gardner on affect: a reply to gardner

7
EXCHANGE GARDNER ON AFFECT: A REPLY TO GARDNER’ John W. Oller, Jr. University of New Mexico In the lead article of the second issue of this journal in 1980, Robert C. .Gardner offers a defense of “the proposition that achievement in a second language is related to attitudinal and motivational variables.” It is claimed that “Oller and his colleagues have questioned the existence of such a relationship” (1980:268). Further, that we have “given the impression that the empirical foundation for it is weak.” In defense of his own previous work, Gardner argues that “it is the responsibility of scientists to investigate phenomena and provide the most parsimonious explanation possible for all the research literature” (p. 256). He writes, “that is precisely what we have done in our investigations of the role of attitudes and motivation in second language learning”(at this point he cites a number of his own previous publications, some with coauthors including W.E. Lambert, L. Gliksman, P. Smythe, C.L. Smythe, and R. Clement). This reply has two major purposes. The first is to show that Gardner misconstrued the arguments he attempted to refute. The position defended by him has never been under attack in any of the publications by “Oller and his colleagues” which are cited. To the contrary, it has always been assumed (and made explicit in the publications cited by Gardner) that affective factors must play a significant role in the acquisition of any language (primary or nonprimary). Never in any of the publications cited by Gardner, have 1 or my colleagues “questioned the existence of such a re la t ionship. ”* The second purpose of this reply is to try to clarify the fact that it is the measurement of affective variables which has been questioned (not their existence, nor their relationship to learning a second language). However, if’the measures of affect are suspect, rhen to this extent, won’t the empirical basis of certain popular theories of how affective factors I R.C. Gardner, “On the Validity of Affective Variables in Second Language Acquisltion: Conceptual, Contextual, and Statistical Considerations,” Language Learning 30:255-270 (December 1980). * The main articles cited were Oller and Perkins (197th. 1978b) and Oller( 1977, incorrectly given as 1978). Updates of the entire argument can be found in Oller (1981a and 1981b). 183

Upload: john-w-oller-jr

Post on 30-Sep-2016

229 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: GARDNER ON AFFECT: A REPLY TO GARDNER

EXCHANGE

GARDNER ON AFFECT: A REPLY TO GARDNER’

John W. Oller, Jr. University of New Mexico

In the lead article of the second issue of this journal in 1980, Robert C. .Gardner offers a defense of “the proposition that achievement in a second language is related to attitudinal and motivational variables.” It is claimed that “Oller and his colleagues have questioned the existence of such a relationship” (1980:268). Further, that we have “given the impression that the empirical foundation for it is weak.” In defense of his own previous work, Gardner argues that “it is the responsibility of scientists to investigate phenomena and provide the most parsimonious explanation possible for all the research literature” (p. 256). He writes, “that is precisely what we have done in our investigations of the role of attitudes and motivation in second language learning”(at this point he cites a number of his own previous publications, some with coauthors including W.E. Lambert, L. Gliksman, P. Smythe, C.L. Smythe, and R. Clement).

This reply has two major purposes. The first is to show that Gardner misconstrued the arguments he attempted to refute. The position defended by him has never been under attack in any of the publications by “Oller and his colleagues” which are cited. To the contrary, it has always been assumed (and made explicit in the publications cited by Gardner) that affective factors must play a significant role in the acquisition of any language (primary or nonprimary). Never in any of the publications cited by Gardner, have 1 or my colleagues “questioned the existence of such a re la t ions hip. ”*

The second purpose of this reply is to try to clarify the fact that it is the measurement of affective variables which has been questioned (not their existence, nor their relationship to learning a second language). However, if’the measures of affect are suspect, rhen to this extent, won’t the empirical basis of certain popular theories of how affective factors

I R.C. Gardner, “On the Validity of Affective Variables in Second Language Acquisltion: Conceptual, Contextual, and Statistical Considerations,” Language Learning 30:255-270 (December 1980).

* The main articles cited were Oller and Perkins (197th. 1978b) and Oller( 1977, incorrectly given as 1978). Updates of the entire argument can be found in Oller (1981a and 1981b).

183

Page 2: GARDNER ON AFFECT: A REPLY TO GARDNER

184 Language Learning Vol. 32, No. I

influence second language learning necessarily be drawn into question? On the other hand, it has not been claimed that the popular theories are necessarily incorrect. They may or may not be on the right track. But how can this be determined so long as the instruments used in the measurement of affect are of undetermined validity? This is the trsubling question that we have pondered.

To sustain the first point, consider one of the concluding remarks to Oller and Perkins (1978a; one of the main references cited by Gardner 1980):

None of this, of course, is to deny our own belief that affective variables play a n important role in learning generally, and language learning specifically. (p. 95)

Then, again in Oller and Perkins (1978b; also cited by Gardner 1980) we wrote:

We believe that affective variables are indeed important to learning in a great variety of ways, but we d o not believe that the validity of the commonly used measures has been adequately shown. (p. 422)

Missing the main point of the argument, the Gardner article fends off an assault that has never been mounted. It sets out to prove that there exists a relationship between second language achievement and affective variables, but this proposition has never been questioned. As evidence, certain correlations between an affective inventory and measures of attained proficiency in French as a second language are cited. But how does the demonstration of such correlations simultaneously prove the validity of the affective measures?

Perkins and I (1978a, 1978b) showed that suchcorrelationscould bedue largely to extraneous factors that may enter into responses on affective measures in general. Can this possibility be ignored? More importantly, the demonstration that a correlation exists between responses to the AM1 and learning FSL is not even directly relevant to the question about the validity of the affective measures themselves.

What would be relevant is a demonstration that affective items aimed at the same construct(s) are strongly correlated with each other and not with

’ Neither is the sort of correlation that Gardner points out sufficient to demonstrate the validity of aptitude measures. As suggested by Krashen (1977) it is important to determine concerning the aptitude measures whether they are in fact measuringa propensity to acquire language(s) or possibly only previous learning.

Page 3: GARDNER ON AFFECT: A REPLY TO GARDNER

Exchange I85

affective measures aimed a t distinct constructs. Wouldn’t it be more useful to examine the convergent and divergent validity of the affective measures themselves? Also wouldn’t it be necessary a t some stage to find a method to demonstrate that the extraneous variables discussed in previous research are negligible (if indeed they are negligible)? Or is it safe to assume that the affective measures are valid and continue to interpret the existing empirical work as if this were necessarily SO?^

The idea that affect is difficult t o assess is not a new one. Back in the 1920s, no less a scholar than Sigmund Freud commented that affects are often very intractable. Can it be anticipated that affective factors are less difficult t o get a t in self-report questionnaires than they are in psycho- analysis? Is it not t o be expected that true feelings and beliefs may be more difficult t o elicit through predetermined agreement/ disagreement scales than through detailed psychiatric interviews? If so, how can the present- day confidence in affective measures be sustained without a n adequate research basis?

As Perkins and I attempted to show, among the extraneous variables which may influence the validity of affective measures (and their tendency to correlate with measures of first or second language proficiency) are the tendency to flatter oneself, t o seek social approval (as demonstrated by Crowne and Marlow 1964), and the aim to merely be consistent in responding to affective questionnaires.

Gardner (1980)asks, “But why stop here?To this we could add deliberate faking . . . , etc.” (p, 262). Indeed, why stop with just those factors! Does it improve the prospects for affective measures to lengthen the list of possible extraneous factors which may account for substantial portions of the variance generated by those measures?

Actually, the reason that we brought up just the three factors that we did was because they are all linked up with language proficiency (or verbal intelligence) it would seem, from the very start. T o give responses that are self-flattering, o r that make one look good in the eyes of others, or that are merely consistent with what one has said in previous responses on the same questionnaire, one must comprehend and use the language of the questionnaire a t fairly deep levels. For instance, in a self-report measure of test anxiety the respondent might wonder (though perhaps not in just these words), “If I say that my palms sweat when taking tests, will it fit the image of how I want to look? Or will it help me get the job I am applying

These issues are dealt with somewhat more extensively in Chapter 5 of L U I I ~ , U ~ P 7p.\/.\ (I/

School, 1979, pp. 105-147.

Page 4: GARDNER ON AFFECT: A REPLY TO GARDNER

186 Language Learning Vol. 32, No. 1

for? Or will it be consistent with what I said in response to the question about losing sleep the night before a big test? Am 1 presenting myself as a sane and desirable sort of employee?” and so forth.

Thus, all three of the extraneous factors that we focused our attention on would seem to offer possible explanations for the sbserved correlations between affective measures and measures of second language proficiency. That is, simply put, if the affective variables are measuring deep language proficiency to begin with (through the influence of such extraneous factors), then those measures would be expected to correlate a fortiori (but on a very different basis than the popular theories suggest) with attained proficiency in a second language. This follows in part from the well demonstrated fact that first and second language proficiency are substan- tially correlated (see Oller and Perkins f978c: I 18, and Cummins 1979). Thus, if we measure the one, to some extent we are also measuring (or indirectly predicting success in) the other.

In fact, Perkins and I were able to demonstrate empirically that the self- flattery factor may account for as much as 25% of the variance in certain commonly used affective questions which are presented in the form of Likert-type agreement/ disagreement scales. From this we reasoned that such factors may reduce the validity of affective measures (Oller and Perkins 1978a). However, Gardner (1980) writes, “I failed to uncover any such demonstration in my reading of their paper” (p. 263). Could this be because of the abstractness of the accompanying argument?

The empirical demonstration concerning the strength of the self-flattery factor was carried out in two steps. First, we asked 57 ESL learners (drawn from the University of Michigan’s English Language lnstituteand from the Center for English as a Second Language at Southern Illinois University)’ to indicate which of two opposite personal attributes they considered more desirable on each of eight bipolar semantic differential scales. The scales included contrasting personal traits such as being “outgoing” or “re- served,” “quiet” or “talkative,” “nervous” or “calm,”“cautious” or “happy- go4 uc k y , ” “sociable ” or “solitary , ” “serious ” or “carefree,” “a leader” or “a follower,” “an initiator” or “an observer.”

The second step, then, was to ask them to rate themselves on each of the same scales. To the extent that their ratings of themselves correlated with their valuation of the traits, we inferred that their self-ratings could be

We were assisted by H . Douglas Brown in obtaining the first sample and by Richard Daesh and others in getting the second.

Page 5: GARDNER ON AFFECT: A REPLY TO GARDNER

Exchange 187

judged to be self-flattering. The pertinent correlation matrix was presented as Table I in Oller and Perkins 1978a and as Table 5-3 in the 1978c paper which was a fuller version of the same article.

Since the correlations on the diagonal (those elements showing the relationship between the self-rating and the corresponding valuation on the same scale) were larger in every instance than any of the off-diagonal elements (the correlations between the eight self-ratings on the one hand and the eight scale valuations on the other), it was possible to infer that self- flattery was operating rather strongly. In fact, it seemed to be the main factor determining the magnitude of the various elements of the correlation matrix. Otherwise, scales such as “outgoing” or “reserved,” “quiet” or “talkative,” ‘‘sociable’’ or “solitary” might have been expected to correlate more strongly with each other in the case of either the self-ratings or valuation of the traits. However, it was the corresponding valuations and self-ratings on each scale that always correlated most strongly. This could be taken to indicate that the semantic content of the scales themselves was somewhat less significant than the way the subjects valued that content and wanted themselves to look in relation to that valuation.

The diagonal correlations ranged from .31 to .69 with a mean of S O . Therefore, we estimated that about 25% of the total variance in all of the scales might be attributed to a self-flattery factor. It may be noted that this figure is somewhat larger than the 14% of variance in attained second language proficiency which Gardner (1980:261) believes can be attributed to affective factors. Is it possible that the observed 14% of estimated common variance between language measures and affective measures could be swallowed up in the 25% or so that might be attributed to extraneous variables such as self-flattery, the approval motive, or mere consistency in response to affective questionnaires?

Gardner also reported that he was puzzled by our reference to “strange reversals” in some of the published research findings. Among the unexpected outcomes which we mentioned were the findings on the popular “instrumental/ integrative”dichotomy. Hasn’t the Gardner/ Lam- bert theory generally predicted that an integrative orientation should

’ In any event, even if the problems in affective measures could be satisfactorily resolved, what sorts of educational decisions could be made o n the basis of only 14%, of the total variance in second language achievement? Or would there be enough to work with if the 17% or so that might be attributed to aptitude could be added in? Even if the two sources can be added together (i.e., if they are in fact uncorrelated with each other), only 31%) of the total variance in achievement would be accounted for. This much could generally be predicted by examining measures of previous attainment in the primary language.

Page 6: GARDNER ON AFFECT: A REPLY TO GARDNER

188 Language Learning Vol. 32, No. I

produce faster progress and a higher level of eventual attainment in a second language‘? The research reveals, however (see Oller, Hudson, and Liu 1977, also Oller 1979 for the pertinent references), that an integrative orientation (as measured by various methods suggested by Gardner and Lambert 1972) sometimes seems to be more strongly correlated with attainment in a nonprimary language, sometimes equally, and sometimes less strongly than a n instrumental orientation. True, explanations have always been offered in each new case, but are these post hoc rationaliza- tions completely satisfying? If the theory has merit, shouldn’t it give us advance warning about the outcome of the appropriate kinds of empirical research? Or must we always take our chances and embrace all outcomes equally after the fact?

Aside from “strange reversals” in the research findings, there have been some interesting contrasts in theoretical perspectives advocated by leading authorities in the affective ’domain. For instance, Gardner and Lambert (1972) predicted that higher scores on their “Anomie Scale” should be associated with higher attainment in second language study. However, Leo Srole, who introduced the concept of “anomia”in 1956, wrote that it “is a factor related to the formation of negative rejective attitudes toward minority groups” (p. 7 14). Would such “rejective attitudes” be expected to enhance second language acquisition? Srole also contended that anomia o r “lostness is one of the basic conditions out of which some types of political authoritarianism emerge” (Srole 19567 14, f.n.). But haven’t Gardner and Lambert consistently ,and quite reasonably) maintained that authoritarian and ethnocentric attitudes should be negatively correlated with attainment in a second language? To complicate matters, however, Christie and Geis (l970), of Machiavellian fame, have predicted that there should be no relationship at all between anomia and authoritarianism (p. 360).

Can a consistent theory be made of the foregoing? Isn’t it probable that any explanation which can embrace all of these positions will run into logical difficulties? Apparently, Gardner holds out hope that consistent explanations can a t least be offered for the discrepancies in empirical findings which have been noted in earlier publications. In fact, he offers a paragraph of some thirty-eight lines in his 1980 article (pp. 265-266) showing that possible interpretations for some of the surprising outcomes of empirical studies d o in fact exist. However, he cautions:

I am certainly not proposing that the above interpretations of the so-called negative findings are correct., . I am suggesting simply that there are interpretations which do not

Page 7: GARDNER ON AFFECT: A REPLY TO GARDNER

Exchange 189

invalidate the general proposition that achievement in a second language is mediated by attitudinal and motivational variables. (p. 266)

But the existence of a relationship between second language acquisition and affective factors has not been questioned. That proposition needs no defense. On the other hand, isn’t the correctness or incorrectness of the “interpretations of the so-called negative findings”the very issue that needs to be addressed? Shouldn’t we at some point be concerned to find correct interpretations? And isn’t the possible lack of validity of affective measures one of the interpretations that needs to be considered and, hopefully, ruled out by appropriate research?

REFERENCES

Christie, R., and F. Geis. 1970. Studies in Machiavellianism. New York: Academic Press. Crowne, D.P., and D. Marlow. 1964. The Approval Motive. New York: Wiley. Cummins, J.P. 1979. Cognitiveiacademic language proficiency, linguistic interdependence.

and some other matters. Working Papers on Bilingualism 19: 197-205. Gardner, R.C. 1980. O n the validity of affective variables in second language acquisition:

conceptual. contextual, and statistical considerations. Language Learning 30:255-270. Gardner, R.C., and W.E. Lambert. 1972. Arritudes and Morivarion in Second Language

Learning. Rowley. Mass.: Newbury House. Krashen. S.D. 1977. Second language acquisition and second language testing. Paper pre-

sented at the Annual Meeting of the American Council of Teachers of Foreign Languages, San Francisco.

Oller. J.W., Jr. 1977. Attitude variables in second language learning. In Burt, M.K.. H.C. Dulay, and M. Finocchiaro (eds.). VieH.point.s on English asa Second Language, pp. 172- 184. New York: Regents.

Oller, J . W., Jr . 1979. Language Tesrs ar Sc,hool: A Pragmatic Approach. London: Longman. Oller, J.W., Jr . 198 la. Can affect be measured’? lnternarional Review oJ’Applied Linguisrics. Oller, J.W.. Jr . 1981b. Research on the measurement of affective variables: some remaining

questions. In Andersen, R. (ed.). New Dimen.sion.s in Second Language Acqui.yition Re- search, pp. 14-28. Rowley. Mass.: Newbury House.

Oller, J.W., Jr., A. Hudson, and P. Liu. 1977. Attitudes and attained proficiency in ESL. . Language teorning 27: 1-27. Oller, J.W., Jr. and K. Perkins. 1978a. Intelligence and language proficiency as sources of

variance in self-reported affective variables. Language Learning 2835-97. Oller, J.W., Jr., and K. Perkins. 1978b. A further comment on language proficiency as a

source of variance in certain affective measures. Language Learning 28:417-423. Oller, J.W., Jr.. and K. Perkins. 1978~. Language proficiency as a source of variance in self-

reported affective variables. In Oller and Perkins (eds.), Language in Education: Tesring rhe Tests. pp. 103-125. Rowley. Mass.: Newbury House.

Srole, L. 1956. Social integration and certain corollaries: an exploratory study. Americ,an Sociological Reviena 2 I :709-7 16.