1
INTRODUCTION
This book recounts a history of two sets of ideas. One set of ideas has at its centre the term âuniversal grammarâ, a name enduringly applied in western linguistic tradition to the notion that human languages necessarily share important, formative, properties. The other set of ideas has at its centre an experience, namely the experience of acquisition of a second or foreign language. In the innumerable generations during which humans have participated in varieties of this experience, or observed others doing so, many people have been moved to speculate about its nature. These speculationsâsometimes explicitly worked out as proposals about adult language learning, sometimes implicit in records of how languages have been learned and taughtâform a second set of ideas, a history of which is recounted in this text. I am in particular interested in points of intersection between universal grammar and foreign language learning. From the ancient world up to around the sixteenth century, the focus is on Europe and the Mediterranean area. Since the 1500s, work historically rooted in Europe has been carried out in various parts of the world, so the geographical locus becomes more diffuse from that time up to the present day.
Some basic definitions
One of the ambitions of this text is to describe the several incarnations of universal grammar over the course of many centuries. Another is to describe the western heritage of ideas about second language acquisition. In neither case is there a continuous thread of development, with the discoveries and the reflections of one age always building neatly on those of the previous generation. In particular, the concept of universal grammar has fallen out of intellectual fashion several times, and on being revived, emerged in different guises. During some periods competing concepts of universal grammar have existed side-by-side. Second language learning has likewise been variously construed. But even granted this diversity, it is possible (and, in fact, necessary) to have from the start a working sense of what these two terms have referred to.
âUniversal grammarâ
The fundamental insight captured by the expression âuniversal grammarâ is that human languages have significant properties in common despite their obvious differences. The term âuniversal grammarâ, along with a related term (sometimes used synonymously), âgeneral grammarâ, did not have much currency before the early seventeenth century.1 However, versions of the idea which this term labelsâversions more or less developed, more or less explicitly articulatedâgo farther back in western linguistic tradition. It is useful to start with some illustrations of how participants in this tradition have talked about commonalities among languages.
In 1270 Roger Bacon wrote that âgrammar is substantially one and the same in all languages, despite its accidental variationsâ.2 Although Baconâs words are frequently quoted as an early exposition of universal grammar, several recent historiographical studies have shown them in a different light, an issue addressed in Chapter 4. I cite Bacon here because, legitimately or illegitimately, many people since his day have taken his statement to present âmost clearly and unconditionally the basic principle of universal grammarâ (Hovdhaugen 1990: 118). During the Middle Ages following Baconâs day, one languageâLatinâachieved a preeminent intellectual and social position. Medieval grammarians valued Latin as a unique reflection of human cognition and the structure of reality, and so projected the westâs first explicit notion of what is essential to language from the categories and features of Latin.
After a period of dormancy during the Renaissance, an array of concepts of universal grammar emerged which were no longer yoked to Latin. Rather, grammarians opposed what is universal in human languages to the âparticularâ or âspecialâ properties which distinguish one language from another. Johann Heinrich Alsted defined âgeneral grammarâ in his 1630 Encyclopedia as âthe pattern [norma] of every particular grammarâ (Salmon 1969: 171). The British grammarian John Wilkins (1668: 297â8) contrasted âNatural Grammar (which may likewise be stiled Philosophical, Rational, and Universal)â, with âInstituted and Particular Grammarâ. The former âshould contain all such Grounds and Rules, as do naturally and necessarily belong to the Philosophy of letters and speech in the General â, whereas the latter âdoth deliver the rules which are proper and peculiar to any one Language in Particularâ. To Wilkins, the job of grammarians was to evade being âprejudiced by the common Theory of the languages they [are] acquainted with [so that they may] abstract their rules according to Natureâ. In France, scholars speculated about language commonalities and differences, most famously in the 1660 Grammaire GĂ©nĂ©rale et RaisonnĂ©e. This text, and others of its genre, argued for a rationalist basis for universal grammar, and tried to work out what it would mean to learn a language granted the roots of general grammar in human cognition.
By the end of the eighteenth century, scholars under the influence of nationalism and romanticism shifted their attention away from similarities across languages to the characteristics of individual languages. Universal grammar remained in retreat as comparative-historical and typological studies of languages flourished in the 1800s. Then at the beginning of the twentieth century, Ferdinand de Saussure asserted that languages could, and should, be studied as autonomous systems independent of their histories. Historians conventionally take Saussure as the starting-point of autonomous or structuralist linguistics, which extends to the present day. In the United States, structuralism is usually represented as having developed a decidedly anti-universalist trend in the first half of the twentieth century. To this era belongs Martin Joosâ famous representation of the âAmerican (Boas) traditionâ as one which holds that âlanguages could differ from each other without limit and in unpredictable waysâ (1957: 96)âan assertion which seems to be cited about equally as often as Baconâs seminal thirteenth-century assertion to the contrary. (And, as in the case of Baconâs words, we will return to whether the conventional interpretation of Joosâ statement is justified.)
The kind of extreme anti-universalism which Joos is taken to attribute to early American structuralism has been reconsidered in the last half of the twentieth century. By the 1970s, there were two schools of research centrally concerned with discovering and specifying language universals. One school is associated with the linguistic anthropologist Joseph Greenberg. Greenberg and his colleagues have endeavoured to identify characteristics which are shared across human languages through comparison of (for example) the array of their phonemes, word order facts, or morphological properties. The approach is to analyse the features of many typologically different languages, to discern their common features and the relationships which hold across those features. This represents one modern concept of how to define universal grammar.
A second school is identified with Noam Chomsky. Beginning in the mid-1960s, Chomsky started reconstructing a concept of universal grammar within generative linguistics as a âsystem of principles, conditions, and rules that are elements or properties of all human languages not merely by accident but by necessityâ (1975: 29). According to Chomsky, universal grammar is a theory of innate principles that constitute the starting point of language acquisition, âthe âinitial stateâ of the language learner, hence the basis on which knowledge of language developsâ (1980: 69). âAll languages are cut to the same patternâ (Chomsky 1965: 30), because universal grammar defines what is linguistically possible and impossible. The job of the linguist is to discover the pattern, that is, to recognize and specify the contents of universal grammar. Chomsky and others working with him have developed this version of universal grammar to an unprecedented level of formal precision. Research into child language plays a key role in generative linguistics, because it is argued that children exhibit knowledge of language which cannot have been derived by observation of, or by induction from, their linguistic environment. To paraphrase Chomsky, such knowledge is not merely accidental, but necessary; and therefore study of what child learners bring to the task of language learning provides insight into the contents of universal grammar.
It is appropriate to have concluded this sketch of the history of universal grammar by citing Chomskyâs version of the idea. This is the perspective of much modern discussion of universal grammar as well as second language learning, and thus familiar to some of the readership of this text. One of the tasks of this text is to examine the historical context of Chomskyan universal grammar. It will sometimes be convenient to compare generative and other versions of universal grammar, even at the risk of distorting earlier language science. Such comparison does not assume that Chomskyan linguistics in some way culminates inquiry into universal grammar. Rather, it simply recognizes that generative universal grammar is a useful shared starting point for many readers of this text, as for its author.
In any case, despite the evident differences in views of universal grammar from Bacon to Chomsky, one striking motif is the persistent contrast between the terms ânecessaryâ (or âessentialâ) and âaccidentalâ, perhaps echoing Aristotleâs use of these words in discussing the nature of scientific inquiry. Attributed to universal grammar are those parts or properties of human language that are necessarily present, that is, which form the definitive or essential features of language, granted that such properties have been conceptualized differently in different periods. Features of a given language not governed by universal grammar are characterized as âaccidentalâ, with their presence or absence not crucial to the general definition of a human language.3
There is, of course, much more to the interlocking sets of hypotheses, speculations and assertions which have gone by the name âuniversal grammarâ, but a distinction between necessary and accidental properties is intrinsic to many of them. It is also a distinction whose status within generative linguistics is sometimes misunderstood. For example, G. Steiner (1992: 104) characterizes âChomskian universalismâ as holding that âDifferences between languages represent differences of âsurface structureâ only. They are accidents of the terrain which impress the eye but tell us scarcely anything of the underlying âdeep structureâ â.4 On these grounds Steiner claims that âThe axiom of universal deep structures ⊠entails inevitably a relegation to accident, to superficiality, of the facts of linguistic multiplicity and differenceâ (pp. xivâxv). What Steiner misconstrues is that from a generative perspective, differences between languages may be âaccidentsâ with respect to universal grammar, without making them uninformative or theoretically âsuperficialâ. How languages differ can reveal a lot about how they are alike. A version of generative grammar prevalent in the 1980s, the âprinciples and parametersâ framework, attempts to specify invariant properties (the âprinciplesâ) of grammar in relation to properties that exhibit constrained cross-linguistic variation (properties following from the setting of âparametersâ). âAccidentalâ properties need to be distinguished from those that are ânecessaryâ, but what is âaccidentalâ is not unnecessary to the construction of a universal grammar.
For the benefit of readers encountering the concept of universal grammar for the first time, it may be useful to specify what the term does not refer to. First, it does not refer exclusively to âgrammarâ in its narrow sense, that is, to syntax. Rather, âgrammarâ is conventionally extended to include many domains of language: phonology, the lexicon, morphology and semantics, as well as syntax. Second, although âgrammarâ and âlanguageâ are sometimes assigned partially overlapping senses, âuniversal grammarâ is conceptually distinct from âuniversal languageâ. This text narrates a history of ideas about what has been assumed (or postulated) to be shared across human languages, using this descriptor as a minimal common denominator among different senses of the term âuniversal grammarâ. In contrast, Eco (1995) chronicles how the term âuniversal languageâ has been used to refer to linguistic systems created, or reconstructed, as âperfectâ in one of several ways. For example, invented languages like Esperanto, which aspire to overcome the social or political consequences of linguistic diversity, have sometimes been called âuniversal languagesâ (Salmon 1996a). Another sense of the term âuniversal languageâ is that which some Judeo-Christian traditions consider to have been employed in the biblical act of creation, or by humans before the collapse of the Tower of Babel. Philosophers have also sought to find or to create a âuniversal languageâ in a third sense, namely, a language able âto express ideas perfectly and to discover thereby new connections between the diverse aspects of realityâ (Eco 1995: 3). In some instances, the history of universal grammar abuts or intersects with the history of universal languages in these senses. But the two concepts are distinct.
âSecond language acquisitionâ
The other object of inquiry in this book, the conceptualization of foreign language acquisition, also needs to be specified in some preliminary manner. The experience of learning a language other than oneâs native tongue has probably been available to at least some people within every western culture from the ancient world to the present day (Lewis 1976). But in the last twenty years, terms for describing that experience have been refined to point out several purported distinctions. One such distinction is between adult versus child learners, with questions about the significance of this difference having been debated for centuries and continuing to be controversial.5 I observe this distinction, assuming that there may indeed be essential differences between adult and child language learning. This book is concerned most centrally with adult learners, although information about the acquisition of second languages in childhood, as well as about childrenâs acquisition of a first language, often enters the narrative. A second distinction which contemporary research has sometimes maintained is that between âsecond languagesâ (or âL2sâ) and âforeign languagesâ. A âsecond languageâ refers to a language learned in a context where it is used natively in the surrounding culture (regardless of whether it is a true second language [L2] to the learner, or actually a third [L3], fourth [L4], etc.). In contrast, a âforeign languageâ refers to a language learned, for example, in an educational setting where some other language is employed natively outside the classroom. It is unclear exactly what, if any, consequences follow from this difference (R. Ellis 1994: 214â216). In this text I will refer indiscriminately to a non-natively acquired language as a âforeignâ or a âsecond languageâ (abbreviated as âL2â).
Definitions of the terms âacquisitionâ and âlearningâ have likewise been debated. I use them as synonyms, setting aside Krashenâs (1981) distinction between naturalistic, untutored, âacquisitionâ and the instructed âlearningâ which takes place in a classroom, because I am concerned broadly with the experience of adults encountering a non-native language and with how that experience has been understood, regardless of the context in which the encounter takes place. (I also set aside the somewhat overlapping issue of implicit versus explicit learning (N. C. Ellis 1994).) In research on children, some generativists in the last twenty years have rejected both the terms âlearningâ and âacquisitionâ in favour of âlanguage growthâ as better representing Chomskyâs view that control over a native language appears as âa âlanguage organâ ⊠that develops under strong, discrete genetic and epigenetic constraintsâ within the childâs mind (Marshall 1987: 47).6 Whether a second language can also be best represented as âgrowingâ in the mind of an adult is controversial. Some writers (e.g. Strozer 1994: xii) find neither âlearningâ nor âacquisitionâ fully satisfactory with respect to adults, but I will prescind from the debate to employ both terms interchangeably.
Three caveats
Continuity and discontinuity in notions of universal grammar
Three caveats are in order. First, the sketch given above demonstrates that the fundamental insight captured by the term âuniversal grammarâ has been elaborated variously, weaving in and out of western language science. The source of universal grammar has also been construed variously. Medieval grammarians believed universal grammar to follow from the facts of the phenomenal world; they presumed it to be âdependent upon the structure of the world of realityâ (Bursill-Hall 1963: 46). Some scholars in the seventeenth century identified universal grammar with mental concepts to which language applies linguistic labels. Chomsky (1980: 56â65) locates the seat of universal grammar in an innate computational, not conceptual, faculty, and argues that (generative) universal grammar is what it is because of human genetics.
In these ways, universal grammar has been conceived differently along several dimensions. Moreover, not all periods have been preoccupied with universal grammar. People reflecting on linguistic issues in some eras have simply denied that languages have significant shared properties, or they have assumed that universal grammar exists but neglected it in favour of attending to differences among languages. Nevertheless, the idea of universal grammar arose early in western language science, and there is enough continuity to trace its several metamorphoses. Robins (1973: 17) characterized âthe problem of linguistic universals in relation to the obvious surface differences of language structureâ as one which âin different forms has beset linguists ever since [the Middle Ages]â. Itkonen (1991: 165) goes farther, declaring that âthe history of Western linguistics has to a very large extent been a history of the notion of universal grammarâ (emphasis in the original). Coseriu (1974: 47) gives maximum scope to the idea, arguing that all linguistic inquiry implicitly presupposes some notion of language universals. To ask about any language âWhat are its phonemes?â or âWhat grammatical categories appear?â is, to Coseriu, to accept in advance that that language, like all others, has phonemes and grammatical categories. But however central a role one might posit for universal grammarâor whether one attends more to rejection of universal grammarâit is important to recognize a certain instability of concepts of its nature, contents and source through the history of western linguistics. Brekle (1986) points out that the notion of âlinguisticsâ has to be historically relativized to make the study of its past possible. Similarly, âuniversal grammarâ has to be repeatedly re-defined to reflect the interests of different periods.
Explicit and implicit concepts of second language learning
A second caveat concerns the other focus of this text, second language acquisition. Foreign language learning has taken place in one way or another in most cultures. At the very least, virtually all cultures have recognized the existence of foreign languages.7 But not every society has sufficiently valued foreign language learning to reflect on its own assumptions about what it means to learn another language, or what constitutes knowledge of an L2. Moreover, not every society left records of whatever such reflection did take place. Some of this information can be pieced together from references to language learning and learners, or deduced from pedagogical records. But pedagogy does not always faithfully represent contemporary thinking (or even unconscious assumptions) about the nature o...