In the course of the following essay I have striven to be accurate in acknowledging its obligations to my numerous teachers. In some places however I may have unconsciously reproduced what I have heard or read, —a mistake almost inevitable to those whose information is derived partly from lectures and partly from books, or in others equally unconsciously I may be reproducing what I have not heard or read but what has been already suggested, and accepted or refuted long ago. And further than this, some of the principles on which my small superstructure is based I may have regarded, perhaps prematurely, as the common property of scholars and needing no formal acknowledgment. This is especially the case with borrowings from the Grundriss der Vergleichenden Grammatik. To this references are generally given in specific points, but no number of references could in the least express my sense of the debt which all students of language owe to its distinguished author, Dr Karl Brugmann. And I am anxious, in view of the one or two points in which I have ventured to differ from his authority, to acknowledge beforehand my share in the stimulating influence his teaching has everywhere exerted on the study of linguistic science.

In all that concerns Latin I owe very much to Dr Roby's lists of parallel forms which have been constantly before me. The evidence in Umbrian, Oscan and the Minor Dialects could hardly have been found except in Bücheler's Umbrica and Zvétaieff's two handbooks, the Sylloge Inscriptionum Oscarum, and the Inscriptiones Italiae Mediae Dialecticae. Osthoff's Geschichte des Perfects is of course indispensable, however much one may differ from his conclusions.

The essay was written in March last as a dissertation for the Language Section of the Classical Tripos, Part II, 1887, and has since been thoroughly revised and enlarged by the Appendices. The last of these is of course only printed as a suggestion.

GONVILLE AND CAIUS COLLEGE,

CAMBRIDGE, September, 1887.

VERNER'S LAW IN ITALY.

INTRODUCTION.

1. THE attempt made in the following essay to reduce to rule a series of phenomena hitherto neglected or First princiunnoticed is one which, by this time, hardly needs ples. an apology even in England. It is now generally admitted that the modern view of phonetic change is fully justified by its results. The more or less a priori considerations by which its supporters at first sought to defend it were of a kind whose convincing power varies greatly with different minds and even with the same mind at different times; but the mass of evidence they have since accumulated is surely sufficient to establish at least this principle, that sound-change, so far as we know it historically, whatever possibilities we may reserve for it in the abstract, happens only in accordance with certain definite sequences which we call Phonetic Laws;-to establish it, that is, as far as any doctrine can be by purely inductive evidence, a basis, however, which has proved sufficient for the whole fabric of Natural Science. And we may perhaps notice that the power of prediction, which is popularly regarded as the crucial test of all scientific theories, may be said in a sense to have been exercised successfully on behalf of the new principles of the Science of Language. Fresh instances occur every day of stray words that have at length been brought home to their correlatives in other languages after having for long resisted identification through some difficult change of form or meaning, simply because we have been led to expect,

that is, we have predicted, that the form which the original sound would take in that language was the one which afterwards has been recognised in this particular word. Conversely, when we know precisely what origin or origins a particular group of sounds existing in any word can have in the language to which it belongs, and precisely what their correlatives are in kindred languages, our field of search for cognates is immensely narrowed, and if they have survived, always provided our phonetic generalisations are correctly made, they are sure to be sooner or later discovered.

2. The endeavour therefore to arrive at further generaliScope of the sations of this kind, whether in any particular Essay. instance it succeed or fail, may be fairly regarded as a legitimate method of work, and single explanations and inferences, which while strictly in accordance with the rules of sound-change that we have already recognised, might nevertheless, if advanced for their own sake, be considered over fanciful or unduly emphasised, may perhaps claim a more generous indulgence if they help in any way to throw light on the possibility of such a result. And in this case, whether the general rule is finally accepted in the form in which it presents itself in this essay is a matter of small importance; I shall be more than content if I succeed in achieving two things; if I can render any clearer the probability that there is some rule to be discovered, and any easier for more experienced hands the task of determining its final form. It is in view of the first of these objects that I have endeavoured to rearrange under a new method of grouping many classes of facts already well known; as for example in dealing with the Latin changes, most of the words I have discussed will be found somewhere in the collections of Roby, Corssen, Brugmann, Stolz, or Mommsen'; for the Romance languages I have depended entirely on Diez's Grammatik der Romanischen Sprachen, except for Italian. My second and more important object was to collect new evidence on the question where it

1 Roby's Latin Grammar, Vol. 1. Corssen, Aussprache des Lat. Brugmann, Grundriss. Stolz, Lat. Grammatik. Mommsen, Unteritalischen Dialekten.

had not been sought before,-not at least with any approach to exhaustiveness, namely, in the various Italic dialects, especially of course in Umbrian and Oscan. This I have endeavoured to carry out as thoroughly as possible, that is in such a way as to give with equal fullness what evidence there is on both sides, against, as well as in favour of the conclusions which commend themselves to me.

Title.

3. The title 'Verner's Law in Italy' suggests the origin of the enquiry undertaken in what follows. The apparent irregularities of the change of s to r in the Teutonic languages were explained by Verner as due to difference of accent in the different words; the Latin accent we know to have prevailingly the same exspiratory character as the Teutonic; and an attempt to apply Verner's method of explanation to the Latin irregularities led me to the conclusion that the exceptions to the rule were governed by a new set of special conditions closely analogous to those which determine the change of s to r in Sanskrit. My impression that in Latin it was largely governed by accent was strengthened by the parallelism of one or two of the Umbrian forms, and this naturally demanded a detailed investigation'. The Appendix on the history of s in Aryan and Romance is really a necessary piece of evidence only separated for convenience. The change of s to r at the end of words in later Umbrian is discussed in the

་

1 In speaking of the title I may perhaps deprecate an objection that Verner's Law does not hold in the Italic languages for the other spirants (ƒ) and therefore presumably not for the sibilants.' If no positive evidence were to be had on one side or the other this a priori argument might be allowed some weight, but it can hardly count for much as balanced against such evidence. Besides it is clearly unscientific to demand that any particular phonetic cause shall have exactly the same apparent effect in all the languages in which it has any effect at all: it is, e.g. no evidence against the loss of original σ between vowels at some period in the history of the Greek spoken in Attica, that in Laconian every a whether original or hysterogen equally fell away; yet Attic and Laconian in other respects are obviously far more closely related than Italic and Teutonic. And, after all, in the still confused state of our knowledge of the history of the aspirates in Italic (v. Br. Gds. § 389 Anm. which appears to be a mere slip) it would need considerable boldness to assert that something analogous to Verner's Law had never been in operation at some period of their development.

that is, we have predicted, that the form which the original sound would take in that language was the one which afterwards has been recognised in this particular word. Conversely, when we know precisely what origin or origins a particular group of sounds existing in any word can have in the language to which it belongs, and precisely what their correlatives are in kindred languages, our field of search for cognates is immensely narrowed, and if they have survived, always provided our phonetic generalisations are correctly made, they are sure to be sooner or later discovered.

2. The endeavour therefore to arrive at further generaliScope of the sations of this kind, whether in any particular Essay. instance it succeed or fail, may be fairly regarded as a legitimate method of work, and single explanations and inferences, which while strictly in accordance with the rules of sound-change that we have already recognised, might nevertheless, if advanced for their own sake, be considered over fanciful or unduly emphasised, may perhaps claim a more generous indulgence if they help in any way to throw light on the possibility of such a result. And in this case, whether the general rule is finally accepted in the form in which it presents itself in this essay is a matter of small importance; I shall be more than content if I succeed in achieving two things; if I can render any clearer the probability that there is some rule to be discovered, and any easier for more experienced hands the task of determining its final form. It is in view of the first of these objects that I have endeavoured to rearrange under a new method of grouping many classes of facts already well known; as for example in dealing with the Latin changes, most of the words I have discussed will be found somewhere in the collections of Roby, Corssen, Brugmann, Stolz, or Mommsen1; for the Romance languages I have depended entirely on Diez's Grammatik der Romanischen Sprachen, except for Italian. My second and more important object was to collect new evidence on the question where it

1 Roby's Latin Grammar, Vol. 1. Corssen, Aussprache des Lat. Brugmann, Grundriss. Stolz, Lat. Grammatik. Mommsen, Unteritalischen Dialekten.