Semantic Differential

Semantic Differential Definition

Semantic DifferentialThe semantic differential is a method of measurement that uses subjective ratings of a concept or an object by means of scaling opposite adjectives to study connotative meaning of the concept or object. For example, the first level meaning of a car is that of a transportation device; the second level meaning of a car can also be its value as a status symbol. The semantic differential is designed to measure these second levels—in other words, connotative meanings of an object. The semantic differential is mostly used for measuring attitudes toward social and nonsocial objects, but also to assess quality and type of interactions between people. The method was developed by Charles Osgood in the 1950s and has been broadly used in and outside of psychology.

The semantic differential usually consists of 20 to 30 bipolar rating scales (i.e., the scale is anchored by an adjective on each side, for example warm-cold) on which the target object or concept is judged. Basis for the judgment is not so much the denotative or objective relation of the object and the adjective anchors of the bipolar scales (because it may not be given at first glance given our car example earlier and the rugged warm-cold adjective pair) but, rather, the metaphoric or connotative closeness of the object and the anchors of the bipolar scales. For example, on a metaphorical or connotative level, a family car might be judged as warm, whereas a delivery truck might be judged as more cold. The denotative meaning, that is, firsthand meaning, might be quite similar, in terms of being an adequate transportation device in both cases.

Semantic Differential Background

Social psychologists, but also market researchers or public pollsters, are often interested in the subjective (i.e., somewhat hidden and varying between individuals) definition of meaning that an object or concept has beyond its mere brute facts, as well as in the attitude of a certain group of people concerning a certain object or concept.

Meaning can be divided into four different dimensions: structural (a possible higher-level similarity to other objects, e.g., a sports car and a truck are different, but structurally similar because they are both means of transportation), contextual (depending on the current context, e.g., a truck serves as a transportation device, but can also be an vintage car later on), denotative (objective, brute facts of the car, such as horsepower), and connotative (more metaphoric, second-level associations). Osgood was particularly interested in this fourth dimension of meaning. His scaling method was meant to measure individual differences in the connotation of a word describing an object or a concept.

Construction and Use of Semantic Differentials

The actual questionnaire consists of a set of bipolar scales with contrasting adjectives at each end. The positions on the scale in between can be numbered or labeled. Note that the neutral middle position is usually marked by zero and the other positions by numbers increasing equally in both directions. Thus, each scale measures the directionality of a reaction (e.g., good vs. bad) and its intensity (from neutral via slight to extreme). In most cases, the universal adjective pairs are used because translations in many languages are available. Besides universal semantic differentials, object- or concept-specific sets of adjective pairs can be used. For the latter, great care while constructing the respective semantic differentials is necessary to avoid problems (outlined in the next section). For the universal semantic differential, cross-cultural comparisons revealed that three basic dimensions of response account for most of the covariation. These three dimensions have been labeled “evaluation, potency, and activity” (EPA) and constitute the semantic space (i.e., the set of descriptive attributes) of the target to be judged. Some of the adjective pairs are direct measures of the dimensions (e.g., good-bad for evaluation, powerful-powerless for potency, and fast-slow for activity); others rather indirectly relate to the single dimensions of the EPA structure. Given the research conducted, for each new case meaning of the scales should not just be inferred from previous results. Dimensionality should be checked so that scales that do not represent a unidimensional factor are not summed up.

Analysis of Data

At first glance, analysis of semantic differential data seems easy, but actually, it is a rather complex procedure. It is not sufficient to simply average scale ratings for each individual and to use mean differences on a judged object or concept. In fact, the underlying factor structure must be determined and correlations of similarity between the profiles must be computed. Data from semantic differentials contain three levels or modes: the target objects or concepts, the scales themselves, and the responding individuals. Thus, before factor analysis, these three-mode data need to be collapsed into a two-mode structure. This can be done either by summing over targets for each individual and scale or by averaging over individuals for each scale-concept combination. Also, one can deal with target objects separately, likewise with individuals. Finally, each individual target object-concept response can be transferred in a new matrix and inter-scale correlations can be computed. Note that different methods of collapsing modes can produce rather different correlation patterns.

The original semantic differential is currently rarely used in social psychology (but widely outside this field). Yet, a lot of related measurement methods in social psychology have been influenced by it. Almost every stereotype rating using, for example, competence or warmth as its basic dimensions follows the idea of the original concept. The use of the original concept is not without pitfalls and problems. This is especially crucial because many researchers outside of social psychology are not aware of these issues. First, the method is partly self-contradictory: For some words (in this case, the concepts to be measured), people’s connotations are assumed to differ, but for other words (in this case the adjectives used as endpoints of the single scales), this assumption should not hold. Second, scales may be relevant to the target objects or concepts to a different degree. These concept-scale interactions are to be treated carefully by determining the structure of the dimensions by using a factor analysis instead of the blind adoption of the EPA structure. Third, a number of problems arise during the administration itself. For some individuals, judging objects on the given scales is hard because the adjective pairs seem unrelated to the target object. In addition, respondents may give socially desirable answers, or can develop a so-called response set, meaning that they would consistently give moderate or very extreme answers. Some of these problems can be overcome by anonymity of the respondents, inclusion of irrelevant target words to disguise the true purpose of the semantic differential, or by checking for response sets. Finally, some problems with the semantic differential arise from a thoughtless use, administration of the method, and analysis of its data. Not every set of bipolar scales and given adjective pairs constitute a semantic differential. The underlying dimensions and possible overlap of the adjective pairs are not assessed in many cases and consequences resulting from it are ignored.

The semantic differential can be an informative and economic measure for the connotation of objects or concepts. However, the user should be fully aware of the complexity of the method and reflect its value carefully.


  1. Heise, D. R. (1970). The semantic differential and attitude research. In G. F. Summers (Ed.), Attitude measurement (pp. 235-253). Chicago: Rand McNally.
  2. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Urbana: University of Illinois Press.

Read More: