In the social sciences, it is often important to measure attributes of individuals that are not readily observable, such as beliefs, attitudes, emotional experiences, and personality traits. Because such attributes are not readily observable, social scientists rely on scales that allow individuals to report the extent to which they possess or experience them.
SCALES OF MEASUREMENT
A scale provides both qualitative information and quantitative information about an attribute. The most basic type of scale has an individual simply indicate whether he or she possesses a certain attribute. This type of scale provides only a qualitative indication of the presence or absence of the attribute and is considered to have a nominal scale of measurement. A second type of scale has the individual indicate the presence or absence of the attribute in question and the quantitative amount of the attribute he or she possesses. For example, a scale of this sort could ask individuals to indicate how much they like ice cream by selecting a number from one to three, where one represents not at all, two represents moderately, and three represents extremely. This type of a scale has an ordinal scale of measurement, which provides a rank-order quantitative value of the degree to which the attribute is present. Note that even if these values are related to one another in a rank ordering, the psychological distance between neighboring values may not be equivalent (e.g., in the example above, the psychological difference between not at all and moderately may not be the same as the psychological difference between moderately and extremely). When a scale provides rank-order values of an attribute and equal psychological distances between neighboring values, the scale has an interval scale of measurement. Finally if a scale has equal psychological distances between neighboring values and a value that reflects the complete absence of the attribute in question, then the scale has a ratio scale of measurement. Most scales used by social scientists are nominal or ordinal in nature because it is very difficult to objectively determine if the psychological distance between neighboring scale values is truly equivalent.
TYPES OF SIMPLE SCALES
Two general types of non-nominal scales are used in the social sciences. The first type is the Likert scale, which consists of labeled discrete values located between two labeled endpoint values. The ice cream scale described above is an example of a Likert scale. The second type is the visual analog scale, which consists of either discrete unlabeled values or a continuous unlabeled line between labeled endpoints. Likert scales are ordinal and provide only a limited number of values from which to choose, but they yield information about the qualitative degree to which an individual possesses the attribute in question. In contrast, continuous visual analog scales provide a large number of potential values from which to choose, but they yield little information about the qualitative degree to which an individual possesses the attribute in question. Some believe that visual analog scales have an interval scale of measurement, whereas others argue that they possess only an ordinal scale of measurement.
COMBINING SCALES INTO INVENTORIES
Ratings on individual scales are often combined to form an inventory. Inventories are essential when assessing complex attributes. Creating an inventory ensures that an attribute is measured in its entirety, which may be difficult to accomplish with a single scale. In an inventory, individuals respond to a number of statements using the same scale of measurement. For example, an anger inventory may be created by asking respondents to rate how angry, irritated, and mad they feel, each on a discrete visual analog scale. Each statement in the inventory must refer to a single object in order to minimize confusion. It is also useful to include statements that require responses that are the opposite of other statements in the inventory (e.g., How much do you like ice cream? vs. How much do you hate ice cream?) in order to reduce the likelihood that individuals are reporting responses without paying close attention to the scale items.
RELIABILITY AND VALIDITY
A scale or inventory must be both reliable and valid. Reliability is the extent to which an individual makes similar ratings across time using the same scale or, in the case of an inventory, the similarity of an individual’s responses to related items within the inventory. Assuming that a scale or inventory is reliable, it must also have high construct validity. Construct validity is the extent to which the scale or inventory accurately measures the attribute in question. High construct validity can be demonstrated in a number of ways. First, one can establish convergent validity by showing that responses on the scale or inventory are positively associated with patterns of behavior or responses on another scale or inventory believed to measure the attribute or a similar attribute. Second, one can establish discriminant validity by showing that responses on the scale or inventory are not positively associated with patterns of behavior or responses on another scale or inventory believed to reflect an unrelated attribute.
