599 
of Edinburgh, Session 1874 - 75 . 
sella. Balbus and barbarus seem in like manner to be connected, 
the meaning of barbarus being one who speaks unintelligibly. 
We may here give an example of the same radical word appear- 
ing in two different forms in the same language with diversified 
but kindred meanings. The G-reek d/xipyco has the general meaning 
of pressing or squeezing, while afxeXy co has the special meaning of 
pressing out milk from the udder. The first of these has not been 
adopted by the other European languages, but d/reAyco is very 
widely diffused as mulgeo in Latin, and milk in the Teutonic lan- 
guages. 
A somewhat similar example may be found in the Greek words 
ypacfju) and yXacfxD. These two words mean different methods of a 
kindred operation, that of marking intelligible forms by some 
sharp or cutting instrument, the one designating the process of 
writing or painting, and the other that of carving or modelling. 
Another cognate seems to be y \vcj>a). But of these words, and 
some others connected with them, I shall have occasion afterwards 
to speak more fully. 
Examples of the interchange of r and l might be further multi- 
plied, but those already given will sufficiently illustrate the subject, 
and direct attention to this mode of discovering latent affinities in 
words. 
It has often been surmised that a similar relation subsists be- 
tween other liquids, as between \vfMf>a and nympha, and it seems 
clear that the Attic dialect frequently changed an v into A, as in 
mpov , At rpov, &c. But this hypothesis has not as yet been suffi- 
ciently matured, and I refrain from entering on it. 
The next point that I shall notice is the peculiarity that un- 
doubtedly exists, of leaving out or adding an initial sibilant in 
cognate words, so as partially to disguise them. In some lan- 
guages combinations of letters are found to be admissible, and 
even frequent, which are not found in other languages, though 
nearly allied. Neither G-reek nor Latin seems to admit of the 
initial combinations of si or sn, and, accordingly, it is probable 
that an affinity subsists between words in other European languages 
which show these combinations, and Latin and Greek words that 
show no sibilant. Thus laxus may be a cognate of slack, limus 
of slime, &c. Nix, undoubtedly, is identical with snow, and 
