The del technique

The del technique (Drazin & Kazanjian, 1993) was used to examine Hypothesis 4 for the two levels of goal-performance discrepancy, minor and major, and the three levels of cognitive structures. The del technique permits an analysis of the fit of research data with an a priori prediction of the distribution of frequencies of a cross-classification matrix. Matrix cells for which a researcher expects frequencies to he 0, or error cells, are assigned a weight of 1. Cells for which the frequencies are expected to be greater than 0—predicted cells—are assigned weights of 0. Weights between 0 and 1 are used for indicating a researcher’s degree of confidence that the frequency of a cell is greater than 0. Hypothesis 4 predicted that the minor goal- performance discrepancy row of the two-hy-three matrix would have an empty cell for strategies (W :̂, = 1), and the frequency for new script tracks would be slightly greater than 0 [W^z = -5)- The cell for the previously accessed script tracks of this row and each of the three cells for the major goal-performatice discrepaticies row were expected to have frequencies greater than 0 (W^. W^j, W22, and W^^ = 0). Using these weights, the del technique results support Hypothesis 4 for hoth frequently occurring job prohlems [V = .62. Z – 2.22, n = 72. p < .05; x̂ ^ 6.35, df = 2. p < .05) and rarely occurring prohlems [V = .65, Z = 3.28, n = 71, p < .01; x^ = 6.35, df = 2, p < .05). x̂ values are provided for comparison with V values.

Chi-square tests were used to examine the prediction of Hypothesis 5 that the most accessible script track, the one each subject identified first, is the most likely to he identified as the track that the employee would actually use to handle the joh problem. The chi-square analysis provided strong support for this hypothesis for both the frequently occurring job problems (X̂ = 23.03, df = 1, p < .001) and the rarely occurring job problems (x^ = 23.84, d/ = 1, p < .001). Even though employees mentioned up to seven tracks, 76 percent of the tracks that were the most accessible in their mem- ories were also the ones most likely to he used.

Cognitive complexity is litiearly correlated with the number of script tracks for both the frequently occurring job prohlems [r̂ ^ = .19. p < .05) and the rarely occurring job prohlems [r-^^ = .23, p < .05), supporting Hypo- thesis 6. Promotability is also correlated with the number of script tracks for rarely occurring joh prohlems (rĝ = .43, p < .01), but not for frequently occurring joh problems [rga = .01. p > .05); thus. Hypothesis 7 is partially

1994 Wofford 189