Abstract
Researchers commonly conceptualize forgiveness as a rich complex of psychological changes involving attitudes, emotions, and behaviors. Psychometric work with the measures developed to capture this conceptual richness, however, often points to a simpler picture of the psychological dimensions in which forgiveness takes place. In an effort to better unite forgiveness theory and measurement, we evaluate several psychometric models for common measures of forgiveness. In doing so, we study people from the United States and Japan to understand forgiveness in both nonclose and close relationships. In addition, we assess the predictive utility of these models for several behavioral outcomes that traditionally have been linked to forgiveness motives. Finally, we use the methods of item response theory, which place person abilities and item responses on the same metric and, thus, help us draw psychological inferences from the ordering of item difficulties. Our results highlight models based on correlated factors models and bifactor (S-1) models. The bifactor (S-1) model evinced particular utility: Its general factor consistently predicts variation in relevant criterion measures, including 4 different experimental economic games (when played with a transgressor), and also suffuses a second self-report measure of forgiveness. Moreover, the general factor of the bifactor (S-1) model identifies a single psychological dimension that runs from hostility to friendliness while also pointing to other sources of variance that may be conceived of as method factors. Taken together, these results suggest that forgiveness can be usefully conceptualized as prosocial change along a single attitudinal continuum that ranges from hostility to friendliness.
Introduction
For decades, social psychology has treated forgiveness as a "rich complex" of psychological shifts. When we forgive, we are thought to undergo a multi-faceted transformation involving our emotions, cognitions, and motivations. This theoretical complexity is mirrored in popular measurement tools like the Transgression-Related Interpersonal Motivations (TRIM) inventory, which segments forgiveness into distinct subscales: Revenge (the desire to harm), Avoidance (the desire to distance oneself), and Benevolence (the desire for goodwill).
However, a persistent gap exists between this multi-dimensional theory and the resulting data. In practice, these subscales are often so highly correlated that they appear to measure the same underlying construct. This paper, published in the Journal of Personality and Social Psychology, argues for a "leaner" model. By combining advanced psychometrics with experimental economic games, the authors demonstrate that forgiveness is best understood as a single attitudinal continuum ranging from hostility to friendliness.
Methodology: A Multi-Method Approach
To test whether forgiveness is truly multi-dimensional or a single continuum, the researchers conducted five studies involving approximately 1,750 participants across the United States and Japan. The research design integrated three sophisticated methodologies:
1. Bifactor (S-1) Modeling
Traditional models treat subscales as separate-but-related entities. In contrast, the bifactor (S-1) model extracts a "General Factor"—the core psychological essence shared across all items—while treating the remaining variance in specific subscales (like Revenge or Benevolence) as "specific factors" or method effects. This allowed the team to see if the core "General Factor" was the primary driver of behavior.
2. Experimental Economics
To move beyond the limitations of self-report surveys, the authors utilized staged transgressions. Participants were defected against in a Prisoner’s Dilemma Game (PDG) and then given opportunities to interact with the transgressor in various behavioral "criterion measures," including the Trust Game, Dictator Game, and Pay-to-Punish Game.
3. Item Response Theory (IRT)
The researchers applied IRT to calculate "Location Indices," which map the psychological "difficulty" of different items. This helped determine if renouncing revenge, reducing avoidance, and endorsing benevolence are distinct categories or simply different points on the same sliding scale.
Key Results: The Dominance of the General Factor
The findings consistently supported a uni-dimensional view of forgiveness. The bifactor (S-1) model showed superior fit (CFI/TLI > 0.95) across all samples. Crucially, the General Factor was the only consistent predictor of actual prosocial behavior. In Study 1, for example, the General Factor significantly predicted lower propensities to defect ($b = -0.493$) and higher transfers in the Trust Game ($b = 0.293$) and Dictator Game ($b = 0.350$).
The Forgiveness Ladder
IRT analysis revealed a consistent hierarchy of difficulty, which the authors liken to a ladder. It is psychologically "easiest" to renounce revenge, "harder" to reduce avoidance, and "hardest" to endorse benevolence. Because these items fall on the same metric, they represent different intensities of the same attitude rather than different psychological motivations.
Cross-Cultural and Relational Invariance
One of the study's most striking findings was the consistency of this structure across cultures. Whether in the individualistic United States or the collectivistic Japan, the single-continuum model held firm. Furthermore, while avoidance and benevolence become almost indistinguishable in close relationships compared to interactions with strangers, the underlying General Factor remained the primary predictor of forgiveness outcomes.
Discussion and Implications
By conceptualizing forgiveness as a simple attitude change along a hostility-to-friendliness continuum, researchers can tap into decades of established social psychology literature regarding attitude formation and modification. This "leaner" model suggests that for most practical and research purposes, a simple mean score of forgiveness items is a highly effective proxy for the complex latent construct ($r = 0.94$ to $0.98$ correlation with the General Factor).
While the "specific factors" (the variance left over in subscales like Revenge) occasionally predicted behavior, they were inconsistent across studies. This suggests they may often be "method effects"—artifacts of how questions are worded—rather than meaningful psychological nuances. Future work should continue to explore these nuances, particularly in the context of long-term relationship dynamics where the distinction between avoidance and benevolence may blur further.
Summary updated