BY C. H. TOLL

THE purpose of this investigation, of which the following gives a preliminary report, was to compare the tendency to associate by contiguity, with the tendency to associate by similarity.

In every series of stimuli to which one gives attention there is tendency to association by contiguity. But some similarity among certain elements of the series may produce a dissociation of the given elements into two series with some bond of similarity in each. This is a matter of common experience, as when you find you can read your newspaper and listen to your neighbors' conversation at the same time, understanding both, although the actual order in which the several words are perceived would form a meaningless mixture.

We may say dissociation is always accomplished by a tendency to association by similarity overcoming the constant tendency to association by contiguity. Study of the relative efficacy of the two may therefore be called a study of dissociation. The tendency to associate by contiguity might be measured in two ways.

First, when one attempts to learn a series in exactly the given order, the number of errors in the series as recollected may be taken as an inverse indication of the strength of association by contiguity. The three kinds of error possible in nearly all of the experiments were Omissions, Displacements, and Imperfections. All of these three have been tabulated. But the number of elements omitted seems considerably the most reliable as an indication of the degree of inadequacy of the associative tendency. The cases of displaced or imperfect elements are comparatively few: moreover, Displacements and Imperfections are not mutually exclusive categories. A single element may be both imperfectly recollected and wrongly placed in the recollected series. On the whole, it seems that the number of given elements which were omitted in the recalled series is the most positive and reliable of the errors. Our conclusions are based on the Omis

Second, when one makes no attempt to learn the series, simply giving attention to each element as it comes, and afterward lets the elements recur spontaneously, the number of cases in which a recollected element is followed by an element given contiguously may be taken as a direct indication of the strength of association by contiguity.

Tendency to association by similarity can evidently be measured in the same two ways, by counting errors when one purposes to learn the series as two groups of similar elements, and by counting sequences of similar elements when one avoids any effort to learn the series and recollection is spontaneous.

In the first seven experiments we used the first method. The errors made when the purpose is to associate by contiguity can then be compared with the errors made when the purpose is to associate by similarity, an equal number of series, given under the same conditions, and of identical character, being given in each case.

In the last four experiments we have used the second method. The number of sequences of elements given contiguously can then be compared with the number of sequences of similar elements.

Five subjects have coöperated in this, but the experiments were strictly individual, one observer being alone in the room with the experimenter. Each test lasted about an hour. As a matter of course, the results have been calculated for each of the five subjects and their agreements and disagreements have been carefully considered. But as this first report is to indicate merely the general tendency, we give here at first only the average of the five persons.

The experiments have varied as to the kind of elements used, the manner of presentation, the time allowed, and the manner of recording the recollected series. But throughout each experiment the series were of one identical type, while the individual elements were altered in each series.

In the experiments where the series were to be learned, some in the given order, some dissociated by similarity, it was found rather confusing to turn from one method to the other; so several consecutive series were learned by one method, and then several by the other, four alternations being made each hour to neutralize any effect of practice or of fatigue.

The series were of course different in kind in the several experiments, but were usually of eight or of ten elements. Half of this number had some distinct characteristic in common, the other half some other characteristic. In some experiments these elements were alternated, in some arranged irregularly.

In the first eight experiments the subject wrote down the elements recalled, as soon as the series had been given. In the last three the subject spoke the elements recalled.

In all cases where the first method of measurement was used, the time allowed for learning the series was made a little too short to permit of learning the series perfectly. Since comparison of the number of mistakes was our method, we naturally had to make sure there would be mistakes to compare.

The details of the several experiments were as follows:

(1) The elements were letters and numbers. They were about 12 × 8 mm. in size and were printed on white cards 15 × 30 mm.

Five letters and five numbers were placed, alternately, in a straight row on a sheet of white cardboard. The series was then exposed to the subject by turning up the small tin shutter of a screen that was clamped to the table-edge.

The time of exposure was measured with a stop-watch and was constant throughout the hour for each individual subject. Four seconds proved the best time for most of them, but in one case it was necessary to allow only three seconds. Twenty series were presented during each hour, ten for each method of memorizing. There were duplicates of all the numbers, and of eight letters, but not more than two of any element. Selection in forming the series was by chance. In dissociating, the letters were separated from the numbers.

As soon as the exposure was ended, the subject wrote down the elements recollected, trying to preserve their relative order. This recollected list was then copied beside the operator's record of the given series, so making the errors apparent.

(2) The elements were all letters, printed as before, and the alternate cards were placed half their length out of alignment with the original row.

The method of presentation, the length of exposure, the number of elements presented, etc., were as in no. 1. In dissociating, the letters on one level were separated from those on the other.

(3) The elements were all letters, printed as before, and five of the ten elements presented were placed out of alignment. But the disaligned cards were at irregular intervals and often in groups, and were only a quarter of an inch out of alignment. This order was varied each time, but without any system.

The other details were as in no. 1.

During this experiment I came to notice the effect produced by the natural tendency to learn the five elements of the dissociated series

in a rhythmical form, thereby increasing the ability to retain them; while there appeared to be no natural tendency to apply any such inclusive rhythm to the ten elements of the series when learned in the given order. To counteract this effect the subjects were instructed to consider the series, when learned in the given order, as two consecutive series of five elements each, and to use the same natural rhythm in learning these as they did in the dissociating. But this correction was not made in the first two hours, nor very perfectly in the rest.

(4) The elements were all numbers, printed as before, five of the ten being placed a quarter of an inch out of alignment, and in irregular groups, precisely as in the last experiment.

The time was reduced to three seconds for some and two seconds for the others. Details of presentation were as described in no. 1.

This time all the subjects tried to neutralize the effect of the instinctive rhythm for the five-element series by learning the tenelement series in two groups of five elements each.

(5) The elements were all nonsense syllables, each consisting of a vowel between two consonants, printed on white cards 20 × 20 mm. Eight of these were placed in an even row on a sheet of white cardboard, and four of them were marked by laying a quarter-inch strip of blue paper over the bottom of the card. The serial position of the marked cards was irregular, and was altered each time.

Ten seconds was given to some subjects, eight to the others. Other details of exposure, etc., were as in no. 1.

In learning the series in the given order, the blue markings were ignored; but in dissociating, the marked and unmarked syllables were learned in separate groups.

There seemed to be no rhythmical tendency; but to be safe the subjects were instructed to learn the straight series in groups of fours. Seven series were given to be learned in each method during the hour with each subject.

(6) The elements were one-syllable nouns, alternated with nonsense syllables, all spoken by the operator. The nonsense syllables were all different from those used in the preceding experiment: the nouns were ordinary words, and were so arranged as to avoid any obvious sequence or relation among them. Very few, if any, were used twice in one hour. Five nouns and five syllables were given in each series.

The elements were spoken at the rate of forty-six a minute, timed by a metronome which was muffled in a heavily padded box so that