1-5 rating scale wording: the labels for each point

1-5 rating scale wording that people actually use

Twelve label sets, counted from a published library rather than invented for a blog post.

A rating question can be built two ways, and the choice decides how much wording you need. The first way is a numbered row with a word under each end, which is how all 644 rating questions in our template library are built. The second is a list of five options that each carry their own words, which the same library uses 141 times. Only the second kind needs all five labels written out, and those are the ones below.

Agreement wording dominates. It is the Likert family's home ground, compared with plain rating questions further down this page, so it takes two rows here and then gets out of the way. The other constructs are the ones you reach for when the thing you are measuring is not agreement. Where a construct has more than one published wording, the alternatives sit together; pick one and keep it for the whole survey.

Every set with a count is published in the library, written here low to high. The twelve published sets cover 133 of the 141 fully labelled five-option questions; the other eight are one-offs, and a set published in both directions is counted once at its combined total. The example prompts and the set marked editorial are ours, not the library's. Source: SuperSurvey Question Corpus, 109 live templates and 1,422 questions, read 11 September 2026.

Eleven published sets plus the revised fixed-period set. The flagged calendar wording is left out on purpose.

What the number means

The code behind each button counts how much of the attribute the respondent reports: 1 is the least, 5 is the most. That is not the same as worst to best. On eight of the nine constructs above the two coincide, so 5 is the answer you hope for. On the Amount set they part company: Much too little is 1, Much too much is 5, and the answer you hope for is the 3 in the middle, Just about right.

Averaging a scale that peaks in the middle rewards nobody, because a room split between too little and too much scores the same as a room that is content. Report the share at 3 instead, or split the scale into "too little" and "too much" and report both.

Direction is the other decision nobody writes down. Every numbered row in the library runs low on the left and high on the right, but 102 of the 141 labelled lists are published with the positive option first, because a list is read top down and the good news goes at the top. Both are defensible. Running both inside one survey is not, and it is the fastest way to collect answers about your layout.

Frequency wording has its own trap, which is why one published set is flagged above. Never, Rarely, Monthly, Weekly, Daily mixes two kinds of word: the first two describe an impression, the last three describe a calendar, and they overlap, because someone who uses a thing every Monday is both Weekly and, over a year, Monthly. Fix it by naming the recall period in the question, the last 4 weeks, and offering categories that cannot both be true: Not at all, On 1 to 3 days, On 4 to 9 days, On 10 to 19 days, On 20 or more days. Nobody has to guess what Rarely means, and every respondent falls into exactly one box.

The concentration is the useful part. Five pairs cover 553 of the 644 questions, so a reader who has answered any survey in the last year has seen your end words before and does not have to stop and work them out. That is worth more than a bespoke phrase. It also has a cost: the agreement pair invites people to agree. To avoid biased answer options where the thing you want is an attribute, how clear, how easy, how fast, ask for the attribute instead and the pull disappears.

1 to 5 or 0 to 10: which rating scale to use

Of the 644 rating questions in the library, 637 run 1 to 5 and seven run 0 to 10. There is no third option and no middle ground: nothing runs 1 to 7 or 1 to 10, and the ranges were read from the rendered buttons rather than from the question text.

What makes the seven interesting is that they are the same question. Every 0 to 10 scale in the library asks how likely somebody is to recommend something, anchored Not at all likely to Extremely likely, one per template. That is the honest answer to when a 0 to 10 earns its extra six buttons: when one number is going to be tracked on its own, over time, against an outside benchmark, and needs room to move.

All seven carry the same end anchors, Not at all likely to Extremely likely, and each of these templates also asks 1 to 5 questions. Source: SuperSurvey Question Corpus, 109 live templates and 1,422 questions, read 11 September 2026.

The evidence behind the range is older than any of this. Preston and Colman had 149 people rate a shop or restaurant they had recently visited, on scales that differed only in how many categories they offered, from two up to eleven.² Two, three and four categories did poorly on reliability, validity and discriminating power, and the indices were significantly higher for scales with more categories, up to about seven; test-retest reliability tended to fall again past ten. Respondents themselves liked the 10-point scale best, then the seven and the nine. So the scale people enjoy most and the scale that measures best are not quite the same scale.

A longer scale is not a repair, either. Dell-Kuster and colleagues had the same hospital patients answer five questions twice, once on a short labelled scale and once on an 11-point end-anchored numeric one. Ceiling effects were high on both, and the longer scale did not substantially reduce them.⁴ Their conclusion was that the questions needed the work, not the scale. They also found three response choices too few, which is the floor worth remembering.

The working rule: 1 to 5 with words when the answer has to mean something specific, 0 to 10 when one headline number is being tracked and compared. Whether that headline number should start at 0 or at 1 is a separate decision, and a smaller one than it looks.

1-10 versus 0-10: ten choices or eleven

A 1 to 10 scale offers ten choices and has no exact middle; 0 to 10 offers eleven, with 5 in the centre. That is the whole mechanical difference. The one that matters in practice is compatibility: the Net Promoter Score is defined on 0 to 10, with 9 and 10 counted as promoters, 7 and 8 as passives and 0 to 6 as detractors, so a recommendation question that wants to be compared against any NPS benchmark has to use the eleven-button version. That is why all seven of the library's long scales run 0 to 10 and none runs 1 to 10, and why the NPS survey template ships the question, the anchors and the follow-up already wired.

Use 1 to 10 when the number is yours alone and nobody will benchmark it: an internal quality score, a rating that stakeholders already read out of ten, a form that has always used it. It is a rating question, not an NPS question, and its results cannot be compared with NPS results even if the end words match.

Rating scale formats beyond 1 to 5

Numbers and words are two of five formats a rating question can take. The library ships no stars, no faces and no sliders, so the counts on this page say nothing about them; what follows is what each control does to the answer, which is a matter of mechanism rather than measurement.

Five ways to put the same question on screen. The 644 figure in the first row is the count of rating questions in the library; the last three rows carry no counts because the library publishes none of those formats. Source: SuperSurvey Question Corpus, 109 live templates and 1,422 questions, read 11 September 2026.

A fully labelled list is not really a rating widget at all: it is a single-select question whose options happen to be in order, which is why option order and option count matter to it in the way answer option design describes. That is also why the wording table above exists. Nothing forces you to write five labels when you use a numbered row, and almost nobody does.

Whether you label every point or only the ends is a real choice with real consequences. Weijters, Cabooter and Schillewaert tested both the number of response categories and whether those categories carried labels, and found that scale format moves response style: how often people take an end, how often they sit in the middle.³ de Rezende and de Medeiros examined the same four outcomes: reliability, use of the extreme points, use of the middle point, and which scale respondents said they preferred.¹ The practical reading is not that one format wins. It is that changing format mid-programme changes your numbers while nothing has changed outside.

Likert scale vs rating scale

A Likert scale is one kind of rating scale. Rating scale is the parent term: any question that asks for a judgement on an ordered set of options, whether those options are numbers, words, stars or faces.

What makes something Likert is not the word "agree". It is that several statements share one answer set and are scored together as a single index. Likert-type items are written for satisfaction, frequency and likelihood as readily as for agreement; the answer set just has to stay identical down the block so the items can be summed. A one-off question with its own bespoke labels is a rating question and nothing more, however much it looks like the others. If you need to score multiple items together, that is the battery to build, and the wording on this page is only its first row.

So the difference is scope, not vocabulary. Every Likert item is a rating question. Most rating questions are not Likert items, because they stand alone and nothing is summed. The test is whether removing one item from the set would change the score you report.

Survey templates that use a 1 to 5 rating scale

Eighty-eight of the 109 templates in the library ask at least one rating question, and 81 of those never leave the 1 to 5 scale. The split by category is uneven in a way that matches what the questions are for: 11 of 11 technology templates use one, 27 of 33 employee templates, 17 of 20 business, 15 of 22 satisfaction, 10 of 13 marketing and 8 of 10 training.

If a template is the fastest route, the survey template library is sorted by those six categories and every one of them is editable before you send it. The nine questions below are lifted from it word for word, one per end-anchor family, so they can be pasted straight into a draft.

How to report a 1 to 5 rating scale

Lead with the distribution: the percentage who chose each of the five points, with the number of people it is based on. Then add the top-two-box, the share choosing 4 or 5. Those two together answer the question a mean cannot, which is whether a middling result is a room full of people who feel lukewarm or two camps who disagree. The steps that follow, coding, cleaning and cross-breaks, are set out where we analyse rating-scale answers as quantitative data.

Whether you are entitled to average the numbers at all has been argued for twenty years, since Jamieson set out the case in Medical Education that answers of this kind are ordinal, so the gaps between points cannot be assumed equal.⁷ Uher goes further and asks whether a rating scale encodes the property it claims to measure closely enough for the arithmetic to mean anything.⁶

You do not have to settle that argument to publish a defensible number. A share needs no assumption about spacing, so top-two-box is safe in a way a mean is not, and a room full of people understands it without a footnote. Publish the mean beside it if your organisation expects one, never instead of it, and never for a scale whose best answer sits in the middle.

Track the shape, not just the headline. A top-two-box that holds steady while the bottom box doubles is a result moving in two directions at once, and a mean shows you neither half. Put one text box straight after the rating and you get the reason as well as the number, which is what open-ended questions are for.

Once a scale is in the field, leave it alone. Relabelling the points or switching the range breaks the series, and afterwards you cannot separate the change in wording from the change in opinion. The neighbouring decisions, who to ask and how many replies you need before the number holds, are indexed in the survey learning centre.

Frequently asked questions

References

1. de Rezende, N. A., and de Medeiros, D. D. (2022). How rating scales influence responses' reliability, extreme points, middle point and respondent's preferences. Journal of Business Research, 138, 266-274. doi.org/10.1016/j.jbusres.2021.09.031
2. Preston, C. C., and Colman, A. M. (2000). Optimal number of response categories in rating scales: reliability, validity, discriminating power, and respondent preferences. Acta Psychologica, 104(1), 1-15. doi.org/10.1016/S0001-6918(99)00050-5
3. Weijters, B., Cabooter, E., and Schillewaert, N. (2010). The effect of rating scale format on response styles: the number of response categories and response category labels. International Journal of Research in Marketing, 27(3), 236-247. doi.org/10.1016/j.ijresmar.2010.02.004
4. Dell-Kuster, S., Sanjuan, E., Todorov, A., Weber, H., Heberer, M., and Rosenthal, R. (2014). Designing questionnaires: healthcare survey to compare two different response scales. BMC Medical Research Methodology, 14, 96. doi.org/10.1186/1471-2288-14-96
5. Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in surveys. Applied Cognitive Psychology, 5(3), 213-236. doi.org/10.1002/acp.2350050305
6. Uher, J. (2018). Quantitative data from rating scales: an epistemological and methodological enquiry. Frontiers in Psychology, 9, 2599. doi.org/10.3389/fpsyg.2018.02599
7. Jamieson, S. (2004). Likert scales: how to (ab)use them. Medical Education, 38(12), 1217-1218. doi.org/10.1111/j.1365-2929.2004.02012.x
8. SuperSurvey Question Corpus, 109 live templates and 1,422 questions, read 11 September 2026.

Correction, 10 September 2026: an earlier version of this page credited reference 1 to four authors. The paper has two. Every reference above was checked against the publisher's deposited record before publication.

Further reading

• Likert scale questionsAgreement blocks, how many points, and scoring several items as one index.
• Multiple choice questionsHow many options to offer, and what order to put them in.
• Open-ended questionsThe follow-up that explains why a rating moved.