The UCLA Loneliness Scale

The UCLA Loneliness Scale is the questionnaire most loneliness research runs on: 20 statements about ordinary social experience, each answered on a four-point frequency scale, adding up to one number (Russell, 1996). It asks how often you feel left out, or without company you can count on. It never asks whether you are lonely, and the omission is deliberate.

Key Takeaways:

• Version 3 of the scale contains 20 statements answered by frequency rather than by agreement, so the output is how often, not yes or no (Russell, 1996)
• Most items describe a social experience instead of naming the state, which sidesteps the reporting problem that the word “lonely” creates
• Hughes and colleagues built a three-item short scale so that large surveys could carry a loneliness measure at all, and much of the population data since comes from it (Hughes et al., 2004)
• A score is not a diagnosis. Loneliness appears in no diagnostic manual, and no validated cutoff on this scale marks anyone as clinically lonely
• Every prevalence figure you have read, in any headline, is a count of people above a line that a researcher chose, on an instrument like this one

What the UCLA Loneliness Scale Actually Asks

It asks about frequency of specific social experiences, one statement at a time. Version 3, published by Daniel Russell in the Journal of Personality Assessment in 1996, presents 20 statements and asks how often each applies to you, with four options running from never to often (Russell, 1996). The published item set is Russell’s, and reproducing all 20 here would be lifting a copyrighted instrument, so take two as representative: one asks how often you feel left out, another how often you feel there are people you can really talk to.

Two design details do most of the work. The items are worded in both directions, some toward company and some toward its absence, which stops a person who agrees with everything from producing a high score by temperament alone. And the response is frequency rather than agreement, so the instrument records how often an experience shows up rather than whether you endorse a description of yourself.

Version 3 exists because the earlier versions were hard for some respondents to complete accurately. Russell simplified the wording so the scale could be read aloud, used with older adults, and used with people for whom double negatives in a questionnaire are a genuine obstacle. That is an unglamorous kind of revision and it is the reason the instrument travels as widely as it does.

If you have ever taken an online “loneliness test,” check what it asked. Statements plus a frequency response means it borrowed from this family, while a set of yes/no questions with a cheerful verdict at the end borrowed nothing except the reputation.

Why the Word “Lonely” Is Mostly Missing

Asking people directly whether they are lonely produces a worse measurement than asking about the experiences underneath it. Loneliness carries a social penalty. It reads as an admission of failure at something adults are assumed to manage, so a direct question invites the answer the respondent would prefer to be true, and it invites it hardest from anyone answering aloud to a stranger on a phone.

The scale routes around this. You are not asked to accept a label, you are asked how often a specific ordinary thing happens to you, which is a much smaller thing to admit. Enough of those small admissions add up to a picture the same person might have declined to give in a single word.

There is a cost, and the instrument’s users are open about it. A questionnaire that avoids the word cannot check whether the person would have used it themselves, so it measures a construct defined by the researchers rather than the reader’s own account of their life. Someone can score high and reject the description entirely. That is not a malfunction, but it is worth knowing about a number reported as though it settled something.

Take the same lesson into ordinary conversation, where it is more useful than it looks. Asking a friend “are you lonely” reliably gets you “no.” Asking who they talked to this week that was not work gets you an answer with information in it.

The Three-Item Version Built for Large Surveys

The short scale exists because 20 items will not fit into a national survey. Hughes and colleagues published a three-item version in Research on Aging in 2004, derived from the longer scale and designed for telephone administration, with three response options rather than four (Hughes et al., 2004). The items ask how often a person feels they lack companionship, how often they feel left out, and how often they feel isolated from other people.

That is a lot of measurement compressed into 30 seconds, and the compression has consequences. Three items cannot separate the person who is short of a group from the person who is short of one particular relationship, a distinction between two kinds of loneliness Weiss drew as far back as 1973 (Weiss, 1973). Both come back as a single loneliness score. The instrument was built for population estimates, where averages across thousands of people are the point, and it does that job well.

Where it gets misused is in the retail direction, when a three-item result about a population gets talked about as though it described an individual’s inner life. It does not have the resolution. Nobody designed it to.

A Score Is Not a Diagnosis

No score on this scale, or any loneliness scale, is a diagnosis of anything. Loneliness is not a clinical disorder, no diagnostic manual contains it, and there is no validated cutoff on the UCLA range above which a person is officially lonely. What researchers do instead is compare group averages, or split their own sample, so a score called “high” is high relative to the other people in that particular study.

Which means the number has no meaning on its own. A score of 48 out of a possible range from 20 to 80 tells you nothing until you know what the rest of the sample scored, and reading a single result as a personal verdict is a category error, however official the questionnaire looked. The arithmetic cannot tell you what loneliness is, what is driving yours, or what helps.

What the scale genuinely gives an individual is a second reading. Take it twice, six months apart, and the change is informative in a way the absolute number is not. Same instrument, same person, different point in a life: that comparison is the one the design supports.

The other boundary matters more, and no version of this instrument was built to hold it. A questionnaire cannot tell loneliness from depression. The two overlap in their symptoms, the scale was never designed to separate them, and a person sitting alone with a total in the 50s has no way to work out from that number which of the two they are looking at. That question belongs with a clinician, not with arithmetic.

Every Loneliness Headline Rests on Something Like This

The statistic in any loneliness article you have read is not a fact about the world in the way the sentence implies. It is a count of survey respondents who landed above a threshold that somebody chose, on an instrument that somebody designed, in a population that somebody sampled. Change any of the three and the figure moves, often by a lot. That instability is why whether the loneliness epidemic is a real trend or a measurement artifact is a live question and not a rhetorical one.

Consider what varies. A study using the 20-item scale and a median split reports a different share than one using the three-item version and counting anyone who answered “often” to a single item. One asks by phone, one asks online, and answering aloud to a human being suppresses the admission. One samples adults over 65, an age where loneliness is a different problem than at 25 or 45. Another samples whoever has internet access and a free evening. All four can be competently run and still produce headline shares that are nowhere near each other, which is why two loneliness statistics published in the same month can look like they describe different countries.

The 2023 U.S. Surgeon General advisory brought this measurement into public argument, and rightly, because the health associations underneath it replicate (Holt-Lunstad et al., 2010). Read that word association strictly. The mortality work is built on observational samples, where health and money sort people into their relationships as much as the other way round, so the finding is that weak social ties travel with worse outcomes, not that repairing ties repairs the risk. The cigarette comparison people quote from it was a way to convey effect size, never a measured equivalence. What travels worst of all is the precision: the advisory’s numbers came from instruments like these, with every choice above baked in, and by the time a figure reaches a headline the instrument has been stripped off it.

One question is worth carrying from here into any statistic you meet: measured with what, and above which line? If the article cannot say, the number is decoration.

FAQ

Where can I take the UCLA Loneliness Scale myself? Versions circulate widely in research papers and public health materials, and the three-item short scale appears in government survey documentation, so both are findable. Taking it changes very little on its own, since no score means anything without a comparison group. If you do take it, note the date and take it again in six months, because your own two readings are the only comparison that applies to you.

What is a high score on the UCLA loneliness scale? There is no official high. The 20-item version produces a score between 20 and 80 by arithmetic, and researchers interpret it by comparing people within a sample rather than against a fixed cutoff. Any website that hands you a threshold and a verdict has invented the threshold, whatever the questionnaire it borrowed.

Why do loneliness statistics differ so much between studies? Because they measure with different instruments and cut them at different points. A three-item phone survey counting anyone who says “often” produces a much larger share than a 20-item questionnaire split at the sample median, and neither is wrong. Before comparing two loneliness figures, check that both used the same scale and the same threshold, which they usually did not.

Does the scale distinguish being alone from feeling lonely? Yes, and that separation is the reason it exists. None of the items ask how many hours you spend by yourself or how many people you live with, because those are measures of social isolation and get counted separately. The scale only asks how your relationships feel to you, which is why a person with a full house can score high on it.

Instruments shape what a society can notice. Before Russell’s work there was no standard way to put loneliness into a survey next to income and blood pressure, and a thing that cannot be counted does not get funded, studied, or argued about in public. That is the real contribution, and it comes with a warning attached. The number the scale produces is a research tool that works on groups, and it was never built to tell one person at a kitchen table what their life is like. Your own answers to those 20 statements are more useful read as a list than added up. Read as a list, those answers point at specific gaps rather than a grade. Where one of them is simply having nobody around on an ordinary day, Lona can fill some of that hour while the real contacts get rebuilt. It is a stopgap, the kind the scale was never designed to recommend or rule out.

Sources

• Russell, D., “UCLA Loneliness Scale (Version 3),” Journal of Personality Assessment, 1996
• Hughes, M. E. et al., “A Short Scale for Measuring Loneliness in Large Surveys,” Research on Aging, 2004
• Weiss, R. S., “Loneliness: The Experience of Emotional and Social Isolation,” 1973
• Holt-Lunstad, J. et al., “Social Relationships and Mortality Risk: A Meta-analytic Review,” PLoS Medicine, 2010
• U.S. Surgeon General, “Our Epidemic of Loneliness and Isolation,” 2023