LINGUISTIC SOCIETY OF AMERICA EDITED BY GEORGE MELVILLE BOLLING Ohio State University AURELIO M. ESPINOSA EDWARD SAPIR Stanford University University of Chicago NUMBER I NOVEMBER, 1925 A TECHNIQUE FOR THE EXPERIMENTAL INVESTIGATION OF ASSOCIATIVE INTERFERENCE IN ARTIFICIAL LINGUISTIC MATERIAL BY ERWIN ALLEN ESPER Assistant Professor of Psychology, University of Illinois. LINGUISTIC SOCIETY OF AMERICA 204 St. Mark’s SQUARE, PHILADELPHIA 1925 92 1956 as) PELL r sen ries saaee at FOREWORD The experiments here reported, which were carried out in the Psychological Laboratory of the Ohio State University dur- ing the academic year 1921-1922, represent a further attempt in the development of a technique for the experimental inves- tigation of the sensorimotor conditions involved in the process of analogic change in language. My interest in this problem dates from 1917, at which time I first became acquainted with the pioneer work of Thumb and Marbe in this field. The expe- riment which I reported in the Psychological Review in Novem- ber 1918 was a repetition of the work of Thumb and Marbe; the method employed in the present investigation was sug- gested by Thumb in 1907. The conception of linguistic change as an experimental problem is one for which Thumb deserves the gratitude of all students of language. I wish to express my deep sense of obligation to my teach- ers, George Melville Bolling and Albert P. Weiss. To the former, I owe my interest in language problems ; to the latter, my training in psychological methods as well as guidance and advice throughout this investigation. I am indebted also to Professor Leonard Bloomfield for valuable suggestions in the early stages of this work. |Sisie ed Be Ohio State University January 1925 J aaa 4! eval bt at Ben Fe dae i ; Peery ts Lona tae gk ieee ( ; Pa Alay oy poy ak eae” hae Oar PAC PE eee rh Che Rr Hs Cw : tl ak! “ q Si at's . 3 ‘ yi TABEEY OF COMNLENTS PPINTROUUCIION Sirentc ar antes gots sa aiee Vcc es Aetas Gee e eM eie a bees DIM STATEMENT TOP NTR Ee ROSLEM Mie ce. sta oe ais tales Cae cates ce eee fins BREE MREHOR ANUS A PPARALU Ste enamel Wiss ghia aed inure a hyie's Gese.y doe PSS ORM Ree ete hear ha a ae ais ana: sim wd Uo win aad oh ah se eevee DR ADPAAVUS eas ce ele en es Wey eet de ae eat sea Rene 3 5 Coyahe Miniatore: Linguistic Systems yes, eee eh eG ay, Le General bs perimental | Procedures 214) 2 ish) ipa ede ib ice oY DUS Me DeOae te TE SICTIOS aii ois let lalate as wha a We Wale tates Ko 6 Bold GL CCORIMUICTE NENIOG ca sha (Ry arate yie'y git Gains nels tie igidclw Ras g.) Dhe Distribution of the Repetitions. (0. vce sca e eee ee Bay LHe AUGAt hed: FIgUteso ar. Pas Guha) ois ba. ite ease e ees E. The Material and Special Conditions of ExperimentI......... ie F, The Material and Special Conditions of Experiment II......... G. The Material and Special Conditions of Experiment III......... MUM RSUL LSet cS Gris ce Mialbcickh, id RMR tide Pee LY ota) a liar e'@ataldia > Wid cg le Mat A. The Rates of Learning in Experiments I, II, and III........... PER OPUITRG Siac tret sis eee ra tier ag Fo Oa Sis CORK MO ac saad a th Sas Bi PRMOFUNCUG Ts oral Art ray ee ee Mista ds ec Abie € c's saa bee ais oy ee MRP MOETIMCUE Unni. ces fale eee ee Gee we eae oe a ee weeks ESHIE ey UEP SNe Ray Wie e wit Me gy PROTA LAN, tas bet B. Analysis of the Responses in Experiment I................... C. Analysis of the Responses in Experiment II................... D. Analysis of the Responses in Experiment III.................. WP OUMMAR Ys crcrcr co's halo he Be ates see Raa Na area a ule Beats locstis tone Sears chat erate y ils wtp on a a oo ad t ‘ : I, INTRODUCTION Aside from changes in the meaning of words, linguistic science has distinguished two main classes of changes in language. The one class is called phonetic change, and includes all those cases where a given sound undergoes a certain modification wherever it occurs under the same phonetic conditions. The other class is called analogic change, and includes all those cases where words or groups of words, associated semantically or formally, become assimilated to each other in form. This second class has been theoretically explained as due to ‘ associations of ideas’ or ‘ word-associations’.: That is, if a word a is strongly associated with a word 4, then it may often happen that the articulation of a will be interfered with by the associated word ), with a resultant mixture of the two words. In rgor the philologist A. Thumb, in collaboration with the psy- chologist K. Marbe, made the first attempt at an experimental study, in the psychological laboratory, of the conditions underlying analogic change.' Their specific problem was to investigate what characteristics a word- association must have in order to result in an actual analogic change. A list of words was chosen from a number of categories which in the history of the Indo-European languages had shown a strong tendency toward analogic change, with a consequent assimilation in form of the different words of each category to each other. These selected words were used as stimuli in a free-association experiment. The results were as follows : (1) the more frequent a response was (i.e., the larger the number of subjects who gave the same response to a given stimulus word), the shorter was the reaction time; (2) words of a given cate- gory were associated predominantly with words of the same category ; (3) in all the categories, with the exception of numerals, reciprocal associations were found ; (4) numerals were associated predominantly with higher numerals ; (5) adjectives were associated predominantly with adjectives of opposed meaning. Further, since the most frequent 1. A.Thumb und K. Marbe : Experimentelle Untersuchungen ueber die psychologi- schen Grundlagen der sprachlichen Analogiebildung, Leipzig, 1901. 5 6 LANGUAGE MONOGRAPH NO. I, 1925 associations, which were also those occurring most rapidly, showed a substantial agreement with the courses of analogic change and assimi- lation revealed in the history of the various Indo-European languages, Thumb concluded that (6) analogic changes result from those associa- tions which predominate in a group of individuals and which occur with the greatest rapidity in the individual members of the group.? Thumb himself regarded this experiment as but a beginning.3 He believed that very important results might be obtained by extensive studies of various languages; more particularly, by investigating the associational tendencies in a given dialect and comparing these with actually developing analogic changes. From a behavioristic point of view, we may say that the method of Thumb and Marbe is of value in indicating the interconnections of groups of verbal reactions ; and that, by giving a measure (reaction- time and frequency) of the relative strength of these interconnections, it may aid us in predicting what modifications in the verbal reactions are likely to take place. On the conditions under which the intercon- nections themselves are developed, however, the method throws no light. It necessarily deals always with systems of verbal reactions which have been long and firmly established.* 2. Further experiments with similar method, material, and resuits were reported by Thumb, Indogermanische Forschungen, 22. 1-55 (1907), together with a reply to the criticisms of Wundt and others. Oertel, Am. J. Philology, 22. 261-267 (1901), reported an experiment with widely diverging method and consequently diverging results, for a discussion of which see the above article by Thumb and also Marbe, Am. J. Psych., 13. 450f. (1902). The writer in 1917 repeated the experiment of Thumb and Marbe with a large number of subjects, including a group of children and one of uneducated adults (4 Contribution to the Experimental Study of Analogy in Psych. Rev., 25. 468-487 [1918]). Since the results of this repetition with American subjects were in substantial agreement with those of the German investigation, the range of applicability of the latter was greatly extended. 3. Cf. Experimentelle Untersuchungen, pp. 84 ff., and the article cited in the preceed- ing note. 4. This limitation was recognized by Thumb. Cf. Experimentelle Untersuchungen, p. 79 :‘*Aber nehmen wir einmal an, dass die Frage beantwortet ware, welche Asso- ciationen vorhanden waren, und warum sie eine lautliche Wirkung hervorgerufen haben, so miissten wir doch zur Feststellung der Gesetzmassigkeit analogischer Neue- rungen auch noch wissen, unter welchen Bedingungen bestimmte Wort- oder Form- associationen lberhaupt eintreten. Das sind jedoch Fragen, die von der Sprach- wissenschaft allein noch weniger gelést werden konnen — aber auch die Psychologie ist in diesen Dingen noch nicht weit gekommen”. I. STATEMENT OF THE PROBLEM Thumb suggested a new method of investigating the conditions under which analogic changes take place, in the following words : ** ,... ist es vielleicht méglich, kiinstlich bei Versuchspersonen Ana- logiebildungen (Kontaminationen) hervorzurufen ?..... Wenn es gelingen wird, Analogiebildungen experimentell zu erzeugen — natiir- lich an einem sprachlich véllig neutralen Material, d.h. an kiinstlichen Lautgebilden —, so werden wir imstande sein, die bis jetzt erkannten Bedingungen des Vorganges in ihrem Wirken qualitativ und quantitativ zu studieren und weiteren Faktoren der Analogiebildung auf die Spur zu kommen.” 5 The present series of experiments represents an attempt to develop such a method. When the problem is approached in this manner, however, its scope becomes ultimately much broader than the original question of the conditions under which analogic changes take place in the historic languages. It becomes a problem of how language reactions become organized into the systems which we call grammatical categories, and how the component reactions of these systems come to be marked by common elements ; in other words, under what conditions the ‘ asso- ciations’’ which have been held to be responsible for linguistic assimi- lation are themselves established. A limitation of such an investigation is at once evident : we can experiment only with subjects who are already possessed of a highly complex and firmly fixed set of language habits ; a‘ primitive man’ or isolated infants are unfortunately not to be obtained for experi- mental purposes. This limitation, however, is rather general in psy- chological experimentation.® We can at least expect that the manner in which artificial linguistic material becomes organized into categories 5. Indogermanische Forschungen, 22. 48-49 (1907). 6. Dr. A.P. Weiss reminds me that such limitations are not restricted to psycho- logy : ‘‘ The ideal scientific method, ‘to vary only ome factor while all the other factors remain constant ’, is only an ideal. What we actually do is to keep as many conditions constant as we can and by repetition and variation try to segregate the relative potencies of the elements (often unknown) ofthe combination which cannot be physically broken up.” “I 8 LANGUAGE MONOGRAPH NO. I, 1925 will bear a definite relationship to the tendencies inherent in the speech- habits of the subjects, and that a comparison of the results gained from subjects of widely different language habits may reveal certain uniformities and general laws according to which the organization of language habits takes place. A program of investigation is thus sug- gested for which the present study aims merely at making a begin- ning in the development of an adequate technique. Briefly stated, my problem was to arrange a number of experiment- al conditions for the léarning of artificial linguistic material, and to compare the results under each of these conditions with reference to the development of linguistic categories. Il. METHOD AND APPARATUS A. GENERAL. The method consisted essentially of joint visual, auditory, and kinesthetic stimulation. The visual stimuli were irregular figures of equal area (6 sq. cm.) and of different colors (red, green, blue, yellow). The auditory stimuli were nonsense syllables, pronounced by the experimenter as the visual stimuli were exposed. The kinesthetic stimuli resulted from the subject’s being required to pronounce the nonsense syllables after the experimenter. B. APPARATUS. The apparatus used for exposing the visual stimuli is a form ot tachistoscope designed for this experiment by Dr. A.P. Weiss. It con- sists of a drum 20.6cm. in diameter mounted in slots at the top of a wooden frame 1.52 m. in height. On the axle of the drum is a five- pointed starwheel, the points of which are tripped by a trigger on a pulley-wheel. With each revolution of the pulley-wheel, the drum is therefore turned through one-fifth of a revolution. The apparatus is driven by a 1/to h. p. electric motor. A series of pulleys reduces the rate of the drum to one-fifth revolution every 2.7 seconds. During each of these intervals the drum is in motion approximately 0.7 sec. and at rest 2.0 sec. The figures which constitute the visual stimuli were cut out of standard (Milton-Bradley) colored papers and mounted in two vertical rows on acontinuous paper belt 21.6 cm. in width. This belt was then hung on the drum of the tachistoscope, the figures being so spaced on the belt that with each one-fifth revolution of the drum one figure moves into a given area. The frame of the apparatus was cover- ed with grey cardboard so as to conceal the paper belt except for a window of 22.8 by 12.7 cm. at a height of 1.07 m. above the floor. In the opening thus left was placed, in horizontal grooves, a grey card- board slide 11.4 by 12.7 cm., which made possible the exposure of either the right or the lett vertical row of figures. There is thus an exposure area of 11.4 by 12.7 cm. on either the right or the left side according to the position of the slide. A shutter, also operated by the trigger on the pulley-wheel, closes over the exposure area while the drum and belt are in motion and opens when the next figure is in position. 9 IO LANGUAGE MONOGRAPH NO. I, 1925 No. I. No. 2. No. 3. No. 4. FIGURE I. The four shapes used in all experiments, reduced to 1/2 diameter. ESPER, ASSOCIATIVE INTERFERENCE TI The experiment was conducted in a dark-room, the exposure area of the apparatus being illuminated by a daylight bulb. The subject was seated 1.4 m. from the apparatus and directly in front of the exposure area. The experimenter sat at a table to the right and slightly to the rear of the subject, from which position he was able to control the apparatus either by a switch stopping and starting the motor, or by means of a cord which, running through a series of pulleys, controlled an idler-wheel which acted asa belt-tightener. This idler-wheel made it possible to stop the drum without stopping the motor. C. Tue Mintature Lincuistic Systems: The visual and auditory stimuli of each experiment constitute together an artificial miniature linguistic system. The shapes of the figures (Fig. 1) were devised with a view tomaking them unlike any familiar object for . which the subject could find a name. Ingenious subjects will of course always find some object with which to compare a nonsense figure, just as they will always contrive to associate a nonsense syllable with some word of actual language. The attempt was made, however, to reduce the possibility of such associations foreign to the experiment. The four shapes (numbered 1, 2, 3, 4) and the four colors (red, green, blue, yellow) made possible 16 combinations. These combinations will be symbolized hereafter by numerals indicating their shape and initial letters indicating their color; thus, 1R = shape No. 1 of red color ; 2G =shape No. 2 of green color ; etc. In what will be describ- ed below as the ‘ learning series’, only 14 of the combinations were used, 2 being omitted (2Y and 4R); while in the ‘recognition series’ all combinations were used. For each of the combinations of shape and color, a nonsense name was devised. In the construction of these names only English sounds were used, although the sequence of sounds deviates to some extent from the English system. An especial effort was made to prevent a resemblance of the artificial words to any specific words of actual lan- guage. The syllables of the artificial names are correlated in differing degrees in the three experiments with the two factors of color and shape in the figures. D. THe GENERAL ExpERIMENTAL PROCEDURE. In Experiment I, seven subjects were used; in Experiments II and II, eight each. All were undergraduate students in the Ohio State University. No individual acted as subject in more than one of the experiments. 12 LANGUAGE MONOGRAPH NO. I, 1925 The subject having been seated before the apparatus, the motor (but not the exposure drum) was started and allowed to run while the instructions were given, in order that the subject might become adapted to the sound of the motor. It may be remarked here that the sound of the motor and of the apparatus together did not interfere with the understanding of speech in ordinary conversational tones. The sound of the apparatus, being relatively constant, was actually an advantage, inasmuch as it eliminated disturbing noises arising outside the dark- room. 1. The Learning Series. The Jeft vertical row of figures constituted the learning series, the right row being concealed by the slide. The following instructions were read to each subject : This is not an intelligence test. It isan experiment to determine how quickly you can learn the names of certain sacrificial objects in the Morgavian language, a language spoken on the northern slopes of the Himalaya Mountains. As each object is shown, I shall pronounce its name. You will immediately repeat the name after me aloud. If you are not sure of my pronunciation of any word, do not ask for a repetition, but do your best. Each time the series is repeated, try to make your pronunciation more like mine. Do not make any comments until the end of the hour, and do not set yourself any other task than looking at the objects and repeating their names after me. 7 2. The Recognition Series. After the learning series had been repeated four times, the drum was stopped, and the slide pushed over so as to expose the right half of the belt. The row of figures on this side of the belt constituted the recognition series, which differed from the learning series in that the figures were in an entirely different order, and in including all of the possible 16 combinations of shape and color, while in the learning series 2 combinations (2Y and 4R) had been omitted. These two com- binations will be referred to hereafter as the ‘ unlearned figures’. The following instructions were now read to the subject: I shall now show you the objects in a different order, without pronouncing their names. As soon as you see each object, call out its name in a loud distinct voice. If you are not sure of the name, guess ; but be sure to call out promptly the instant each object is exposed, without waiting to think about it. It is not expected that you will learn the series immediately. Be sure to call out loudly and distinctly. Do not set yourself any other task than calling out the names of the objects as they appear. 7. The reference to ‘sacrificial objects’ and the ‘ Morgavian language’ is of course fictitious. It was desired to avoid a nonsense setting as far as possible. te ESPER, ASSOCIATIVE INTERFERENCE 13 The drum having again been started, the experimenter recorded the subject’s responses. Those responses which varied from the ‘correct’ responses (i.e., those taught in the learning series) were recorded as accurately as possible in the symbols of the International Phonetic Association. After the recognition series had been given once, four repetitions of the learning series were again given in the same way as before ; then a recognition series, followed again by four repetitions of the learning series, and so on. 3. The Distribution of the Repetitions. Each subject appeared on seven different days distributed over a period of two months. The sequence of days was arranged in geomet- rical progression as follows : 1st day, 2nd day, 4th day, 8th day, 16th day, 32nd day, 62nd day. The recognition series was given after every four repetitions of the learning series, except on the last three days when the total number of repetitions of the learning series became less than four. On the 1st day, 32 repetitions of the learning series were given ; on the 2nd day, 16; on the 4th, 8; onthe 8th, 4; on the 16th,2 ; on the 32nd,1; on the 62nd, none (only the recognition series being given). On each day after the first, the subjects were given a recognition series to begin with, to test their remembrance of the material from the work of the preceding days. Table I summarizes the distribution of repetitions. 4. The ‘ Unlearned’ Figures. In each of the three experiments, figures 2Y and 4R with their cor- responding names were omitted from the learning series, but included in the recognition series. The subject accordingly never saw these figures or heard their names in the learning series. Whatever responses he made to them when they appeared in the recognition series were necessarily dependent on analogy with the other figures and names which he had been taught. The subjects were not told that there were two more figures in the recognition series than in the learning series. Some subjects, more especially in Experiments II and III, became sus- picious after they had reached a certain stage of learning and inquired whether there were not more figures in the recognition series. The invariable reply of the experimenter was that ‘‘the recognition series 1s the same as the learning series except that the objects are in a diffe- rent order.” I4 LANGUAGE MONOGRAPH NO. I, 1925 Age 3) cnt SuMMaRY OF ReEpeTITIONS, ExperIMENTS, I, II, anv III. On the upper horizontal line, rst, 2nd, etc. refer to first day, second day, etc. L= learning series; R = recognition series. Numerals indicate number of repetitions. Ist 2nd 4th 8th 16th 32nd 62nd IR tR IR IR IR 2R 4L 4L 4L 4L OE. IL IR IR IR IR IR IR 4L 4L 4L IR IR IR 4L 4L IR IR 4L 4L IR IR 4L IR 4L IR 4L IR 4L - IR Totals’ 32L,6R 16L,5R"8Li3Ro0 qLieR sole Reyri ek 2R TABLE I THe MATERIAL OF EXPERIMENT [. Com. = shape-color combination ; the numbers refer to the shapes of the figures (see Fig. 1); the letters refer to the colors red, green, blue, and yellow. The two ‘unlearned ’ figures and their names are starred. Com. Name Com. Name Com. Name Com. Name IR naslil) 2R nascaw 3R__snasdeg 4R* __ naskop* 1G weclil) 2G wecttaw 3G wetdeg 4G weckop 1B Jownlil) 2B fowncaw 3B fowndeg 4B Jownkop 1Y rojliy ed rojcaw* aX rojdeg 4Y rojkop ESPER, ASSOCIATIVE INTERFERENCE 15 TABLE Ill Tue MateriaL oF Experiment II. Explanation as under Table II, page 14. Com. Name Com. Name Com. Name Com. Name IR nulgen 2R dojlgen 3R pelgen 4R* wilgen* 1G nugdet 2G dojgdet 3G pegdet 4G wigdet 1B nuzgub 2B dojzgub 3B pezgub 4B wizgub rY numbow 2Y* dojmbow* 3Y pembow 4Y wimbow E. THe MATERIAL AND SPECIAL CoNDITIONS OF ExPERIMENT I. The nonsense names assigned to each of the combinations of shape and color in Experiment I are indicated in Table II in the symbols of the International Phonetic Association.® The special conditions of Experiment I, as the table in part shows, are as as follows : 1. The nonsense syllables are correlated throughout with the two factors of shape and color in the figures. For example, to shape No. 1 corresponds the syllable -/in ; to the color red corresponds the syllable nas-. 2. The order of the linguistic elements is color-shape, conforming to the English construction adjective-noun (e. g. ‘ red square’). 3. Each of the linguistic elements is a natural syllable. 4. The two syllables of each name were spoken by the experimenter with even stress and without pause. F. THe MATERIAL AND SPECIAL CONDITIONS OF ExPERIMENT II. Table III will indicate the nonsense names assigned to each of the combinations of shape and color in Experiment II (see footnote 8). The special conditions of Experiment II are as follows : 1. Specific sound-sequences are correlated throughout with the two factors of shape and color in the figures. For example, to shape No. 1 corresponds the sound-sequence nu- ; to the color red corresponds the sound-sequence -/gen. 8. The stimuli were of course not presented to the subject in the order of the Table. They were arranged in one order for the learning series and in another order for the recognition series; in both cases in such a manner that like figures did not follow one another. a 16 LANGUAGE MONOGRAPH NO. I, 1925 No. I. No. 2. I II Il I II Ill R_ naslitj nulgen éelhib R_ nascaw dojlgen nuzgub G weclil) nugdet vit G wettaw dojgdet kulren B fownlit) nuzgub zib B fownltaw dojzgub sotmil} Y rojliy numbow tekbem Y *rojcaw “dojmbow *. No. 3. No. 4. I II Ill I II Il R_nasdeg pelgen jowndeg R_ *naskop *wilgen * G wetdeg pegdet awa G_ weckop wigdet feb B fowndeg pezgub faZa B fownkop wizgub wut Y rojdeg pembow jonéim Y rojkop wimbow pel Ficure Ia Showing the names assigned to the figures in Experiments I, II, and III. Figures are reduced one-half. Under each figure, the vertical column at the left indicates the colors, and the Roman numerals at the top refer to the three experiments. The two shape-color combinations which did not occur in the learning series are starred. ESPER, ASSOCIATIVE INTERFERENCE 17 TABLE IV THe MarertaL OF EXPERIMENT III. Explanation as under Table II above. Com. Name Com. Name Com. Name Com. Name IR élhib 2R nuzgub 3R jowndeg 4R* — 1G vit 2G kulren 3G awa 4G Jeb 1B zib 2B sotmil) 3B faza 4B wuc Bs tekbem 2Y* oo 3Y jonfim 4Y pel 2. The order of the linguistic elements is shape-color, reversing the English order of adjective-noun. 3. The linguistic elements are not separated by a natural syllable division ; e. g. in mulgen the two elements are nu-/gen, an impossible syllable division in English. 4. The experimenter pronounced each name without pause, stressing the last syllable. It will be seen that conditions 2 and 3 of Experiment II greatly lessen the parallelism of our artificial system with the English con- struction. G. THe MATERIAL AND SPECIAL CONDITIONS OF EXPERIMENT III. Table IV will indicate the nonsense names assigned to the combi- nations of shape and color in Experiment III (see footnote 8). The special conditions of Experiment III are as follows : 1. No correlation whatever is provided between the visual and the auditory stimuli, except that, since most of the auditory stimuli contain two syllables, a division into two elements corresponding to the two factors of the visual stimuli is suggested. Five of the auditory stimuli, however, contain only one syllable. 2. The experimenter pronounced each name without pause, and in the case of those names containing two syllables stressed the second. 3. Since there is no correlation between visual and auditory stimuli, the names of the other figures provide no pattern for the naming of the two ‘unlearned’ figures (2Y and 4R). The material of Experiments [, I], and III may be compared in Figure Ia. IV. RESULTS A. Tue Rates oF Learnine IN Experiments I, II, anp III. In a comparison of the linguistic results of Experiments I, II, and III, it will be important to have a measure of the relative difficulty of the learning under each of these three sets of conditions. In Exper- iments I and II, where the two visual factors of shape and color are exactly correlated with specificand uniformly recurring sound-sequences, it is also possible to compare the rates and degrees of learning of (1) the syllables corresponding to shape and color, and.(2) the names corresponding to the 14 figures of the learning series and those cor- responding to the 2 figures which occurred only in the recognition series. By correct response is meant a response to a figure substantially identical with the name which was taught to the subject in the learn- ing series, or, in the case of the two figures which did not occur in the learning series, a response showing a correct analysis of the factors of shape and color in accordance with the categories of the linguistic system of the experiment. Many responses have been,counted as correct, for the purposes of these comparisons, which deviate phonet- ically to some extent from the words taught, but which are never- theless unmistakeable variants of these words. They are due to the subject’s misunderstanding of certain sounds in the learning series or his inability to reproduce them exactly in the recognition series, and ultimately are to be referred to three causes: first, some of the sound-sequences of the experimental material differed from very com- mon sequences in English ; second, the dialect of some of the subjects differed from that of the experimenter ; and third, the unfamiliarity of material made possible confusion between such sounds as voiced and voiceless stops, m and, thand f, etc. Such variant pronunciations were inevitable under the conditions of the experiments, where adult individuals were orally taught unfamiliar material without prelimi- nary training and under definite time restrictions. To eliminate such responses from the tables would, in the writer’s opinion, give a dis- torted picture of the rate of learning. In the curves of Figures 2 to 7, the upper row of figures along the x-axis indicates the cumulative number of repetitions of the 18 ESPER, ASSOCIATIVE INTERFERENCE 19 learning series, and the lower row the order of the days on which these series were given, as shown in Table I. The figures along the y-axis indicate the percentage of correct responses. In both tables and curves, a value of 100 would mean that, for the recognition series in question, every subject responded correctly to every figure. The first recognition series of each day after the first, and the two series of the 62nd day, are not indicated in the curves, because these were given without additional learning, and hence represent only the degree of retention from the previous days. The percentages for these series are however shown in the tables. EXPERIMENT I. Table V and Figures 2, 3, and 4 give the results for Experiment I. Columns I, II, and III of Table V and the curves of Fig. 2 show the rates of learning of the complete names corresponding to the figures, and of the syllables corresponding to shape and to color. The curves show the same general character; a rapid rise during the first day’s work, with a falling off in the last series of this day ; there- after, on the second day, a more gradual rise, and, beginning with the fourth day, a levelling off. Noteworthy is the amount of retention on the 62nd day. The two recognition series of this day were given, without additional learning, 30 days after the last previous series. They show practically no losses due to forgetting. The curve of learning for the shapes rises very rapidly for the first three series of the first day, but thereafter levels off, and on the seventh series is crossed by the curve for color. The relationship between these two curves is however obscured by the responses of one subject who, while responding correctly to the colors, consist- ently confused the shapes. Fig. 3 shows the curves for shape and color as they appear when the responses of this subject have been eliminated. Here, while the same levelling off of the curve for shape occurs after the third series, with a crossing of the curves on the seventh series, yet in general, in addition to its more rapid initial rise, the curve for shape remains above that for color. Columns IV and V of Table V and the curves of Fig. 4 show the rates of learning of the names of the 14 figures of the learning series and of the 2 figures which occurred only in the recognition series, respectively. There are no correct responses to the unlearned figures until the fourth series of the first day, but thereafter the curve for these figures rises rapidly, following the general form of the curve 20 LANGUAGE MONOGRAPH NO. I, 1925 TABLE V EXPERIMENT I Column I : the percentage of responses correct as to both shape and color (inclu- sive of the responses to the 2 unlearned figures). Column II: the percentage of responses correct as to shape. Column III : the percentage of responses correct as to color. Column IV: the percentage of correct responses to the 14 learned figures. Column V : the percentage of correct responses to the 2 unlearned figures. DAY SERIES both shape color learned unlearned I I III IV V Ist I I 16 7 I fe) 2 TZ 43 edi 14 fe) 3 2 50 24 14 ) 4 20 50 34 21 7 5 24 st 41 27 7 6 38 57 56 41 21 7 61 70 77 63 43 8 56 69 76 56 57 2nd I 48 66 62 52 21 2 69 ah 81 71 50 3 69 78 89 70 57 4 80 84 94 82 71 5 82 385 96 84 71 4th I 79 84 92 83 57 86 87 98 86 86 3 88 go 98 go 78 8th I 82 85 95 83 78 88 90 96 89 78 16th I 80 88 92 83 64 2 88 89 99 go 78 32nd I 82 87 94 85 64 2 85 89 96 86 78 62nd I 82 86 94 85 64 2 88 88 97 88 86 for the learned figures, and reaches its maximum on the first series of the fourth day. It then falls off again, but as the table shows, this maximum is again reached on the 62nd day. 21 ESPER, ASSOCIATIVE INTERFERENCE es a1 3 Yt fap pve <9 “7 099 v ib: = tb" 09 229s US. Et ae ov Selio PUIUIBIT] JO SU0I}1QS Oy §239AA09 BAWUGS °° sor ere yrr4a0e> J0l09 —— yoauae> WRYUS Pu s0jo7 43°Q Ge yurawaraadXg e S14 Bese TETAS © Sow ese as a“ eh ee 0 on °o oe? s PY hs Lae / C\ Assam nd tee 209 a en aa we “a a é ee ~ wemncee aes “ OF 09 03 PERCENTAGE OF CorRRECT RESPONSES I, 1925 LANGUAGE MONOGRAPH NO. 22 qorsaoa adeug "rn / a 4924109 40909 —— a wo poalans a0 Josasuodseas burpaqoxs Aime ; : oF Ky wy J peauredyg A ! mA : S “> f B & a) if fy gta / fee Ve ; S a SAS / ae | “er am lS ; “ = oa eS rod “y ue 03 ae 2 Ye gens \ ma soaks ® ne me ey wage pion on 9 9 ESPER, ASSOCIATIVE INTERFERENCE y 8 4a 79 0 19 sh th ob °7 sesuodsa4 JIIs40 VS ‘$34nb9} pouavaqzun, Jo vazen}2xa saslodsea Joe4110 ‘—T ys sursedX ‘1 by Ra ere yr .,! ~:) ’ Q * oY | | A wecreece - oi a? SUE REO RRANESR SS ’ y> *e, Se. 2° *, ° bt 32? hep pug 6Cyl 0 ieee $2IAIC Burusesy jo sudrqrqzedoy ee “ee tse eKe ews? a Px Lag ¢ 48 Pike) ~Y » rd ” »» \ ne , x4 =: A. ro Re » ue P) . > iz.) 7% Gq hep 3st 8 4% 0v ov 08 PERCENTAGE OF CoRRECT RESPONSES 24 LANGUAGE MONOGRAPH NO. I, 1925 TABLE VI EXPERIMENT II. . Column I : the percentage of responses correct as to both shape and color (inclu- sive of the responses to the 2 unlearned figures). Column II : the percentage of responses correct as to shape. Column III : the percentage of responses correct as to color. Column IV : the percentage of correct responses to the 14 learned figures. Column V : the percentage of correct responses to the 2 unlearned figures. DAY SERIES both shape color learned unlearned I II Il IV V Ist I 8 19 16 9 O 2 8 16 20 9 o 3 8 27 23 9 f°) 4 13 35 22 T5 Oo 5 16 37 28 19 fe) 6 22 50 37 2 o 7 17 44 34 19 6 8 14 29 29 16 fe) 2nd I 20 28 31 at 6 2 25 48 41 29 o 3 30 59 4l 34 6 4 34 39 39 39 v 5 34 63 41 39 eS) 4th I 27 48 34 29 6 2 50 7° 51 54 19 3 48 66 49 54 6 8th I 38 56 43 43 2 52 70 58 59 6 16th I 4I 61 47 47 oO 2 48 68 54 54 6 32nd I 44 65 53 49 6 2 55 72 60 61 12 62nd I 46 62 53 53 re) 2 50 62 55 54 19 EXPERIMENT II. Table VI and Figures 5 and 6 give the results for Experiment II. Columns I, II, and III of Table VI and the curves of Fig. 5 show the rate of learning of the complete names of the figures and of the sound- sequences corresponding to shape and, to color respectively. The three 25 ESPER, ASSOCIATIVE INTERFERENCE “a1 8 udp hep pur Rep ist Ne 1 ee a eS 8 E Pl ee eel el hele Guyusvay fo 5407177 2499 “oO yP24409 adeus ete PI ae 2? £8 P4409 419799 rere saa oen.- 00° 3224403 sdous} 4 pur 40702 4304 rr 4-¢ Bay 08 PERCENTAGE OF Correct RESPONSES LANGUAGE MONOGRAPH NO. I, 1925 26 Fie senile an ©} sesucdse4 qo21409 ies ‘gs panbay peusvayun so} AISA JIxa‘sesucdses 4P25409 Tr dx 4°98 Rep 3s} V 0b PERCENTAGE OF CORRECT RESPONSES ESPER, ASSOCIATIVE INTERFERENCE 27 TABLE VII EXPERIMENT III. The percentage of correct responses to the 14 figures of the learning series. SERIES Ist day 2nd day 4th day 8th day 16th day 32nd day 62nd day ro) I 19 36 55 51 45 47 rs 7 35 57 66 68 62 48 3 8 46 56 4 13 49 5 15 yt 6 19 7 22 8 23 curves are, as in Experiment I, roughly similar in configuration. As in Experiment I, the rate of learning of shape is more rapid than that of color, but the difference between these two curves is more pro- nounced than in Experiment I, as would be expected from the differ- ence in the material. In Experiment II, the sound-sequences corre- sponding to color were rendered more difficult of learning by the fact that they did not constitute natural syllables. Columns IV and V of Table VI and the curves of Fig. 6 show the rates of learningofthe 14 learned and the 2 unlearned figures respect- ively. The correct responses to the unlearned figures reach a maximum of 19 percent on the fourth day; this maximum is not again reached until the 62nd day. ExpERIMENT III. Table VII and the curve in Fig. 7 show the rate of learning of the names of the 14 figures of the learning series in Experiment III. In this experiment we cannot speak of correct responses to the two figures which occurred only in the recognition series, since the lin- guistic material provides no correlation between the visual and the auditory stimuli, and hence no phonetic elements corresponding to shape and color. For the same reason, we cannot analyse the re- sponses as being correct in color or in shape. The curves of learning of the 14 learned figures of each experiment may be compared in Fig. 7. If we compare the speed and extent of learning in the three experiments, Experiment I falls in one class, as being much more rapidly and completely learned, and Experiments II and III fall together in another class with respect to the general level of their curves and the lower degree of learning attained. If however we compare the shapes of the curves, we get a different classification: I, 1925 LANGUAGE MONOGRAPH NO. 28 1g 9g 79 09 ua . hep put 96 1$ sb ty OP 4 Fos RY — PU ee Ul be hes $0742G buquaesy fo Suo Pyede yY TITY ew qaedx yw" dT qs saad ys = T qpomnadxy 1,624 hep ST S74 0z Ov OO}: PERCENTAGE OF CORRECT RESPONSES ESPER, ASSOCIATIVE INTERFERENCE 29 The curve of Experiment III is smoother ; those of Experiments I and II more irregular. This difference reflects the difference in the mate- rials of the three experiments. In Experiment III, we taught dissim- ilar, uncorrelated names for the figures; learning could accordingly proceed only by the development of separate specific habits for each of the 14 figures. In other words, we are dealing here with 14 sepa- rate and distinct elements; we should expect the addition of one after another of these to give a gradually rising curve. In Experiments I and IJ, on the other hand, we taught linguistic systems exactly cor- related with the two variable factors of the visual stimuli. The learn- ing was therefore not of isolated elements but of categories, and hence was of a type which we should expect to be characterized by alternating plateaus and sharp rises. The difference in the rate and degree of learning between Experiments I and II is of course to be explained by the wide deviation of the material of the latter from English speech habits. RESUME OF RATES OF LEARNING IN ExpERIMENTS I, II, anp III. 1. ExperimentI showed the most rapid and complete learning ; Experiment II was learned somewhat less rapidly and thoroughly then Experiment III. 2. Learning in Experiments I and II was characterized by alternate plateaus and sharp rises ; in Experiment III, learning proceeded at a more even rate. 3. The retention of the material learned was greater in Experi- ments I and II than in Experiment II]. Of the maximum number of correct responses given on the previous days, the following percent- ages were given correctly on the 62nd day, 30 days after the last pre- vious series : Experiment I, 98 percent ; Experiment II, 90 percent ; Experiment III, 71 percent. 4. Thelearning of the two figures which occurred only in the recog- nition series proceeded at about the same rate as the learning of the other figures in Experiment I, reaching a maximum of 86 percent ot correct responses on the fourth day, with no loss in retention in the 30 days which elapsed between the 32nd and the 62nd days. In Experi- ment II these two figures were only slightly learned; the maximum of correct responses was 19 percent, which was however also retained on the 62nd day. 5- In both Experiment I and Experiment II, the sound-sequences corresponding to shape were learned more rapidly and completely than those corresponding to color. LANGUAGE MONOGRAPH NO. I, 1925 30 ‘poxleys ov sandy , pauivayuy , ‘(esuodsa1 1) se = J ‘(suoreyuasaid ]ej01) sasuodsa1 jo sJaquinu aqissod ]v103 Jo saSequasied ut passaidxa are saruanbai.y *puodsais0s Aayi yoy 0} samy oy pue sasuodsas ay ‘mor yeyozt0y Jaddn { 1009 pue adeys sy3 pue saindy JO sawieU SayvIIPUT Io] UO UUUINOD [vOTII A *‘] LNAWIadxXy NI SASNOdSHY TVNOLLNHANOD JO NOLLOAGIUISIGA FHT, 69 I L I 9 tT: Kv doylo1 I v9 «Ot I L Ab z ba L ¢ qr doyamof t I Lo I I L ¥ $ L z gb doyzaa E L 9 1$ ar bs I L as 9 = yr doyseu , eI Bee Rees Vy Speen I VE in eed xe Baplor I € a v9 z z t g qé sapumof C4 t L ¢ v 1$ I g 45 v zt I Hf dap2om I a i zt $ Sg9=e tT € L ¢ we gapseu $ - z I 1$ S z I ZI I z Take Mezlor, v L te L 2. t z S . qz@ avgumof z I Ge I 99 I z $ T. Hz MPIIIM L T Vv ¢ <3 I9 L v Wz AMPQSeU € z v I z LZ I z aie Ciyfos ae £ ¢ L I ¢ vL qt (irumof E I z 9 S S 6ST oe OS (ryy20a. rT L z L z z z 79) OY {it[seu doyumo| Ciyseu doy2om ,doyseu soplos Zapumo{ Soppom Sapseu AMALQIOM AvQSeU Ciyfos Citjuaof Criyppaan TASS TST ESPER, ASSOCIATIVE INTERFERENCE 31 TABLE VII a The conventional responses of Experiment I classified according to their color cate- gories. The vertical column on the left indicates the color category of the stimuli ; the horizontal row at the top, the color category of the responses. Frequencies are expressed in percentages of total number of presentations. nas- wet- fown- roj- (red) (green) (blue) (yellow) nas- (red) 71 2 5 3 wec- (green) 3 74 5 5 fown- (blue) 2 3 78 4 roj- (yellow) 2 4 3 79 TABLE VIII 3 The conventional responses of Experiment I classified according to their shape categories. The vertical column on the left indicates the shape category of the stim- uli ; the horizontal row at the top, the shape category of the responses. The num- bers of the shapes are indicated in parentheses (see Fig. 1). Frequencies are expressed in percentages of total number of presentations. -lin -caw -deg -kop (1) (2) (3) (4) -lij) (1) 76 5 3 4 -caw (2) 10 70 4 5 -deg (3) 7 4 73 9 -kop (4) 6 3 8 74 B. ANALYSIS OF THE RESPONSES IN EXPERIMENT I. Table VIII shows the distribution of the conventional responses of all subjects and for all the recognition series of Experiment I taken together. By ‘conventional’ responses are meant those responses which conformed to the sound-sequences taught in the learning series. The responses of Table VIII are classified according to their color category in Table VIIIa, and according to their shape category in Table VIIIb. Table VIIIa shows that there is no strong tendency for any one color category to be confused with a specific other color category. In Table VIIIb, however, we find a rather marked tendency for shape No. 2 to be confused with shape No. 1, and for shapes Nos. 3 and 4 to be confused with each other. It is noteworthy that the omission of two combinations from the learning series does not seem to have influenced the degree of learning of the categories to which these combinations belonged. Thus, a red and a yellow figure were omitted ; but the yellow category shows the highest percentage (79) of correct responses. Of the shapes, a No. 2 and a No. 4 were omitted, but the No. 4 category shows the second highest. percentage (74) of correct responses. RK 32 LANGUAGE MONOGRAPH NO. I, 1925 TABLE IX Tue ASSOCIATIVE VARIANTS IN EXPERIMENT I. Column I : the names of the stimuli, with the shape and color indicated in paren- theses. Column II : the responses showing associative interference; in parentheses, the lower-case letters indicate the subjects who gave these responses, and the numerals following these letters indicate the number of occurrences. Column III : the names to which the interference is most probably to be ascribed, and in parentheses the shape and color to which these names correspond. I II Ill Stimulus Variant Associate naslif} (1R) nasdel) (Str) nasdeg (3R) » » nojli} (sr) rojlin (1Y) fownlit) (1B) Jownslif} (st) nasli) (1R) rojlig (1Y) rojslij. (rr) nasli) (1R) nascaw (2R) neccaw (tI) wettaw (2G) wettaw (2G) necctaw (tI) nascaw (2R) *rojcaw (2Y) cit (v1) rojlin. (1Y) nasdeg (3R) nasde!) = (s1) nasli) (1R) » » nojsdeg (ri, si) rojdeg (3Y) » » nojdeg (11) » » wetdeg (3G) natdeg (x1) nasdeg (3R) fowndeg (3B) rowndeg (vt) rojdeg (3Y) » » fojdeg (x1) » » *naskop (4R) nojkop (sr) rojkop (4Y) In Table VIII we find few cases of a strong tendency to react to any figure with the name of any specific other figure. wectaw (2G) shows a tendency to be confused chiefly with weélin (1G) ; fowndeg (3B) with fownlin (1B) ; rojdeg GY) with rojkop (4Y); Jownkop (4B) with fown- deg (3B). Of the two ‘unlearned’ figures, *rojcaw (2Y) is most often confused with rojlia (1Y); the responses to *naskop (4R) are more scattered. Table IX indicates the responses in Experiment I which showed associative interference (column II), together with the figures to which these responses were given (column I) and the associated names to which the interference is most probably to be ascribed (column III). It will be seen at once that the number of such variants is very small, and that no example occurs more than twice. Most of the examples show interference in the syllable correspond- ing to color. Thus, mas- shows the influence of roj- five times (two subjects), appearing as nojs- twice and as noj- three times ; nas- and ESPER, ASSOCIATIVE INTERFERENCE 33 wec- show a reciprocal interference in the forms neccaw, naideg (two subjects) ; the responses rowndeg and fojdeg show two different results of an association of fowndeg with rojdeg (two subjects) ; there are two examples of the carrying over of the medial s of masliv to other forms, fownslin and rojsliy. In only three cases do the syllables correspond ing to shape show interference ; masdew occurs as a response to both nasliy and nasdeg; civ, which occurs as a response to the ‘ unlearned’ figure 2Y (*rojfaw) shows an interference between -caw and -lin. In general, both the distribution of the conventional responses (Table VIII) and the small number of analogic variants in Experiment I show that the linguistic system remained relatively stable. As was shown in the section on the rate of learning, the material of this experiment was learned very rapidly. Most of the variants of Table IX occurred in the first day’s work ; four occurred on the second day, and one on the sixteenth day. The correct responses therefore soon became fixed ; and the possibility of the development of further lin- guistic categories in the directions suggested by the variant responses of Table IX was obviated by repeated learning. C. ANALYSIS OF THE RESPONSES IN EXPERIMENT II. Table X shows the distribution of the conventional responses in Experiment II. These responses are classified according to the color categories to which they belong in Table Xa, and according to the shape categories in Table Xb. Table Xa shows that there is a tendency to react to blue figures with the sound-sequence corresponding to yellow. Table Xb shows a tendency to confuse shapes Nos. 1 and 2, and Nos. 3 and 4. In Table X we find that in each shape category the red figure shows the highest percentage of correct responses, except in the cate- gory of shape No. 4, where 4R was one of the ‘ unlearned’ figures ; here 4Y shows the highest percentage of correct responses. In the color categories, we find that in the red and the green categories, shape No. 3 shows the highest percentage of correct responses, but in the blue and the yellow categories it is shape No. 4. An examination of the responses incorrect as to either shape or color will reveal certain general tendencies within each category. If we consider those responses to the figures within each shape category which are incorrect as to color we find that : a)in the category of shape No. 1, nulgen (1R) is the most frequent variant color response to nugdet (1G), nuzgub (1B) and numbow (1Y). * LANGUAGE MONOGRAPH NO. I, 1925 34 wv ~ mee HO RH ROHR HH TOO +N AN on tah RnR Agnhism = HH yopsia ‘T] LNAWIadXY NI SHSNOdSHY TVNOILNHANOD JO NOILAGIULSIG SH], eT RF $ U3] aBaHhagqgaononHnaed Land a Avoquuad <6 Si qnézod cal Ha aRH wah +H sha es yapsad ma Htnonnankh aan ™~ BH ‘E I 4 LE LI I ry; ¥ I Res de Be ie) Cc a lon ‘XY ATEV], ‘polivys oie sainsy ,pausesjug , ‘(asuodsas 1) d0v13 = JF ‘(suonequasaid [vj0}) sasuodsa1 jo saquinu aqissod ]e101 jo saSeyuaoied ur passoidxa are sapuanbe1ry *puodseisios Aayi yoy 0} sainsy ay pue sasuodsas ayn ‘aor jeyuozt1oy Jaddn ‘ 1oyoo puv adeys moy3 pur soan3y jo sate sayeorput yay UO UUIN]OD TBO A $ Aoquinu — monk hHHRna Ht a qnsznu BO tH aAMaR OHM ahh AOQUUITA qnszim JOpSIM ussTIM , Moquiad qnszad japsad uaspod Aoquilop , qn8zlop jopslop usstlop moquinu qnsznu ‘jopsnu u2sinu ESPER, ASSOCIATIVE INTERFERENCE Wo mM TABLE Xa The conventional responses of Experiment II classified according to their color cate- gories. The vertical column on the left indicates the color category of the stimuli ; the horizontal row atthe top, the color category of the responses. Frequencies are ex- pressed in percentages of total number of presentations. -lgen -gdet -zgub -mbow (red) (green) (blue) (yellow) -Igen (red) 52 5 3 6 -gdet (green) 5 36 4 5 -zgub (blue) 4 3 34 7 -mbow (yellow) 6 4 2 39 TABLE Xs The conventional responses of Experiment II classified according to their shape category. The vertical column on the left indicates the shape category of the stimuli; the horizontal row at the top, the shape category of the responses. The numbers ot the shapes are indicated in parentheses (see Fig. 1). Frequencies are expressed in percentages of total number of presentations. nu- doj- pe- wi- @) (2) G3) (4) nu- (1) 50 8 6 5 doj- (2) 10 48 6 7 pe (3) 4 4 47 9 wi- (4) 2 4 10 57 b) in the category of shape No. 2, dojgdet (2G) is the most frequent variant color response to doj/gen (2R) and dojzgub (2B) ; the responses to *dojmbow (2Y) are more scattered, but dojgdet is among the most frequent. c) in the category of shape No. 3, pembow (3Y) is the most frequent variant color response to pegdet (3G) and pexgub (3B) and the next most frequent response to pelgen (3R). d) in the category of shape No. 4, wimbow (4Y) is the most fre- quent variant color response to *wilgen (4R) and wizgub (4B) ; wim- bow and wizgub (4B) are the two most frequent responses to wigdef (4G). Similarly, if we consider those responses to the figures within each color category which are incorrect as to shape we find that : a) in the category red, pelgen (3R) is the most frequent variant shape response to nulgen (1R), dojlgen (2R), and *wilgen (4R). b) in the category green, dojgdet (2G) is the most frequent variant shape response to nugdet (1G), pegdet (3G). and wigdet (4G). LANGUAGE MONOGRAPH NO. I, 1925 TABLE XI Tue ASSOCIATIVE VARIANTS IN EXPERIMENT II. Explanation as under Table IX, Experiment I. I Stimulus nulgen (1R) » » » » » » » » nuzgub (1B) » » » » dojlgen (2R) » » dojgdet (2G) » » » » » dojzgub (2B) » » pelgen (3R) Il Variant nulget (f1) nulget (ar) nulden (f1, gr) nuzgen (e€3) nojlgen (c2) nulget (ar, f3) nuldet (f2, g2, h8) nuzdet (e4, gi, h4) nojgdet (c4, di, f1) dugdet (d1) nulgub (f1, h2) nugeub (f1) nojzgub (d3) nulbow (f2, g3, h1o) nulgow (f3) nugdow (d1) nuzbow (h3) dojzgen (cI, e1) delgen (cr) dojldet (f5, h4) dojlget (c1, f2) dojzdet (e5, 27, h4) nojgdet (dr) dojlgub (f1, h2) dojggub (e3) dojgzgub (gr) nojzgub (dr) duzgub (f1) dojlbow (e1, fg, k8) dojlgow (f2) dojgbow (e2) dojzbow (er, h2) delgen (gr) Ill Associate nugdet (1G) » » » » nuzgub (1B dojlgen (2R) nulgen (1R) » » nuzgub (1B) dojgdet (2G) » » nulgen (1R) nugdet (1G) dojzgub (2B) nulgen (1R) » » nugdet (1G) nuzgub (1B) dojzgub (2B) pelgen (3R) dojlgen (2R) » » dojzgub (2B) nugdet (1G) doj!gen (2R) dojgdet (2G) » » nuzgub (1B) » » dojlgen (2R) » » dojgdet (2G) dojzgub (2B) dojlgen (2R) ESPER, ASSOCIATIVE INTERFERENCE TABLE XI, continued. I II III Stimulus Variant Associate pegdet (3G) pelget (fr, hr) pelgen (3R) » » pezget (h1) pezgub (3B) » » » » pezgub (3B) pembow (3Y) pezdet (f1, h7) pugdet (f1) pelgub (er) pelgow (hr) pelbow (h2) pezbow (f6, h4) » » nugdet (1G) pelgen (3R) pelgen (3R) » » pezgub (3B) » » dembow (c1) dojmbow (2Y) *wilgen (4R) wiggen (cI) wigdet (4G) » » wizgen (f14) wizgub (4B) » » wimgen (e4, h2) wimbow (4Y) wigdet (4G) wizget (hr) wizgub (4B) » » wizdet (a1) » » » » wimdet (e9, g2) wimbow(4Y) wizgub (4B) wiggub (h2) wigdet (4G) » » wigzgub (dr) » » wimbow (4Y) wigdow (f1) wigdet (4G) wizgow (f1) wizbow (f4) pimbow (al, g)) wizgub (4B) » » pembow (3Y) 37 c) in the category blue, wizgub (4B) is the most frequent variant shape response to nuzgub (1B) and pezgub (3B). The most frequent response to dojzgub (2B) is nuggub (1B) ; the next most frequent is again wizgub. d) in the category yellow, wimbow (4Y) isthe most frequent variant shape response to numbow (1Y), “dojmbow (2Y), and pembow (3Y). We should expect, as a result of the above analysis, that any ana- logic changes in the linguistic system of Experiment II would show definite tendencies within each category. Table XI indicates the responses in Experiment II which showed associative interference. Jt will at once be evident that these variants are much more numerous than in Experiment I, and that many of them were given by two or three subjects and by each subject a num- ber of times in spite of repeated learning. 38 LANGUAGE MONOGRAPH NO. I, 1925 In the category of shape No. 1, the influence of nulgen (1R) is most marked ; thus we have : for nugdet — nulget, nuldet ; for nuzgub — nulgub ; for numbow — nulbow, nulgow. The influence of nuzgub, nug- det is also seen. Association of shape No. 1 (nu-) with shape No. 2 doj-) is seen in nojlgen (for nulgen) ; nojgdet, dugdet (for nugdet) ; nojzgub (for nuzgub). In the category of shape No. 2, dojlgen (2R) interferes most strong- ly with the other forms, although the influence of dojzgub (2B) also is evident. Thus we have : for dojgdet — dojldet, dojlget ; for dojzgub — dojlgub ; for *“dojmbow — dojlbow, dojlgow. Association of shape No. 2 (doj-) with shape No. 1 (mu-) is seen in nojgdet (for dojgdet) ; nojzgub, duzgub (for dojzgub). In the category of shape No. 3, both pelgen (3R) and pezgub (3B) influence the other forms ; so that we have : pelget : pezget, pezdet (for pegdet) ; pelgub (for pexgub) ; pelgow, pelbow : pexbow (for pembow). An association of shape No. 3 (fe-) with shape No. 1 (nu-) is seen in pugdet (for pegdet) ; and with shape No. 2 (doj-) in delgen (for pelgen) and dembow (for pembow). In the category of shape No. 4, both wimbow (4Y) and wizgub (4B) influence the other forms. Thus we find : wimgen : wizgen (for *wil- gen); wimdet : wizget, wizdet (for wigdet) ; wizgow, wizbow (for wimbow). The influence of wigdet (4G) is seen in wiggub, wigzgub (for wizgub). An association of shape No. 4 (wi-) with shape No. 3 (fe-) is seen in pimbow (for wimbow). In general, we find a tendency is modify the non-English syllable divisions of Experiment II in accordance with English speech habits. In the categories of shapes Nos. 1, 2, and 3, this tendency takes the direction of extending the natural syllables nul-, dojl-, and pel-, occur- ring in nulgen, dojlgen, pelgen, respectively, to the other words. These natural syllables thus become semantic units corresponding to shape, while the syllables -gen, -det, -gub, -bow similarly become semantic units corresponding to color. But nul-, dojl-, pel-, tend to carry with them the following g of nulgen, etc., as for example in the variants nulgow, dojlget ; this is no doubt favored by the occurrence of medial g also in nuzgub, dojzgub, etc. In the category of shape No. 4, *wilgen (4R), being one of the unlearned figures, could not serve as a pattern. Here the syllables wim- (wimbow) or wiz- (wizgub) were extended to the other words. These tendencies are seen most clearly in the re- sponses to the unlearned figures, where a ‘ proportional analogy’ was called for. Thus *dojmbow appears as dojlbow ; *wilgen as wimgen, wizgen. ESPER, ASSOCIATIVE INTERFERENCE 39 There appears to be a rather strong tendency to assimilate the syllable for shape No. 1 (nu-) to that for shape No. 2 (doj-), and there is some indication of a reciprocal influence between the two. No definite tend- ency toward associative modification of the syllables for shapes No. 3 (pe-) and No. 4 (wi-) appears. We find, then, in the variant responses of Experiment II certain general tendencies ; but it is clear that the processes involved have not gone far enough to permit us to conjecture in what manner the lin- guistic system would reach an equilibrium. Whether the patterns sug- gested by the analysis of the conventional responses (pp. 33 ff.) would eventually appear in this system cannot be determined from the mate- rial at hand. D. ANALYSIS OF THE RESPONSES IN EXPERIMENT III. Table XII shows the distribution of the conventional responses in Experiment III. Table XIla classifies these responses according to the color of the figures to which they correspond, and Table XIIb, according to the shape of the figures. The analysis in the latter two Tables is not of the same kind as in the corresponding Tables of Experiments I and II. In Experiment III the linguistic material is not organized into form- al categories, so that there are no sound-sequences corresponding specifically to red, green, shape No. 1, etc. Tables XIla and XIIb only indicate how many times the names of, e.g., red figures were given as responses to red, green, blue, and yellow figures ; and how many times the names of figures of, e.g., shape No. I were given as re- sponses to figures of shapes Nos. 1, 2, 3, 4. Table XIIa shows a tendency to confuse the names of red, green, and yellow figures with those of blue figures, and the names of blue figures with those of green figures. Table XIIb shows a tendency to confuse the names of figures of shape No. 2 with those of shape No. 1, and those of shape No. 3 with shape No. 4. In Table XII, we find that the figures of shape No. 1 were confused chiefly with figures of the same shape ; thus, vit (1G) and zib (1B) are confused with each other, and tekbem (1Y) is confused with vit (1G). éslhib (1R) shows little confusion. The figures of shape No. 2 were confused partly with figures of the same shape, partly with figures of shape No. 1 of the same color. Thus, nuzgub (2R) is confused with éelhib (xR) but also with sotmin LANGUAGE MONOGRAPH NO. I, 1925 40 ‘T]] LNANINadxX| 1$ I ¢ t (A it 9 € I SV it L t I ¢¢ c t <3 t 9 Zt I if Zz BR mt ONO SO ecey : "paxrieys ore somsy , pausvatug , ‘(asuodsas 1) aor] = J, *(suoneyuasoid e103) sosuodsai jo s9quunu ayqissod [e101 jo so8viusoi1ad ut passeadxa sorouanbaiq *puodsasios Asy yotyAr 02 saansy ayy pue sasuodsaz oui ‘mor [eIUOztoy Jaddn { 1009 pue adeys soy) pur sain8y JO sowWeuU soyv2IpUl Ya] UO ULUINTOS [edIIIA NI SHSNOdSaUY TIVNOLLNAANOD JO NOILAAMLSIG AAT, oN eA co DOE se Mo 100 aM ao es PAP sopumol Ke TX Ap ED £ qb, (a I 9 ey c v tv ¢ ¢ AE oT; eK rOX4 oI = 3. a 5S =) ATaV J, L ah tH ahah qnsznu tuagyeo} hi? HHH R yA Ox o UT Jad on q2) wniguol eles BML Sopumol esee * (jruuj0s ual]ny qnsznu waqyoy qiz WA qu? ESPER, ASSOCIATIVE INTERFERENCE 41 TABLE XIIa. The conventional responses of Experiment III classified according to color. The vert- ical column on the left indicates the color of the stimuli ; the horizontal row at the top indicates the color to which the responses corresponded according to the series as tdught. Frequencies are expressed in percentages of total presentations. red green blue yellow red 1 3 4 2 green 2 47 7 3 blue I 5 35 I yellow 2 3 5 32 TABLE XII The conventional responses of Experiment III classified according to shape. The vertical column on the left indicates the shape of the stimuli (see Fig. 1); the horizon- tal row at the topindicates the shapes to which the responses corresponded according to the series as taught. Frequencies are expressed in percentages of total presenta- tions. T = trace (less than 1 percent). shape 1 shape 2 shape 3. shape 4 shape No. 1 42 I I ay shape No. 2 7 27 2 oO shape No. 3 I 2 46 4 shape No. 4 3 2 3 40 (2B) ; kulren (2G) is confused with sotmin (2B); sotmin(2B) is confused with nuzgub (2R)and kulren (2G). The responses to the ‘ unlearned’ figure 2Y show a strong association with rY (tekbem), and a less strong association with 2B (sotmin). The figures of shape No. 3 were confused chiefly with figures of the same shape. Thus, jowndeg (3R) is confused with awa (3G) and faza (3B) ; awa(3G) is confused with faza (3B) and jowndeg (3R), but also with feb (4G) ; faza (3B) is confused with awa (3G); jonbim (3Y) is confused with awa (3G). Thus we find jowndeg, faza, and jondim each confused chiefly with awa. The figures of shape No. 4 were confused partly with figures of the same shape, partly with figures of shape No. 3 (to a less extent with shape No. 1). Thus, the responses to the ‘unlearned’ figure 4R show an association with 4B (wuc) ; feb (4G) is confused with pel (4Y) but also with vit (1G); wué(4B)is confused with pel (4Y) and awa (3G); pel (4Y) is confused with jondim (3Y). Table XIII indicates the examples of associative interference in Experiment III. The number of these is very small, and the individual forms are neither frequent for a given subject nor common to a number of subjects, with two exceptions. 42 LANGUAGE MONOGRAPH NO. I, 1925 TaBLe XIII Tue AssociIATIVE VARIANTS OF EXPERIMENT III. Explanation as under Table IX, Experiment I. ‘ Unlearned’ figures are starred. I I Ill Stimulus Variant Associate vit (1G) zit (i2) zib (1B) zib (1B) vib (j1,pr). vit (1G) tekbem GY) éekbem (it) éelhib CR) Soe *(2Y) fim (or) jonfim (3Y) » sotéil) Gri sotmil) (2B) -jonfim (3Y) » tekpim (o1) tekbem (1Y) -jon#im (3Y) jowndeg (3R) jondeg — (j,k, q3, p9) jonfim (3Y) faga (3B) aa (qn) awa (3G) jonfim (3Y) jownlim = (j1) jowndeg (3R) » » jonfem (p2) tekbem (1Y) A reciprocal interference of vit (1G) and zib (1B) is seen in the forms zit and vib. The response 6im to the ‘ unlearned’ figure 2Y illustrates a process which it had been expected would find more extensive expression in this experiment; namely, the development of new semantic units from the dissyllabic names of objects which exhi- bit two variable attributes. Here we see the possibility of 6im coming to correspond to the attribute yellow. In general, the examples of table XIII show a tendency toward the assimilation of the forms within each shape category to each other. V. SUMMARY The problem was that of developing a technique for investigating (1) to what extent and in what manner associative interferences occur in artificial linguistic material, and (2) to what extent and in what manner such interferences tend toward the development of linguistic categories. For this purpose, nonsense figures of four shapes and tour colors were presented visually, and nonsense names for the figures were presented orally. A geometrically decreasing number of repeti- tions was distributed over a geometrical sequence of days, beginning with 32 repetitions on the first day, and ending with 1 repetition on the thirty-second day. A recognition series was given after every four repetitions of the learning series; it was also given (twice) thirty days after the last learning series (62nd day). Three degrees of correlation between the figures and names were provided in different experiments with different sets of subjects : Experiment I. — The names consisted of two elements which were exactly correlated with the factors of color and shape in the figures, and which conformed to English speech habits in word order and syllable division (e.g., nas-liy = red + shape 1). Under these condi- tions : 1. Learning was very rapid and there was practically no forgetting 30 days after the last learning. 2. Sound-sequences corresponding to shape were learned more rapid- ly than sound-sequences corresponding to color. 3. The names of two figures omitted from the learning series but presented in the recognition series were learned (by analogy with the other names and figures) at about the same rate as the names of the other figures. 4. There was a tendency to confuse shapes Nos. 1 and 2, and Nos 3 and 4. 5. The cases of associative interference were very few ; no variant form occurs more than once fora given subject, and only one is com- mon to two subjects. 6. No definite tendencies toward the development of further lin- guistic categories appear. 43 44 LANGUAGE MONOGRAPH NO. I, 1925 Experiment II. — The names consisted of two elements which were exactly correlated with the factors of color and shape in the figures, but which deviated from English speech habits in word order and syllable division (e.g., nu-lgen == shape 1 + red). Under these conditions: 1. Learning was very slow, and the material was incompletely learned, the maximum percentage of correct responses attained (for all subjects taken together) being 61; but of this maximum, 90 percent was retained 30 days after the last learning. 2. Sound-sequences corresponding to shape were learned more rapidly than sound-sequences corresponding to color. 3. The names of two figures omitted from the learning series but presented in the recognition series were learned but slightly (by ana- logy with the other figures and names). 4. There was a tendency to confuse shapes Nos. 1 and 2, and Nos.3 and 4; and within each shape and color category, to confuse the names of the figures with the name of a particular figure of the category. 5. There were numerous cases of associative interference; many of these variant forms were given repeatedly by the same subject and were common to two or three subjects. 6. There was a general tendency to modify the non-English syllable division in accordance with English speech habits. This tendency took the form of assimilating the words within each shape category tc a partic-' ular word of the category (e. g.,nu-gdet > nul-det, nu-mbow > nul-bow, after nu-lgen). 7. There was a tendency to assimilate the syllable for shape No. 1 to that for shape No. 2 (e.g., nulgen > nojlgen, after dojlgen). There would thus arise a further linguistic category which would include shapes Nos. 1 and 2, the difference between the two being marked by the initial consonant (e.g., nojlgen : dojlgen). , Experiment III. —The names showed no correlation whatever with the factors of color and shape of the figures, each figure having an entirely different name. Nine of the names were dissyllabic ; five were monosyllabic. Under these conditions : 1. Learning was very slow, and the material was incompletely learned, the maximum percentage of correct responses attained (for all subjects taken together) being 68; of this maximum, 71 percent was retained 30 days after last learning. 2. The responses to two figures omitted from the learning series but presented in the recognition series (for naming which the material ESPER, ASSOCIATIVE INTERFERENCE 45 of the experiment provided no analogies) were the names of figures either of the same shape but different color, or of the same color but different shape, the name of a particular figure being favored in each case. 3. There was a tendency to confuse shapes Nos. 1 and 2, and Nos. 3 and 4. The names of the figures of shapes Nos. 1 and 3 were howev- er confused chiefly with those of figures of the same shape but differ- ent colors. 4. There were very few cases of associative interference, and most of these were given only once by one subject. 5. The cases of associative interference show a slight tendency toward assimilation of the words of each shape category to each other. \ VI. CONCLUSION. The present investigation was designed as a beginning in the devel- opment of the optimum technique for the study of associative inter- ferences and the development of linguistic categories in artificial lin- guistic material. The results obtained during the learning period with three types of material suggest the following points : 1. The names of those figures which are confused with one another during learning tend to interfere with one another so as to produce mixtures of articulation. In other words, when a stimulus tends to evoke either verbal reaction a or verbal reaction b according to the sensorimotor conditions existing at a given moment, a fluctuation of the sensorimotor conditions during the process of articulation will result in a mixture of the two sequences of articulation. 2. Of the three types of experimental conditions investigated, the strongest tendencies toward associative interference were found when the linguistic material, although exactly correlated with the two factors of the visual stimuli, deviated from English habits of syllable division (Experiment II). 3. The names of figures of a given category tend to be confused with and assimilated to the name ofa particular figure of that category. This appears most clearly in Experiments II and II. 4. In further experimental work on this problem it will be neces- sary fo continue the investigation for a long period of time after com- plete learning of the material. Only trom the end-results of a long period of observatican can we draw conclusions as to the regularity and uniformity of the process of associative interference and the result- ant development of linguistic categories. While the material described under Experiment II (p. 15) gave the most extensive results during the learning period, it would seem that for alonger period of investi- gation material of the type of Experiment III (p. 17) would yield the least ambiguous results. APPENDIX THE VALUES OF THE PHONETIC SYMBOLS. The values of the phonetic symbols used in these experiments are those represented by the Middle Western pronunciation of the itali- cized letters in the following words : py pet S son €) Sat b bet Jf shun u_—spaut bean tip Zoszedl oO son d dip 3 azure von k kid h hat a father Bw *get Eeap aw house ¢ choke m man ow show } joke n not uw moon iG fat ) sing aj my Vv vat f “red ej way § thing Pepin oj boy 0 this Gre pet ij mean PROTAT BROTHERS, PRINTERS, MACON (FRANCE). — MCMXXV Ne RE OP ie Vee VPC yt sre sla int basi icoattaeeainin aneelateean eae ASG Sy AY TORRE, MONE ERP OER ETI \ . } rod , * es oo My “4 Li we Pot