« 上一頁繼續 »
words. We should therefore expect that the mean intensity of tone in class TG would be greater than in class G and, inasmuch as this tone is positive, that the mean memory value would also be greater.
The former is the case as will be seen later. That the latter is not may, I think, be explained as follows: complex-indicators, other than those here analysed, are rare in my material, but no fewer than nine are to be found in class TG; the mean memory value of these is 5.2 and if they are eliminated from the class the mean memory value of the remainder is 7.4 which raises it to the position of equal first with class G. (Actually, if we take the mean to another place of decimals we have mean for class TG 7.425, mean for class G 7.41.) Whether this alteration is legitimate is a matter of opinion into which subjective factors enter largely. I, personally, think that it is, and I am strengthened in this view by considerations of the relative intensity of the tone in different classes as shown by the mean magnitude of the complex-indicators.
Since the Probable Means of the reaction times and galvanometer deflections vary considerably in different subjects it might be unwise simply to calculate the arithmetic means of the times and deviations in the different classes and to use these as measures of the intensity of affective tone; to do so would involve a danger of the classes being dominated by a few subjects whose probable means are unusually high. I prefer to express each value as a percentage of the corresponding Probable Mean and to use the mean of these percentages as the measure of intensity; I think this plan might profitably be adopted in all similar work which may need to be compared with results obtained by other experimenters. It is equivalent to reducing all subjects to terms of a 'standard subject' whose Probable Mean is 100 units.
In this particular case it makes no difference which method we use. The results obtained by both are shown below:
This shows that class TG is more intensely toned than either of classes T or G, and class TGR than either TR or GR; this applies both to reaction time and galvanometer deflection; which is just what we should expect. The discrepancy between the percentage of the Probable Mean and the Arithmetic Mean in the case of the reaction time in classes T and TR is negligible.
It is possible to apply a further check to the results. If the words learned form a reasonably representative sample of the whole of the material available there ought to be some degree of correspondence between the mean values of the reaction time and the galvanometer deflection, in the various classes, when calculated from all the reactions, and the values yielded by the learned words only. The correspondence will not be quantitatively exact because in the case of my first 25 subjects, of whom 15 are included in the 18 here concerned, I selected to be learned the 15 words giving the largest galvanometer deflection and the 15 giving the smallest. This was done with the idea of giving affective tone the best possible chance of exhibiting any influence on memory which it might have, and this circumstance affects the quantitative relations between the 'sample' and the remainder of the material in a somewhat complicated way.
The comparatively large number of data here available makes it unnecessary to use the percentage method. The values of the arithmetic means for the whole material are:
The correspondence between these values and those given in the preceding table is obviously very close.
I have entered into these details because I want to show how very concordant the results are and how those obtained by one method of treating the data harmonise with those obtained by another.
When we remember how rough a test of the quality of affective tone the 'memory value' of a word must necessarily be in practice, and how
many accidental causes may distort and obscure its indications, it will
In the 1741 reactions given by the 18 subjects dealt with above there are 460 disturbances in reproduction; this is equal to 26-9 % as compared with Jung's "33 % not reproduced1." It is not clear whether this last figure includes associations reproduced with great hesitation, presumably it does. I attribute the difference between these two values to the fact that among the 28 subjects dealt with by Jung there were 25 nervous and mental patients of different kinds, whereas all my subjects were normal.
Jung found that "on the avarage, 62.2 % of the absent reproductions lie, as regards the reaction-times, above the probable mean"; I find only 49.2 %. This difference is probably due to the same cause. Abnormal subjects will, in general, possess more numerous and stronger complexes than my normal subjects, and the more intense tone aroused by the complex-striking stimulus words-indicated in both cases by the disturbance of reproduction-will tend to prolong the reaction time more frequently in the case of the abnormal subjects.
In view of the evidence I have brought forward above which shows that disturbance in reproduction is more intimately associated with negative affective tone than are either of the other two complex-indicators discussed, I do not think it is necessary to reproduce from my data the figures analogous to those which Jung gives in favour of regarding this phenomenon as significant. It may be pointed out, however, that even among his so largely abnormal subjects, it is probable that a certain number of reaction times were prolonged on account of positive affective tone aroused by the stimulus word, or on account of intellectual difficulties. If Jung had been able to distinguish between such prolongations and those due to negative tone his figures would, presumably, have borne out his contention even more strongly than they did.
There can be no doubt whatever that for quantitative work the galvanometer deflection is a far more valuable indicator than the reaction time. It is not under voluntary control and is not affected to any appreciable extent by non-significant intellectual factors such as sometimes prolong reaction time. Moreover, the absolute magnitude of
1 Loc. cit. p. 401.
the deflections can, in general, be magnified to any extent desired and
Still more important is the fact that the magnitude of the galvano-
4 Divorce 50-8 24 Head 31.7
time ceases to be proportionally significant. No one would suggest, for example, that a time of one minute, say, in a series whose Probable Mean is two seconds, is likely to be the result of an affective state 15 times as intense as that responsible for a time of four seconds. But such considerations cannot be extended to the galvanometer deflections. Table VI shows the 100 words of my list arranged in the order of magnitude of their mean reaction times, calculated for the whole of the 50 subjects examined; Table VII shows the words similarly arranged on a basis of their mean galvanometer deflections.
There can be no doubt that the order of words given by the galvanometer represents their relative affective value far more accurately than that given by the reaction time.
The following points may be noted:
(i) The highest value in the galvanometer series is 5.12 times as great as the lowest; in the time series it is only 1.98 times as great. The 'resolving power' of the galvanometer is, therefore, rather more than 24 times that of the reaction time.
(ii) In accordance with this we find in the reaction time series seven pairs of words whose mean time is the same, seven such groups of three words each, five of four words each, and two of seven words each. In the galvanometer series there are only eight such pairs and one group of three.
The galvanometer therefore differentiates gradations of affective tone with much greater delicacy than does the reaction time. (iii) The first six words on the galvanometer list are Kiss, Love, Marry, Divorce, Name, Woman. Of these, five are obviously closely connected with sex-life and the other, Name, is probably constellated by the same ideas. These six words stand out head and shoulders above the remainder of the series, as I pointed out in my paper on memory. (N.B. The effect is very noticeable if the series is represented graphically.) Their mean value is 145% of that of the seventh word and 220 % of that of the Probable Mean of the series.
Compare with these the first six words of the time series, Name1, Friend, Despise, Make, Sad, Proud. This is not nearly so homogeneous a group; its mean value is only 110% of the seventh word and only 141 % of the Probable Mean of the series.
1 For the probable reason of the very long time for this word, compare page 254. The first name to occur is likely to be that of a wife, fiancée, lover or other person of sexual significance to the subject.