Session Summary Understanding Consumer Decision Processes Using Verbalization Data: Substantive and Methodological Perspectives

Gabriel Biehal, University of Maryland
Dipankar Chakravarti, University of Arizona
[ to cite ]:
Gabriel Biehal and Dipankar Chakravarti (1995) ,"Session Summary Understanding Consumer Decision Processes Using Verbalization Data: Substantive and Methodological Perspectives", in NA - Advances in Consumer Research Volume 22, eds. Frank R. Kardes and Mita Sujan, Provo, UT : Association for Consumer Research, Pages: 269-270.

Advances in Consumer Research Volume 22, 1995      Pages 269-270



Gabriel Biehal, University of Maryland

Dipankar Chakravarti, University of Arizona

Thought verbalization data are used extensively in consumer decision process research (Bettman and Park 1980; Bettman, Johnson and Payne 1990; Biehal and Chakravarti 1986) and researchers have developed various methods for collecting such data. Many researchers use retrospective protocols which require that subjects recount a prior decision episode verbally or in writing. Others ask subjects for introspective reports, i.e., verbalized thoughts about the reasons for current/past decisions. Still others use concurrent verbal protocols (verbalization simultaneous with task performance).

There is controversy about the validity of verbal protocol data. Since verbalization places demands on subjects' mental capabilities, it may distort underlying processes. Some psychologists (Nisbett and Wilson 1977) dismiss retrospective protocols as introspectional data of limited value. Others (e.g., Ericsson and Simon 1984) provide empirically-based rebuttals, mainly regarding the validity of concurrent protocols. Recent research shows that even concurrent verbalizing may subtly influence consumer decision processes (Biehal and Chakravarti 1989; Russo, Johnson and Stephens 1989). Hence, the strengths and weaknesses of various thought verbalization methodologies deserved more examination.

Session Overview

This session included three presentations that used different verbalization methods and contexts to study substantive consumer decision making issues. The first (Shah, Gilbert and Park) explored how introspection and concurrent verbalization may not only influence framing, processing and memory representations in the focal task, but also in subsequent tasks. In the second, Bhoovaraghava and Mitchell showed how retrospective and concurrent protocols along with process tracing data can provide converging evidence on base rate information use in consumer decisions. Kuusela, Ahtola and Chakravarti examined how self-perceived and actual knowledge influence decision framing and processing as well as choice times and outcomes. They showed how concurrent verbalization data can link task performance to process variables, noting that concurrent verbalization may create differential demand characteristics for low versus high knowledge (both actual and self-perceived) subjects. The discussant, Eric Johnson, provided a summary of the strengths and limitations of verbalization methods.

The Shah, Gilbert and Park Presentation

Reshma Shah, Robert Gilbert and C.W. Park (Pittsburgh) presented, "The Impact of Concurrent Verbal Protocols on Encoding and Retrieval." They reiterated the controversy about the validity of protocol data, noting recent research showing (a) that concurrent verbalization can interfere with information processing, memory and problem solving because it draws on limited attentional capacity (Biehal and Chakravarti 1989; Schooler and Schooler 1990; Schooler, Ohlsson and Brooks 1994); and (b) that introspection (through verbalization) can lower the quality of decisions and attenuate satisfaction with choices (Wilson and Schooler 1991; Wilson et al 1994). They also noted that verbal overshadowing (verbal disruption of non-verbal processes) may impair problem solving and learning (Fallshore and Schooler 1993).

The authors pointed out that whether concurrent verbal protocols facilitate or inhibit information processing may depend upon the task. Verbalization may induce more analytical processing relative to nonverbalizing conditions in which consumers may process more holistically. Due to greater analytical processing, verbalizing subjects should find technical, functional and utilitarian product attributes more salient and important in choice. Moreover, analytical processing should lower recall and recognition and also reduce liking and satisfaction with the task and the choice. In contrast, the absence of verbalization should generate more holistic processing. Hence, such subjects should deem artistic, aesthetic and luxurious product attributes as more salient and important. Moreover, recall and recognition should be facilitated and these subjects should exhibit greater liking and satisfaction with their choices and the task. Verbalization should also facilitate problem framing strategies and thus influence encoding/retrieval processes in subsequent tasks.

The empirical study considered the impact of concurrent verbalization instructions on consumer decision making in two sequential choice tasks. In Task 1, subjects made choices from among four homes, based upon pictures and descriptions of their location/surroundings and other features. Some subjects verbalized their choice process, whereas others did not. Following choice, measures were taken of recall, recognition, liking, satisfaction, attribute importance, and processing style. In Task 2, subjects once again chose from among a set of homes. Measures were taken to investigate the differential impact of verbalization on problem framing and subsequent decision making. The authors discussed the implications of the findings for the use of concurrent verbalization data in the study of consumer decisions.

The Bhoovaraghava and Mitchell Presentation

This presentation by Sriraman Bhoovaragava and Deborah Mitchell (Temple University) was entitled, "Point-of-view and Related Factors Affecting the Use of Base Rates: Verbalization Data as Converging Evidence on Decision Processes." The authors noted previous research showing that utilization of base rates and other information in decision-making is dependent both on the perceived relevance/diagnosticity of the information (Feldman and Lynch 1988) and on whether or not it conflicts with other environmental cues (Lynch and Ofir 1989). They then reported three studies examining how contextual variables and reference-dependence influence individuals' use of base rates in various decisions. A variable termed "point-of-view" (Kahneman and Lovallo 1993), as well as the reference point implicit in the valence of the potential decision outcome, directly affected base-rate utilization and predictive performance. Interestingly, decision makers' confidence in their predictions was negatively correlated with base-rate utilization and performance.

The three experiments manipulated 'point-of-view' (inside versus outside), outcome valence, and information consistency (Lynch and Ofir 1989). Consumers were asked to predict the likely outcome of service encounters, based on the information provided to them. Process data was collected in each study to investigate the cognitive processes underlying consumers' decision-making. In Study 1, written retrospective protocols were collected and analyzed. In Study 2, consumers made decisions based on information acquired from a computer-based display (Mouselab). Finally, in Study 3 a small sample of consumers provided concurrent verbal protocols as part of their decision task.

All three studies demonstrated significant effects of 'point-of-view' as well as outcome valence on predictive performance. The findings replicated those of Lynch and Ofir (1989) regarding information consistency and base-rate utilization. Convergent process data collected in Studies 1-3 indicated the specific manner in which base rates were utilized. The presentation emphasized how different types of process data can be used for investigating biases in decision-making and the importance of convergent process evidence for a more complete understanding of consumer decision processes.

The Kuusela, Ahtola and Chakravarti Presentation

Hannu Kuusela (Tampere), Olli Ahtola (Helsinki School of Economics) and Dipankar Chakravarti (Arizona) presented, "The Effects of Self-Perceived and Actual Knowledge on Choice Processing: A Verbal Protocol Analysis. They began by noting prior research showing that the actual (objective) consumption relevant knowledge that consumers possess influences decision processes and outcomes (e.g., Bettman and Park 1980; Brucks 1985; Johnson and Russo 1984). They also noted the importance of self-perceived (subjective) knowledge (Park, Gardner and Thukral 1988) and argued that the effects may be motivational (influence task interest) or cognitive (influence information acquisition/use). Since self-perceived knowledge reflects a consumer's prior assessment of the decision task, it may influence problem-framing and guide processing. However, the correspondence between self-perceived and actual knowledge is often less than perfect and consumers must deal with unanticipated contingencies during choice. Consequently, the levels of, and correspondence between, self-perceived and actual knowledge should influence choice processes and outcomes.

Ecological correlations between self-perceived and actual knowledge make it difficult to study their separate and interactive effects (Brucks 1985). The authors reported a 2 x 2 randomized block experiment, manipulating the level of self-perceived knowledge (high/low) and blocking on actual knowledge (high/low). Sixty-four business student subjects from two Finnish universities chose among multi-attribute descriptions of alternative homeowners' insurance policies. The students varied in their actual knowledge of homeowners' policies and were blocked on this factor. Self-perceived knowledge was manipulated by asking subjects to take multiple choice tests that varied in difficulty and in the feedback provided. Verbal protocols and other relevant measures were collected during choice. The protocols were coded for evidence of problem framing, use of processing operations of different complexity (e.g., compensatory, pairwise comparison, holistic) and decision time.

The findings show that higher levels of both self-perceived and actual knowledge reduce explicit problem framing and lower both the number of elementary processing operations and decision time. Also, problem framing and elementary processing activities are highest when both self-perceived and actual knowledge are low. Decisions are made fastest when both self-perceived and actual knowledge are high. The protocol data in this study provide both process and outcome evidence regarding the effects of self-perceived and actual knowledge on choice. However, the collection and analysis of such data is not straightforward and the insights obtained vary by the level (macro or micro) of protocol coding and analysis (Biehal and Chakravarti 1986, 1989). The authors suggested useful rules of thumb for choosing coding schemes and the level of aggregation of code categories.

Discussant's Comments

Eric Johnson summarized the papers with an integrative analysis of the strengths and limitations of verbalization data, comparing such data with data from other process tracing methods such as Mouselab.