INTRODUCTION: PROBLEMS AND TESTS

ly, with only one of these parts. In practice, however, this is seldom done in writing about foreign language testing. One does not generally write about techniques or operations without consideration of variables those techniques should measure; one does not generally write about variables without consideration of their relevance to some other behavior; one does not generally write about problems with no thoughts as to their solutions; one does not generally write about validity except in the context of a problem and the means of accumulating information for its solution; and one does not write and publish anything about testing without, in a sense, adding to a kind of information storage system. Although none of the papers in this volume deals exclusively and abstractly with single ftpartsr7 of the problem solving by testing procedure, each has its own emphasis. They are ordered according to their emphasis on problems, variables, techniques, validation, and information about tests. The first paper, Hopkins' "Language Testing of North American Indians," provides an historical context of testing that group-if it can be properly referred to as a single group. It suggests that some early testing may not have been directed towards solving any particular Indian problems, that those efforts may have been largely wasted. The most pressing current problem, to the solution of which second language information is needed, is curriculum development: placement of students, the identification of student problems, and determination of student progress. Brihre's paper touches on a problem, variables and techniques. Beginning with a problem common to every foreign language teacher-pacing and control of his own teaching-and with a prescribed set of variables-defined by the text and course aims-he shows how instruments are developed to help solve the problem. Thus Briere offers an overview of the problem solving approach to testing. White's paper on the "Second Language Testing Requirements of the Canadian Public Service Commission" shows a vast range of problems, including that of a teacher's control of his own instruction, but also going beyond to complex problems with extremely important consequences, such as personnel selection and awarding of bonuses. White invites suggestions about the nature of variables which will be significant for solving the problems of the Canadian Public Service Commission. Variables to be tested are not "drawn from thin air" by a budding test writer. They are generally fvelementstt or more "elementary processes" of theories about the kinds of behavior which would provide good or bad solutions to a test user's problems. If, INTRODUCTION: PROBLEMS AND TESTS xi for example, a businessman's problem is to select a tgood secretary," the variables he is interested in are those related to good and bad performance of secretaries. A "theory" or "model7q of secretarial behavior might include such "elements" or 'Ielementary processes" as typing speed, spelling ability, regularity of attendance, etc. These then become the variables in a testing program for the selection of secretaries. The variables of most foreign, or second language testing are still "elements" of structural linguistic theories of language. But this set of variables is inadequate for reasonably good solution of many problems faced by those concerned with second language behavior of bilinguals. Whether this inadequacy is due more to the limited compass or to inherent shortcomings in the linguistic theories from which the variables are drawn was not at issue in the conference reported here. All of the participants were, however, conscious of the severe limitations of current language testing prac tice, and recognized that what is being measured is too often inadequate or irrelevant. Two of the conference papers present models of language behavior. From an elaboration of the Ervin and Osgood concept of compound and coordinate bilingualism Jakobovits has developed a psycholinguistic model of bilingual language function. The variables implied by the model's ~~elementsTt and processing operations are reflected in some twenty-seven tests, less than half of which are also based upon elements of a structural linguistic theory. Cooper's "Elaborated Language Testing Model" makes provision for adding "communicative competence" to "linguistic competence." Rejecting the notion of a language as a single code applicable in all social contexts, Cooper makes explicit that acceptable and effective language behavior presupposes not only an implicit knowledge of linguistic rules, but also another set of selection rules which determines the variety of language to be used in a given social context, i.e., which, when applied to linguistic rules, determines what is said, when, to whom and where. The three papers which follow (Pimsleur, Spolsky et al., and Plaister) also reflect foreign language testing variables. The variables, although complex, a re conceived of as essentially unidimensional, in contrast to the multidimensional models of Jakobovits and Cooper. There is also in these papers a much greater emphasis on instrumentation, on the development of tests to provide reliable and t'pure97 measures of specified variables. In addition to providing a comprehensive picture of the development of a testing instrument, Spolsky et al. indicate in their conclusion how a new test for the measurement for one variable may embody useful techniques for measuring other variables as well. xii PROBLEMS IN FOREIGN LANGUAGE TESTING Valette, Lewis and Ingram are all concerned with, among other things, test validity. Valette's paper addresses the problem of establishing criterion measures for foreign language learning. In the absence of external criteria the language teacher can never be certain of the values and efficiency of his own course. That is, if students learn just what they have been taught, their teachers must be concerned also to know the value of that learning. Lewis' paper, although covering many aspects of foreign language testing, is here placed with "validation" papers because the author makes explicit how a subset of variables are selected to be measured in a testing program when one of a larger set of problems is to be solvedvalidity involves the significance of variables for problem solving. Ingram presents the persistent problem (unresolved in this volume) of evaluating the results of educational experiments when data are test scores. When teaching materials or instruction produce meae urable changes in students' performance, do these changes indicate only that the students have been subjected to "experimental manipulation," or do they mean that a desired goal has been achieved? If no changes are observed, is this apparent stability in performance a true indication of no change, or is it due to an instrument which does not record changes which have occurred? Mackey presents arguments for the establishment of a test information repository and retrieval system, and suggests the kind of information that should be included. Savard explicates how such a system is already being established at the International Center for Research on Bilingualism at Lava1 University. Palmer's conference summary, with the discussion it provoked, is especially useful in pointing out the many problems in foreign language testing which received short shrift or were entirely ignored at the conference. There was great concern over misapplication of tests and test information, especially when that information contributes to decisions about the careers of examinees. The caveat that test users cannot wait for "perfect tests" notwithstanding, much of the misapplication problem is attributable to a paucity of tests which measure significant variables and which are suitable for large bilingual populations heretofore largely ignored by test writers. I must apologize to authors whose intentions have been misconstrued or misrepresented. Apologies are due also to participants whose "prize comments" have been omitted or whose dialogue may have been unfairly altered during the process of transcription and edition. For all of these e r rors the editor must assume full responsibility.