organic processes simply follow suit. And even sensory rhythms can often be varied at the subject's own choice, or on the suggestion of another; and then again the breathing may be altered in consequence. Familiar instances of such procedure are to be found in the 'tunes' so readily attributed to the ticking of a clock, the puff of a locomotive, the churning of a steamer's screw, and the like.

Psychical rhythm, then, we may conclude, is due to attention or apperception, but the conditions determining this are many, and their relations very complex. If the presentations to be 'rhythmized' (the Rhythmizomena, as the Germans say) succeed each other slowly, the length (or shall we say the breadth?) of the 'psychical present' tells one way: the first impression is nearer the threshold when the third appears. If they arrive rapidly, their intensity and duration and the span of prehension tell another way; for it is essential that they retain their individual distinctness, and only so many can be grasped at once. But if the series continue long enough, or be frequently experienced, sub-groups may be treated as individuals; and indeed till some facility is acquired, the subject attending is aware of no rhythm. In the act of attention itself there are phases, in so far as expectation involves preadjustment to what is coming: usually the first members of a tact are predominant, and the rhythm tends to 'fall'; several alternations of accent within a complex rhythmic whole are of course still compatible with this. But it is important to note that, whether simple or complex, the rhythm is an intuited unity as truly as a geometrical figure may be. Unlike a geometrical figure, however, it rarely or never has symmetry. We cannot reverse a tune and obtain an effect comparable with that obtained by reprinting the score backwards in line with the original. We now pass to a question in which the psychological bearing of this fact becomes apparent1.

But first a new method of dealing with memory-problems must be mentioned, in which the connexion between rhythmizing

66

1 The following are among the more important papers on rhythm: T. L. Bolton, "Rhythm," Am. Journ. of Psychology, vi. (1893), pp. 145 ff.; E. E. Meumann, Untersuchungen z. Psychologie u. Aesthetik des Rhythmus," Phil. Studien (1894), x. 249 ff., 393 ff.; M. K. Smith, "Rhythmus und Arbeit," Phil. Studien (1900), xvi. 71 ff. 197 ff.; Arbeit und Rhythmus (1899), by K. Bucher, a well-known economist, bringing out the teleological aspects of rhythm; K. Koffke, “Exp. Untersuchungen z. Lehre v. Rhythmus," Zeitschr. f. Psych. (1909), pp. 1 ff.

and memorising has been turned to account by the Göttingen psychologists. The method of Ebbinghaus consisted in-at least, it depended on-ascertaining the repetitions saved in consequence of previous repetitions, when the verse perfectly learnt before, was relearnt some fixed time later. Hence this method is called the learning method or the method of saving. But now, using verses in trochaic measure, let the subject, a given time after a fixed number of repetitions (insufficient for perfect reproduction) be confronted with one of the accented syllables; then let him be asked to name the unaccented syllable that belongs to it. He will answer sometimes rightly, sometimes wrongly, and sometimes be unable to answer at all. This, the new method, is therefore named die Treffermethode, the method of 'shots,' or, as it has been called, the scoring method. It enables the experimenter to obtain far more insight into details than was possible before, for the 'misses' as well as the 'hits' are instructive. Moreover, by measuring the time of each answer (Trefferzeit) and comparing these times together, much can be learnt. In stronger or more recent associations, for example, the answers are made quicker than in weaker or older ones.

'Regressive Association.

§3. Does association work forwards only or does it work backwards also, as the middle link of a chain, when lifted, raises the contiguous links on either side of it? This is the question mentioned above to which we now pass, and it is one of firstrate importance. For empirical psychology must be radically wrong, if it be a fact that-even though attention only moves forwards-association may nevertheless 'run backwards,' as the Germans say. Such is certainly not the case when the forward direction makes sense, but with nonsense verses, if the mechanical analogy is a sound one, such reversal might be expected. For here there are none of the 'obstructing associations' which ' rhyme and reason' imply; and Ebbinghaus actually found in relearning a verse backwards that there was a saving of 124% of the time originally taken up in learning it forwards. Even when relearning backwards and skipping one syllable, the order of syllables, that is to say, being 16, 14, 12...2, 15, 13, II...I, Ebbinghaus found a saving of 5%. But the number of his

experiments in this case (four) was too few to give this result much value, as he fully admits. The variation in the time saved was also in both cases suspiciously great, ranging between 8" and 236" for mere reversal, and between 15" and 91" for reversal with omission of alternate syllables. Still these experiments as a whole might incline us to suppose that association -left to itself, so to say-can work in both directions, though the connexions backwards are considerably weaker. But if so, the associations both ways should be alike at least in formcontinuous, that is to say, backwards, dc ba, as well as forwards, a b c d. In that case, however, d would revive a more c frequently than b, and b more frequently than a Such a connexion between strength of association and proximity is invariable in so-called 'mediate association' when the direction is forwards.

In favour of 'regressive association' there is, in fact, no consistent evidence forthcoming. Quite the contrary. For example, in two or three hundred experiments by Müller and Pilzecker, verses of twelve syllables were repeated a set number of times in anapaestic measure-accented, that is to say, on the 3rd, 6th, 9th and 12th. After a fixed interval the subject, confronted with one of the accented syllables, mentioned any of the other syllables which he called to mind. Now the cases in which the second syllable of a foot (that immediately preceding) was revived were only about half as frequent as those in which the first syllable of a foot (the next but one preceding) was revived, not more frequent, as we should naturally have expected. Moreover the scoring time (Trefferzeit) for the first but remoter syllable was shorter than that for the second and nearer1. Such results are incompatible with the theory of continuous backward association, but they are readily explained by the fact that the group of three syllables had become one complex whole: it was a single foot in a rhythm. Hence the tendency to reinstate the initial member of the group was stronger than that to reinstate the middle". The saving effected in Ebbinghaus's experiment is also thus explained. In short, the conclusion to which these results seem to point is that they

1 Müller and Pilzecker, "Experimentelle Beiträge zur Lehre von Gedächtniss,” Zeitschr. f. Psych., Ergänzungsband i. (1900), § 39.

2 Cf. above, § 1, p. 225.

immediately involve only relations of coexistence. With temporal order either forwards or backwards they are not concerned: the term 'regressive association'' is thus inappropriate. They seem to be cases of redintegration not of pure association at all.

་

Unfortunately, beside the scanty experiments of Ebbinghaus just mentioned there are no others specially devoted to this problem. Müller and Pilzecker, however, bring together what they regard as conclusive evidence of 'genuine regressive association' incidentally furnished by some of their experiments3. A large part of this evidence is derived from the 'misses' or 'false cases' yielded by the scoring method as above described. A greater number of these wrong answers, that is to say, than chance would explain,' consisted in naming not the syllable following the stimulus-word but the syllable preceding it. In one series of experiments where chance would account for only 23 cases there were actually 7 in another the corresponding numbers were 1 and 5. Further evidence is adduced from experiments in which a different method was used. The subject, confronted with an accented syllable, instead of being directed to name only the following syllable or what he took to be such, was left free to name whatever syllable the stimulusword first evoked and to add a second, if such occurred. This method, calling for 'free associations,' required, as it turned out, greater psychological savoir faire on the part of the subject than the old one restricted to 'intentional associations.' In fact of the only two series of experiments dealing at all directly with regressive association-in both of which this method was adopted-one was disallowed because of the subject's incompetence; and even the subject of the other appears to have been new to the work. Anyhow the summary of his answers is as follows (the lines consisting of twelve syllables in trochaic measure): An accented syllable being presented, the following, i.e. the unaccented syllable of the same foot, was named first in 50% of the cases. The preceding syllable, i.e. the unaccented

1 Cf. ch. vii, § 3.

* With this Müller and Schumann fully agree: cf. their "Experimentelle Beiträge zur Untersuchung des Gedächtnisses," Zeitschr. f. Psych. vi. (1894), p. 308 fin. Cf. also A. Wreschner, "Die Reproduction und Assoziation von Vorstellungen," op. cit. Ergänzungsband iii. (1907), p. 578.

Cf. op. cit. § 41, pp. 207-12.

syllable of the previous foot, was named first in 4% of the cases and second in 6% or only 10% in all. The remaining (9) syllables together secured but 13% of the answers. Clearly then both methods bring to light some sort of connexion backwards as well as forwards: not simply from a to b but also from b to a. But is the latter genuine regressive association? An examination of the whole situation seems to render such an interpretation exceedingly doubtful.

Well, in the first place the almost invariable recency of this so-called association is remarkable. In the experiments with 'free association' just described the testing began 3' after the line had been learnt, and the effect of longer intervals was not investigated. But in the earlier experiments, where the subject was restricted to 'intentional association' and the relevant cases were all misses, it was found that in one series of 48 experiments in which the stimulus-words were shewn 20" after the reading, there were 15 cases in which the preceding syllable was named in mistake in two other series of 216 experiments in all, tested after an interval of 24 hours, there was only one such case.

In the second place, in consequence of this recency, the line as a whole was in a state of 'preparation' (Bereitschaft) such that no syllable was far from the threshold of consciousness. In the terminology of the writers their Perseverationstendenz, or readiness to reappear, was still so strong that with every syllable a very slight reinforcement of this tendency sufficed for the syllable's actual reproduction'. Further the subject frequently knew the place of the presented syllable in the line, and this knowledge often enabled him to find the syllable wanted. Sometimes, when the whole line was 'firmly imprinted' he would run through it as far as the presented syllable, the accumulating efficiency of revival due to the whole securing what the single syllable could not effect. This resource would obviously be specially available where, as in the cases we are considering, the repetitions had all been recent.

In the third place there were signs in all these cases of a certain embarrassment or contretemps akin more or less to what the writers happily styled Verlegenheitsnennungen. Thus, in the

1 Cf. op. cit. p. 66.

2 Op. cit. p. 16. Cf. also F. Arnold, "The Initial Tendency in Ideal Revival," Am. J. of Psych. xviii, 1907, pp. 239 ff.