The obsolescence which plays a part in many semantic changes, need not present any characteristics other than those of ordinary loss of frequency; what little we know of fluctuations in this direction will apply here. The expansion of a form into new meanings, however, is a special case of rise in frequency, and a very difficult one, since, strictly speaking, almost any utter-ance of aform is prompted by a novel situation, and the degree of novelty is not subject to precise measurement. Older students accepted the rise of marginal meanings without seeking specific factors. Probably they took for granted the particular transferences which had occurred in languages familiar to them (foot of a mountain, neck of a bottle, and the like). Actually, languages differ in this respect, and it is precisely the spread of a form into a new meaning that concerns us in the study of semantic change.
The shift into a new meaning is intelligible when it merely reproduces a shift in the practical world. A form like ship or hat or hose designates a shifting series of objects because of changes in the practical world. If cattle were used as a medium of exchange, the word fee 'cattle' would naturally be used in the meaning 'money,' and if one wrote with a goose-feather, the word for 'feather' would naturally be used of this writing-implement. At this point, however, there has been no shift in the lexical structure of the language. This comes only when a learned loan-word pen is distinct from feather, or when fee on the one hand is no longer used of cattle and, on the other hand, loses ground in the domain of 'money' until it retains only the specialized value of 'sum of money paid for a service or privilege.'
The only type of semantic expansion that is relatively well understood, is what we may call the accidental type: some formal change - sound-change, analogic re-shaping, or borrowing - results in a locution which coincides with some old form of not too remote meaning. Thus, Primitive Germanic *['awzo:] denoted the 'ear' of a person or animal; it appears as Gothic ['awso:], Old Norse eyra, Old German ora ( > modern Dutch oor [o:r]), Old English ['e:are], and is cognate with Latin auris, Old Bulgarian [uxo], in the same meaning. Primitive Germanic *['ahuz] denoted the grain of a plant with the husk on it; it appears in Gothic ahs, Old Norse ax, Old German ah and, with an analogic nominative form due to oblique case-forms, Old German ahir ( > modern Dutch aar [a:r]), Old English ['ehher] and ['e:ar], and is cognate with Latin acus 'husk of grain, chaff.' The loss of [h] and of unstressed vowels in English has made the two forms phonetically alike, and, since the meanings have some resemblance, ear of grain has become a marginal (transferred) meaning of ear of an animal. Since Old English [we:od] 'weed' and [we:d] 'garment' have coincided through sound-change, the surviving use of the latter, in widow's weeds, is now a marginal meaning of the former. Of course, the degree of nearness of the meanings is not subject to precise measurement; the lexicographer or historian who knows the origins will insist on describing such forms as pairs of homonyms. Nevertheless, for many speakers, doubtless, a corn on the foot represents merely a marginal meaning of corn 'grain.' The latter is a continuation of an old native word; the former a borrowing from Old French corn ( modern bound [bawnd], past participle of bind), that a new-formation bound [bawnd] replaced it. The result is that bound in such phrases as bound for England, bound to see it figures as a marginal meaning of the past participle bound. Both the word law and its compound by-law are loan-words from Scandinavian. The first member of the latter was Old Norse [by:r] 'manor, town'-witness the older English forms bir-law, bur-law - but the re-shaping by-law turned it into a marginal use of the preposition and adverb by.
Beside the central meaning please 'to give pleasure or satisfaction,' we have the marginal meaning 'be willing' in if you please. This phrase meant in Middle English 'if it pleases you.' The obsolescence of the use of finite verbs without actors, and of the postponement of the finite verb in clauses, the near-obsolescence of the subjunctive (if it please you), and the analogic loss of case-distinction (nominative ye : dative-accusative you), have left if you please as an actor-action clause with you as the actor and an anomalous marginal use of please. The same factors, acting in phrases of the type if you like, seem to have led to a complete turn-about in the meaning of the verb like, which used to mean 'suit, please,' e.g. Old English [he: me: 'wel 'li:ka?] 'he pleases me well, I like him.'
Partial obsolescence of a form may leave a queer marginal meaning. To the examples already given (e.g. meat, board) we may add a few where this feature has led to further shifts. The Latin-French loan-word favor had formerly in English two well-separated meanings. The more original one, 'kindly attitude, inclination,' with its offshoot, 'kindly action,' is still central; the other, 'cast of countenance,' is in general obsolete, but survives as a marginal meaning in ill-favored 'ugly'. In the aphoristic sentence Kissing goes by favor, our word had formerly this marginal value (that is, 'one prefers to kiss good-looking people'), but now has the central value ('is a matter of inclination'). Similarly, prove, proof had a central meaning 'test' which survives in the aphorism The proof of the pudding is in the eating; this was the meaning also in "The exception proves the rule", but now that prove, proof have been shifted to the meaning '(give) conclusive evidence (for),' the latter phrase has become a paradox.
The old Indo-European and Germanic negative adverb *[ne] 'not' has left a trace in words like no, not, never, which reflect old phrasal combinations, but has been supplanted in independent use. Its loss in the various Germanic languages was due partly to sound-change and led to some peculiar semantic situations. In Norse it left a trace in a form which, owing to its original phrasal make-up, was not negative: *[ne 'wajt ek hwerr] 'not know I who,' that is, ' I don't know who,' resulted, by phonetic change, in Old Norse ['n?kurr, 'nekkwer] 'someone, anyone.' In other phonetic surroundings, in pre-Norse, *[ne] was entirely lost. Some forms which were habitually used with the negation must have got in this way two opposite meanings: thus, an *['ajnan] 'once' and a *[ne 'ajnan] 'not once, not' must have led to the same phonetic result. Actually, in Old Norse, various such expressions have survived in the negative value: *[ne 'ajnan] gives Old Norse a 'not'; *[ne 'ajnato:n] 'not one thing' gives Old Norse at 'not'; *[ne 'ajnaz ge] 'not even one' gives Old