Introduction to experimental education

Introduction to experimental education

3 1924 013 082 098

The original of tliis book is in the Cornell University Library.

There are no known copyright restrictions in the United States on the use of the text.

INTRODUCTION TO EXPERIMENTAL EDUCATION

Introduction

TO

ROBERT R. RUSK

M.A. (gLASC), B.A. (caNTAB.), PH.D. (jENa)

LONGMANS, GREEN, AND CO.,

39, PATERNOSTER ROW, LONDON

NEW YORK, BOMBAY, AND CALCUTTA

All rights reserved

Preface

THIS work seeks to make accessible in convenient form for English readers the main results of investigations in the new subject of Experimental Education. It is based on E. Meumann's "Vorlesungen zur Einflihrung in die experimentelle Padagogik." ' The material available since the publication of Meumann's work has been incorporated, and emphasis has throughout been laid on the results of English investigations. The publication of G. M. Whipple's " Manual of Mental and Physical Tests " ^ has enabled the writer to dispense with full descriptions and illustrations of apparatus. Appended to each chapter will be found a brief list of publications suitable for further reading : these references, which are in no sense exhaustive, are confined to English writings or works available in English, and indicate what may be regarded as most suitable and most readily accessible.

The thanks of the writer are due to Mr. J. R. Cameron, M.A., High School, Arbroath, for reading typescript and proofs.

" Engelmann, Leipzig, ist ed., 1907, 2 vols., pp. xviii + 555,467. " Warwick and York, Baltimore, 1910, pp. xiii + 534.

Contents

CHAP. PAGE

I. THE STANDPOINT OF EXPERIMENTAL EDUCATION I

II. THE METHODS OF EXPERIMENTAL EDUCATION . lO

Chapter I The Standpoint Of Experimental Education

THE most important stage in the development of a subject of study arrives when its data and conclusions can be presented in quantitative as well as in merely qualitative terms ; its claim to be recognised as a science can then no longer be disputed, for, as Herschel says, " numerical precision is the soul of science." The subject of Education is at length entering this definitely quantitative stage, and when the methods of investigation described as " Experimental Education" become generally applied, and the results universally accepted, Education will forthwith be recognised as an exact science.

That the traditional pedagogy cannot be termed scientific will generally be admitted ; and the words in which Hume described the metaphysical sciences of his day might not inaptly be applied to the condition of current Education : " Even the rabble without doors may judge from the noise and clamour which they hear that all goes not well within. There

2 1 is nothing which is not the subject of debate, and in which men of learning are not of contrary opinions. The most trivial question escapes not our controversy and in the most momentous we are not able to give any certain decisions. Disputes are multiplied as if everything were uncertain ; and these disputes are managed with the greatest warmth, as if everything were certain. Amidst all this bustle 'tis not reason which carries the prize, but eloquence ; and no man needs ever despair of gaining proselytes to the most extravagant hypothesis, who has art enough to represent it in any favourable colours. The victory is not gained by the men-at-arms, but by the trumpeters, drummers and musicians of the army."

The older pedagogy, when it did not rely wholly on experience without attaining to principles, accepted principles without verifying them by appeal to ..experience. Many schemes of education have, for example, assumed the " recapitulation " principle without attempting to prove the parallelism, or to determine the points of correspondence, between racial and individual development. The classical school of educationists likewise adopted the doctrine of formal training, which implied that the power derived from pursuing a certain course of study could be applied indifferently in any sphere of mental life ; but they omitted to determine to what extent a general improvement is secured by training a specific function, whether this general improvement is automatically transferred to other specific functions, or whether the improvement of the second

STANDPOINT 3

specific function could not be attained more economically by direct training.

The new pedagogy, however, claims to be scientific. It seeks to present its data in exact quantitative terms. It is inductive rather than deductive. Instead of following the old d priori methods, it adopts the sure method of experiment.

The claim that Education should be founded on experiment is not new. In the Appendix to " Practical Education " ' by Maria and R. L. Edgeworth (1798) there is quoted Mrs. Honora Edgeworth's opinion that the art of education should be considered as an experimental science, and that many authors of great abilities had mistaken their road by following theory instead of practice. Pestalozzi2 maintained that all branches of instruction demand analysis of their methods, and that the age at which each branch of instruction may be imparted to the child, should be exactly determined. Kant 3 also asserted that experimental schools must first be established before we can establish normal schools.

The recent advance in Education has only been made possible by the progress of such other sciences as anthropology, psychology, etc., which have helped to provide Experimental Education with its method. The way for this advance has also to some extent been paved by the results of the Child-Study movement, which, however, are themselves inadequate as a basis for the new method. Excellent studies of

■ First Edition (1798), vol. ii. p. 734.

" " How Gertrude Teaches," English trans., p. 126.

s " Kant on Education," trans, by Annette Churton, p. 21.

individual children have been made by Preyer, Perez, Darwin, Miss Shinn, and Dr. Dearborn, but these stop before school age — just when they would have been of most value to the educator. The other department of Child Study is mainly statistical, yet its results cannot always be regarded as scientific. In most Child-Study investigations the questions employed are ambiguous, owing to inadequate analysis, and the statistics are frequently vitiated by the factor of "selection," for example, the answers recorded are given mainly by those to whom the questions happen to appeal. It is sometimes assumed that these objections do not hold when the range of replies is very extensive ; but to this Dr. C. S. Myers says, " I want to protest as strongly as I can against the notion that any useful purpose can be served, so far as psychology is concerned, by collecting masses of psychological data with the help of an army of untrained observers. I have heard it confidently asserted that the gross errors inevitably arising from inaccuracies and inconsistencies of procedure among different observers cancel one another in the long run of such vast numbers of measurements. Nothing, I think, can be more dangerous or false than this idea that the untrustworthiness of crude methods diminishes as the number of observers increases. It involves the assumption that in the long run errors occur to an equal extent in opposite directions — a most unlikely hypothesis." '

' Myers, C. S., "The Pitfalls of Mental Tests"— a paper communicated to the British Association at the Sheffield Meeting, 1910.

STANDPOINT 5

The vagueness of the Child-Study methods can be illustrated by the variety of the views expressed on such a subject as the reasoning powers of children. One writer asserts, "Before a bright child can talk, he is already a reasoning human being, who will rapidly develop from his inner consciousness such a power of logic as passes understanding." ' King, in " The Psychology of Child Development," states that some investigators conclude, from a study in children's inferences and reasonings, that there is at the end of the ninth year a marked increase in the logical faculty.2 Mrs. Barnes, from her investigation on "The Historic Sense among Children," declares that the results of her test go to show that the number of inferences made rises decidedly at the age of twelve for boys and thirteen for girls ; 3 whereas another writer on " The Reasoning Powers of Children" maintains that a considerable portion of those who have received a good education and have reached the age of fifteen, fail to show anything but the germs of thinking of a logical kind.4 Such contradictory conclusions indicate that the methods employed are unscientific ; the results are at best mere generalisations, suggesting hypotheses which require verification by more exact methods. Unfortunately, however, they are frequently set forth in statistical form, thus acquiring an absoluteness to which they have no claim.

3 Studies in Education, vol. i. pp. 47-53.

The new Education consequently turns to Child Psychology rather than to Child Study for its methods. Experimental Education is not, however, to be regarded merely as applied psychology. It is without doubt an independent science ; for, although to some extent deriving its data from other sciences, it regards such data from its own special standpoint. In this respect it is best comparable with the science of geography, which, although dependent on astronomy, geology, etc., has nevertheless its own peculiar point of view. The distinguishing feature of Experimental Education is its practice of approaching all problems from the standpoint of the child of school age.

The subject includes the investigation of the physical and mental development of the child and the discovery and improvement of appropriate means of measuring such development. In addition to weighing and measuring the child. Experimental Education treats of sensory acuity and sensory discrimination, the types of observation, the various forms of memory, the nature of the child's mental imagery, and the part which such imagery plays in the child's thinking. It even attempts to estimate quantitatively a function so complex as general intelligence and one so subtle as suggestibility. The individual differences in children, disclosed by investigations in these directions, together with the extent to which intellectual and moral differences are dependent on the original nature of the child, and the limitations which such endowment sets to the work of education — all pro-

STANDPOINT 7

vide fruitful fields of inquiry. The most economical modes of learning, the fatigue involved in various forms of mental work, and the effect of school organisation on the child, form a special section ; and the psychological analysis of the efforts of the child in the various school subjects — reading, writing, etc., — is also a branch of Experimental Education of great practical importance."

Valuable as are the results already obtained, the justification for Experimental Education does not solely depend on these ; rather is it to be found in the questions being raised in spheres where formerly there was apparently nothing in dispute.

It may be asked how the new method will affect the position of the teacher. It will deliver him from the tyranny of tradition and the caprice of the faddist, and bring him under the servitude of his science, since the scientific worker must ever submit to the method of his subject. If it removes the domination of an arbitrary authority, it demands submission to a rational authority, one that can be questioned and tested.

Doubtless teachers will have but few opportunities of accomplishing original research work in Experimental Education. This requires a training which cannot form part of the ordinary professional course, and the time demanded by research work can hardly be given by one engaged in the routine duties of teacher. But, as Professor Dewey says,^

' Cf. Thorndike, Journal of Educational Psychology, vol. i.

pp. S-I2.

= " Educational Essays," p. 158,

" unless our laboratory results are to give us artificialties, mere scientific curiosities, they must be subjected to interpretation by gradual reapproximation to conditions of life. . . . Now the school, for psychological purposes, stands in many respects midway between the extreme simplifications of the laboratory and the confused complexities of ordinary life. Its conditions are those of life at large ; they are social and practical. But it approaches the laboratory in so far as the ends aimed at are reduced in number, are definite, and thus simplify the conditions ; and their psychological phase is uppermost, while in ordinary life these are secondary and swallowed up." It is in the school, then, that the laboratory results must be put to the test. " The task of reviewing them is clearly one of great delicacy," and here the teachers will be able to render invaluable assistance to the new science.

Experimental Education, it may be added, does not profess to settle all the problems of education. It does not, for example, attempt to determine the general aim of education. Such determination falls within the sphere of social science or social philosophy.i Professor Ward has said in his article on Educational Values : " The first thing the educator should be clear about is what he intends, what his end and aim is, or rather should be. To ascertain this ideal, he must turn not to psychology, but to life : it is a social and ethical, rather than a psychological problem." 2 It is, however, for Experimental

■ See Mackenzie's " Social Philosophy," ist ed., pp. 352-59. " Journal of Education, November, 1890.

STANDPOINT 9

Education to decide whether the aims are compatible with the child's nature and how these aims can best be attained. Experimental Education, then, although it does not comprehend the whole of Education, provides the empirical groundwork of the subject,

REFERENCES FOR FURTHER READING.

MuNSTERBERG, H. Psychology and the Teacher, oh. xi, xii.

Thorndike, E. L. Educational Psychology (ist edition), ch. xiv, XV. "The Contribution of Psychology to Education " (Journal of Educational Psychology, vol. i.

Spearman, C. " The Way to Develop Experimental Pedagogy " (Journal of Experimental Pedagogy, vol. i, pp. 1-3).

Myers, C. S. "The Pitfalls of Mental Tests" (British Medical Journal, January 28, 1911. No. 2613, pp. 195-97).

Green, J. A. " Experimental Psychology and Education " (The School World, March-June, 1909).

Rusk, R. R. " Experimental Education " (Journal of Education, December, 1910).

Chapter II The Methods Of Experimental Education

EXPERIMENTAL Education employs the methods common to all the exact sciences, namely, systematic observation and experiment : their advantages over mere casual observation and opinion do not require enumeration here.

Experimental Education consists largely of the application of the methods of Experimental Psychology to the school child, and the very existence of the subject, therefore, depends on the possibility of an Experimental Psychology. The validity of the latter has frequently been questioned, but the time has probably arrived when we can apply the pragmatic test, and, in justification of Experimental Psychology, maintain that the results are a proof of the possibility of the subject and of the correctness of its principles. It now remains to be considered how far the methods of Experimental Psychology are capable of application to school children. Frequently it has been maintained that children are incapable of introspection ; but recent tests in child psychology show that the school child, at least from six years of age upwards, is able to introspect quite satisfactorily in such simple experiments as the reproduction of presentations, and that, owing to the individual and concrete nature of his mental imagery, he can give more definite introspective detail than the adult. The high degree of suggestibility in children is also urged as an objection to the application of the methods of Experimental Psychology, but with proper care this can be overcome, or it can be estimated and allowed for. It is maintained, too, that the subject of an experiment should have some idea of the purpose of the test and that this is impossible with young children. Undoubtedly, by reason of the age of the subjects, certain limits are set to the application of experimental methods with very young children, but in such cases simple experiments only need be attempted ; on the other hand, there may be certain advantages in the simplicity of the child's mental life and the natvetd of his outlook. Indeed, these factors sometimes make experiment easier with children than with adults, who are apt to introduce reflections and guesses as to the outcome of the test and thereby to modify their responses. An open mind, whatever it may be politically, can be of advantage psychologically.

No objection on methodological grounds is usually urged against the statistical methods employed by Experimental Education, provided the data are accurate.

The general methods can be divided into the two main classes, analytic and synthetic. The various educative processes may be analysed into their several factors and each considered in isolation ; thereafter the factors may be combined by synthetic experiment into the original complex process. For example, the reading process may be analysed into the visual apprehension of the words in continuous reading and the eye movements involved therein; the comprehension of the v/ords whereby we attain the meaning of what is visually apprehended ; and the vocal process whereby the meaning is expressed. These factors may then be investigated independently, and thereafter the task is laid upon synthetic experiment of explaining how these partial processes co-operate in producing the original process of continuous reading.

For educational practice, synthetic experiment is undoubtedly the more important of the two methods ; but it presupposes adequate analytic experimentation, and it is a misfortune, inevitable at this early stage of the development of the new science, that the investigations should be mainly analytic and consequently not so productive in practical applications as might otherwise be expected. Consequently, many of the methods and results at first sight appear to have but little bearing on school practice ; but they should not on that account be ignored. In the early stage of the development of experimental psychology it was protested that nothing of value could possibly come from the simple reaction experiments then practised ; yet it was by the development of these methods that a highly complex process like thinking at length yielded to experimental treatment. The seemingly unproductive process of memorising of nonsense syllables likewise provides us with clues and methods by which we may hope to solve some of the difficulties in the training of the child's will — difficulties which the teacher at present cannot hope to overcome, if he even recognises their existence. Practical results alone should not at this stage be demanded, since hasty conclusions and applications will only bring the subject into discredit.

Experiments may be " individual," when the tests are applied separately, or " mass " or " group " experiments conducted in class. Each type has advantages and disadvantages. Individual experiments enable us not only to investigate individual differences but also to obtain the subject's introspection, and thus to check the objective results of tests. The value of such control has been emphasised by Dr. Myers in the article previously referred to : " To neglect introspection," he says, " is usually to court certain disaster. If we are in total ignorance of what has been going on in the mind of the subject during the experiment, it is rarely possible to argue from the objective data — from the measurements which it yields. For example, we may be trying to determine whether any correlation exists between sensory discrimination and general intelligence. A positive result may be simply due to the fact that the very nature of the test has compelled the subject to use his intelligence while carrying out sensory discriminations. We may be correlating mental ability with mental fatigue, and neglect the fact that sometimes we may not be measuring fatigue at all, that in some subjects the task becomes automatic, in others tedious, or that boredom may be in others overcome by motives of duty or ambition. We may be testing the visual acuity of two persons, and obtain a different result from each, despite the fact that really they have the same visual acuity. The result may be due to the fact that the one subject strains every effort to interpret what he but dimly sees, while the other only reads what he believes he can clearly see. Thus again we merely obtain a blurred or erroneous result from the blind applications of statistical methods to measurements which are really meaningless owing to our failure to analyse the conditions determining the character we are measuring."

Mass experiments, apart from the economy in time, may disclose uniformities which would otherwise escape notice. They should not, however, be undertaken without previous individual experiments to render the questions given in class quite unambiguous. Experimental Education has to determine what questions can be most appropriately investigated by the respective methods and how a proper balance between the two may be maintained.'

It is also important to know the number of cases necessary to be examined in order to obtain a result which may be regarded as typical. The danger of too few cases is obvious : we may get

' For determination of the relative value of the two methods in certain typical tests, see "Experimental Tests of Higher Mental Processes and their Relation to General Intelligence," by C. Burt {journal of Experimental Pedagogy, vol. i. pp. 104-5).

contradictory answers to the same questions owing to the influence of individual differences. There is likewise danger with too many subjects, for by massing results obtained from subjects of different ages or nationalities or of varied social status — even within the same school — essential differences may be quite concealed.^

Galton states that from his experience with composite portraiture he believes it may attain statistical constancy within limits not easily distinguished by the eye, after some thirty haphazard portraits of the same class have been combined ; ^ and he maintains that, generally, " numerical averages usually begin to agree pretty fairly when we deal with twenty or thirty cases." 3 It may also be remarked that as women exhibit fewer variations than men, a reliable average can probably be obtained with fewer subjects ; 4 but if typical results are desired, experiments should not be confined to girls at the puberty stage, as they present great variations of development.

A method which has frequently been adopted by Winch.s and which is likely to be more extensively applied in educational experiments, is that known as the method of equivalent or parallel groups. When, for example, it is required to determine the effect of

" See Myers' article for illustrations of both forms of error. " " Inquiries into Human Faculty," Everyman Edition,

3 Ibid., p. 239. Cf. Brown, " The Essentials of Mental Measurement," p. 62.

■• Cf. Thorndike, " Educational Psychology,'' 2nd edition, oh. iii.

5 E.g., British Journal of Psychology, ii, pp. 284-93.

training a particular mental function, the following procedure is adopted. A test is applied to a class before the period of training is entered upon, and, on the basis of the results of this test, the class is divided into two groups, each containing pupils of equal ability. One group is then trained in the exercise of the function in question, while the other is engaged upon different work. At the conclusion of the period of training, both groups are again subjected to a test. Should both show improvement over the original performance, and should the pupils of the trained group exhibit a greater improvement than their colleagues, we may ascribe the advance of the untrained group to natural development and endowment ; and the difference between this and the degree of improvement of the practised group will give the measure of improvement due to training alone. When, however, a whole class is tested before training and again after training — without the control of an untrained group — it is impossible to estimate how much improvement is due to training and how much to natural development, or to determine whether the difference might not be accounted for by the final test being less difficult than the initial test.

The statistical methods common to Experimental Psychology and Experimental Education also call for brief mention. Results of tests expressed in numerical terms can be reduced in various ways, and for each series we must determine what is the most representative value by which it can be characterised.

There are three measures commonly employed to denote a series : —

I. The ordinary arithmetical mean or average is obtained by simply dividing the sum of the values by the number of cases, e.g., the average height of a class is the sum of the individual heights divided by the number of pupils. This is the measure most commonly applied, but it is not always the most representative. In assessing the damages in a libel suit, for instance, the jury might agree on the average of the amounts suggested by the various members, say £60, £60, £ys, ;6"ioo, ^^125, .^150, ^'150, ;^iSo, £2(X>, ;^200, ;^250, and ;£'i,ooo. The average in this case, ;^2io, is clearly not representative, since ten of the twelve jurymen would have awarded less. In such circumstances a crank carries weight in proportion to his eccentricity, whereas a just estimate should, as far as possible, discount the verdict of such a person.

An average is also valueless unless the range of variations is given. Two places might, for example, have the same mean annual temperature and yet entirely different climates, one presenting extremes of heat and cold while the other had an equable temperature. To indicate the variability of a series of values represented by the arithmetical mean it is necessary to state what is termed the mean variation, symbolised by m.v. This is ascertained by finding the average of the differences between the arithmetical mean and the different values representing the series. Thus in the case quoted above the series runs 60, 60, 75, 100, 125, 150, 150, 150, 200, 200, 250, 1,000 and the arithmetical mean is 210.

The differences between the a.m. and the values 3 are: 150, 150, 135, no, 85, 60, 60, 60, 10, 10, 40, 790 and the average of these differences, the mean

variation, or m.v., is i^° = 1 38-3, which represents

12 the range of variability of the series.

2. When it is evident that the extremes of a series are due to the presence of extraneous factors in the test, and it is considered desirable to eliminate these to obtain a representative value, the measure employed is that termed the median. To ascertain the median, the values are arranged in order of magnitude and the middle one selected as representative of the series ; in the event of an even number of values in the series, the average of the two middle ones is regarded as the median. In the case already quoted the series runs: 60, 60, 75, 100, 125 150, 150, 150, 200, 200, 250, 1,000, and, as the series comprises twelve values, the median would be the average of the sixth and seventh, i.e., 1 50. By the employment of this measure the representative value is not, like the arithmetical mean, affected by the extreme and abnormal value of 1,000.

To denote the range of variability in a series represented by the median, the semi-interquartile range is usually given ; it is calculated by taking half the difference between the quartiles above and below the median, i.e., half the difference between 100 and 200, or so. Here again the extremes are eliminated.

3. The third measure used, the mode, is simply the value which occurs most frequently in the series. In the illustration we have been using, 1 50 occurs three times, and smce no other value occurs more than twice, 150 would be taken as the representative measure.

In experiments with numerous examples in the series the arithmetical mean may be employed and the mean variation, or some other measure of variability, also given. When the examples are not numerous and it is conjectured that disturbing factors may be influencing some of the results, the median should be employed, in which case the series should comprise an odd number of examples in order to simplify reduction. When the semi-interquartile range has to be calculated, 13 or 17 will be found more convenient than 11 or 15. The mode is used to characterise a type, that is, to indicate the case most frequently recurring.^

When a series of different tests has been applied to a number of subjects and each series reduced by any of the foregoing methods, for example, the arithmetical mean {a.m.) and mean variation (m.v.), it may be considered desirable to determine also the degree of difficulty of the various tests by comparing one series with another for all the subjects taken together. This may be accomplished by calculating the arithmetical mean and the mean variation of the representative values already computed ; and to distinguish these measures from the ordinary a.m. and m.v. it has been suggested by Dr. Rivers that they should be symbolised by A.M. and M. V. respectively.^

' For further amplifications and qualifications of these measures reference is suggested to works treating specifically of statistical methods.

= British Journal of Psychology , vol. i. pp. 3S4-5S-

Statistical methods have nowhere been more fruitfully applied than in the case of correlation ; but the very ease with which the formula may be employed is apt to produce misleading results, owing to its application to data which have not been scientifically collected. Opinions on the natural affinities amongst the various mental functions and the various educational subjects have usually been vague and inconsistent, but the application of statistical methods now enables us to state in exact quantitative fashion the extent, if any, of such correlation. Several methods for ascertaining the degree of correlation are employed, but Spearman's Foot-Rule for measuring correlation ' is likely to become the most popular in Experimental Education, since its ease of application more than counterbalances any lack of accuracy attaching to its use.

The following illustration given by Spearman ^ will indicate its simplicity : —

Subject. Rank in Rank in Gains in rank