experiments calling for 'free associations,' the 50% of cases in which the syllable succeeding the one presented was returned first had an average scoring time (T) of 3100 ; in the 4% of cases in which the preceding syllable was returned first this T was 6500 and in the 13% of other cases it was longer still. In that extra three and a half seconds we may reasonably assume that manifold interchanges, sometimes antagonistic, sometimes complementary, occurred between the 'perseveration-tendencies' of some of the eleven barely subliminal syllables, all of them, in consequence of their recent repeated appearances within the focus of consciousness, integrated into a more or less compact whole. At all events in the experiments calling for 'intentional associations' where the answers were all wrong, evidence of such varied interplay is furnished abundantly by the analysis of such cases which the authors provide1. The scoring time in these cases, we may reasonably assume, was as a rule longer than it was in the cases where the answer was right.

Taking all the circumstances concerned into account, then, we may still doubt whether the new facts brought forward in the masterly investigations of Müller and Pilzecker place the existence of a genuine reversal of the temporal order, in which association is first effected, beyond question. The interpretation advocated above when dealing with the facts advanced by Ebbinghaus, seems here also the simplest and best. In both instances we are concerned not with a series but with a tout ensemble -the foot in the one case, the line in the other. The very same tendency to unify and organize which has made out of two syllables a single foot has made out of six feet a line: in both cases the syllables, in addition to their originally temporal order, have acquired the relation of part to part in a coexistent whole; they have added to the seriality of the memory-thread the higher dimensions of the ideational continuum. This way of interpreting the facts will account for the comparative frequency of the wrong answers and the free associations that seem at first to point to genuine regressive association. When, for some reason, what we may call the normal response to the stimulus-syllable fails and the consequent perplexity and delay brings the line as a whole into greater clearness, the probability is that the parts specially related to the given syllable will be quickened the

1 Cf. op. cit. §§ 28, 45-7.

most, and among these, when the succeeding syllable fails, the preceding syllable stands next1.

It may fairly be said that the whole difference between the interpretation here expounded and that of Müller and Pilzecker turns simply on the fact that they sometimes give to 'association' a wider meaning. But that wider meaning, it is here contended, implies a complex of associations or what is better termed redintegration.

'Mediate Association.'

§4. A similar examination of the evidence advanced in favour of what is called 'mediate association' seems to justify the same interpretation of the facts. But 'mediate association' is used in two senses. First, and more commonly, it is used of cases in which prima facie there is no association at all, where, that is to say, an idea seems to 'rise freely' into consciousness -to use Herbart's phrase-though no mediating suggestion whatever is apparent. Of such an experience we have the stock instance of Hamilton, when, thinking of Ben Lomond, "this thought was immediately followed by the thought of the Prussian system of education." The 'intermediate and unawakened links' that explained 'the anomaly' he succeeded in tracing to a conversation about Prussian schools between himself and a certain German whom he chanced to meet on his last visit to the mountain. This and like instances, it is reasonable to assume were really cases of association, not of an idea reviving spontaneously as the Herbartians maintained. There is then no anomaly about them unless it be this absence of direct evidence. But, where not even indirect evidence is forthcoming, it would be rash too confidently to assert the impossibility of any spontaneous revival of a presentation (freisteigende Vorstellung), especially so in view of such facts as 'recurrent sensations','' perseveration,' and

1 The characteristic of the call for free association is that the subject is directed to the line as a whole, and we have seen already that when the first or accentuated syllable of a foot was given the last syllable of the preceding foot was named in 10% of the cases. It was also found that when the second or unaccented syllable was named the first syllable of the next foot was named in 9% of all the cases. Two comparable adjacencies had comparable strengths.

2 Cf. ch. vii, § 3.

delirium. Nevertheless if Herbart's 'spontaneous revival' or G. E. Müller's 'perseveration' were to be taken so 'atomistically' as to imply the complete rupture of the continuity of the memory-thread or the ideational tissue, it would be still more rash to assert that it was possible. But the mediate association we have here specially to consider is quite different from all this. In relearning verses forwards but omitting alternate syllables Ebbinghaus found a saving in time of 108%; by omitting two syllables, the saving effected was 70%; and by omitting three, 5.8%. This he explained by assuming that in memorising a series a b c d e ... there was formed not only a 'principal' or primary association of each term with its immediate successor, of a with b, of b with c, &c., but also subsidiary or mediate associations of each term with all the rest, of a with c, a with d, &c.; likewise of b with d, b with e, and so on. To these mediate associations he referred the savings obtained on relearning-the more distant associations being naturally the weaker and the saving therefore less. Such a series he rightly regarded as involving not merely a memory-thread but also an ideational 'plexus.' But the two, as we have seen, are of different dimensions.

The simpler process, as such, cannot then yield the more complex any more than a spinning wheel can do the work of a loom. Again mediate connexion between the members of the linear series is, of course, implied in its continuity, but this connexion presupposes association and cannot therefore constitute it. When the primary association of a with b begins, there can be no subsidiary association of a with c or d, or any subsequent member, for these members are not yet present. When this process is merely repeated, we can readily understand that the 'thread' is strengthened, but not that a whole tissue consisting of distinct threads begins to be formed 'associating every term with every other '—a tissue, that is to say, which in a verse of sixteen syllables would involve 105 subsidiary associations altogether in the forward direction alone! But after several repetitions, when the primary associations have begun to be familiar, the subject's attitude may change; and it does, and does so with some persons sooner and more frequently than with others. It is then possible to note various relations between the members of the series beside their serial order. The tendency to do this

distinguishes what Kant called the 'judicious' from the merely 'mechanical' memory. These two processes are not only distinct; they are also incompatible, in so far as an increase of the tempo, which favours the more mechanical process1, is a bar to the more intellectual one. It is true that the rate of learning which Ebbinghaus found 'convenient' was an unusually rapid one-150 syllables a minute. For all that, he could more than double it when learning 'sensible' material; so that at his usual rate there would be time for side glances; and in fact his remarks concerning the sources of error, to which he felt liable, shew that he was not altogether mechanically absorbed.

Indeed the ample experimental records now available shew unmistakably that even the least intelligent subjects are something more than mechanical registers. As G. E. Müller, the master in this department of psychology, has said, we should form but a very poor idea of the learning process if we assumed that no associations are actually effected between the different members of a series but such as would result if attention were confined to the one monotonous routine of linking item to item as each filed past. "The subject's activity in relation to the series to be learnt displays far more freedom and spontaneity than that." But the point is that unless such further subjective initiative is present nothing more is achieved. As the result of that initiative, however, a supplementary process of 'interrelating' (Zuordnung) comes into play, whereby "certain elements of the series, standing far apart, are often associated together, which would never be appreciably related at all, if the reading were nothing but an uninterrupted transition from one item to the next." This secondary interrelating is the distinguishing feature of Kant's 'judicious memorising' and implies the more complex process of redintegration. We may conclude then by saying with Müller that for experimental psychology it "still remains an open question how far, apart from all interrelating, direct associations between the mediate members of a series can be formed3." At present we may fairly say that there is no clear evidence for such mediate' association, as Ebbinghaus

1 Cf. Ebbinghaus, Grundzüge der Psychologie, 3rd ed. i. 672 f.

2 Ueber das Gedächtnis, p. 58.

3 Cf. G. E. Müller, "Zur Analyse der Gedächtnistätigkeit und des Vorstellungsverlauf," Zeitschr. f. Psych. Ergänzungsband v. 1ster Teil, 1911, pp. 315-7.

assumed, but rather a strong presumption on general grounds against it.

Reading.

§ 5. The synthesis or integration of simple linear associations into complex unities of higher dimensions might be fitly called the principle of psychical organization par excellence1. We have appealed to it incidentally in the above discussion; but now we have in the recent experimental investigations into the psychology of reading a favourable opportunity of studying it directly on its own account. For this process-unlike the earlier processes of building up our temporal and spatial perception and our intuition of real things-falls entirely within the domain of social intercourse, and is therefore throughout amenable to observation and control.

The earliest stage in the process of reading—that of learning the several letters-may be here regarded as merely a series of simple assimilations. In beginning the next stage, spelling, the child at first takes longer to recognise a monosyllable than to recognise a letter; for the monosyllable is still directly apprehended as a series of two or more letters. But after sufficient practice a short word is recognised directly as a unity, and is then recognised as soon as, or even sooner than, a single letter. But a word of three or four syllables may still have to be painfully spelled. Presently, however, when greater fluency is attained, it is found that a passage of sense, consisting of longer but fewer words is read more quickly and easily than one of equal length consisting entirely of monosyllables. For at this stage words are the units attended to, not syllables. Finally we come to read not by an almost continuous movement of the eyes as is generally supposed-taking in syllable by syllable or even word by word; but we compass a whole line of print like the present by three or four fixations of the eye, separated by pauses too brief to allow of the recognition of each separate syllable. When, however, this is requisite, as in reading nonsense 1 Cf. above, ch. iii, § 3.

2 The letters, that is to say, in reading apart from writing, being recognised merely as wholes.

3 Cf. M. Beer, "Die Abhängigkeit der Lesezeit von psychologischen und sprachlichen Faktoren," Zeitschr. f. Psychol. Bd. Ivi. (1910), pp. 271 ff.