Scales of Measurement

As described above in the discussion of the concept of intelligence, how a researcher measures the dependent
variable in a study can be important. Different types of measures allow different questions to be asked about
the behaviors. In addition, some types of measures are more reliable than others. The choice of scale also
constrains the types of statistical tests that are used to analyze the data. There are four primary scales of
measurement for dependent variables: (1) nominal, (2) ordinal, (3) interval, and (4) ratio. Each of these scales
is described below.

Nominal Scales. The simplest scale of measurement is a nominal scale, where nonordered categorical responses
are collected. For example, if researchers are measuring a person’s current mood, they may ask the respondent
to choose from categories such as anxious, happy, depressed, and angry. These categories are not ordered on
any continuum; they are merely different categories that reflect a person’s mood. Another example of a
nominal scale is a student’s selection of a college major. Values on this scale are categories (e.g., English,
education, psychology, philosophy) that are not ordered in any particular way. Gender is another common
nominal variable (e.g., male, female). The responses on a nominal scale do not involve numerical values; thus,
nominal scales are considered qualitative data rather than quantitative data.

Ordinal Scales. When the response categories of a measure contain an ordering (i.e., can be ordered from
lowest to highest) on a continuum of measurement, the measure is considered an ordinal scale. However,
response categories on an ordinal scale are not assumed to be equally spaced on the continuum. They are
merely ordered categories on the scale. For example, in measuring a person’s mood, a researcher may ask
respondents to rate their current anxiety level with possible response categories of “not at all anxious,” “a little
anxious,” “fairly anxious,” and “very anxious.” These categories are ordered from lowest amount of anxiety to
highest, but there is no implied equivalence in the difference between the “not at all anxious” and “a little
anxious” response categories and the “a little anxious” and “fairly anxious” response categories. Thus, the
response categories on this scale are not necessarily equally spaced. Another example of an ordinal scale might
be asking students to indicate how often they consume alcohol during a typical week with response categories
of “none,” “once,” “two to five times,” and “more than five times.” Again, these response categories are ordered
from smallest to largest amount, but the difference between “none” and “once” is not the same as the
difference between “once” and “two to five times.”

Rank orderings of stimuli are also considered ordinal scales. For example, if preference for different products
is being measured, the respondents may be asked to rank order (e.g., first, second, third, etc.) the products
according to their preference. In rank orders, the difference between the first and second choices may be much
larger or smaller than the difference between the second and third choices. Like nominal scales, ordinal scales
involve qualitative data because the response values are not numerical.

Interval and Ratio Scales. Interval and ratio scales involve numerical response values; thus, they involve
quantitative data. There are many measurement scales in psychology that are considered interval and ratio
scales. Interval scales involve numerical categories that are equally spaced. Thus, a common interval scale is

180
the Likert scale, where respondents are asked how much they agree or disagree with a statement on a 1 to 5 or
a 1 to 7 scale (where possible responses are any whole value between these endpoints). Responses from surveys
are often measured on a Likert scale with qualitative anchors provided to give the numerical values meaning
(e.g., 1 = item is least like me, 3 = item is somewhat like me, 5 = item is most like me). However, the responses
given by participants are the numerical ratings, with equal spacing between the different numbers on the scale.

Ratio scales also involve numerical measurements, but ratio scales allow a ratio comparison of values between
individual scores. For example, a score of 50 on a ratio scale is twice as high as the score of 25. The amount of
time to complete a task (sometimes called reaction time) is a common ratio scale used in psychological
research. If one participant takes 250 milliseconds to complete a task and another participant takes 500
milliseconds to complete the same task, it can be said that the second participant took twice as long as the first
participant. Accuracy for a task is also a ratio scale. In fact, measurement scales for distance, time, accuracy,
height, or weight are ratio scales. For many ratio measurement scales (such as many of those listed above), 0 is
the lowest value possible; however, some ratio scales do involve values lower than 0 (e.g., reaction times
measured in a study can be negative, if the participant responds before the stimulus appears).

Temperature scales provide a good example of the difference between interval and ratio scales. Common
temperature scales (Fahrenheit, Celsius) involve values of 0 that are not the coldest temperatures possible
(anyone who has lived through a cold winter can tell you that negative temperatures are possible and can be
unpleasant). In addition, Fahrenheit and Celsius temperature scales are interval scales because 20° is not twice
as cold as 10° or four times as cold as 5°. The Kelvin scale for temperature, however, does include a 0 value
that is the lowest possible value. The value of 0 on this scale indicates that it is the lowest temperature
possible, and scores on the Kelvin scale are ratios of each other (e.g., 100 K is twice as hot as 50 K). Thus, the
Kelvin scale of temperature is a ratio scale. See Table 5.1 for a comparison of the different measurement scales
and some additional examples of each type of scale.

Correctly identifying the scale of measurement used for the dependent variable is an important step in
conducting appropriate statistical analyses. Different scales may require different types of analyses in some
designs. This issue will be discussed further in Chapters 7 and 15.

Dependent/Response Variable: a variable that is measured or observed from an individual

Reliability: the degree to which the results of a study can be replicated under similar conditions

Operational Definition: the definition of an abstract concept used by a researcher to measure or manipulate the concept in a research

181
study

Nominal Scale: a scale of data measurement that involves nonordered categorical responses

Qualitative Data: nonnumerical participant responses

Quantitative Data: numerical data

Ordinal Scale: a scale of data measurement that involves ordered categorical responses

Interval Scale: a scale of data measurement that involves numerical responses that are equally spaced, but scores are not ratios of each
other

Ratio Scale: a scale of data measurement that involves numerical responses, where scores are ratios of each other

Likert Scale: a scale of responses that measures a participant’s agreement or disagreement with different types of statements, often
with a rating from 1 to 5 or 1 to 7

Reaction Time: measurement of the length of time to complete a task

182