Interval data is a level of measurement in which values are ordered, the gaps between adjacent values are equal, and there is no true (absolute) zero. Because the zero point is arbitrary, you can meaningfully add and subtract interval values, but you cannot multiply or divide them — so 20°C is not ‘twice as warm’ as 10°C. Classic examples include temperature in Celsius or Fahrenheit, IQ scores and calendar years.
Interval data sits one step below ratio data on the measurement hierarchy: it shares the ordered categories and equal spacing of ratio data but lacks ratio data’s meaningful zero. That single missing property changes how the data is collected, what instruments produce it, and how you are allowed to interpret your results. This guide focuses on the characteristics of interval data and how it is measured and collected — for a deeper run-through of analysis and additional worked examples, see our companion guide on levels of measurement in statistics.
Defining Interval Measurement & Its Characteristics
Interval data is defined by a small set of characteristics. Together they place it third on the four-rung ladder of measurement, above nominal and ordinal data but below ratio data.
- Ordered values — like ordinal data, the values follow a meaningful rank order (30°C is warmer than 20°C).
- Equal intervals — the difference between any two adjacent points represents the same quantity across the whole scale. The gap from 10°C to 20°C is exactly the same amount of heat as the gap from 70°C to 80°C.
- No true zero — the zero point is arbitrary or conventional, not an absence of the measured attribute. 0°C does not mean ‘no temperature’, it is simply the freezing point of water.
- Values can be negative — because zero is just a point on the scale, interval data can fall below it. −5°C is a perfectly valid temperature; a calendar year can be expressed as a BC (negative) date.
The defining contrast with ratio data is the missing true zero. Because zero is arbitrary, ratio statements are meaningless: you can add and subtract interval values, but multiplying or dividing them produces nonsense. You can correctly say 20°C is 10 degrees warmer than 10°C, but not that it is ‘twice as warm’.
“Interval — differences between measurements are meaningful, but there is no true zero point… Ratios of interval data are not meaningful.” — Adapted from S. S. Stevens, ‘On the Theory of Scales of Measurement’, Science (1946)
Unsure which measurement level your data is?
ResearchProspect to the rescue!
Our statisticians will identify your variables, pick the right tests and run the full analysis — explore our statistical analysis service.
Classic Examples of Interval Data
The table below lists familiar measurements and explains why each is interval rather than ratio — the deciding factor is always whether zero represents a genuine absence of the attribute.
| Variable | Level | Why Not Ratio? |
|---|---|---|
| Temperature in Celsius | Interval | 0°C is the freezing point of water, not ‘no temperature’ |
| Temperature in Fahrenheit | Interval | 0°F is equally arbitrary — based on a saltwater solution |
| Temperature in Kelvin | Ratio | 0 K is absolute zero — complete absence of thermal energy |
| IQ score | Interval | An IQ of 0 does not mean ‘no intelligence’ — it is outside the scale range |
| Calendar year (AD) | Interval | Year 0 is a convention, not the beginning of time |
| Standardised test score (mean = 100) | Interval | A score of 0 means ‘no marks on this test’, not ‘no ability’ |
| Reaction time (ms) | Ratio | 0 ms = no time elapsed — a genuine absence |
What Statistics Can You Use
Interval data unlocks the full range of parametric statistical methods. This is a significant jump from ordinal data, which is restricted to non-parametric alternatives.
Appropriate for interval data
- Arithmetic mean — equal intervals make the mean a meaningful summary statistic.
- Standard deviation and variability — measure how spread out the data is around the mean.
- Pearson’s r — correlation between two interval (or ratio) variables.
- Independent-samples t-test — compare means between two groups.
- Paired-samples t-test — compare means from the same group at two time points.
- One-way ANOVA — compare means across three or more groups.
- Linear regression — predict an interval outcome from one or more predictors.
Not appropriate for interval data
- Ratio statements — you cannot say 60°C is ‘twice as warm’ as 30°C.
- Coefficient of variation — this requires a meaningful zero to be interpretable.
The Likert Scale Debate
This issue comes up in almost every research-methods class. Single Likert-scale items — strongly agree / agree / neutral / disagree / strongly disagree — are technically ordinal: the psychological gap between ‘agree’ and ‘neutral’ may not equal the gap between ‘neutral’ and ‘disagree’, so the equal-interval characteristic does not strictly hold.
Despite this, many researchers treat composite Likert scales (several items summed into a total score) as interval data, arguing that summing produces an approximately continuous, roughly normally distributed outcome. This is a pragmatic position that enables more powerful parametric statistics.
| Approach | Treats Likert as | Implications |
|---|---|---|
| Conservative / orthodox | Ordinal | Use the median and non-parametric tests (Mann–Whitney U, Spearman’s rho) |
| Common in practice | Interval (composite scales) | Use the mean, t-tests, ANOVA, Pearson’s r |
| Recommended | Check discipline norms | Be explicit about the assumption; cite a methodological justification |
Interval Vs Ratio In Practice
For most common statistical tests — t-tests, ANOVA, regression and Pearson’s correlation — the interval/ratio distinction does not change your choice of method. Both levels support the same parametric toolkit.
The distinction matters for interpretation. With ratio data you can say ‘Group A’s mean was twice that of Group B’. With interval data you can only say ‘Group A’s mean was 15 points higher’. Being precise in how you describe findings signals methodological literacy. For the full picture of ratio analysis, see our companion guide to ratio data: definition, examples and analysis, and revisit the ladder in levels of measurement in statistics.
How Interval Data Is Collected & Measured
What sets interval data apart in practice is that it is almost always produced by a calibrated instrument or a standardised scale — a measuring device whose zero point has been fixed by convention rather than by nature. Understanding the collection method helps you decide, at the design stage, whether a variable will be interval, and how reliable it will be.
The hallmark of interval collection is that the measuring tool, not nature, fixes where zero sits. A thermometer maker decided that water freezes at ‘0’ on the Celsius scale; a test publisher decided that the average IQ should read ‘100’. Neither zero point marks a true absence of the underlying property, and that is precisely what makes the resulting numbers interval rather than ratio. Knowing this in advance lets you plan the right analysis and avoid claiming ratios your data cannot support.
Instruments and scales that produce interval data
- Thermometers (Celsius / Fahrenheit) — the scale’s zero is anchored to a reference point (water freezing) rather than to the absence of heat.
- Standardised psychometric tests — IQ tests and many aptitude tests are normed so the population mean is fixed (often 100) with a set standard deviation, creating equal-interval scores from raw responses.
- Calendar and clock systems — dates and clock times are recorded against an agreed origin (the start of the AD era, midnight) that is conventional, not absolute.
- Composite rating scales — summing several Likert-type items into a total score is often treated as interval (see the debate below).
A practical workflow for collecting interval data
- Choose a calibrated instrument or normed scale. Confirm its zero is conventional and its units are equally spaced.
- Standardise the measurement conditions. Use the same instrument, units and procedure for every case so intervals stay comparable.
- Check reliability and validity. Calibrate devices and pilot test scales — see our guide on reliability and validity — before full data collection in statistics.
- Record the unit, never just the number. ‘20’ is ambiguous; ‘20°C’ is interval data you can analyse.
Because interval measurements come from instruments, much of it is gathered through structured quantitative observation and forms part of the wider family of numerical data. For more on the metrics that depend on a true zero (and so do not work here), see ratio data: definition, examples and analysis.
Collecting interval data
The Four Levels of Measurement
Interval data is best understood against the other three measurement levels. The table summarises what each level supports; a fuller treatment is in our guide to levels of measurement in statistics.
| Level | Ordered? | Equal intervals? | True zero? | Example |
|---|---|---|---|---|
| Nominal | No | No | No | Blood type, nationality |
| Ordinal | Yes | No | No | Satisfaction ranking, race position |
| Interval | Yes | Yes | No | Celsius temperature, IQ, calendar year |
| Ratio | Yes | Yes | Yes | Height, weight, reaction time, income |
