Deciphering Scales of Measurement: Nominal to Ratio

Every time a researcher designs a survey – whether studying media consumption habits, audience trust in news, or public opinion on a political issue – one fundamental question must be answered before a single number is crunched: what kind of data am I actually working with? The answer lies in understanding scales of measurement. These scales classify the nature of values assigned to variables, and they determine which statistical operations are valid, which are misleading, and which are outright meaningless. Get this wrong, and even a beautifully designed study can produce conclusions that don’t hold up. There are four scales to know: nominal, ordinal, interval, and ratio – often remembered by the acronym NOIR.

Why scales of measurement matter in research 🔗

Numbers can be deceiving. Assign the number “1” to male respondents and “2” to female respondents in a dataset, and it looks like numerical data – but calculating the average of 1.5 tells you nothing meaningful. That’s the core problem scales of measurement solve. The level of measurement of a variable determines the appropriate statistical tools and techniques that can be applied for analysis. In other words, the scale you choose when collecting data sets the ceiling on what you can legitimately say about that data later.

Psychologist Stanley Smith Stevens introduced this four-level framework in a landmark 1946 article in Science, titled “On the Theory of Scales of Measurement,” arguing that all scientific measurement falls into one of these four types. Decades later, it remains the standard classification system in social science, behavioral research, and communication studies.

The nominal scale: labeling without ranking 🔗

A nominal scale is used to label variables that have no natural order or quantifiable difference between values. The word “nominal” comes from the Latin nomen, meaning name – and that’s exactly what this scale does: it names categories. Think of variables like religion, political party affiliation, type of media consumed (print, digital, broadcast), or a respondent’s city of residence. These categories are mutually exclusive and have no hierarchy. One is not greater than the other; they are simply different.

Numbers can be assigned to nominal categories purely for convenience – for instance, 1 = Print, 2 = Television, 3 = Online. But these numbers carry no mathematical meaning. No arithmetic computation – addition, subtraction, multiplication – may be performed on nominal measures. The only valid statistical operation is counting frequency and identifying the mode (the most common category). Bar charts and pie charts are the standard visualization tools for nominal data.

Nominal data in communication research 🔗

In journalism and media studies, nominal data appears constantly. Content analysis involves coding and categorizing elements of communication messages – such as news articles, social media posts, and advertisements – based on predefined nominal categories. For example, classifying news stories by topic (crime, politics, health, entertainment) is nominal-scale work. So is recording whether a source in a story is male or female, or whether a media outlet is public or private.

The ordinal scale: ranking without equal intervals 🔗

The ordinal scale takes us one step further. With ordinal data, categories can be ranked, but the distance between the ranks cannot be measured. The order is meaningful; the gaps between ranks are not assumed to be equal.

A practical example: if respondents are asked to rank their satisfaction with a news channel as “Very Dissatisfied, Dissatisfied, Neutral, Satisfied, Very Satisfied,” we know that “Satisfied” is better than “Neutral.” But we cannot claim the emotional jump from Neutral to Satisfied is the same size as the jump from Satisfied to Very Satisfied. The ordinal scale places events in order, but makes no attempt to ensure equal intervals between those positions.

For ordinal data, the median is the appropriate measure of central tendency, since it identifies the middle-ranked value. Statistical tests like the Mann-Whitney U test and the Kruskal-Wallis H test are used to compare two or more ordinal groups.

Likert scales: ordinal or interval? 🔗

This is one of the most debated questions in social science research. According to the SAGE Encyclopedia of Communication Research Methods, researchers in the social sciences generally consider Likert scales to be an interval-level measure, meaning equal distances exist between consecutive points – though there is ongoing disagreement about whether the scales are more appropriately defined as ordinal measures. Some scholars argue respondents don’t perceive the psychological distance between “Disagree” and “Neutral” as identical to the distance between “Neutral” and “Agree.”

In practice, researchers often treat summed Likert scores as interval data if the scale has multiple items and behaves like a continuous measure – enabling the use of means, standard deviations, and parametric statistical tests. When reporting such analysis, researchers are advised to note that the underlying items are technically ordinal, even if the composite score is analyzed as interval data.

The interval scale: equal distances, no true zero 🔗

The interval scale adds a critical new property: equal, measurable distances between values. The interval scale labels and orders variables with a known, evenly spaced interval between each value. This means arithmetic on the differences is valid – you can meaningfully say that the gap between two values is twice the gap between two other values.

The most commonly cited example is temperature in Celsius or Fahrenheit. The difference between 20°C and 30°C is exactly the same as between 40°C and 50°C. In interval measurement, the distance between 1 and 2 is equal to the distance between 9 and 10. This allows researchers to calculate the mean and standard deviation – something not valid with ordinal or nominal data.

However, interval scales have a crucial limitation: they lack a true zero point. Zero degrees Celsius does not mean the complete absence of temperature – it’s simply a point on the scale. Because of this, ratio comparisons are not valid. You cannot say 40°C is “twice as hot” as 20°C in any physically meaningful sense. Unlike nominal or ordinal scales, the interval scale quantifies the difference between variables, but it cannot calculate a “true zero” value.

Interval data in practice 🔗

In communication and social research, interval-scale data often appears in standardized psychological assessments, credit scores, or IQ tests. IQ scores represent an ordinal scale – or at best interval – where there is no zero point representing a complete absence of intelligence, and a 10-point difference may carry different meanings at different parts of the scale. Attitude measurement instruments that have been rigorously validated also often operate at the interval level.

The ratio scale: the gold standard of measurement 🔗

The ratio scale is the most informative of the four. It has all the properties of the interval scale – order, labeled categories, and equal intervals – plus one defining feature: a true zero point that represents the complete absence of the measured attribute. A good example of ratio data is weight in kilograms. If something weighs zero kilograms, it truly weighs nothing – unlike 0°C, which doesn’t mean zero heat energy.

Because a true zero exists, ratio comparisons are fully valid. A person who earns ₹80,000 per month earns exactly twice as much as someone earning ₹40,000. A news broadcast that airs for 60 minutes is exactly three times as long as one that airs for 20 minutes. The ratio scale accommodates the characteristics of all three other scales – labeling, order, and equal intervals – while also enabling the establishment of absolute zero.

In the physical sciences and engineering, most measurement is done on ratio scales – examples include mass, length, duration, plane angle, energy, and electric charge. In communication research, ratio-scale variables include the number of social media followers a news account has, the duration of a broadcast, the number of times an article was shared, or the word count of a press release.

Statistical richness of ratio data 🔗

With ratio data, researchers can apply the full suite of statistical techniques – measures of central tendency (mean, median, mode), measures of dispersion (standard deviation, range), and more advanced methods like the geometric mean and coefficient of variation. This is why ratio data is highly preferred when study design permits its collection.

Choosing the right scale: a practical decision 🔗

The choice of measurement scale is not always fully in a researcher’s hands – it often depends on the nature of the variable itself. If you’re recording gender, eye color, or marital status, those are definitively nominal. If you’re recording time or temperature values, those are interval or ratio. If you’re ranking preferences or opinions, it’s ordinal data.

Where researchers do have a choice – say, measuring income – they can opt for a richer scale. Recording exact income figures gives ratio-level data. Grouping respondents into “Low, Middle, High” income brackets reduces that to ordinal. The more precise the scale, the more analytical flexibility available later. A general rule: collect at the highest scale possible, since higher-level data can always be downgraded (a ratio variable can be converted to ordinal), but you cannot upgrade lower-level data after the fact.

Misapplying scales leads directly to flawed conclusions. Treating ordinal data as interval – assuming equal intervals between Likert scale points – or analyzing nominal data with techniques that require higher-level measurements, are among the most common errors in communication research. Statistical software like SPSS flags measurement levels precisely to prevent such errors.

Bringing it together: NOIR in a single research project 🔗

Consider a study examining how much Indian audiences trust digital news sources. A single survey could easily use all four scales simultaneously:

• Nominal: Type of device used to access news (smartphone, laptop, tablet, TV)
• Ordinal: Trust level in news sources (Very Low, Low, Moderate, High, Very High)
• Interval: Score on a validated media literacy scale (e.g., 0-100 composite score)
• Ratio: Number of minutes spent consuming news per day

Each variable requires a different analytical approach. Comparing trust levels across device types uses frequency tables and chi-square tests (nominal). Ranking trust scores uses medians and non-parametric tests (ordinal). Calculating the average media literacy score uses means and standard deviations (interval). Calculating the average news consumption time and comparing it across groups uses the full range of descriptive and inferential statistics (ratio). Understanding these levels enables researchers to select appropriate statistical tests and interpret results accurately – and to avoid presenting misleading findings as credible science.

Scales of measurement are not a bureaucratic formality to be memorized for an exam and forgotten. They are the logical infrastructure of every quantitative study. Misidentifying your scale doesn’t just produce wrong numbers – it produces wrong stories. And in research, as in journalism, getting the story right is everything.

What do you think? When a researcher assigns numbers to Likert-scale responses in a survey about media trust, are they justified in treating that data as interval-level for the sake of richer analysis – or does doing so introduce a distortion that undermines the study’s validity? And if you were designing a study on social media news consumption among young adults, which combination of scales would you choose for your key variables, and why?

References

1. https://en.wikipedia.org/wiki/Level_of_measurement
2. https://researcher.life/blog/article/levels-of-measurement-nominal-ordinal-interval-ratio-examples/
3. https://www.statology.org/levels-of-measurement-nominal-ordinal-interval-and-ratio/
4. https://fiveable.me/communication-research-methods/unit-7
5. https://www.questionpro.com/blog/nominal-ordinal-interval-ratio/
6. https://sk.sagepub.com/ency/edvol/the-sage-encyclopedia-of-communication-research-methods/chpt/scales-likert-statement
7. https://careerfoundry.com/en/blog/data-analytics/data-levels-of-measurement/
8. https://web.pdx.edu/~newsomj/pa551/lecture1.htm
9. https://statisticsbyjim.com/basics/nominal-ordinal-interval-ratio-scales/
10. https://library.fiveable.me/communication-research-methods/unit-7/levels-measurement/study-guide/8v6OKBf5bJGFQtGT