assumptions is reasonable, in the light of the other proportions, Estes considers the empirical results incompatible with the predictions from the incremental model. The data appear to favor an all-or-none model supplemented by an assumption of imperfect retention.
Let us suppose that the assumptions underlying the analysis of joint probabilities are valid. It then becomes important to determine over what range of conditions the empirical facts fail to meet the requirements of the incremental model. The predicted equality of the proportions p10 and p01 also fails to hold in the studies reported in Estes' earlier papers ( Estes, 1960; Estes, Hopkins, and Crothers, 1960). In the remaining columns of Table 8-5 we present the appropriate proportions obtained in our recent experiments. There are eight independent determinations of these proportions--four groups in Experiment I, two groups in Experiment II, and two groups in Miss Schwartz's experiment. In each of these eight cases the observed proportions p10 and p01, are very nearly equal or differ only by a very small amount. This uniformity is especially noteworthy since different degrees of learning and different experimental procedures are represented. Thus, the requirements of the incremental model are clearly met by our data.
It is not possible at this time to offer a satisfactory explanation of the discrepancy between Estes' findings and our own. From what has been reported, it is clear that there are important differences in materials and procedures. The conditions of intra-list and interlist interference deserve especial attention since Estes used counterbalanced designs in some of his studies and a continually changing list in the most recently reported experiment. It is clear that the empirical inadequacy of the incremental model is still far from being a generally established fact. The discrepancies in experimental findings must be resolved before a final decision about the model is made.
Examination of the available evidence has failed to convince us that the incremental conception of associative strength has been challenged decisively. The results of Rock's studies have been shown to suffer from a fatal methodological flaw and can therefore be considered as inconclusive. There are compelling grounds for questioning both the generality and the interpretation of Estes' findings. Our own studies using the RTT paradigm give consistent support to the incremental interpretation.
The methodological obstacles to a clear-cut test between the incremental and all-or-none positions remain formidable. A persistent difficulty appears to stem from the problem of item selection which has intruded itself into all the current experimental procedures aimed at a decision between the opposing points of view. This fact is not surprising since experimental analysis has focused on the question of whether learning