can supply us with any scale of sounds, ranged accordingly to the difficulty of their production, which may throw light on the changes which we shall find to have taken place in Greek and Latin. I do not propose to lay down any absolute and invariable scale, and to say, this sound is universally more difficult than that, and therefore this or that change took place in Greek and Latin. I shall hereafter describe the process of change in each language as it is historically traceable. But I believe that each and every change had a reason; whether that reason was some peculiarity (as no doubt it generally was) of the special people, or whether it was some physiological fact which is of much wider and more general application. Let us see whether such facts can be obtained from the details given above. It is quite possible that we may interpret those details wrongly, or apply right conclusions erroneously. If so, it is only to be hoped that subsequent research may set us right.

CH. IV.

General rules to the strength

determine

of sounds.

tary

sounds

stronger

than protracted.

First of all, we may assert with confidence that a mo- Momenmentary sound is stronger than a protracted one, and therefore we may expect to find, as we actually do, that a momentary sound passes into a protracted one, but not vice versa, except from some assimilating influence which is sufficient to explain the apparent irregularity. It is clear from the nature of the sounds that the complete check given for a moment to the breath must require a stronger effort on the part of the organs of speech, than is needed when there is no perfect stoppage, but the stream of air is suffered to flow on in a slightly altered current until it is exhausted; just as the mill-dam endures a more violent pressure than the breakwater over which the stream rushes.

sounds

stronger than soft,

Next, among the momentary sounds, the hard will be Hard stronger than the soft, each in its own class. It is true that the difference of effort in producing the two sounds is hardly perceptible. Physiologically we have seen that the difference between them is this: a greater rush of air

each in

their own

class.

CH. IV.

Unaspirated sounds

stronger than the corre

sponding aspirates.

passes through the glottis for the hard than for the soft sound, because the glottis is at least more open for the first class; whether it be entirely open or not (see page 64, note) is not material. Accordingly greater effort is needed to check the fuller column of air, though not so much greater as to be necessarily perceptible. It is not necessary that man should consciously economise his labour at each utterance: it is sufficient for my principle that he should consciously or unconsciously form such habits as actually involve the least expenditure of muscular energy, even where the saving may be infinitesimally small. There must be some physiological reason to explain the fact that in almost every language the hard passes into the soft, so far as change between them takes place at all. The only exceptions I know are that part of the Teutonic Lautverschiebung, or "shifting of sound," by which g becomes k, d becomes t, and b becomes p: and a certain tendency in the same direction in modern Welsh, not, I think, in the older Keltic, e.g. to pronounce "God" as Cot. Sometimes we find savages who pronounce no soft sounds at all, only hard ones: but this is different. I grant the difficulty of the Teutonic change, and can only account for it as an instance of the striving for distinctness', which sometimes acts counter to the principle of phonetic change. It may possibly be due to admixture

of race.

Next, the aspirate is weaker than the corresponding unaspirated letter. This follows from the nature of the aspirates, of which I have already said something2, and shall have more to say hereafter: the breath heard in each case follows upon less permanent, that is, less strong, contact. On this theory it no longer seems unnatural that the more voluminous kh should be weaker than k, or gh than g. But when the sound denoted here by h became at last, as I believe it did, not a subsidiary breath, but an independent sound, the spiritus asper; then kh, 1 See Chapter 1. p. 4. 2 See p. 60.

gh, &c. must be treated as compounds, subject to the ordinary influences which affect compounds, such as loss of one of the members, or assimilation of one member by the other. This is the reason why, though gh be weaker than g, we can yet find in Latin, and perhaps in Greek, g in the place of original gh: e.g. ang-ustus from IndoEuropean AGH: gh has become a double sound, and the h has been dropped from the end of the compound. That the aspirates were in their origin later than the unaspirated letters, can be best seen in Sanskrit'-a language which especially deserves our thanks for performing within historical times so many of the oldest processes of language.

CH. IV.

Of the dif ferent classes the

are the

The hard then is naturally stronger than the soft, and the unaspirated than the aspirated letter. What now is the relation of one hard to another hard? What gutturals law of strength governs the exchange which we some-strongest. times find between one class and another, guttural and dental, dental and labial, &c.? Where there is a complete check, the rule that their strength varies as their distance from the lungs seems to be absolute; that is, that the gutturals are stronger than the palatals (in those languages which possess palatals), the palatals than the dentals and the labials: and this is the order of the consonants given by the acute Indian grammarians, doubtless intentionally. The current of air is strongest at the outset, and gradually grows weaker. No doubt an additional impulse may be given to it at any point of its passage; but if no such impulse be given, it naturally is feebler at the lips than in the throat. Therefore the minimum effort required to stop it at the lips is less than in the throat: in other words, a labial is naturally a weaker sound than a guttural. The merest closing and opening the lips is sufficient to produce the sound p with hardly more additional effort than is required for the mere passing out of the air within the mouth; but it is impossible to sound. 1 See p. 60.

CH. IV.

General

rules hardly possible for protracted sounds.

k without conscious effort. Therefore we must expect k
to pass
into P, but never without strong reason allow that
p can pass into k. Here again the Sanskrit stands us in
good stead by its greater number of consonants. The
theory, that the guttural k is naturally the strongest of all
sounds, is borne out by the fact that k actually passes into
the palatal ch but not ch into k: when we have the dou-
ble form in a group of Sanskrit words, we find regularly k
in corresponding words of other languages; so that ch is
clearly a Sanskrit weakening. Again, the greater strength
of the gutturals is shewn by the difficulty which children
find in pronouncing them; also by their hardly appearing
in terminations, or when they do, yet never as the second
element of a consonantal group, not tk, dg but kt, gd.
This last argument indeed cannot be pressed to its full
result, for we find in terminations pt and bd, as well as kt
and gd, so that by this reasoning p and bought to be
stronger sounds than t and d. But the reason here seems
to be that labials, which require perfect closing of the
lips, are ill suited for the end of a word, where we in-
stinctively prefer those sounds in which the breath is not
articulated by the lips, as among the vowels e rather
than either o or u. Probably indeed dentals and labials
do not differ much in strength, but still, in the few cases of
exchange, it is the dental that seems to pass into the labial.

We have thus got a general idea of the sort of changes we must expect to find among momentary sounds. For the protracted sounds it is less easy to lay down rules. They are in their nature much less definite than the momentary; and much depends on the length of time during Among the which they are sounded. The nasals, as we have seen, are to some extent dependent on other stronger consonants; the guttural nasal indeed not often standing single. Curtius thinks that where we find m and n in corresponding words the m is the stronger. But most of his examples (e.g. Sópov but domum, Sanskrit damam, and the German Faden for old Fadem) are of final m, which, like other

nasals m

may be stronger

than n.

CH. IV.

the strong

labials, is inconvenient at the end of a word. If we assume the strength of the spirants in the order of their Of the spipronunciation, we should get y, s, v, which is probably cor-rants, y is rect, but they do not seem to interchange much. Cer- est; tainly neither of the last two ever passes into y; and Curtius thinks even the few cases where we find F on inscriptions instead of original y, e.g. Fórt, are pure mistakes in writing; it being known that some letter had dropped, and more trace of v having been left than of y. The history of h differs for different languages. In Greekh is the it is always the remnant of one of the spirants, and weakest in weaker than any of them; in Latin it has replaced gh, and seems to have been pretty strongly sounded. S and r are pronounced alike with the point of the tongue; but in s the back of the tongue is also employed; in r (at least in English r) the point of the tongue alone is raised to the front of the palate: for the vibrated r, the tongue is in very much the same position as for s, but (ex hypothesi) is held much more loosely. Whichever then was the sound of the Latin r, or the Laconian p, we can understand the transition from s to those sounds.

Greek.

Sis strong

er than r.

sounds

more difficult than lateral.

Central sounds appear to be more difficult than Central lateral. Here we should not feel sure of the fact from physiology, but the regular change of r into l, and s into th, is convincing. Where has become r in modern Italian or French, the cause is probably dissimilation, as in old Latin: e.g. in "pellegrino," for peregrino, or in "apôtre," for apostole; here the l, a lateral sound, which requires a central closure, was more difficult than a central one immediately after the central opening of t. R and are pronounced at the same part of the palate; s and th are not so; but a looser th can be sounded without putting the tongue against the edge of the teeth, by laying the back of the tongue against the roof of the palatenot the mere point, which would produce l1. In this th,

It is this th probably which is oftenest substituted for s in lisping; and for which s is substituted, if such substitution ever takes place, which