amples of analogical -en plurals of nouns that were not weak in Old English. There are also some -en plurals that may be interpreted as the Middle English development of the Old English strong dative plural in -um, but the difficulty of discriminating between dative and accusative in this text makes impossible any more definite statement as to the relative frequency of -en and -e in the dative plural of strong nouns. In verb forms final n is nearly always retained.

In the London English of Chaucer there is complete loss of final n in the singular of weak nouns. There is also complete loss of the nasal in both the strong and weak adjective inflection, except for the survival of the Old English ending -an or -um in -self compounds such as my selven, etc. In the great majority of weak nouns the analogical plural ending -es has displaced the historical -en, but in the very small number of weak nouns that retain the historical inflection the ending is always -en and never -e; there are a few nouns that have analogical -en plurals. With regard to verb forms it would be rash in the light of our present knowledge to make a much more definite statement than that neither loss nor retention of final n was complete in any form of the verb.2

Two conclusions, I believe, can be drawn (tentatively, at least) from the facts summarised in the preceding paragraphs. From the fact that some loss of n occurred in all grammatical forms and that in texts like the Owl and the Nightingale there was an approximation to complete loss of n we may infer that we have to do here, in part at least, with the results of sound-change. But from the fact that in all texts, including texts like the Owl and the Nightingale, the loss of n was more complete in some grammatical forms than in others we may infer that the distribution of forms with and without n was not the result of sound change alone. It seems reasonable to expect that sound change alone would result in actual speech either in: (1) complete loss of n; or in (2) a distribution of forms with and without n that would correspond to phonetic

1 The same exception must be made for the Owl and the Nightingale, the Ancren Riwle, and all the other earlier Middle English texts in which loss of final ʼn in the strong and weak adjective inflection is almost or quite complete.

2 I can only partly agree with Wild's opinion "dass die n-losen formen bei Chaucer eigentlich die normalformen sind, die n-formen aber zur beseitigung des hiatus von dem dichter fakultativ verwendet wurden" (Die sprachlichen eigentümlichkeiten der wichtigeren Chaucer-handschriften und die sprache Chaucers, p. 296), but it seems fairly clear that the Ellesmere scribe's partiality for the -en forms should not be accepted as representing Chaucer's own usage.

› Orm has no loss of final n in verb forms except the present subjunctive plural, but his spellings cannot be relied on as proving more than that n was usually retained in verb forms.

categories, not grammatical categories; in written documents, however, the phonetic distribution occurring in speech might very probably be so imperfectly represented that the texts would show (3) a distribution of forms with and without n that would correspond neither to phonetic nor grammatical categories. The distribution that actually occurs is most reasonably accounted for on the hypothesis that some other non-phonetic process operated along with or subsequent to sound-change. This factor we may call in general terms analogy.

When, however, we inquire as to what analogy operated, it is evident that the data we have been considering do not supply the answer. must know when analogy began to affect the distribution of forms with and without n before we can form an opinion as to what these analogical processes actually were. Analogy must have begun to operate and must have established a definite trend of linguistic development at a period earlier than that reflected in the literary texts of the thirteenth century, for even the earliest thirteenth century texts reflect a late stage of development with respect to loss of n. We must appeal, therefore, to texts that reflect the speech habits of the twelfth rather than the thirteenth century.

The texts that are available are twelfth century transcriptions of works that were originally composed in the tenth or eleventh century, contained in such MSS as Cotton Vespasian D 14, Bodley 343, Lambeth 487, Laud 636, Hatton 38, and Harleian 6258. Such a text is not ideal material for linguistic investigation, for its written form cannot be depended on as reflecting consistently the speech habits of the twelfth century scribe who copied it, but is probably determined in some degree by the written form of the earlier text it was copied from. Nevertheless, these texts, when subjected to adequate linguistic analysis, are capable of yielding information as to twelfth century speech habits that we cannot obtain from other sources.

The frequent failure of these twelfth century scribes to reproduce the final nasals that occurred in the Old English texts they transcribed is evidence (especially when taken in connection with what we know of the later history of final n in Middle English) that final n was not always pronounced in the spoken language of the scribes. We may

The Worcester Fragments of the Address of the Soul to the Body are not transcriptions so far as we know; the Winteney Rule, though a transcription, is in a thirteenth century MS; the "Peterborough additions" in the Laud MS are not necessarily transcriptions.

assume that the scribes did not pronounce the final n's they failed to write, but we are not justified in assuming that they pronounced all the final n's they did write. Since these transcriptions were copies of older MSS in which final nasals were generally retained, it is safer to assume that a scribe sometimes copied forms with final nasals as they were written in his original and sometimes wrote the forms that he pronounced himself. Now if loss of final n had occurred indiscriminately and to an equal extent in all the grammatical forms of his speech, the proportion of forms with and without n in his transcription would tend to be approximately the same in the different grammatical categories and the variation in proportions would be the result of his happening to copy the final n relatively oftener in some forms than in others. The distribution of forms with and without n among the different grammatical categories would be a chance distribution, but the differences would not be very great. And as the distribution was the result of chance we should expect to find when we compared a number of these texts with respect to the relative frequency of n-forms and n-less forms in the different 'grammatical categories that the distribution would not be the same in all but would vary from text to text. Moreover, if we consolidated our statistics so as to show the proportion of n-forms and n-less forms in the different grammatical categories for all our texts, we should expect to find that the proportions in the consolidated tabulation would be much more nearly equal than those that we found in the individual texts. Now this is both what we do and what we do not find.

Table I, printed below, shows the distribution of forms with and without n in the following texts:5

1. Peterborough Chronicle, ed. Plummer, Two of the Saxon Chronicles Parallel, A.D. 1048-1121, I, 171–250.

MS Laud 636; 1122; (538) 6

2. Gospel of Saint Matthew, ed. Skeat.

MS Hatton 38; 1154-1189; (903)

'The date of the MS or the century to which it is assigned is given after the MS notation. This date is usually that given by the editor of the text and is of course only approximate. The number in parenthesis at the end is the total number of forms, with and without n, of all five grammatical categories in each of the texts. The total for all the texts together is 6760.

This MS is in one hand up to the year 1121 and thereafter in various hands. I examined the text from the year 1023 but found no examples of loss of n until 1048. My tabulation of the data does not include the "Peterborough additions", for which see No. 13 below.

3. Medicina De Quadrupedibus, ed. Delcourt, Anglistische Forschungen, XL. MS Harleian 6258; c. 1150; (188)7

4. Peri Didaxeon, ed. Löweneck, Erlanger Beiträge, XII.

MS Harleian 6258; XII; (450)7

5. Early English Homilies, ed. Warner, E.E.T.S., pp. 1–33.

MS Cotton Vespasian D 14; XII; (680)

6. Passio Beatae Margaretae, ed. Assmann, Bibliothek der angelsächsischen Prosa, III, 170-180.

MS Corpus Christi College Cambridge 303; XII; (244)

7. Winteney Version of Rule of St. Benedict, ed. Schröer.

MS Cotton Claudius D 3; XIII, first quarter; (1347)

8. Nativitas Sancte Marie, ed. Assmann, Bibliothek der angelsächsischen Prosa, III, 117–137.

MS Bodley 343; XII; (171)3

9. Twelfth Century Homilies I, ed. Belfour, E.E.T.S., pp. 2–76.

MS Podley 343; XII; (794)

10. In Die Pentecosten; De Octo Viciis; ed. Morris, Early English Homilies, I, 87-119.

MS Lambeth 487; XII; (286)

11. De Initio Creature, ed. Morris, Early English Homilies, I, 217-231.

MS Cotton Vespasian A 22; c. 1200; (124)

12. History of the Holy Rood Tree, ed. Napier, E.E.T.S.

MS Fodley 343; XII; (190) 10

13. Peterborough Chronicle, ed. Plummer, op. cit., "Peterborough additions," I, 29-33, 35-37, 39, 52f., 65, 71, 115-117, 127, 144, 163, 183, 198f., 202, 203, 205-207, 209, 234, 238, 241, 245f., 247.11

MS Laud 636; c. 1122; (96)

14. Fragments of Address of Soul to Body, ed. Buchholz, Erlanger Beiträge, VI, 1-10.

MS Worcester; XII; (123)

15. Twelfth Century Homilies II, ed. Eelfour, E.E.T.S., pp. 78–140.1 MS Podley 343; XII; (626)

'Texts 3 and 4 appear, according to the information kindly furnished me by J. P. Gilson, Esq., of the MSS Department of the British Museum, to be in the same hand.

Texts 8, 9, 12, and 15 are all, according to information kindly furnished by Mr. George Watson of Oxford, in the same hand.

'J. P. Gilson, Fsq., of the Iritish Museum wrote in reply to my inquiry as to the date of these homilies that he would be disposed to put them "about the beginning of the thirteenth century”.

10 See note 8 above.

11 As to these see Plummer, op. cit., II, xlv ff., esp. liv. These additions may be readily recognized from their content and their linguistic characteristics.

12 This part of the MS is, as stated above in note 8, in the same hand as texts 8, 9, and 12, but the distribution of forms with and without n in this part is notably different from that in the other three texts.

Table I is divided into three parts, A, B, and C. In A is given the percentage of loss of final n for each of the fifteen texts in five grammatical categories. The grammatical categories used are the singular of weak nouns, the weak adjective, the dative singular and dative plural of the strong adjective, the plural of weak nouns, and the dative plural of strong nouns. The percentage was obtained by dividing the total number of forms of a given grammatical category into the number of forms of that category that had no final n. E.g., text 1 has 122 examples of the singular of weak nouns in which the n is retained and 7 in which the n is lost; the percentage was obtained by dividing 129 (122 plus 7) into 7, giving .054, which was counted as .05. The bottom line of A, marked "Total", shows the percentage of loss of final n in each of the texts for all five grammatical categories taken together. The last column of A, marked "Total", shows the percentage of loss of final n for each of the five grammatical categories in all the texts taken together. The percentages were obtained as before, by dividing the total number of forms with and without n into the number of forms without n.

In B the data on which A is based are presented so as to show the percentage of loss of n in the singular of weak nouns as in A and in the strong and weak adjective taken together and in the plural of weak and strong nouns taken together, the percentages being obtained by the same method as was used for A.

In C the same data are presented but further assembled so as to show the percentage of loss of final n in the singular of weak nouns, the weak adjective inflection, and the strong adjective inflection on the one hand and in the plural of weak and strong nouns on the other, the percentages being obtained by the same method as was used before.

The loss of n in verb forms in these texts is not shown in the table. The majority of texts show no loss of n in verb forms (or a single example that may be due to miswriting). The Hatton Gospels and Medicina de Quadrupedibus show a very small percentage of loss of n in verbs and Peri Didaxeon, the Winteney Rule, De Initio Creature, and the Peterborough Additions show a considerable loss of n. It is evident that the frequency of loss of final n in verbs has no consistent relation to the frequency of loss of final n in the noun and adjective inflection.

I have stated above that the distribution of forms with and without n in the twelfth century texts both is and is not of the kind that we should expect to result from chance. When we examine the percentage of loss of n in the first three grammatical categories (singular of weak nouns, weak adjective, and strong adjective) we see that the percentages are