Social Science Computer Review

This article uses a computerized method to score news coverage of the 1995 referendum in Quebec, Canada, asking whether that province should become sovereign. The news for English-speaking Canada was represented by the Canadian Press (CP) newswire, whereas la Presse canadienne (PC), the sister news service, was used as the surrogate for the French language press in Quebec. The CP was analyzed for paragraphs supporting the Yes and the No positions, and the PC was scored for the equivalent positions of the Oui and the Non. Both news services were also analyzed for the major campaign themes and coverage of the four major political leaders. The Oui and the Non paragraphs in the PC were good predictors of the time trend of opinion polls leading up to the referendum. The campaign could be separated into four distinct phases based on the prominence of coverage of different political leaders.

This article outlines a rigorous method for studying the ways that news media frame contentious issues. The method is based on the VBPro family of computer programs for content analysis. Output from the VBPro mapping program for multidimensional scaling based on co-occurence of key terms is cluster analyzed to discern the frames or points of view in texts that can unambiguously be attributed to competing stakeholders. These frames can be used to investigate propositions about news stories. An example is presented from a study of 1,465 Associated Press articles on wetlands dispatched across an 11-year period beginning in 1982. The results demonstrate that the method provides an objective means of investigating stakeholder influence on news and patterns of change in frames across time.

This article presents a computer content analysis approach to the problem of assessing the validity of Likert scales used in survey research. The wording of questions that make up a scale communicates meaning to the respondent in order to acquire relevant responses that, when combined, serve to measure some concept. Traditionally, beyond informal judgments of face validity, investigators examine responses of a sample of subjects to scale questions in order to determine if expected relationships and structure are found. The content analysis procedure proposed here directly examines the wording of scale questions to see if expected meanings, relationships, and structure are found in order to assess the validity of a scale. Possibilities of this approach are illustrated using two previously analyzed sets of scales: alienation and perception of police and crime.

Category systems extracted from textual material are an important part of the scientific process and qualitative analysis but are frequently developed on an ad-hoc basis. Their development can be improved by a stronger reliance on linguistic and semantic principles. Tracing and recapitulating the development of semantic principles from the 1950s to the 1990s in linguistics and text analysis show how the principled use of information from lexical resources can facilitate the development and analysis of category systems. These principles are demonstrated in examining the category systems of Minnesota Contextual Content Analysis, a technique for analyzing textual material from sentences, answers to open-ended questions on questionnaires, expository texts, and verbatim transcripts. These principles are then extended to show how to abstract category assignments for this type of textual material. The discussion identifies computerized systems and data resources used in the application of these principles.