed-on points for designation of intensity, for example, to indicate absence of, mild, moderate or high levels of loneliness (Russell et al., 1978, 1980; Weeks et al., 1980). Given the number of different versions of the UCLA-LS that have been used by researchers, it is also difficult to compare results across studies.

As with the UCLA-LS, different versions of the DJG-LS have been employed in bereavement research, with most using the full 11-item scale, while some used the six-item version (see Table 2). Like the UCLA-LS, this scale does not include the word loneliness at all, covering such aspects as having no one to share problems with, experiencing a sense of emptiness, feeling rejected, and having no one to trust completely. Thus, the scope seems somewhat broader than the definitions provided by the authors, raising questions about content validity. Instructions were slightly more elaborate than those for the UCLA-LS, explaining that the statements were made by individuals who had previously shared their experience with the researchers, and asking respondents to indicate the extent to which they applied to their own situation, the way they currently felt. However, as with the UCLA-LS, no specific time frame was included. In contrast to the use of a four-point Likert scale for the UCLA-LS, answers for DJG-LS items are scored by circling the statements that applied to the respondent's situation. Notably, a cut-off point relating to severity was provided (De Jong Gierveld and Van Tilburg, 1999). Furthermore, the DJG-LS and its shorter versions enable overall loneliness (from the total scale) as well as emotional and social loneliness (from subscales) to be calculated, facilitating comparison across studies.

More than a third of the studies used single- or few-item measures to assess loneliness (N = 19, 37%). Such items were used across more countries than either of the above validated scales: Half of the studies (N = 9, 50%) were conducted in various European countries, a lesser proportion was conducted in the U.S. (N = 7, 39%). The remainder were either conducted in China (N = 2, 11%) or internationally-administered (N = 1, 6%). A number of the short measures were derived from existing scales (N = 7, 14%; see Table 3). Instructions (not listed here) varied according to the overall focus of the existing scales from which they were derived, and/or the interests of the particular researcher who developed the short measure.

Other studies—accounting for ~one-quarter of the total (N = 12, 24%)—used items that were self-constructed by the authors (see Table 4). A noteworthy difference between these single- or few-item measures and the two validated scales discussed above is that their items all refer explicitly to loneliness, with each of them also employing Likert-type scales. For example, Eckholdt et al. (2018) used a 2-item, 7-point scale of emotional loneliness: “I feel lonely even when I am with other people” and social loneliness: “I have no really close friends.” Savikko et al. (2006), used a 1-item, 3-point one: “Do you suffer from loneliness?” with answer categories from 1 = seldom or never, 2 = sometimes, 3 = often or always.

First, it would be advisable to come to a clearer agreement on the definition of what loneliness is—and is not. The widely-used definition by Valtorta and Hanratty (2012) reflects basic components identified in both theorizing and empirical research. Details of the measures included in our review seem in line with this general definition: They focus on subjective negative feelings and some enable distinction of social from emotional loneliness. The latter distinction seems critical when considering loneliness after bereavement, in line with theoretical formulations. In bereavement, one has lost an attachment figure (Weiss, 1973; Bowlby, 1980) a person to whom one was closely bonded, suggesting the centrality of emotional rather than social loneliness (although social loneliness may be particularly felt among older bereaved persons as their networks decrease). Indeed, there is some empirical support for this supposition (Stroebe et al., 1996; Guiaux, 2010). De Jong Gierveld et al. (2006), provide further discussion of the need for distinction of social from emotional loneliness.

Thus, it may be helpful to use the description by Valtorta and Hanratty (2012) as a working definition for further research; others have endorsed this strategy for research on loneliness (Fakoya et al., 2020). However, openness to alternatives may also be called for. Cramer and Barry (1999) have provided a valuable evaluation and comparison of the various self-rating scales (including the UCLA-LS and DJG-LS). They recommend use of the Social and Emotional Loneliness Scale for Adults (SELSA; DiTommaso and Spinner, 1993, 1997). They add the factor “family loneliness,” a feature that is well in-line with our argument that emotional loneliness is of particular interest when studying bereavement. Furthermore, Bandari et al. (2019) are currently reviewing definitions of loneliness and, on the basis of the accumulated knowledge, are planning to propose a new definition that may better reflect the scientific construction of what loneliness incorporates.

There is need to avoid conceptual overlap with related constructs such as social isolation and social support. Yanguas et al. (2018) state that “…various researchers have referred to “loneliness” and “social isolation” indistinctly. Others, however, find both terms very different from each other. Making accurate evaluations depends on a clear definition of the concept of loneliness, with special awareness of its multidimensionality and its differences with respect to related concepts (social isolations or a lack of social support)” (p. 302–303). Our review showed that some measures cover the latter dimensions in their items, while others do not. Researchers may want to adapt the content of the scales to conform to the narrower definition of loneliness, in order to better distinguish it from related—yet distinct—constructs.

Our results showed that two scales, the UCLA-LS and the DJG-LS, have each been quite widely adopted. A comparative examination of the UCLA-LS with the DJG-LS concluded that the latter has better utility, at least for use among middle-aged and older persons (Penning et al., 2014). However, some of the DJG-LS items border on different domains (e.g., feeling rejected or not having people one can trust). Such items might best be deleted, and/or items from the UCLA-LS that are complementary to those on the DJG-LS added (requiring new validation). The DJG-LS has the advantage of enabling examination of emotional loneliness and social loneliness as well as overall loneliness.

Deciding whether to assess loneliness using a validated instrument or a single/few-item measure requires careful weighing of pros and cons. There is much to support the use of an established scale. Multiple-item usage increases reliability and validity; errors and the specificity that are inherent in single items are averaged out (Bowling, 2005). For such reasons, instruments with carefully selected items are almost always considered superior to single item measures in scientific research. However, the qualification almost may be particularly appropriate for assessing loneliness in bereavement. For instance, the average age of participants in the reviewed studies was 65 years, suggesting a wide interest in loneliness among older bereaved persons, for some of whom research participation may be more challenging than for younger persons. When study participants are overwhelmed by the length of an interview or the cognitive demand of responding to multiple scale items, using multi-item scale measures may not be the better choice: One well-answered item clearly has superiority over a number of unanswered or incorrectly answered ones. In addition to feasibility, when the goal is to assess the general feeling of loneliness rather than distinguish different dimensions of it, one or a few items may be sufficient. The fact that approximately half of the reviewed studies used ten items or fewer may reflect this dilemma. For some purposes, few items may be able to capture loneliness as well as the score on a longer loneliness scale. Finally, the choice depends on the broader aims of the study (e.g., symptomatology in general vs. loneliness in particular).

In summary: Choice of one of the validated scales or a single/few-item measure likely reflects the distinct aims and scopes of the studies, as well as the targeted participants. Indeed, different types of assessment may be called for under specific circumstances. There is no one-size-fits-all solution to the measurement of loneliness in bereavement.

Another issue that the scope of the reviewed studies raises is whether one should ask directly about being lonely, or whether some proxy is preferable¹. On the one hand, one might argue that people in general—and perhaps bereaved people in particular—are likely to know whether or not they feel lonely (and/or how intensely); thus, there is no clear detriment to asking directly about loneliness. However, on the other hand there is risk of social desirability bias, and demand characteristics or perceived expectations may prevail (e.g., a societal norm may prevent someone from admitting to being lonely). In this case, indirect questions may garner more accurate information about loneliness from respondents than direct questions. An example of an indirect item (from DJG-LS) is: “I miss having a really close friend.”

Apart from the content of the items, we noted that response categories used in the scales and single/few-item measures varied among the reviewed studies, ranging from simple “yes”/“no” options to Likert scale usage and to further sub-categorization (e.g., collapsing a larger range of scores into three categories, such as “not lonely,” “moderately lonely” and “strongly lonely”). Again, there would be advantages to coming to agreements where possible on this point, to enable comparison of results across studies. More research is needed to evaluate the pros and cons of different response option formats.

To work toward selection of the most appropriate, reliable, and valid measurement of loneliness for any particular investigation, one strategy would be to compare results on the existing instruments with single- or few-item measures. None of the studies we reviewed used both single/few-item measures in addition to validated scales.². For practical purposes, if a short measure is indicated, the best-loading items of emotional and social loneliness could be selected. However, this may result in a further expansion of loneliness measures. On the other hand, a validated pool of items could be established from which clinicians and researchers may select. In addition, we refer above to the review of loneliness scales by Cramer and Barry (1999), a valuable source for considering the relative merits of different assessment instruments.³

It would also be useful to examine the benefits of the self-constructed items, that is, ones created by the investigators themselves, based on their understanding of “loneliness” (Table 4) compared with those selected from the established, more generic scales (Table 3), such as the Center for Epidemiologic Studies Depression Scale (CES-D) or Symptom Checklist-90 (SCL-90), ones which have also already been used in bereavement research. This would help to improve standardization and optimization of single/few-item measurement of loneliness across studies, allowing for comparison across study contexts and (sub)populations.

Additional limitations to the data set and our review process need to be mentioned. First, our criteria limited the scope of investigation: notably, they restricted the review to adults, excluding bereaved children; only pre-COVID-19 studies were covered; only English language articles were included. Second, our chosen focus was on bereavement and not on other situations or diagnostic categories; comparisons across such domains would also be enlightening (for review of relationships between loneliness and psychiatric as well as physical disorders: Mushtaq et al., 2014). As a result of these selection criteria, the number of participants in the studies reviewed ranged from 30 to 2,018, and most participants were bereaved through loss of their spouse. The average age of sample participants, as noted above, was 65, limiting our ability to examine loneliness in bereavement across different age groups. However, there was no preferred use of specific assessments with regards to study-size or target group. Finally, regarding the available studies: to the best of our knowledge, no scientific investigation has yet been undertaken to compare the relative usefulness of the different measures to assess loneliness in bereavement. For recommendations to be made, such systematic, comparative investigations of the measures are needed.