1 Department of Marketing and Statistics, Aarhus School of Business, Aarhus BSS, Aarhus University2 MAPP - Centre for Research on Customer Relations in the Food Sector, Aarhus School of Business, Aarhus BSS, Aarhus University3 Department of Economics and Business Economics, Aarhus BSS, Aarhus University4 Department of Management, Aarhus BSS, Aarhus University5 Department of Economics and Business Economics, Aarhus BSS, Aarhus University6 Department of Management, Aarhus BSS, Aarhus University
Despite its popularity in consumer research, means-end chain theory suffers from problems of unconfirmed validity. In particular, it is unknown if its central construct, the means-end chain (MEC), is an identifiable memory structure or not. Theoretically, means-end chains can be cast as associative networks with a three-layered structure, consisting of attributes (A), consequences (C) and values (V) that are hierarchically linked. This yields two predictions when operationalized in a lexical decision task: means-end chains should display spreading activation (direct as well as mediated priming) and automatic activation (persistence of priming effects under short inter-stimulus intervals and high proportions of non-words). The predictions were tested in two experiments. Both experiments used a single-presentation lexical decision task with eleven classes of stimuli: AC, CV and AV pairs collected from the respective participant by means of a laddering interview in a pilot session (on average, 85 days prior to the experiment); AC, CV and AV pairs from a laddering interview with a different participant (checked for non-overlap); word pairs with direct, mediated and no associations from a standardized word list; and finally, non-words and fillers serving as distractors. In Experiment 1, long inter-stimulus intervals (750 msec) and a low proportion of non-words (25%) were used. 44 undergraduates participated. Significant priming effects were observed for AC and CV pairs, but not for AV pairs. Surprisingly, the priming effects did not differ depending on whether the word pairs had been taken from the participant's own laddering interview, or a different participant's interview. In Experiment 2, short inter-stimulus intervals (250 msec) and a high proportion of non-words (40%) were used. 46 undergraduates participated. Significant priming effects were observed for AC, CV and AV pairs. Priming effects for CV and AV pairs were larger than for AC pairs. Again, the size of the priming effects did not differ depending on whether the word pairs had been taken from the participant's own laddering interview, or a different participant's interview. Largely, the pattern of results was in line with the predictions. In Experiment 2, direct priming (AC and CV pairs) as well as mediated priming (AV pairs) could be observed, consistent with the spreading-activation assumption. Furthermore, results in Experiment 2 were obtained under conditions designed to suppress strategic processing, suggesting that the automaticity assumption may hold as well. However, some qualifications have to be made. The absence of mediated priming effects (AV pairs) in Experiment 1 but not in Experiment 2 suggests that post-lexical, strategic processes may have overridden a priming effect. This is a phenomenon not uncommon in the literature, especially when the number of unassociated words in a word list is so low that participants are led to expect a link between adjacent words. More surprising is the uniformity of priming effects with regard to the source of the stimuli, i.e. whether they were taken from a participant's own laddering interview or other participants' laddering interviews. At this point, the only explanation appears to be that the associations were relatively uniform across subjects.