Response productivity: A methodological problem in content analysis studies in psychotherapy.

Abstract

Content analysts of psychotherapy materials have, for the most part, failed to consider the possibility that the frequency of units coded to the categories of the content analysis system might be correlated with the total number of units produced. The existence of such relationships may confound interpretation of the association between content analysis variables and other external variables. When investigators have attempted to control for response productivity, they have usually divided the frequency of units in each category by the total number of units. This procedure does not control for response productivity and is often uninterpretable. The issues involved are discussed and demonstrated, and appropriate techniques of controlling for productivity are described.

Response productivity in verbal content analysis: A critique of Marsden, Kalter, and Ericson.

“…We find particularly objectionable the casual treatment given by Marsden et al (1974) to the development of interpret able measures of a construct. Their only concern is for the specific case in.…”

“…In almost every sample, final measures had low, nonsignificant correlations with total words (Gottschalk & Gleser, 1969, p. 18). Judge of our surprise, then, when Marsden et al (1974) cited the Gottschalk and Glcser measures as typical of the unreflecting erroneous use of percentage scores. Correspondence with Marsden et al has yielded no change in their position.…”

“…The reason they give for choosing T-X to control for productivity is to avoid the "spurious inflation of the correlation" between X and T resulting because X is contained in T. Their concern about the part-whole problem seems to stem from a concept of productivity as a secondary or derived variable rather than a basic variable in its own right, independent of the particular content under consideration. That Marsden et al (1974) see the content score as primary is evident from their statement that "X need not constitute a sizable portion of T to substantially affect the correlation between T and variables with which X is correlated" (p. 226, italics ours). But numerous studies (e.g., Matarazzo, Wiens, & Saslow, 1966) attest to the fact that verbal productivity is an important variable in its own right, usually simpler and more basic than the specific content that is being investigated.…”

Vocal interruptions in dyadic communication as a function of speech and social anxiety.

“…In this light, the stepwise regression of an individual's personality measures to his or her interruptive behavior was performed using the individual's back-channel responses, the speaking time and personality measures of the conversational partner, and the individual's and the dyad's (male, female, mixed) sex as covariates. Marsden, Kalter, and Ericson (1974) have pointed out that this analysis can easily be achieved by forc-ing the order of the predictors. Stepwise multiple regression works by adding one variable at a time to the regression while assessing to determine whether the independent variables significantly increase the prediction of the dependent variable.…”

Evaluation of Brief Intervention Models By the Hill Interaction Matrix

“…Although this methodology is widely advocated and employed, recent work (Marsden, Kalter, and Ericson, 1974) raises questions about the adequacy of this approach. Such statistical critiques point to the necessity of developing improved scoring protocols which generate comparable frequencies.…”