or, at least, the most reliable indication of negative tone; I personally regard these two expressions as synonymous.

(iv) The galvanometer detects positive tone as well as negative and in many cases (the whole of class G) does so when the reaction time does

not.

(v) Intensity of affective tone, whether positive or negative, increases both reaction time and galvanometer deflection. In general the most positively toned words are those with too-long times and too-large deflections; next come those with too-large deflections only. Words with no complex-indicators, or with too-long times only, are mostly indifferent. Words with disturbance in the reproduction are almost invariably negatively toned. Words having too-long times and too-large deflections are, on the whole, more intensely toned, whether positively or negatively, than those having too-long times or too-large deflections only.

(vi) For quantitative work the galvanometer-deflection of the psychogalvanic reflex is markedly superior to the reaction time.

(vii) The 'resolving' power and consequently the scope and utility of the word association method is greatly increased if the galvanometer is used in addition to the reaction time. The experimenter can divide his reactions into eight classes, all possessed of quantitatively and qualitatively distinct attributes, instead of into four only.

(viii) The memory test enables us to determine the more important relative properties of these classes. It is a very laborious method and somewhat crude, but the results it yields show a remarkable concordance and it is probable that the conclusions arrived at are reliable.

297

THE RELATION BETWEEN COMPLEX INDICATORS AND THE FORM OF THE ASSOCIATION.

BY W. WHATELY SMITH.

In the preceding paper I investigated the relations which exist between the affective tone aroused by a stimulus word and the 'complex indicators' which accompany the reaction. I did this with regard to three indicators, namely,

(i) 'too-long' reaction time,

(ii) 'too-large' psycho-galvanic reflex,

(iii) disturbance of reproduction in Jung's reproduction test.

I showed that, if we indicate the presence of a 'too-long' time by T, of a 'too-large' reflex by G, of disturbance in reproduction by R, and the absence of any indicator by O, the relation between the affective tone of words and the various classes into which they can be divided according to their indicators is as follows: Classes G and TG

consist in general of positively toned words.

The question now arises as to whether there is any relation between the affective tone of a word and the form of the association, i.e. by co-ordination, co-existence, predicate, etc.

Sundry attempts have been made by various workers to investigate this point by determining the mean reaction time of the different classes of association, but without leading to any very uniform or satisfactory results. This is not surprising for, as I have shown, prolongation of reaction time alone is likely to be a very unsatisfactory and misleading guide; it may be prolonged on account of negative tone, of comparatively intense positive tone, or of purely intellectual factors which have nothing to do with affective tone at all. It is necessary to discriminate between positively, neutrally and negatively toned words before we can hope to throw any helpful light on the question. I have attempted to do so in this paper.

I wholly agree with Jung's statement that "Everyone who does

practical work in association has found the classification of the results the hardest and most tedious part." Many schemes have been devised, none are wholly satisfactory. If the system used is very elaborate and refined the results are likely to be unduly influenced by subjective factors and an immense mass of material is needed in order to give a reasonably large number of data in the rarer sub-classes; if it is too coarse we are liable to miss interesting points which a more detailed analysis might have brought to light. The additional labour entailed by the use of a very elaborate system also greatly reduces its practical value.

I therefore feel it necessary to give some account of the system which I have adopted and of the principles which have guided me in applying it. I may observe in passing that the first and most important principle which should be remembered throughout all work of this kind is that the classification should be in accordance with the workings of the subject's mind and not the experimenter's. A rigidly formal system based on purely logical or grammatical considerations is likely to ignore just those idiosyncrasies which we wish to study, and so to prove of little value. I shall discuss this question of the proper basis for classification in more detail at a later stage.

The system which I finally adopted, after a few preliminary trials, is based on that given by Jung1.

The primary division is between 'inner' and 'outer' associations. The criterion which I have tried to bear in mind in distinguishing between the two is perhaps best expressed by saying that in the case of 'outer' associations the connection between the ideas in the subject's mind has been formed for him, so to speak as a result of objective experience, whereas 'inner' associations are a result of what I may term the 'digestion' of experiences by the mind itself.

For example, the associations Cow-field, or Wine-bottle, are outer associations; one is accustomed to observe cows in fields and wine in bottles, such associations are given ready-made, so to speak, and do not demand any subjective mental work for their formation. The same applies to verbal associations such as Long-short, Black-white, which are constantly 'given' in conjunction. On the other hand such associations as Cow-animal, Frog-nasty, Child-nice, are to some extent dependent upon processes of analysis, synthesis, systematisation and so forth in our minds. This last idea can be clearly recognised in Jung's classification of associations by co-ordination into:

1 Studies in Word Association, pp. 13-38.

This principle is, I think, reasonably unambiguous and on a priori grounds seems the kind of distinction which is likely to prove helpful.

Its application presents certain difficulties, however, when we come to the consideration of the predicate type of association. Jung classes all varieties of predicate reaction together as inner associations, but I have grave doubts as to whether this is either legitimate or profitable.

I quite agree that predicates containing an element of personal opinion should be so regarded. But it seems to me that such reactions as Wine-red, Water-wet, Tree-green, which I may term 'simple' predicates, are just as much 'outer' associations as Wine bottle, Water —pond, Tree-wood. They are equally 'given ready-made' as a part of objective experience and are equally lacking in any product of subjective mental activity. Similar considerations also apply in some measure to very many cases of 'subject relationships' and 'object relationships,' e.g. Jump-horse, Swim-fish, Make-bread. There are, however, certain border-line cases, such as Speak-explicitly, which are difficult to deal with as they clearly contain a strong personal or truly subjective element. I shall return to this point later, but for the present I conform to Jung's arrangement.

Before proceeding to describe and exemplify the system I have used, I ought to say that I have throughout treated reactions as reversible. That is to say, I have not discriminated between the stimulus and reaction words; Tree-green, for example, has been treated just the same as Green-tree, Horse-ride as if it were Ride-horse, and so forth.

The classes into which I finally divided the words were:

A. INNER ASSOCIATIONS.

I. Co-ordination. This class is substantially identical with that of Jung. It is the vaguest and least satisfactory of the classes and I find a tendency in myself to relegate to it associations which I cannot place with certainty in any other class. But Jung himself allows a certain

elasticity1 and I have reduced this tendency to a minimum by omitting altogether from the classification a few words about which I felt real doubt.

I do not feel it necessary to give examples of this class as my divergence from Jung is inappreciable and even so occurs almost exclusively in his last and vaguest sub-class.

II. Predicates. I recognise here five sub-classes which are easily distinguishable.

(a) Simple predicates. By this I mean reactions in which the stimulus word is qualified by the reaction word, or vice versa, in a way which contains no element of personal opinion or judgment of value. Examples:

(b) Predicates expressive of personal opinion or judgment of value. This class needs no further definition. Examples:

(c) Predicates of 'subject relationship.' In this class the two associated words refer to some activity of which one is the subject. Examples:

(d) Predicates of ‘object relationship.' Here the two words relate to

some activity of which one is the object. Examples:

2 These are good examples of the border-line cases mentioned above.