The scale that measures loneliness without using the word

Daniel Russell, Letitia Anne Peplau and colleagues published in 1978 a scale made of statements about feeling understood, having people to talk to, feeling left out. For each one, the respondent indicates how often it applies to them.

None of those items uses the word loneliness. The choice aims to limit desirability effects, since declaring that one is lonely is socially costly, and the word itself carries connotations that vary greatly from one person to another.

The 1978 version contained only statements worded in the same direction, which exposed it to the tendency of some respondents to agree with whatever they are asked. The 1980 revision introduced positively worded items, and a third version followed in the mid-1990s.

It measures a perceived gap between desired and actual relationships, a definition that Peplau and Daniel Perlman formalised in the early 1980s. It is therefore not a count of relationships, and a well-surrounded person may score highly.

That is also why the score is not strongly correlated with the number of relationships measured separately, a result regularly found and regularly forgotten in coverage. A person with a small circle that matches their expectations obtains a low score, and that is exactly what the instrument is designed to record.

The scale produces a continuous score. The percentages of people reported as lonely that circulate presuppose a threshold, set by the authors of each survey and not by the instrument, which was never designed to sort respondents into two categories.

Two surveys using the same tool with different thresholds produce very different prevalences, which explains most of the gaps between published figures for the same country. Reading a percentage without knowing its threshold is like reading a temperature without knowing the scale.

Three-item versions have been developed for large surveys, notably by Mary Elizabeth Hughes and colleagues in 2004, because space there is counted. They correlate with the long version but do not measure exactly the same thing.

Surveys changing version between waves produce breaks in series that are hard to interpret, a difficulty flagged in several methodological reports. Other instruments, such as Jenny de Jong Gierveld's scale, rest on a different conception and are not directly comparable.

Translating the items alters their reach, and response norms differ: declaring that one feels left out does not carry the same social cost everywhere. The same word can evoke an ordinary experience in one language and an admission in another.

Comparisons of level between countries are therefore the most fragile of the set, whereas changes over time within one country, with a constant instrument, are far more solid.