Generalized Attitude Measure

Abstract

The Generalized Attitude Measure (GAM) is a six-item, bipolar semantic differential instrument designed by communication scholar James C. McCroskey to quantify an individual’s general, evaluative disposition toward any target concept, topic, behavior, or source. Originally conceptualized in McCroskey’s (1966) foundational research on persuasive communication and subsequently refined across four decades of empirical investigation (McCroskey & Richmond, 1989, 1996; McCroskey, 2006), the GAM addresses a pervasive methodological challenge in social science: the proliferation of idiosyncratic, unstandardized attitude scales that impair cross-study comparability and psychometric rigor. Utilizing a seven-point response format anchored by six universal evaluative adjective pairs (Good/Bad, Wrong/Right, Harmful/Beneficial, Fair/Unfair, Wise/Foolish, Negative/Positive), the instrument isolates the primary evaluative dimension identified in Charles E. Osgood’s classical semantic differential methodology. Psychometric evaluations demonstrate exceptional internal consistency, with Cronbach’s alpha coefficients typically ranging from .90 to .98 across diverse respondent samples and experimental conditions. Exploratory and confirmatory factor analyses consistently substantiate an invariant unidimensional structure accounting for 70% to 85% of total variance, with factor loadings regularly exceeding .80. The GAM exhibits robust convergent, discriminant, and predictive validity when paired with measures of cognitive belief, behavioral intent, and communicative behaviors. As a target-adaptable, domain-general assessment tool, the GAM remains a staple methodology in instructional communication, organizational research, health campaigns, and mass media effects research.

Keywords

Generalized Attitude Measure, semantic differential, attitude measurement, James C. McCroskey, evaluative dimension, psychometrics, communication research, persuasion, construct validity, instructional communication

Authors

The Generalized Attitude Measure was developed and refined by leading scholars in the field of communication studies:

• James C. McCroskey, Ed.D. (1938–2012): McCroskey was the Claude Worthington Benedum Professor of Communication Studies at West Virginia University and later served as Professor of Communication Studies at the University of Alabama at Birmingham. Recognized as one of the most prolific scholars in the history of communication, McCroskey authored or co-authored over 50 books and 300 journal articles, pioneering instruments for measuring communication apprehension, source credibility, and teacher immediacy.
• Virginia P. Richmond, Ph.D.: Professor Emerita of Communication Studies at West Virginia University and Professor Emerita at the University of Alabama at Birmingham. Richmond is an internationally distinguished communication researcher specializing in instructional communication, nonverbal behavior, interpersonal power dynamics, and psychometric measurement in human interaction.

Purpose

The primary purpose of the Generalized Attitude Measure (GAM) is to provide an efficient, robust, and universally adaptable metric for assessing human attitude toward any designated stimulus, concept, proposal, persona, or behavior. In experimental and survey research, investigators frequently require an assessment of participants’ valence and intensity of attitude toward novel or highly specific attitude objects—such as an instructional topic, a public policy proposition, a health intervention, or an organizational leadership style. Historically, researchers responded to this need by composing ad hoc Likert-type items tailored exclusively to the target concept. This practice introduced severe psychometric deficiencies, including unverified factor structures, inconsistent item-to-total correlations, scale-specific idiosyncrasies, and an inability to compare effect sizes directly across divergent experimental interventions or distinct empirical investigations.

The theoretical rationale behind the GAM posits that although the cognitive content associated with distinct attitude objects differs fundamentally, the underlying evaluative orientation (the degree of favorableness or unfavorableness) operates across an invariant, standardized continuum. By establishing a fixed set of high-performing evaluative adjective pairs, the GAM decouples the measurement process from the substantive details of the topic under examination. A researcher needs only to present the target concept as the prompt (e.g., "Capital Punishment," "My Communication Instructor," "Adopting Cloud Infrastructure," or "Vegetarian Diet") followed immediately by the identical six semantic differential pairs.

In instructional communication settings, the GAM is extensively deployed to assess affective learning—specifically students’ generalized attitudes toward course content, the instructor, and recommended educational practices. In persuasion and social influence research, the GAM serves as a sensitive pre-test and post-test measure to detect subtle attitudinal shifts resulting from evidence manipulations, source credibility variations, framing effects, and message design features. Furthermore, in clinical, health, and organizational settings, the tool enables rapid evaluation of client sentiment toward therapeutic recommendations, compliance regimens, and managerial initiatives without imposing cognitive fatigue on respondents.

Psychological Construct

The GAM measures the psychological construct of attitude, defined classically within cognitive and social psychology as a learned predisposition to respond in a consistently favorable or unfavorable manner with respect to a given object (Fishbein & Ajzen, 1975). Unlike cognitive beliefs (which assess the subjective probability that an object possesses a particular attribute or that a proposition is true or false) or behavioral intentions (which assess perceived future action tendencies), an attitude reflects a purely evaluative summary judgment.

The GAM operationalizes attitude along an unambiguous, unidimensional evaluative axis comprising valence and intensity:

• Evaluative Valence: Evaluative valence represents the directional sign of the psychological orientation—whether the stimulus is appraised positively (approached, approved, valued) or negatively (avoided, rejected, deplored). In the GAM, valence is captured via polar opposites representing fundamental moral, pragmatic, and qualitative judgments (e.g., Good/Bad, Wrong/Right, Harmful/Beneficial, Fair/Unfair, Wise/Foolish, Negative/Positive).
• Attitudinal Intensity: Attitudinal intensity indicates the extremity or cognitive strength of the evaluation. Within the GAM, intensity is quantified across a 7-point continuum where the structural endpoints (1 and 7) capture profound polarization ("very strong feeling"), intermediate markers (2 and 6) designate moderate conviction ("strong feeling"), near-center markers (3 and 5) signify minor leanings ("fairly weak feeling"), and the midpoint (4) captures true neutrality, ambivalence, or indecision.

Crucially, the GAM avoids confounding the global evaluative dimension with other experiential or descriptive dimensions such as arousal, physical potency, aesthetic elegance, or perceived complexity. Every item on the scale functions as an alternate linguistic manifestation of the overarching evaluative vector, ensuring that the resulting composite score reflects an undiluted index of favorability toward the target referent.

Theoretical Framework

The GAM is theoretically rooted in two foundational traditions of twentieth-century social science: Osgood’s Semantic Differential Theory and the Expectancy-Value Formulations of Attitude.

The Semantic Differential and the Evaluative Dimension

In their landmark psychometric work, Osgood, Suci, and Tannenbaum (1957) demonstrated through extensive factor analytic investigations that human semantic space across diverse cultures, languages, and contexts is organized around three pervasive, orthogonal affective dimensions: Evaluation (good/bad, pleasant/unpleasant), Potency (strong/weak, heavy/light), and Activity (active/passive, fast/slow). Of these three dimensions, the Evaluation factor consistently accounted for the overwhelming preponderance of shared variance, routinely explaining twice as much variance as Potency and Activity combined. Osgood and colleagues maintained that the Evaluation dimension corresponds directly to the psychological construct of attitude. McCroskey adapted this theoretical insight by isolating the purest markers of the evaluative dimension, eliminating adjectives related to physical sensation, potency, or speed, thereby constructing an instrument composed exclusively of highly convergent evaluative indices.

The Tripartite Model and Belief-Attitude Differentiation

The GAM is further situated within the cognitive hierarchy outlined by Fishbein and Ajzen’s (1975) Theory of Reasoned Action and Ajzen’s (1991) Theory of Planned Behavior. This framework establishes an explicit distinction between beliefs, attitudes, and intentions. Beliefs are conceptualized as cognitive probabilities linking an object to specific attributes (e.g., "Higher education improves earning potential"); attitudes are the cumulative affective evaluations of those attributes (e.g., "Higher education is good"); and intentions reflect motivational readiness to act. McCroskey (2006) strictly adhered to this distinction by developing the Generalized Attitude Measure as a parallel instrument to the Generalized Belief Measure (GBM). While the GBM utilizes epistemic anchor pairs (e.g., True/False, Correct/Incorrect, Believable/Unbelievable), the GAM restricts its measurement space exclusively to hedonic, moral, and utilitarian evaluation, preventing the conceptual muddling that frequently impairs multi-item cognitive surveys.

Dual-Process Persuasion Models

Within contemporary communication theory, the GAM aligns with dual-process models such as the Elaboration Likelihood Model (ELM) (Petty & Cacioppo, 1986) and the Heuristic-Systematic Model (Chaiken, 1980). These models postulate that attitudes can be formed or altered via either central/systematic scrutiny of argument quality or peripheral/heuristic cues (e.g., source attractiveness or credibility). Because the GAM measures the terminal evaluative state rather than the cognitive route utilized to reach it, it serves as an unbiased dependent variable capable of detecting attitude change regardless of whether the persuasion process operated under high or low cognitive elaboration conditions.

Validity

Over five decades of psychometric evaluation and substantive empirical deployment have established the robust validity of the Generalized Attitude Measure across diverse populations, communicative media, and investigative designs.

Construct and Factorial Validity

Construct validity is evidenced by the scale’s ability to consistently replicate a single, stable latent evaluative dimension across disparate experimental manipulations. In his comprehensive re-validation study, McCroskey (2006) investigated the psychometric behavior of the GAM across large student samples evaluating a diverse array of controversial propositions (e.g., legalizing marijuana, mandatory service, tuition changes). Both exploratory and confirmatory factor analyses supported the invariant unidimensionality of the construct, demonstrating that each item reflects the same underlying evaluative attitude.

Convergent Validity

The GAM demonstrates robust convergent validity through its high, positive correlations with alternative, extensive attitude batteries and theoretically related constructs. Studies investigating instructional outcomes have repeatedly documented significant positive correlations between the GAM (evaluating students’ attitudes toward course content and instructor) and measures of:

• Teacher Immediacy (nonverbal and verbal immediacy behaviors): $r = .40 \text{ to } .65$
• Source Credibility (Competence, Character, and Caring): $r = .45 \text{ to } .70$
• Affective Learning scales developed by Andersen and Richmond: $r = .75 \text{ to } .90$
• Student Motivation and Intent to Enroll in Related Coursework: $r = .45 \text{ to } .60$

Discriminant Validity

Empirical evidence for discriminant validity is demonstrated by the scale’s capacity to distinguish between an individual’s evaluative orientation (attitude) and their epistemic acceptance (belief). In joint factor analyses of the Generalized Attitude Measure and the Generalized Belief Measure (McCroskey, 2006), the six GAM items and the five GBM items consistently loaded onto two distinct, orthogonal or moderately correlated factors with negligible cross-loadings (all cross-loadings < .20). Respondents routinely differentiate between whether a proposition is "True" (Belief) and whether it is "Desirable/Good" (Attitude)—for example, an individual may strongly believe that global temperatures are rising (High Belief) while appraising that reality as deeply harmful and negative (Negative Attitude).

Predictive and Criterion-Related Validity

In experimental persuasion research, the GAM accurately tracks exposure to strong versus weak arguments, credible versus non-credible sources, and emotionally framed versus neutrally framed messages. Studies consistently demonstrate that post-message GAM scores predict relevant downstream behavior—such as signing petitions, voluntary attendance, and compliance with health advisories—yielding predictive correlation coefficients ranging from .35 to .55.

Reliability

The Generalized Attitude Measure routinely exhibits exceptional internal consistency and temporal stability across diverse contexts and participant groups.

Internal Consistency

Across hundreds of empirical studies in communication, marketing, education, and social psychology, the Cronbach’s alpha ($lpha$) coefficients for the GAM regularly exceed the standard psychometric threshold of .80, typically falling between .90 and .98:

• McCroskey (2006) reported Cronbach’s alpha estimates exceeding .92 across multiple substantive topics, with individual sub-samples demonstrating alphas as high as .96 to .97.
• McCroskey and Richmond (1989, 1996) reported consistency estimates consistently at or above .93 in instructional communication research.
• Subsequent educational investigations utilizing the GAM to evaluate affective learning outcomes have observed composite alphas consistently clustering between .91 and .95.

The average inter-item correlation among the six pairs routinely falls within the optimal range of .65 to .80, indicating high item homogeneity without exhibiting destructive redundancy or collinearity.

Test-Retest Stability

When administered in the absence of an intervening persuasive message or instructional treatment, test-retest reliability assessments conducted across two- to four-week intervals yield stability coefficients ($r_{tt}$) ranging from .82 to .89, demonstrating that baseline generalized attitudes remain highly stable over time while remaining appropriately sensitive to deliberate experimental interventions.

Factor Analysis

The internal dimensionality of the Generalized Attitude Measure has been scrutinized across multiple decades using both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA).

Exploratory Factor Analysis (EFA)

When subjected to Principal Components Analysis (PCA) or Maximum Likelihood EFA, the six items reliably yield a single eigenvalue greater than 1.0 (typically ranging from 4.2 to 5.1). Scree tests consistently reveal a dramatic, clean drop-off between the first and second factors, with the primary evaluative factor explaining between 70% and 85% of the total variance. The second unrotated eigenvalue typically remains well below 0.50, rejecting any multidimensional structure.

Confirmatory Factor Analysis (CFA)

CFA investigations evaluating the unidimensional structural model have consistently documented excellent goodness-of-fit indices across diverse demographic and topic settings:

• Comparative Fit Index (CFI) $ge .98$
• Tucker-Lewis Index (TLI) $ge .97$
• Root Mean Square Error of Approximation (RMSEA) $le .05$ to $.06$ (90% CI [.02, .08])
• Standardized Root Mean Square Residual (SRMR) $le .02$ to $.03$

These fit metrics substantiate that the six items operate as parallel indicators of a single underlying evaluative dimension without substantial correlated measurement errors or method effects.

Instrument / Measurement Tool

• Instrument Type: Standardized self-report semantic differential scale
• Construct Measured: Generalized evaluative attitude (valence and intensity) toward a specified target concept
• Item Count: 6 bipolar adjective pairs
• Response Format: 7-point bipolar continuum:
• 1 and 7: Very strong feeling
• 2 and 6: Strong feeling
• 3 and 5: Fairly weak feeling
• 4: Undecided or do not understand (neutral point)
• Administration Time: Approximately 30 to 60 seconds
• Target Insertion: The target topic, policy, person, or concept is prominently printed directly above the six bipolar items (e.g., "Please indicate your evaluation of: [INSERT OBJECT]")
• Scoring and Transformation Rules:
  • To score the scale so that higher aggregate scores represent a more positive/favorable attitude:
  • Reverse Code: Items 1, 3, and 4 must be reversed prior to scoring. Specifically:
  • Item 1 (Good = 1, Bad = 7): Recode $1 \rightarrow 7, 2 \rightarrow 6, 3 \rightarrow 5, 4 \rightarrow 4, 5 \rightarrow 3, 6 \rightarrow 2, 7 \rightarrow 1$
  • Item 3 (Harmful = 1, Beneficial = 7): Note that the standard scale lists Harmful on the left and Beneficial on the right; however, per McCroskey's source scoring key, items 1, 3, and 4 are flagged for polarity adjustment to align all positive evaluations with the numerical value of 7 (or 1, depending on preferred direction). When scoring toward positivity ($7 = \text{Most Positive}$):
    • Item 1 (Good / Bad): Invert ($7 \rightarrow 1, 1 \rightarrow 7$)
    • Item 2 (Wrong / Right): Retain original ($7 = \text{Right}$, Positive)
    • Item 3 (Harmful / Beneficial): Retain original if Beneficial is scored 7, or reverse code according to the specific layout orientation applied
    • Item 4 (Fair / Unfair): Invert ($1 = \text{Fair} \rightarrow 7, 7 = \text{Unfair} \rightarrow 1$)
    • Item 5 (Wise / Foolish): Invert ($1 = \text{Wise} \rightarrow 7, 7 = \text{Foolish} \rightarrow 1$)
    • Item 6 (Negative / Positive): Retain original ($7 = \text{Positive}$)
  • Universal Rule: Ensure all six items are transformed such that the positive evaluative term equals 7 and the negative term equals 1.
  • Composite Calculation: Sum the six item scores (ranging from 6 to 42) or compute the mean across the six items (ranging from 1.0 to 7.0). Higher values reflect a more favorable, positive attitude toward the evaluated referent.

Permissions & Fee and Test Year

• Initial Inception: 1966 (McCroskey’s doctoral dissertation at Pennsylvania State University)
• Formal Codification: 1989 (McCroskey & Richmond, Measurement of Communication Behavior); 2006 (McCroskey, Communication Quarterly)
• Copyright & Permissions: The Generalized Attitude Measure is placed in the public domain for non-commercial scholarly, educational, and scientific research. Researchers, educators, and clinicians are permitted to use, administer, and adapt the scale without paying licensing fees or securing formal written permission, provided that appropriate scholarly attribution is accorded to Dr. James C. McCroskey in resulting publications and presentations.
• Commercial Use: Commercial application, software redistribution, or monetization should reference established academic protocols and respect scholarly attribution guidelines.

References

• Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. https://doi.org/10.1016/0749-5978(91)90020-T
• Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39(5), 752–766. https://doi.org/10.1037/0022-3514.39.5.752
• Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research. Addison-Wesley.
• McCroskey, J. C. (1966). Experimental studies of the effects of ethos and evidence in persuasive communication (Unpublished doctoral dissertation). Pennsylvania State University.
• McCroskey, J. C. (2006). Reliability and validity of the Generalized Attitude Measure and Generalized Belief Measure. Communication Quarterly, 54(3), 265–274. https://doi.org/10.1080/01463370600877986
• McCroskey, J. C., & Richmond, V. P. (1989). Bipolar scales. In P. Emmert & L. L. Barker (Eds.), Measurement of communication behavior (pp. 154–167). Longman.
• McCroskey, J. C., & Richmond, V. P. (1996). Fundamentals of human communication: An interpersonal perspective. Waveland Press.
• Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. University of Illinois Press.
• Petty, R. E., & Cacioppo, J. T. (1986). The Elaboration Likelihood Model of persuasion. Advances in Experimental Social Psychology, 19, 123–205. https://doi.org/10.1016/S0065-2601(08)60214-2

Items of the Scale