Conclusion

This examination of Lanval and Graelent presents telling evidence of the existence in twelfth century French literature of number systems involving the total of lines used in discrete segments of text. The fundamental similarity between the two poems allows a very precise comparison of the two narratives on the basis of the constituent narrative blocks, whose presence can be verified simply and accurately. Most important are the opening lines of sections which allow the reader to see exactly where the corresponding segments of the two poems begin. The status of these segments as discrete units is then confirmed by the narrative functions they perform, each with an inherent unity of time, place and action, like the scenes in a French classical drama, and sometimes stylistic features set up contrasts with neighbouring sections. Once the essential unity of each section is established underlying arithmetical patterns emerge: Lanval is composed in 17×38, Graelent in 9×84.

In a D. Phil. thesis now in the Bodleian Library, Oxford, (Studies in the Technique of Old French Couplet Verse, Oxford 1981), I have argued that the basis of Marie's lai is an old 72-line tripartite canto structure used in the Brendan (20\s+20\s+32), which has been made bipartite (36\s+36) and then overlaid with the 38-line signature. The Graelent poet, too, draws on the old canto structure, but this time that of 84 lines (20\s+32\s+32), which he has subdivided into three equal parts (28\s+28\s+28) and distributed asymmetrically into two scenes (often 28\s+56, or vice versa). Graelent is thus a systematic expansion of Lanval in the ratio of 6:7.

These numerical systems are moreover, applied with considerable sophistication, with halves, single and double units, and short-term excesses and shortfalls of two or three couplets are carefully balanced out over long spans of text. Structure and number can, therefore, be seen as tightly integrated and the techniques of medieval composition as circumscribed and disciplined, imposing upon the poet the need to express himself, like the craftsman, within a mould whose dimensions were known from the outset.

Linguistics for the most part no longer write grammars or grammatical sketches. Common in the past, these forms have largely been replaced by the journal article. Problems, rather than languages, constitute the unit of research. The reasons for this shift are complex. They reflect, in part, the idea that grammar writing is descriptive and non-theoretical, to be valued chiefly as grist for the theoretician’s mill. Indeed, working at the level of ‘the whole grammar’ or even a sizable fragment, it is very difficult to maintain a consistently theoretical stance. Shorter pieces are obviously better suited for a sharp focus on particular questions. However, large-scale theoretically informed descriptions should have a place in modern linguistic discourse, for there are good reasons for attempting to deal with a language in some breadth. Obviously, principles which can account for larger factual domains are correspondingly more secure. And ultimately, the results of highly focussed studies will be valid only to the extent that they can be integrated into fuller descriptions of the language.

Speech consists of sounds, and writing of visible marks on a surface, and so both are concrete and exist in the material world. But the sounds of speech and the marks of writing are not language itself. Language is abstract or immaterial, something we learned as small children by listening to other people using it, not by being taught. All children, at some stage, say things like,

I’ve drinked my milk all up;

He had his foot bended like this;

He’s drawed on the window ledge;

I put a pillow that way and then sleeped on it;

all authentic recordings of a five-year-old girl. She did not learn drinked, bended, drawed and sleeped from her parents. She has learned for herself the grammatical rule that says ‘to form a past tense, add -ed to the verb’, by hearing words like rubbed, walked, and cooked many times, and then applying the rule without consciously knowing what she is doing. She obviously knows what a verb is too, and you can be sure her parents have never mentioned the words verb or past tense to her. What she has yet to learn is that a large number of verbs in English are irregular, and don’t form a past tense by adding -ed, though she will have heard lots of them, including drank, bent, drew, slept and so on.

The Generalized Quantifier Theory, hencefore “GQT”, initiated with Barwise and Cooper (1981) has elicited promising new lines of research and theorizing in semantics, in particular by the investigation of the algebraic properties of natural language quantifiers. Not debating the general relevance of the conceptions developed in GQT, I will try a critical revision of the empirical claims of the theory.

In order to empirically test the theory of adult L2 acquisition outlined in Chapter 3, we must look at the L2 acquisition of English by adult speakers of two different languages, and relevant correspondences between these two languages and English should vary in meaningful ways. The two language groups chosen for this purpose are Spanish and Japanese, Aspects of these languages that are important for a test of the proposed theory of adult L2 acquisition will be reviewed in this chapter. This review will include a general description of properties of word order, head-direction, and anaphora,¹ as well as a discussion of properties of anaphora in adverbial adjunct clauses in each language.