page 1page 2page 3page 4Journal of Cognitive Psychotherapy: AnInternationalQuarterly,Volume 5, Number 2, 1991Assessment of Self-Efficacyin Articulated Thoughts:"States of Mind" Analysisand Association withSpeech-Anxious BehaviorGerald C. DavisonUniversity of Southern CaliforniaDavidA. F.HaagaThe American UniversityJoel Rosenbaum, Sharon L. Dolezal, andKaren A. WeinsteinUniversity of Southern CaliforniaThis study related articulated thoughts during simulated situations (ATSS) tobehavioral observations. Articulated thoughts of undergraduates in supportiveand stressful simulated public speaking situations were coded for indices ofpositive and negative self-efficacy and outcome expectations.Cognitionsvaried as expected across situations. In the stressful, but not the supportive,situation, a "state of mind"ratio relating (positive) to (positive + negative) self-efficacy was inversely correlated with behaviorally rated speech anxiety andself-reported trait speech anxiety. The results suggest that (1) self-efficacy canbe reliably and validly inferred from articulated thoughts; (2) these thoughts aremeaningfully related to overt behavior; (3) social cognitive theory mightfruitfully incorporate the states of mind model; and (4) "challenging"subjectsor clients by such means as imagining stressful situations may be necessary toactivate cognitions associated with negative emotions and dysfunctional behavior.Testing cognitive-behavioral theories requires valid methods for assessing cog-nitions, and further research is needed to determine the conditions under whicheach method is reliable and valid (Clark, 1988). The articulated thoughts duringsimulated situations (ATSS; Davison, Robins, & Johnson, 1983) paradigm offers84Davison et al.an alternative to the most common cognitive assessment format, structuredquestionnaire ("endorsement") methods. ATSS involves audiotaped presentationof hypothetical situations, interrupted at intervals to allow subjects time to thinkaloud. Subjects describecurrentthoughts, which should limit the impact of recallbiases (Genest & Turk, 1981). Moreover, as ATSS does not constrain responses,it is possible to test whether subjectsspontaneouslyreport hypothesized cogni-tions (Haaga, 1989). Endorsement methods, by contrast, might suggest responsesand thereby sometimes yield distorted information (Segal & Shaw, 1988).Finally, ATSS is hypothesized to permit presentation of stressful scenes in asufficiently vivid manner to activate mood-congruent"hot"cognitions (Safran &Greenberg, 1986), which might otherwise be latent and difficult to detect (Parks& Hollon, 1988).Prior studies have found hypothesized relationships between ATSS measuresand situational manipulations (e.g., Davison, Feldman, & Osborn, 1984) as wellas person variables such as social anxiety (Davison & Zighelboim, 1987) anddepression (White, Davison, Haaga, & White, in press).Thus, preliminaryvalidity data are promising. The present study extended earlier research in severalways.First, an important aspect of criterion-related validity not yet studied is theassociation of articulated thoughts with overt behavior. To be sure, correlationsof ATSS categories with clinical diagnoses, themselves made in part on the basisof behavioral observations, support a link between articulated thoughts andaction.The present study, though, related articulated thoughts to behavior(specifically, a behavioral measure of speech anxiety) sampled under morecontrolled conditions.Second, we evaluated ATSS measures of self-efficacy (SE) and outcomeexpectations (OE), central constructs in Bandura's (1977, 1986) social cognitivetheory. Bandura (1986) defines SE as "people's judgments of their capabilitiesto organize and execute courses of action required to attain designated types ofperformances"(p. 391) and OE as"judgment of the likely consequence suchbehavior will produce" (p. 391). SE and OE are typically assessed with endorse-ment measures. If they can also be validly assessed via coding of verbalizationssuch as articulated thoughts, though, the array of data sources suited for researchon social cognitive theory would be expanded. By analogy, validation of theContent Analysis of Verbatim Explanations coding scheme (CAVE; Peterson,Luborsky, & Seligman, 1983) has facilitated the use of speeches, diaries, and thelike in research on explanatory style (e.g., Burns & Seligman, 1989).Third, we integrated the States of Mind (SOM) model (Schwartz & Garamoni,1986) with social cognitive theory in analyzing articulated thoughts.Whereas SEand OE are typically measured on unidimensional scales (e.g., 0—100 ratings), wemeasured separately positive and negative SE (belief that one has/does not havethe means to accomplish a given task), and positive and negative OE (belief thatthe end result of one's actions will be favorable/unfavorable) because positive and© 1991 Springer Publishing Company83Articulated Thoughts and Self-Efficacy85negative thoughts do not always function as simply two sides of the same coin(e.g., Eifert & Lauterbach, 1987). In particular, the SOM model holds thatpsychological well-being is importantly related to the SOM ratio of Positive toPositive + Negative (P/[P+N]) cognitions. The model predicts that optimal func-tioning will be associated with a "positive dialogue"SOM of .618 (optimal rangefrom .56 to .68), which corresponds to the golden section proportion ("the ratiothat obtains between two parts when the smaller part (a) is to the larger part (b) asthe larger part (b) is to the whole (a+b)") (Schwartz & Garamoni, 1989, p. 272).SOM ratios higher than .68 ("positive monologue") are considered maladap-tive, reflecting inattention to possibly important negative events or negativeaspects of the self. An even balance of positive and negative thoughts (SOM ratiosof .45 to .55; "internal dialogue of conflict") is said to be associated with mildpsychopathology, the "negative dialogue"(.32 to .44) with moderate pathology,and the "negative monologue" (0 to .31) with severe pathology. Previous researchsupports hypotheses regarding the three intermediate SOMs (Schwartz & Garamoni;1989), though little information is available regarding positive and negativemonologic SOMs.Prior studies have not applied the SOM construct to SE and OE. It may beuseful, though, for integrating two strands of research in social cognitive theory.First, there is clearly a general positive relationship between SE and adaptivefunctioning. SE is inversely