ON ETHNOGRAPHICAL PHILOLOGY. 
221 
Id gadi lakiza ~ to let loose, tlie two words sUind separate, and ffadi is 
unaltered in form. All this is a question of euphony rather than of com¬ 
position : the truth being that the amalgaraatioii which distinguishes a com¬ 
pound word from two words is, generally speaking, imperfect in the Galla. 
In all languages where there are but few inflexions, the difference between 
one compound, and two separate words, is frequeudy equivocal; in none 
more so than in our own. Nevertheless, the principle of composition may 
be part and parcel of such a language, and in some dialect, other than the 
particular one under consideration, may appear in au uttcquivocal form. 
1. Tbedifference hetxoeen the details of the African Injlcxions in cases where 
tht general principle is the same .—Between tiie general character of the 
gminoiar in the Maudingo, Howssa, and Yarriba languages, no difference 
has ever been indicated. The inflections are found in the same parts of the 
Ismgnage, to the Nxme extent (there or thereabouts), and formed according 
to the same principle. Vet the particular inflectional elements have little or 
DO likenes. The so-called signs of the tenses, moods, &c. difler. Precisely 
diesarac occur in the American languages, M’lierc the general grammatical 
character is pre-eminently uniform. This is not the case with the Indo- 
European languages, where the sigu of (e.ff.) the first person singular is the 
Mmc throughout the whole class, «•/*», su~m, a-^t, &c. Yet Indo-Kuropean 
criticism should not mislead us. Langviagea which separate from the com¬ 
mon mother-tongue after the evolution and fixation of an inflectional system, 
arc more alike in their grammars than in their vocabularies; whilst lan¬ 
guages that separate from the common mother-tongue before the evolution 
of an inflectional system are more alike in their vocabularies than in their 
grammars. They develope the latter independently. Hence tlie value of 
tlip grammatical and glossarial evidence is different for different languages. 
A question similar to the last is tho followingWhat ia the value of the 
numerals aa compared with that of an equal quantity of oilier words equally 
fundamental, in determining the affinities between languages iu general, and 
those of the African languages in particular ? We have frequently the five 
or lea first numerals as specimens of a language, when we have not a word 
^des; and perhaps wo get this scanty amount of information oftener from 
Africa than elsewhere. 
The value of the numerals as compared with that of an equal quantity of 
other Words equally fundamental in dctcrimniiig the aftinities between lan¬ 
guages is not constant; it is different for diflerent languages. Throughout 
the whole range of the Indo-European languages we find the numerals alike, 
Here the remainder of the language pmsents ditforyncea much more con¬ 
siderable. Throughout the American languages in general we find the nu- 
jnerals unlike, where the remainder of the languages presents differences much 
considerable. Now the evidence of the African numerals Is of the same 
kind as that of the American. Coincidence proves affinity much more than 
iion-coinddcncc proves difference. 
If I ventured to account for this inconstancy in the ethnological value of 
tbc numerals, the hypotheses would be based on the following distinction. 
There arc two sorts of numerals, tlie former of which probably grew out 
of the latter. 
!• Exclusive numeraU; like ftco, Mw, (loenfj/, Src. 
2. Secondary numerals; like couple, brace, leash, score, &c. 
When languages have separated from the common mother-tongue after 
tbe evolution of a system of exclusive numerals, there is little change, and 
uo room for any variety of expression. 
When languages have separated from the common stock bejore the evolu- 
