duction' I include (i) complete failure to remember the original reaction word, (ii) substitution of a different word, and (iii) prolonged hesitation in giving the reproduction.)

I next divided the 518 words learned by the 18 subjects into these eight classes and calculated the mean memory value1 for each class. The results are given in Table I.

Mean memory value for all reactions analysed: 6·70.

It must be remembered that the memory value is only a rough test of whether the affective tone evoked by a given word is positive or negative; there is a marked tendency for negatively toned words to drop out early and consequently to show a low memory value, and conversely; but there are innumerable fortuitous and external causes which may interfere with this and cause a word to be remembered or forgotten for reasons quite other than its intrinsic merits. In spite of this the main indications of the table are quite unmistakable and distinctly striking.

First I would call attention to the fact that the one complex-indicator whose presence is uniformly unfavourable to memory, i.e. which uniformly indicates negative tone, is disturbance in the reproduction test. The four classes in which this indicator figures are the last four on the list as regards memory value.

If we treat these results somewhat after the fashion of a team-race, giving one mark for presence in the class occupying the first position, two for the second, and so on, the indicator getting the most marks will be that which is most closely associated with the presence of the variety of affective tone which tends to impede memory, with negative tone to wit, and conversely. The marks thus gained are:

1 The 'memory value' may range from 0, for words never remembered, to 10 for words remembered without difficulty on each of five occasions. Cf. loc. cit. pp. 241–242.

I conclude therefore that this last phenomenon is not only a complexindicator and even this has been questioned1-but the complexindicator par excellence2.

The appearance of class R (disturbances in reproduction only) at the bottom of the list requires some explanation; one would expect this position to be occupied by class TGR on the ground that the more indicators are present the more intense the affective tone is likely to be. The discrepancy may be due to the large influence on this rather small class of one subject (No. 18) whose learned words included seven characterised by disturbance of reproduction only and scored a total memory value of only 24. This may represent some abnormality on the part of this subject whom, indeed, I rather suspect on other grounds. If these reactions are eliminated from the class its mean memory value rises to 6.3 and the order then becomes:

I, personally, regard this order as more correct than the first, but this is a detail of small importance.

The next point to be noted is that class T shows precisely the same mean memory value as class O. That is to say: Prolongation of reaction time alone is not necessarily a complex-indicator; it is only significant if accompanied by other indicators. This is not at all contrary to accepted views; it is commonly recognised that reaction time may sometimes be prolonged on account of purely 'intellectual' difficulties, arising from the rarity of the stimulus word, etc., without the prolongation being due to a complex.

1 Cf. Jung, loc. cit. p. 396.

2 Using complex' in its common pathological sense and not as synonymous with 'constellation.'

I suspect, however, that this matter is not quite so simple as it might appear at first sight. It will be noticed that the mean memory value of the class TG is markedly above that for class O or for all classes; classes containing 'T' have a memory value below this only when they also contain 'R.' The obvious conclusion is that prolongation of reaction time is a sign of negative affective tone, i.e. a complex-indicator, only when accompanied by disturbance in reproduction. I do not think that this conclusion is sufficiently in conformity with general experience of reaction time as a complex-indicator to pass unchallenged, even if we remember, as we should, that we are here dealing with general tendencies rather than with rigid rules. No one would suggest, of course, that every prolongation of reaction time, however small, is necessarily a complexindicator, for it is universally recognised that only the more salient prolongations are significant. But on the other hand I doubt whether any psycho-therapist accustomed to work with the association test would be willing to admit that all cases in which a significantly too-long time is not accompanied by a disturbance in reproduction are to be regarded as accidental lapses from a general rule. That there is a strong tendency for significant prolongations of reaction time to be accompanied by disturbances in reproduction has, it is true, been shown by Jung1; but it should be conceded, in my judgment, that prolongation of reaction time alone may on occasion be a true complex-indicator, quite apart from the presence, or merely accidental absence, of disturbance in reproduction. If this be correct we should expect to find the mean memory value of class T somewhat below that for class O and it is necessary to account for the fact that it is not.

I think the explanation is that the class T really consists of three sub-classes, namely:

(i) Genuinely 'indifferent' words evoking no appreciable affective tone either positive or negative; these would fall in class O were it not for the fact that their reaction time is prolonged for reasons of intellectual difficulty and the like. Their mean memory value would be 6-7.

(ii) Negatively toned words whose prolonged reaction time is significant, possibly accompanied by some of the miscellaneous complex-indicators enumerated on page 247. If these could be separated out from the remainder their mean memory value would presumably be less than 6.7.

1 Loc. cit. pp. 396, 899.

(iii) Positively toned words, which through lack of intensity or for other reasons do not produce a too-large galvanometer deflection, but whose reaction time is delayed for the same 'intellectual' reasons as are operative in sub-class (i). These words will have a mean memory value greater than that of the indifferent words and will thus counteract the effect of the words in sub-class (ii). In spite of this I am strongly of opinion that the statement at the beginning of this section is in general true, that prolongation of reaction time is not likely to be significant unless accompanied by other indicators, and that the proportion of words belonging to sub-classes (ii) and (iii) is small.

It is also probable that some positively toned words may be accompanied by too-long times for the following reasons:

(a) When a word which evokes markedly agreeable associations, which will as a rule be positively toned, is called out to a subject it seems very possible that his reaction time may be prolonged simply on account of the number of equally acceptable images which crowd in upon him; he suffers, in fact, from an embarras de richesses.

(b) In such circumstances there will also be a tendency for his attention to be diverted from the experiment and to dwell on the pleasing ideas conjured up; this momentary inattention may prolong the reaction time.

(c) The subject may not wish to reply with the first word which occurs to him although it may be intensely positive to him and in no way connected with a complex. For example: the stimulus word 'woman' would be very likely to evoke the image of the subject's fiancée, an image which we may suppose to be accompanied by strong and definitely positive affective tone. The first word to occur to him would naturally be her name; but he might not care to give this as a reaction word in the presence of the experimenter. This would delay the reaction time in spite of the positive tone accompanying the word but it would be ridiculous to suggest that such a prolongation of the time should be considered as a complex-indicator.

This agrees with the form of the curve which I obtained in the course of my memory experiments connecting reaction time with memory value. I found that the mean reaction time for the words least well remembered was greater than that for words better remembered and that there was

a slight tendency for the time to be prolonged in the case of the bestremembered (i.e. most positively toned) words.

In view of the foregoing considerations we may regard the position of class T as quite natural.

Perhaps the most important feature of these results is the position of class G at the head of the list. It is closely followed by class TG, and the fact that each of these classes has a mean memory value handsomely in excess of that for class O (no complex-indicators) proves that they consist mainly of positively toned words. This amply confirms my view that the psycho-galvanic reflex detects and measures positive affective tone as well as negative, and shows further that it does so in circumstances those prevailing with regard to words in class G, to wit—in which other indicators do not.

Class G, in fact, consists mainly of words of comparatively intense positive tone, unaccompanied by prolongation of time or disturbance in reproduction; if the galvanometer had not been used there would have been nothing to distinguish them from indifferent words in class O. In these circumstances O and G, T and TG, R and GR, TR and TGR would have been combined and the results would have been:

Here again the dominance of disturbances in reproduction as indicators of negative tone is very noticeable, as also is the non-significance. of too-long times unaccompanied by other signs.

The superior resolving-power, so to speak, which is gained by the method when the galvanometer is used is obvious if we compare these last results with those given in Table I.

The high memory value of class TG is readily accounted for by, and constitutes a powerful vindication of the suggestions put forward on page 254 above. The class consists of words accompanied by strong positive tone whose reaction time is prolonged for one of the reasons there enumerated.

It seems reasonable to suppose that the influence of positively toned words on the reaction time will be approximately proportional to the intensity of their tone; or, at any rate, that prolongation of the time will occur with more intensely, rather than with less intensely toned.

J. of Psych. (Med. Sect.) I

19