foreign language learning ability1
TRANSCRIPT
Journal of Educational Psychology1962, Vol. 53, No. 1, 15-26
FOREIGN LANGUAGE LEARNING ABILITY1
PAUL PIMSLEUROhio State University
ROBERT P. STOCKWELL, AND ANDREW L. COMREYUniversity of California, Los Angeles
This report on 2 studies of factors involved in learning French isbased on 23 tests administered to 208 students in college French, andon 22 tests administered to 202 additional students a year later. Thefactors extracted and rotated in each study included: Verbal, Rea-soning, Speed of Articulation, Pitch Discrimination, Timbre Discrimi-nation, Interest, and Biographic. Multiple correlation test selectionanalyses resulted in (a) R = .65, using 6 tests to predict Coopera-tive French Test scores; (b) R = .41, using 5 tests to predict auralcomprehension; and (c) R = .41, using 5 tests to predict speakingproficiency. Verbal IQ and Interest (motivation) appear to be themost important factors in college foreign language learning; Reason-ing, Word Fluency, and Pitch Discrimination also contribute.
It is popularly believed that foreignlanguages arc analogous to mathematicsor music in that a special talent is re-quired to learn any of them. This beliefappears fallacious and even harmful inthat it may adversely affect the achieve-ment of those who think that they donot possess this special talent. The in-vestigation of which this paper is a partseeks to reduce the so-called "talent forlanguages" to a set of well-defined,measurable components. Tests of thecomponents will then be used to predictprobable success in foreign languagecourses. The attempt to isolate and de-scribe as many as possible of the com-ponents affecting foreign languagelearning should lead to improved under-standing of the learning processes in-volved. Presumably, the tests developedfor predictive purposes will also befound useful in diagnosing the sorts ofdifficulties encountered by poor lan-guage learners.
As Carroll's (1960) thorough reviewof the literature of foreign language
1 Since June 1960, the research reportedherein has been performed pursuant to acontract with the Office of Education, UnitedStates Department of Health, Education, andWelfare.
prognosis indicates, considerable successhas been achieved in predicting resultsof intensive language courses, such asthose given in the military services.However, attempts to predict achieve-ment in high school and college courseshave generally been less successful.There appear to be influences present inthe typical school course which do notenter into intensive courses, and whichneed to be taken into account to attainhigh predictive validity. Prognostic testscomposed entirely of intellectual tasksprobably owe their rather low validityto their failure to include importantnonintellectual variables. Motivationfactors almost certainly affect achieve-ment, and personality may also be im-portant. Another kind of prognostictest, the work sample, sometimes is agood predictive device, but sheds nolight on the language learning process.The studies reported in this paper at-tempt to deal with a number of fairlyprecise intellectual and motivationalfactors. Personality factors, not includ-ed here, will be introduced in subse-quent research. The previous researchmost closely related to that reportedhere is the work of Carroll (1958).
The two studies which are the subject
15
16 P. PIMSLEUR, R. P. STOCK.WELL, AND A. L. COMREY
of this paper were conducted at UCLAin the spring semesters of 1958 and1959, respectively. In both cases, thesubjects were students in second semes-ter French courses, 208 in the firststudy, and 202 in the second. The twosamples appear closely comparable, andquite typical of a moderately goodgroup of American undergraduates.Their average age was 19 years; mostwere graduates of California public highschools who gained admission to collegeby achieving a B average; most weretaking French to fulfill degree require-ments.
The data consist of scores on a num-ber of predictor variables and severalcriteria. In the light of the currentshift toward teaching the spoken lan-guage, it was of particular interest toattempt differential prediction of tra-ditional (grammar-reading) achieve-ment on the one hand, and oral-auralachievement on the other.
AH tests were administered to groups,under good testing conditions duringregular class hours. Tests requiringspeaking or listening were administeredin the language laboratory. Tests simi-lar in nature (e.g., Reading Aloud I andII) were given at least a week apart.The predictor tests were given early inthe semester, the criterion tests at theend.
The two studies will be reported inthe following order:
Study I—Variables, factor analysis,multiple correlation analysis
Study II—Variables, factor analysis,multiple correlation analysis
Results
STUDY I
VariablesThe data for the first study consisted
of scores on 21 test variables and twocriteria. The test variables were chosento represent factors which, on the basis
of teaching experience and linguistictheory, were hypothesized to be relatedto success in learning a foreign language.The hypothesized factors were associa-tive memory, analytic reasoning, reason-ing by analogy, physical dexterity inarticulation, ability to change linguisticset, auditory discrimination, interest inforeign languages, and several othersTests were either found or constructedto represent the hypothesized factors inas "pure" a manner as possible.
It appeared desirable to try out alarge number of such predictor varia-bles. As there were more tests thancould be given in a single semester, thefirst study omitted several factors ofpotential importance, such as auditorydiscrimination and language interest,which then appeared in the secondstudy.
The 23 variables which provided thedata for the first study are as follows:
1. Modern Language Aptitude Test(MLAT), Part III—Spelling Clues (Carroll& Sapon, 1955). The subject chooses whichof five words has the same meaning as theword represented in abbreviated form Sam-ple: kataklzm—(a) mountain lion, (b) dis-aster, (e) sheep, (d) chemical reagent, (e)population. Highly speeded: 50 items, 8minutes.
2. MLAT, Part I—Number Learning. Bytape recording, the subject is taught a simpleartificial system of number expression utiliz-ing nonsense syllables. He is then asked towrite down the arabic numeral equivalentsof a list of 2- and 3-digit numbers in theartificial system spoken at a fairly rapid paceon the tape This version utilizes the digits0, 1, 2, 3, 4. Fifteen items, total time 10minutes.
3 MLAT, Part IV—Words in Sentences.This test was designed to measure the abilityto understand the function of words andphrases in sentence structure, without callingon knowledge of grammatical terminology.Each item consists of a "key sentence" witha word or phrase capitalized, followed by oneor more sentences with other words orphrases underlined and numbered. The sub-ject is directed to pick the word or phrasein the second sentence (or sentence-group)
FOREIGN LANGUAGE LEARNING 17
which docs the same thing in that sentenceas the capitalized word does in the keysentence.Sample:
He spoke VERY well of you.Suddenly the music became quite loud.
1 2 3 4
The test contains 45 items. The time allowedis 15 minutes.
4. MLAT, Part V—Paired Associates.The subject studies a list of 24 "Kurdish-Eng-lish" vocabulary equivalents for 2 minutes; inthe next 2 minutes he practices recalling theEnglish meanings, and he then has 4 minutesto complete from memory a multiple-choicetest of the presented vocabulary (24 items).
5. Letter Series (adapted from Guilford).A series of seven letters is given, to whichthe subject must add the next two Sample:
D E F D E F - - .
6. Reading Aloud I (speed). The subjectis given time to study a meaningful para-graph of English prose, then is asked to re-cord his reading of it, "speaking as quicklyas possible while still remaining intelligible."Score is number of words read in 30 seconds.
7. Same test (accuracy). Number oferrors made in 6.
8. Reading Aloud II (speed). The sub-ject is given time to study a meaninglessparagraph made up of English words puttogether at random. He then records hisreading of it "speaking as quickly as possiblewhile still remaining intelligible." Score isnumber of words read in 30 seconds.
9. Same test (accuracy). Number of er-rors made in 8.
10. Paraphrase. The subject is to give asmany paraphrases as possible for a givenphrase. Sample: for TO DIE, the subjectsmight give "to kick the bucket," "to cash inone's chips," etc Score is number of para-phrases given in 4 minutes.
11. Rhymes. The subject is to give asmany words as possible that rhyme withfour given words (LAKE, CLOUD, SO, GRASS).Score is number of rhymes given in 2 minutes.
12. Synonyms. The subject is to give asmany synonyms as possible for four givenwords Sample: for GO, the subject mightgive "ride," "drive," "fly," "travel," etc.Score is number of synonyms given in 4minutes.
13. Phonetic Perception. The subjecthears three sounds and must tell which oneis different from the other two. The test
uses distinctions which are nonphonemic inEnglish (e.g., "r" vs. "rr"). Twenty items,5 minutes.
14. Linguistic Analysis I. The subject isgiven a list of 22 foreign forms (adaptedfrom Kabardian) together with their Englishequivalents. By consulting the given forms,he is to deduce how other things are saidin this language. Ten multiple-choice items,12 minutes.
15. Linguistic Analysis II The subject isgiven a list of 19 forms in a foreign language(adapted from Kabardian) together withtheir English equivalents By consulting thegiven forms, he is to deduce the meanings ofa number of other foreign forms given tohim. Ten multiple-choice items, 12 minutes.
16. Verbal Comprehension (Guilford-Zimmerman). A test of English vocabularyknowledge. Forty items, 6 minutes.
17. Age The subject gives his age at lastbirthday and is placed in one of five cate-gories- under 18, 18-20, 20-22, 22-25, over25
18. Sex. Female or male, scored zero orone.
19. Bilingualism. To be rated as "bi-lingual" a subject has to answer either orboth of the following questions affirmatively:Is a language other than English spoken reg-ularly in your home? Is your native lan-guage other than English? Scored zero formono-, one for bilingualism.
20. High School Language Grades. Anaverage of grades in high school languagecourses.
21. High School Math-Science Grades.An average of grades in math and sciencecourses in high school.
22. (Criterion) French II Final Grades.Letter grade assigned by teacher at end ofsecond semester of college French. (A, B, C,D,F = 4 ,3 ,2 ,1 ,0 ) .
23. (Criterion) French Speaking Profi-ciency Test. A five-part recorded test ofability to speak French at end of secondsemester Scored by native judges.
Factor Analysis
A matrix of zero-order Pearson cor-relations was computed from the scoresof the 208 subjects on the 23 variablesjust described. These correlations, to-gether with the means and standarddeviations of the variables, are shown
18 P. PlMSLEUR, R. P. S T O G K W E L L , AND A. L. CoMREY
in Table A.2 From this matrix, eightcentroid factors were extracted and thenrotated, using Kaiser's normal varimaxcriterion (Kaiser, 1958).3 Table B pre-sents the rotated factor matrix. Load-ings of less than .30 were not consideredsignificant. Descriptions of the eightfactors follow:
Factor A: Reasoning. This factor ismade up of seven variables (loadingsshown at right).
2. Number Learning .5721. High School Math-Science
Grades .5014. Linguistic Analysis I .483. Words in Sentences .43
15. Linguistic Analysis II .355. Letter Series .34
22. French II Grades (criterion) .34
This factor resembles others variouslycalled General Reasoning, Verbal Rea-soning, Induction, etc., in previous fac-tor analytic findings. It appears toinclude the ability to induce the orderlyprinciple in a group of examples (e.g.,Letter Series), and to apply this princi-ple to a variety of new examples (e.g.,Number Learning).
Factor B: Word Fluency. The vari-ables with sizable loadings on this factorare:
10. Paraphrase12. Synonyms11. Rhymes
.65
.64
.41
2 Tables A, B, C, and D have been depos-ited with the American Documentation Insti-tute. Order Document No. 6943 from ADIAuxiliary Publications Project, Photoduplica-tion Service, Library of Congress; Washing-ton 25, D. C , remitting in advance $1.25for microfilm or $1.25 for photocopies.Make checks payable to: Chief, Photodupli-cation Service, Library of Congress.
3 The factor analytic computations in StudyI were performed on SWAC, an electroniccomputer located on the campus of UCLAand supported by the Office of Naval Re-search.
The identification as Word Fluencyrequires some amplification. Word Flu-ency is usually defined as the ability toproduce words which conform to cer-tain structural limitations, e.g., four-letter words of die types s E. Inthe present case, the limitations are notstructural but semantic. Both kinds oftests entered into factors called WordFluency by previous investigators (Car-roll, 1941; Thurstone, 1941). The fac-tor found here most nearly resemblesGuilford and Christensen's (1956)Eduction of Conceptual Correlates. Thename Word Fluency has been retainedbecause it is more familiar.
Factor C: Biographic. This factoris made up of:
10 Sex—scoring code:feminine = 0, masculine = 1 .64
20. High School LanguageGrades - . 5 8
17. Age .524 Paired Associates —.343. Words in Sentences —.31
This factor seems to reiterate the well-known fact that girls do better than boysin high school language courses. How-ever, at the college level, sex fails tocorrelate significandy with either cri-terion (—.02 and —.08, respectively).
Factor D: Achievement in French.This factor is made up of the twocriteria:
23. French Speaking Test -.6522. French II Grades -.57
It seems to represent a unique portionof the variance in the criteria. It may-be pointed out that the two criteria werearrived at in different ways, by differentjudges.
Factor E: Speed of Articulation.Two variables have significant loadings:
FOREIGN LANGUAGE LEARNING 19
8. Reading Aloud II (speed) -.716. Reading Aloud I (speed) —.68
The identification seems clear.
Factor F: ? The variables includedin this factor are:
5. Letter Series -.4213. Phonetic Perception -.3719. Bilingualism .3415. Linguistic Analysis I —.32
The meaning of this factor is notclear, beyond the fact the bilinguals per-form poorly in the three other variablesNo interpretation will be attempted, asthese variables appear again in the nextstudy.
Factor G: Response Set. The twovariables composing this factor are:
9. Reading Aloud II (errors) —.487. Reading Aloud I (errors) —.44
This factor, reflecting the number oferrors made while reading a passage attop speed, may be interpreted as Re-sponse Set, i.e., the tendency of someto avoid mistakes by working a littleslower, while others work very rapidlyeven at the risk of making mistakes.
Factor H: Verbal. Three variablesmake up this factor:
16.1.
11.
Verbal ComprehensionSpelling CluesRhymes
.59
.43
.37
The chief loading being that of anEnglish vocabulary test, this may beidentified as the Verbal or VerbalKnowledge factor.
Summarizing the factors whose iden-tification appears clear, they are: Rea-soning, Word Fluency, Biographic,Achievement in French, Speed of Ar-ticulation, Response Set, and Verbal.
Multiple Correlation Analysis
A multiple correlation coefficient, R,
was computed1 for each of the two cri-teria, using as predictors the entire bat-tery of 21 tests. The results, togetherwith the proportion of variance contrib-uted by each variable, are presented inTable 1, Columns A and C.
Reduction was attempted in the sizeof the batteries. The three or four testsmaking the highest contributions werechosen as a basic battery. To these wereadded other tests, singly or several at atime, in many different combinations,and the resulting Rs computed. Thebatteries shown in Table 1, Columns Band D, are those which minimize thenumber of tests while maximizing themultiple correlation coefficient.
For predicting French grades, theminimum battery consists of NumberLearning (or Spelling Clues), Words inSentences, Letter Series, Reading AloudI, Paraphrase, Linguistic Analysis II,age, and high school math-sciencegrades This seven-test battery yields amultiple R of .43. The inclusion ofNumber Learning or Spelling Cluesgives approximately the same results;the inclusion of both does not raise theR appreciably.
For predicting Speaking Test scores,the minimum battery consists of Spell-ing Clues, Letter Series, Reading AloudII, Verbal Comprehension, and Bilin-gualism. This five-test battery yields amultiple R of .42.
In evaluating the multiple correlationcoefficients, several factors which mili-tate against accurate prediction must betaken into account. The first of theseis the unreliability of the teacher gradesused as one of the criteria. If grades areassumed to have a reliability of, say,
4 This part of the analysis used the BIMD-6program supplied by the Biostatistics Lab-oratory, Department of Preventive Medicine,UCLA. The computations were performedon an IBM 709 computer, through the co-operation of the Western Data ProcessingCenter, UCLA.
20 P. PlMSLEUR, R. P. S T O C K W E L L , AND A. L. COMREY
TABLE 1
STUDY I: MULTIPLE CORRELATION ANALYSIS
1.2.3.4.5.6.7.8.9.
10.11.12.13.14.15.16.17.18.19.20.21.
Variable
Spelling CluesNumber LearningWords in SentencesPaired AssociatesLetter SeriesReading Aloud I (speed)Reading Aloud I (errors)Reading Aloud II (speed)Reading Aloud II (errors)ParaphraseRhymesSynonymsPhonetic PerceptionLinguistic Analysis ILinguistic Analysis IIVerbal ComprehensionAgeSex (feminine = 0, masculine = 1)BilingualismHigh School Language GradesHigh School Math-Science Grades
French II
A(21 tests)
.033
.025047.002.008.021.000.002.014.009.001.000.005.000.010.002.012.001.022.004.008
R = .478
Grades
B(7 tests)
.036063
.029
.012
.012
012
024
R - .433
French Speaking Test
C(21 tests)
.058
.002
.007
.003
.019
.023
.003
.010
.001
.001
.000
.000008.000.004.036002.001028000.001
R = .454
D(5 tests)
.058
.024
.027
.044
.023
R = AIS
Note —Proportion of variance contributed by each variable to the prediction of each criterion. A and C—using all tesfe, B and D—using the most economical battery.
.70, then the correction for attenuationraises the multiple R from .43 to .52.The Speaking Test criterion might besimilarly corrected, though there is someevidence to show that it has high relia-bility (Pimsleur, 1961). A second con-sideration is the fact that certain poten-tially important variables were notincluded. The use of additional variablesand more reliable criteria are features ofthe second study. A third considerationis the restriction in range of abilityamong the examinees. The college pop-ulation is highly select; the secondsemester population of French studentsis even more so. This homogeneitymakes accurate prediction very difficult.
STUDY II
Variables
In Study II, a somewhat different setof variables was used. Those which hadproven uninteresting in Study I werediscarded. Others were added to tapadditional factors, e.g., Numbers 9through 12, dealing with an auditoryfactor, and Numbers 13 and 14, treatinginterest or motivation. Despite theirgood results in the first study, theMLAT parts were dropped becausethey contain copyright material and socould not be used in developing a newbattery.
The last three variables are three
FOREIGN LANGUAGE LEARNING 21
criteria, all new to this study. TheCooperative French Test was used as ameasure of achievement in the readingand writing goals. Oral achievementwas rated by a laboratory instructor whohad listened to students and gradedthem in the lab once a week for a fullsemester. Achievement in understand-ing spoken French was measured by thePictorial Auditory Comprehension Test,an objectively scored test.
The variables for Study II are these:1. Verbal Comprehension (Guilford-
Zimmerman). See Study I for description.2. Linguistic Analysis I. See Study I3. Ship Destination Test (Christensen-
Guilford). Find the distance from ship toshore considering the influence of severalvariables. Thirty-three items, 15 minutes.
4. Grammatical Transformations. Thesubject is given a sample transformation (e g.,"They are good. Books are interesting.'"—» "Good books are interesting.") In a newitem, he is to select from among four choicesthe one which represents a transformationsimilar to the sample (e.g., "It is noisy.Telephones are annoying." —> "Noisy tele-phones are annoying.") Twenty items, 9minutes.
5. Paraphrase. See Study I.6. Rhymes. See Study I.7. Reading Aloud I. See Study I.8. Reading Aloud II. See Study I.9. Phonetic Perception. See Study I.
10. Chinese Pitch Perception Subjectis taught three Chinese words which differonly in pitch. These are then imbedded inChinese sentences and subject must tellwhich he hears. Thirty items, 10 minutes.
11 Seashore Pitch Test. Subject hearstwo tones, must tell whether the second ishigher or lower than the first. Thirty-fiveitems, 5 minutes.
12. Seashore Timbre Test Subject hearstwo chords, must tell whether the second isthe same or different than the first. Thirty-five items, 5 minutes.
13. Interest I. Twenty items relating tointerest in language. Subject answers eachon a five-point scale.
14. Interest II. The subject is asked torate on a five-point scale the extent of hisinterest in studying the foreign language heis now studying.
15. Bilingualism. See Study I.
16. Age. See Study I.17. Sex. See Study I.18. High School Language Grades. See
Study I.19. High School Math-Science Grades.
See Study I.20. (Criterion) Cooperative French Test,
Advanced Forms Q and R. A standardizedtest of achievement in reading, grammar, andvocabulary Total scores were used.
21. (Criterion) Lab Grade. An estimateof speaking ability given by laboratory in-structor on basis of listening to student allsemester, plus a final oral test. Grades from0 to 11
22. (Criterion) Pictorial Auditory Com-prehension Test. An objectively scored testof French listening comprehension. Subjectmust select from among four pictures theone which correctly illustrates the sentencehe has just heard. Tape recorded, 50 items,20 minutes.
Factor Analysis
From the scores of 202 subjects on the22 variables a matrix of zero-order cor-relations was computed. It is presented,together with the means and standarddeviations of the variables, in Table C.From this, eight factors were extractedand rotated.5 The rotated factor matrixis presented in Table D.
The analysis yielded eight factors, allof which could be identified, thoughwith varying degrees of assurance.
Factor A: Achievement in French(Verbal). The first factor is composedof seven variables (loadings at right).
20. Cooperative French Test(criterion) .73
21. Lab Oral Grade (criterion) .5522. Aural Test (criterion) .52
1. Verbal Comprehension .4718. High School Language
Grades .4413. Interest I .3517. Sex -.31
This factor is ambiguous. The first threevariables are the three criteria, and
8 This part of the analysis used the BIMD-17 program, on the 709 computer.
22 P. PlMSLEUR, R. P. STOCK.WELL, AND A. L . CoMREY
appear therefore to compose a factorspecific to achievement in learningFrench. Here, they are associated how-ever with a verbal component (Tests 1and 18), and a component of motiva-tion or interest (Test 13).
Factor B: Speed of Articulation.This factor is made up of:
8. Reading Aloud II7. Reading Aloud I
77.71
The identification is clear. However,these variables did not correlate highlywith the criteria, as they did in the firststudy. For example, Reading Aloud Icorrelated .20 (significant beyond the.01 level) with the first criterion in theprevious study, but only .04 (nonsignifi-cant) with the first criterion in thepresent study.
Factor C: Reasoning. The compo-nents of this factor are:
19. High School Math-ScienceGrades .51
3. Ship Destination Test .462. Linguistic Analysis I .374. Grammatical
Transformations .32
This factor seems easily identified asReasoning, or General Reasoning. Thefirst two tests (19 and 3) are nonverbalin nature, while the other two (2 and 4)involve linguistic tasks. This factorshows a close relation between verbaland nonverbal reasoning.
Factor D:consists of:
16.1.
17.18.
AgeVerbalSex
Biogi aphic. This
Comprehension
High School LanguageGrades
fac
.56
.39
.37
-.31
It seems appropriate to consider this abiographic factor, though its interpre-tation is not clear. It is of interest that
sex, which did not correlate significantlywith the criteria in the first study, diddo so in the second.
Factor E: Pilch Discrimination. Thevariables which form this factor are:
11 Seashore Pitch Test .5010. Chinese Pitch Test .444 Grammatical
Transformations .37
The identification as pitch discrimina-tion seems correct. It should be notedthat the two pitch tests (11 and 10) arequite different in nature, the formerinvolving pure tones and the latter in-volving language. These two pitch testscorrelate significantly with the first andthird criteria, but not with the secondone. The presence of Test 4 on thisfactor is not readily explainable.
Factor F: Word Fluency. This fac-tor is made up of:
5.6.
15
ParaphraseRhymesBilingualism
.50
.41- 3 8
The identification is clear. The presenceof Test 15 merely means that foreignsubjects have less word fluency in Eng-lish than do Americans.
Factor G: Interest in Languages.The two variables composing this factorarc:
1413.
Interest IIInterest I
.70
.67
The identification presents no difficulty.Both these variables correlate signifi-cantly with all three criteria.
Factor H: Timbre Discrimination.This factor consists of:
9. Phonetic Perception .3612. Seashore Timbre Test .35
The first of these variables (9) involvesactual language sounds, while the other
FOREIGN LANGUAGE LEARNING 23
(12) involves nonlanguage tones.Though the loadings are low here, it isnevertheless intriguing that pitch andtimbre appear as two distinct factors.These two variables correlate signifi-cantly with the third (aural) criterion,but not with the other two.
Multiple Correlation Analysis
The test selection procedure differssomewhat from that used in the firststudy. Here, a stepwise regression pro-gram was used in conjunction with amultiple correlation program, the for-mer to find the best one-test battery,best two-test battery, etc., up to 9 or 10tests, and the latter to determine the
multiple R and the contributions tovariance of each test in each of the bestbatteries.0
Results of the analysis are presentedin Table 2. As can be seen in Column Aof Table 2, the criterion of CooperativeFrench Test scores can be predicted tothe extent of R = .673 when all 19 testsare used. The number of tests may bereduced considerably, without much lossin predictive accuracy. Column B showsthe best six-test battery for predictingthis criterion. This is a modest batterywhich can be administered in a 50-minute class period; it yielded an R of
6 The programs referred to are BIMD-6and BIMD-9.
TABLE 2
STUDY I I : MULTIPLE CORRELATION ANALYSIS
1.2.3.4.
5.67.8.9.
10.1112.13.1415.16.17.18.19.
Variable
Verbal ComprehensionLinguistic AnalysisShip DestinationGrammatical
TransformationsParaphraseRhymesReading Aloud IReading Aloud IIPhonetic PerceptionChinese PitchSeashore PitchSeashore TimbreInterest IInterest IIBilingualismAgeSexHigh School LanguageHigh School
Math-Science
Criterion 20:Cooperative French Test
A(19 tests) (6
.193
.017
.000
.004
.000010.002.001.001.019.000.004.123.020.007.006.022.019
003R = .673 R =
Btests)
.193
.024
.125
.021
.037
.024
.652 R
Criterion 21.Lab (oral) Grades
C(19 tests)
.021001000
.002000.015.006.018.002.005.000.000040.009.003.008.013.048
.016= .457 R
D(5 tests)
.021
.052
.009
.073
013= .410
Criterion 22:Auditory Comprehension
E(19 tests)
.075
.030
.000
001.004003.000.001.008.011.002.009.032.001.000.000.003005
.005R = .436 R
F(5 tests)
.075
.030
.015
.013031
= .405
Note —Proportion of variance contributed by each variable to the prediction of each of 3 criteria: A,C,E—using all tests; B,D,F—uiing the most economical battery
24 P. PlMSLEUR, R. P. STOCKWELL, AND A. L. CoMREY
.652 (estimated shrinkage, JR = .646) ?Two improvements in the second studyas compared with the first probablyaccount for the more satisfactory result.They are the use of a more reliable cri-terion, i.e., a standardized test in placeof teacher grades, and the inclusion ofnew variables which turned out to bequite important.
The other two criteria can also bepredicted by a small battery of tests, asmay be seen in Columns D and F ofTable 2. However, these multiple i?sare considerably lower than those forthe Cooperative French Test, probablydue to the lower reliability of the othertwo criteria.
RESULTS
Prognosis
By far the best predictive accuracywas obtained in the prediction of Co-operative French Test scores, in StudyII. A multiple correlation coefficient ofR — .652 was arrived at by a six-testbattery that can be administered in lessthan an hour. While this battery hasnot yet been validated on other samples(such work is under way), it appears tobe as good a device for predicting suc-cess in foreign language learning incollege as is now available. The regres-sion equation for predicting CooperativeFrench Test scores is:
Y' = 47.37 + .47Xi + .18X]0 + .09X13
+ 1.21X14-2.15X17+1.82X,» [1]
As for the other two criteria, speakingability and listening comprehension, themultiple correlation coefficient obtainedfor each of them was approximately .41,with a battery of five tests. The regres-sion equations associated with the pre-diction of these two criteria in Study IIare, respectively:
7 Shrinkage estimated by McNemar's For-mula 75. See McNemar (1955, p. 186).
Y' = 3.25 + .04X! + .31X14 + .64X15
+ .98X18 - .37X19 [2]Y' = 17.80 + .25X! + .37X2 + .13X10
+ .16X,2 + .09X13 [3]
Diagnosis
The manner in which prediction ofCooperative French Test scores wasachieved merits attention. The testscontributing to this prediction areVerbal Comprehension, Interest I andII, Chinese Pitch, Sex, and High SchoolLanguage Grades; i.e., they representfactors of verbal intelligence, motiva-tion, pitch discrimination, and certainbiographic elements. While the predic-tive accuracy they afford is satisfying,it does not come about in the way onemight wish. It had been hoped thatforeign language achievement could bepredicted on the basis of intellectualfactors, such as the ability to discrimi-nate sounds, to induce grammaticalprinciples, and so on. Instead, it ap-pears from these studies that the twobiggest factors in such achievement arethe very general ones of verbal IQ andmotivation. This finding says that asfar as language study in college is con-cerned, anyone will do well who isintelligent and wants to learn, regard-less of such concerns as having "a goodear," "a good memory," and "goodreasoning powers."
Nonetheless, it is of interest to exam-ine the relative importance of the factorswhich contribute to the prediction ofeach criterion.
Though the first criterion (the gram-mar-reading goal) was measured inquite different ways in the two studies,this fact will be ignored for the moment,to permit comparison of the two setsof findings. These are the tests whichcontributed 1% or more to predictionof the first criterion:
FOREIGN LANGUAGE LEARNING 25
Study I3. Words in Sentences .0471. Spelling Clues .0332. Number Learning .025
19. Bilingualism .0226. Reading Aloud I .0219. Reading Aloud II
(errors) .01417. Age .01215. Linguistic Analysis II .010
Study II1. Verbal Comprehension .193
13. Interest I .12317. Sex .02214. Interest II .02018. High School Language .01910. Chinese Pitch .0192. Linguistic Analysis I .0176. Rhymes .010
In Study I, three of the tests (3, 2,15) are associated with the Reasoningfactor. Test 1 is associated with theVerbal factor. Test 6 is associated withSpeed of Articulation; Test 9 with Re-sponse Set; Test 17 with the Biographicfactor.
The factorial findings in Study II dif-fer partly from those of Study I, sincedifferent tests are involved. In Study II,Tests 1 and 18 are associated with aVerbal factor; Tests 13 and 14 with anInterest (motivation) factor; Test 17with a Biographic factor; Test 10 witha Pitch Discrimination factor; Test 2with a Reasoning factor; Test 6 with aWord Fluency factor. The Verbal, Rea-soning, Word Fluency, and Biographicelements were found in both studies.The interesting Speed of Articulationfinding of the first study was not con-firmed by the second. Interest and PitchDiscrimination are new additions in thesecond.
These findings permit the tentativeconclusion that achievement of the tra-ditional (grammar, reading, writing)objectives of a college French courseinvolve primarily verbal intelligence andmotivation on the student's part, butmay also involve his ability to reason
both analytically and by analogy, andbis ability to think up words quickly.
In order to examine the two sets offindings with regard to oral (speaking)achievement, the difference between thetwo measures of the criterion will againbe ignored. There follows a summaryof the tests which contribute 1% ormore to the prediction of this criterion:
.058
.036
.028
.023
.019
.010
.048
.040
.021
.018
.016
.015
.013
Study I1.
16.19.6.5.8.
Spelling CluesVerbal ComprehensionBilingualismReading Aloud ILetter SeriesReading Aloud II
Study II18. High School Language13. Interest I
Verbal ComprehensionReading Aloud IIHigh School Math-ScienceRhymesSex
1.8.
19.6.
17.
Represented in the findings of Study Iare the Verbal factor (Tests 1 and 16),the Speed of Articulation factor (6 and8), and the Reasoning factor (5). Thefactors represented in Study II areVerbal (Tests 1 and 18), Interest inLanguage (13), Speed of Articulation(8), Reasoning (19), Word Fluency(6), and Biographic (1, 17, 18).
These findings suggest that in order todo well in learning to speak French incollege the student must possess, aboveall, verbal intelligence and high motiva-tion, and that in addition he may behelped by his analytic reasoning ability,his ability to articulate words rapidly,and his ability to think them up.
As for the third criterion, listeningcomprehension, it is predicted by:
Study1.
13.2.
10.12.
IIVerbal ComprehensionInterest ILinguistic Analysis IChinese PitchSeashore Timbre
.075
.032
.030
.011
.009
26 P. PlMSLEUR, R. P. S T O C K W E I X , AND A. L. CoMREY
The factors involved here are, again,verbal intelligence and motivation, plusreasoning, pitch discrimination, andtimbre discrimination.
It is concluded from the studies thusfar completed (a) that achievement ina traditional (grammar-reading) lan-guage course may be predicted withreasonable validity by a set of tests, eachof which taps only one rather precisecharacteristic of the learner; (b) thatnonintellectual characteristics, notablymotivation, must be included as well asintellectual ones; (c) that oral and auralachievement are less subject to satisfac-tory prediction at the present time,probably due to the lack of adequatecriterion tests for achievement in theseskills; (d) that although better criteriontests will improve prediction somewhat,substantial improvement probably de-mands the inclusion of entirely new fac-tors as predictors; (e) that among suchnew factors, the personality of the stu-dent and the characteristics of theteacher are those which appear mostpromising and are most in need ofresearch attention.
REFERENCES
CARROLL, J. B. A factor analysis of verbalabilities. Psychometrika, 1941,6,279-307.
CARROLL, J. B. A factor analysis of two for-eign language aptitude batteries. / . gen.Psychoi, 1958, 59, 3-19.
CARROLL, J. B. The prediction of success inintensive foreign language training. Har-vard University Graduate School of Edu-cation, Laboratory for Research in In-struction, 1960. (Mimeo)
CARROLL, J. B., & SAPON, S M. ModernLanguage Aptitude Test. New York:Psychological Corporation, 1955.
GUILFORD, J. P., & CHRISTKNSEN, P. R. Afactor-analytic study of verbal fluency.U. Sth. Calif. Psychoi. Lab. Rep., 1956,No. 17.
KAISER, H. F. The varimax criterion foranalytic rotation in factor analysis. Psy-chometrika, 1958, 23, 187-200.
MCNEMAR, Q. Psychological statistics, NewYork: Wiley, 1955.
PIMSLEUR, P. A French Speaking ProficiencyTest. French Rev., 1961, 34, 470-479.
THURSTONE, I. L., & THURSTONE, T. G.Factorial studies of intelligence. Psycho-metr. Monogr., 1941, No 2.
(Received April 7, 1961)