Which of the following best describes the behavior indicating to the patient that their time is valued?

SEMANTIC DIFFERENTIAL SCALES

Semantic differential scales consist of word lists and categories that represent various aspects of the patient's pain experience. The patient is asked to select words from these lists that best describe his or her present experience of pain. These types of scales are designed to collect a broad range of information about the patient's pain experience and to provide quantifiable data for intrasubject and intersubject comparisons. The semantic differential scale included in the McGill Pain Questionnaire, or variations of this scale, are commonly used to assess pain132-134 (Fig. 22-19). This scale includes descriptors of sensory, affective, and evaluative aspects of the patient's pain and groups the words into various categories within each of these aspects. The categories include temporal, spatial, pressure, and thermal to describe the sensory aspects of the pain; fear, anxiety, and tension to describe the affective aspects of the pain; and the cognitive experience of the pain based on past experience and learned behaviors to describe the evaluative aspects of the pain. The patient circles the one word in each of the applicable categories that best describes the present pain.132,134

Semantic differential scales have a number of advantages and disadvantages compared with other types of pain measures. They allow assessment and quantification of various aspects of the pain's scope, quality, and intensity. Counting the total number of words chosen provides a quick gauge of the pain severity. A more sensitive assessment of pain severity can be obtained by adding the rank sums of all the words chosen to produce a pain-rating index. For greater specificity with regard to the most problematic area, an index for the three major categories of the questionnaire can also be calculated.134 The primary disadvantages of this scale are that it is time consuming to administer, and it requires the patient to have an intact cognitive state and a high level of literacy. Given these advantages and limitations, the most appropriate use for this type of scale is when detailed information about a patient's pain is needed such as in a chronic pain treatment program or in clinical research.

Semantic Differential Scale

The semantic differential scale is usually used for psychological measures to assess attitudes and beliefs.13 The researcher develops a series of rating scales in which the respondent is asked to give a judgment about something along an ordered dimension, usually of seven points. Ratings are “bipolar” in that they specify two opposite ends of a continuum (e.g., good-bad, happy-sad). The researcher sums the points across the items.

The Session Evaluation Questionnaire used in a study of Internet-based psychotherapy is an example of a semantic differential scale.9 The scale measures two constructs, “depth” and “smoothness,” related to clients' perceptions of clinical counseling sessions in which they have participated. The scale also measures two postsession mood constructs, “positivity” and “arousal.” The scale includes 24 items, 12 each for the two sections. One item related to the session reads:

The session was:

Bad-Good

One post-session item reads:

Right now, I feel:

Happy-Sad

The respondent is instructed to place an X in the space that most accurately depicts his or her feelings.

Semantic differential scaling is most useful when natural phenomena can be categorized in opposite or contrary positions. However, it limits the range of responses to a linear format.

2.3 High-dimensional Context Spaces

Osgood et al.'s semantic differential can also be seen as a predecessor of contemporary corpus-derived measures of semantics and semantic similarity. For example, Burgess and Lund's (1997) hyperspace analogue to language model (HAL) learns a high-dimensional context from large-scale linguistic corpora that encompass many millions of words of speech or text. The model tracks lexical co-occurrences throughout the corpus and from these derives a high-dimensional representational space. The meaning of a word is conceived of as a vector. Each element of this vector corresponds to another word in the model, with the value of an element representing the number of times that the two words co-occurred within the discourse samples that constitute the corpus. For example, the vector for dog will contain an element reflecting the number of times that the word ‘bone’ was found within a given range of words in the corpus. These vectors can be viewed as the coordinates of points (individual words) in a high-dimensional semantic space. Semantic similarity is then a matter of distance between points in this space.

Several other such usage-based models have been proposed to date; similar in spirit to HAL, for example, is Landauer and Dumais's (1997) latent semantic analysis (see also Semantic Processing: Statistical Approaches). The basic approach might be seen to be taking to its logical consequence Wittgensteir's famous adage of ‘meaning as use.’ Its prime advantage over related approaches such as spatial models of similarity and the semantic differential lies in the ability to derive semantics and thus measures of semantic similarity for arbitrarily large numbers of words without the need for any especially collected behavioral data. The ability of these models to capture a wide variety of phenomena, such as results in semantic priming and effects of semantic context on syntactic processing, has been impressive.

Semantic Differential

Osgood, Suci, and Tannenbaum used the semantic differential method to examine the effects of music on three different interpretations of a film. This technique was used in several of the following studies, measuring participants' reactions using bipolar scales with contrasting adjectives at each end. These adjectives are organized through ‘evaluation,’ ‘potency,’ and ‘activity’ (EPA). For example, evaluation may include the adjectives, ‘good-bad’; potency, ‘strong-weak’; and activity, ‘active-passive.’ The bipolar scales used in EPA studies are usually numbered 3-2-1-0-1-2-3, where 0 is labeled ‘neutral,’ 1 ‘slightly,’ 2 ‘quite,’ and 3 ‘extremely.’ In studies that involve the semantic differential, researchers may assess subjects' confidence in particular choices by means of a 5-point scale, from 1 (I have no idea) to 5 (I am sure that it is). Some scholars ask subjects to rate music on scales that have extreme values of ‘low-high liking,’ ‘very strong disagreement–agreement,’ ‘highly displeasing–pleasing,’ and ‘not familiar at all-extremely familiar.’ Others require subjects to rank emotions according to certain adjectives from 1 (not at all) to 5 (very much).

Convergent/Concurrent

Dasgupta and Greenwald (2001) reported significant correlations between their feeling thermometer and semantic differential scales (similar but not identical to the one discussed in the next section) for attitudes towards White (r=.51) and Black (r=.43) Americans (Experiment 1), and for attitudes towards the young (r=.70) and old (r=.71; Experiment 2). Turner and Feddes (2011) found that their respondents’ outgroup feeling thermometer ratings correlated strongly with their ratings of the same outgroups using a different attitudinal measure (.73 ≤ .91). These same-trait multimethod correlations indicate that the feeling thermometer shows good convergent validity.

1 Overview of Attitudes and Attitude Change

Attitudes are based on some combination of cognitive, behavioral, and affective influences, and are typically measured by self-report scales such as the ‘semantic differential,’ where a person rates the target on bipolar evaluative dimensions such as how good/bad or favorable/unfavorable it is. Increasingly, researchers have appreciated that it is also useful to assess attitudes on dimensions other than their valence, such as their accessibility (how quickly the attitude comes to mind) and ambivalence (how consistent the basis of the attitude is). These indicators of attitude ‘strength’ are useful in determining which attitudes are consequential and which are not. Strong attitudes are those that persist over time, are resistant to change, and predict other judgments and actions (Petty and Krosnick 1995). At any given moment, one's expressed evaluation can be influenced by a variety of contextual factors, but the common assumption is that one's core ‘attitude’ is the underlying evaluation that is capable of guiding behavior (one's actions), cognition (one's thoughts and memories), and affect (emotional reactions).

Attitude change occurs when one's core evaluation shifts from one meaningful value to another, and is typically inferred from a change in a person's scale rating, although behavioral and other indirect or implicit procedures for assessing change are sometimes used. Most studies of attitude change involve exposing individuals to a persuasive communication of some sort but, as noted below, some attitude change techniques do not involve exposure to any message. The earliest work on attitude change attempted to examine which variables and procedures increased and which decreased the likelihood of change (e.g., did more change occur when the message source was described as an expert than when the source lacked expertise even though the message was the same?).

Overview of Attitudes and Attitude Change

Attitudes are based on some combination of cognitive, behavioral, and affective (emotional) influences and are typically measured by self-report scales such as the semantic differential, where a person rates the target on bipolar evaluative dimensions such as how good/bad or favorable/unfavorable it is. Increasingly, in addition to assessing people's explicit or deliberative attitudes, assessments are made of the evaluations that come to mind automatically without much reflection (i.e., quick evaluative associations) using implicit measures. Measures such as Greenwald's Implicit Association Test (IAT) and Fazio’s evaluative priming measure are popular in this regard. Often, implicit and explicit measures yield the same outcome, but sometimes there are discrepancies which might suggest that people are unwilling to report their attitudes due to social desirability concerns or might be unable to report them due to limited access.

Furthermore, in addition to assessing attitudes on a positive to negative continuum, scholars have also found it useful to assess attitudes on dimensions other than their valence such as their accessibility (how quickly the attitude comes to mind), ambivalence (how consistent the basis of the attitude is), and certainty (how confident people are in the validity of their attitudes). These indicators of attitude ‘strength’ are useful in determining which attitudes are consequential and which are not. Strong attitudes are those that persist over time, are resistant to change, and predict other judgments and actions. At any given moment, one's expressed evaluation can be influenced by a variety of contextual factors, but a common assumption is that one's core ‘attitude’ is the underlying evaluation that is capable of guiding behavior (one's actions), cognition (one's thoughts and memories), and affect (emotional reactions).

Attitude change occurs when one's core evaluation shifts from one meaningful value to another, and is typically inferred from a change in either an implicit (automatic) or explicit (deliberative) measure of evaluation. Often, an attitude change induction produces change on both kinds of measures but sometimes change is more likely on one type of measure than another. Most studies of attitude change involve exposing individuals to a persuasive communication of some sort, but as explained shortly, some attitude change techniques do not involve exposure to any message. The earliest work on attitude change attempted to examine which variables and procedures increased and which decreased the likelihood of change.

3 Attitude Measurement

Because attitudes are so important, researchers have developed a variety of tools with which to measure them. Two of the most commonly used techniques are the Likert scale and the semantic differential. As shown in Fig. 2, participants reporting attitudes on a Likert scale are asked to what extent a variety of statements are characteristic of them. In a similar vein, as shown in Fig. 3, participants reporting attitudes on a semantic differential measure are asked to indicate how well a series of adjectives describe the attitude object. In either case, the researcher will typically provide a variety of items for the participant to complete. The scores from the items are combined to create an overall measure of the person’s attitude toward the target.

There are several guidelines that researchers can follow in attempting to tap a person’s attitude with as little error as possible. One guideline is to reverse code some of the items by framing them negatively as done in the third and fifth items in Figs. 2 and 3. Doing so serves two purposes. First, it reduces the likelihood that a person will simply check the leftmost or rightmost box down the entire instrument in attempts to save time or cognitive resources. A person is less likely to do so if he or she realizes that various items are indeed reverse coded. Similarly, if a person does simply check the leftmost or rightmost box down the entire instrument, the researcher may recognize this and treat the data accordingly. Also, some research suggests that providing respondents with five to seven options from which to select provides the most valid data. Scales with fewer than five options lack the precision necessary to measure attitudes accurately, whereas scales with more than seven options tend to suffer from additional noise without benefiting from any additional precision.

Likert scales and semantic differential scales are ubiquitous. Students typically use Likert scales when evaluating their professors at the end of a semester, and departments and universities use these data in part to make personnel decisions. Companies also use such techniques to measure customers’ attitudes toward the products and/or services the companies offer. By learning which aspects of the companies customers like and dislike, managers can direct resources to improve deficient areas of their companies. In short, proper attitude measurement can gauge the likes and dislikes of a person or a group of people. Behavior of a person or a group of people can then be predicted based on the attitude measure.

6.2 Measurement of Meaning

An early research program at the University of Illinois was the attempt to develop an operational definition of the central concept of meaning (Osgood et al. 1957). The semantic differential response scale was developed to track the dimensions along which people orient themselves to external entities, such as people in the news. Three dimensions of meaning were usually found, across a variety of rated objects: evaluation (‘good–bad’), activity (‘active–passive’), and potency (‘strong–weak’). The three-dimensional concept of ‘semantic space’ was explored for its cross-cultural validity, theoretical implications in cognition, and applications to content analysis. The evaluative rating method persists in the field as a versatile and efficient method for measuring opinion change as a communication effect.

Semantic Distance

Cognitive psychologists have developed a number of experimental techniques for investigating hypotheses derived from schema theories. An early approach made use of the idea of semantic distances between words or concepts based on semantic differential ratings. This methodology was pioneered by Charles Osgood and his colleagues as a means of exploring the way meaning is represented in the mind (Osgood, Suci, and Tannenbaum, 1957; Snider and Osgood, 1969). It involved having a large number of research participants rate particular words or concepts on a series of rating scales, each scale representing a dimension between a pair of adjectives with opposite meanings (e.g., hot-cold, fast-slow, likable-unlikable). These ratings were then factor analyzed to identify a smaller number of basic dimensions (factors) that capture most of the variance in the ratings.

Using ratings of a large number of concepts and many different samples of participants, Osgood and his colleagues repeatedly found three basic orthogonal factors, which he labeled Activity (active-passive), Evaluative (good-bad), and Potency (strong-weak). These three factors appear to be basic dimensions by which people mentally organize meanings that they attach to a wide range of concepts. The factors can be conceptualized as dimensions of a hypothetical three-dimensional cognitive “space” in which people store words and concepts in their minds. The factor loadings of a particular word or concept can be used to identify where the concept is stored in this space. Concepts that are similar in meaning are stored closely together in this hypothetical semantic space, since they have similar loadings on the three factors, whereas those that are quite different in meaning have different loadings and are stored at more distant locations. Thus, semantic distances between pairs of words or concepts can be quantified by means of the difference in their loadings on the semantic differential factors. This technique provided cognitive researchers a method for investigating the way knowledge or meanings of concepts are organized in people's minds.

This method was applied to the study of humor by Michael Godkewitsch (1974), at the University of Toronto, using the semantic distances between pairs of words as a method of quantifying incongruity. Participants were presented with a number of adjective-noun pairs and asked to rate them for funniness and wittiness. The degree of smiling and laughter of participants was also observed. The semantic distance between the words in each pair was computed on the basis of their loadings on the semantic differential factors. As predicted by incongruity theory, adjective-noun pairs that were more discrepant from one another in semantic space were judged to be funnier and evoked more smiles. For example, the adjective-noun pair “happy child,” in which both words load similarly on the semantic differential factors, was not seen as very funny. In contrast, “wise egg,” with an intermediate distance, was funnier, and “hot poet,” with a high semantic distance, was even funnier. Although, admittedly, the humor evoked by these word pairs was not very great, it was systematically related to the semantic distance between the two words in each pair, providing support for incongruity theories of humor.

Tim Hillson and I also employed a semantic distance procedure to model the concept of resolution as well as incongruity in such simple verbal stimuli (Hillson and Martin, 1994). We hypothesized that word pairs that are quite distant on some dimensions of semantic space (incongruity) but are also quite close on other dimensions (resolution of incongruity) might be funnier than those that are either distant or close on all dimensions. We employed a methodology, called the domain-interaction approach, which had previously been used in the study of metaphors by other researchers (e.g., Trick and Katz, 1986). As humor stimuli, we used simple metaphor-like statements combining two concepts in the form “A is the B of A's domain” (e.g., “George Bush is the buzzard of world leaders”). The domains used were actors, world leaders, birds, makes of cars, foods, and magazines, and within each domain we used four nouns (e.g., Sylvester Stallone and Woody Allen were two of the actors).

Semantic differential ratings provided by a group of subjects on these nouns and domain names were factor analyzed, yielding four factors. We identified two of the factors as domain-distinguishing (i.e., different nouns within a given domain were found to have very similar loadings on these two factors, while nouns from different domains had more distant loadings). The other two factors were identified as domain-insensitive (i.e., different nouns within the same domain could have quite different loadings on these two factors). On the basis of these factor loadings, two types of semantic distance between the nouns were computed: a within-domain distance (using the domain-insensitive factor loadings), and a between-domain distance (using the domain-distinguishing factor loadings). We considered between-domain distance to be a way of operationally defining incongruity (greater distance = greater incongruity), and within-domain distance to be a way of operationalizing resolution (less distance = greater resolution).

We then created metaphor-like sentences using pairs of nouns from different domains and asked a second group of participants to rate them for funniness. As predicted (and consistent with the findings of Godkewitsch, 1974), the between-domain distance (incongruity) of the noun pairs in each sentence showed a significant positive correlation with the funniness ratings of the jokes. That is, noun pairs with greater between-domain distance were rated as more funny. Also as predicted, within-domain distance (resolution) showed no simple correlation with funniness, but did produce a significant interaction with between-domain distance in predicting funniness ratings. In particular, sentences that were rated as most funny were those that showed both high between-domain distance (incongruity) and low within-domain distance (resolution). To illustrate, a sentence that received a relatively high mean humor rating was “Woody Allen is the quiche of actors.” The between-domain semantic distance between Woody Allen and quiche was large (actors are quite different from foods on some dimensions), but the within-domain distance was small (Woody Allen and quiche are quite similar in some ways within their respective domains). Thus, there is incongruity but also some sort of resolution to the incongruity (i.e., the incongruity “makes sense” in some way).

The semantic distance approach did seem to capture some relevant dimensions of humor, as it was able to systematically predict funniness ratings of simple verbal material. It could still be a useful method for exploring various additional parameters that may be relevant to some types of humor. However, this technique has several limitations. It provides only a static picture of the organization of semantic meaning, and is therefore not useful for examining the processes whereby cognitive structures (schemas) are activated over time in processing humorous information. It also assumes that cognitive organization is the same in all people, and, because mean ratings are averaged across large numbers of participants, it is not amenable to studying individual differences in humor comprehension. In addition, it allows only for the study of simple “pseudo-jokes” made up of word pairs, rather than more complex real jokes and other natural forms of humorous material.