The prefixes and their corresponding particles have varied greatly in form from the original syllables, as the various Bantu dialects became more and more corrupt. Assuming these prefixes to have consisted once of two distinct particles, such as, for example, Nos. 1 and 3, ñgu-mu-, or the 6th plural prefix Nga-ma-, the first syllable seems to have been of the nature of a demonstrative pronoun, and the second more like a numeral or an adjective. Mu- probably meant "one," and Ma- a collective numeral of indefinite number, applied to liquids (especially water), a tribe of men, a herd of beasts - anything in the mass.[14] In the corresponding particles of the concord as applied to adjectives, verbs and pronouns, sometimes the first syllable, ñgu or ñga was taken for the concord and sometimes the second mu or ma. This would account for the seemingly inexplicable lack of correspondence between the modern prefix and its accompanying particle, which so much puzzled Bleek and other early writers on the Bantu languages.

In many of these tongues, for example, the particle which corresponds at the present day to the plural prefix Ma- is not always Ma, but more often Ga-, Ya-, A-; while to Mu- (Classes 1 and 3) the corresponding particle besides -mu- is gu-, gw-, u-, wu-, yu-, ñ-, etc.

The second prefix. Ba- or Aba-, is, in the most archaic Bantu speech (the languages of Mt. Elgon), Baba- in its definite form (ñgaba sometimes in Zulu-Kaffir). The concord is -ba- in all the less corrupt Bantu tongues, but this plural prefix degenerates into Va-, Wa-, Ma-, and A-. The concord of the 4th prefix, Mi-, is gi-, -i-, -ji-, and sometimes -mi-. The commonest form of the 5th prefix at the present day is Li- (the older and more correct is Di-), and its concord is the same; this 5th prefix is often dropped (the concord remaining) or becomes Ri-, I-, Ji-, and Ni-. The 7th prefix, Ki-, in many non-related dialects pursues a parallel course through Ci- into Si- (=Shi) and Si- and its concord resembles it. The 8th prefix is still more variable. In its oldest form this is Ibi- or Mbibi-. It is invariably the plural of the 7th. It becomes in different forms of Bantu speech Vi-, Pi-, Fi-, Fy-, Pši-, ši-, I-, By-, Bzi-, Psi-, Zwi-, Zi- and Ri-, with a concord that is similar. The 10th prefix, which was originally Ti- or Tin-, or Zi- or Zin-, becomes Jin-, Rin-, Din-, Lin-, θin-, θon-. etc. The n in this prefix is really the singular prefix No. 9, which is sometimes retained in the plural, and sometimes omitted.

In the case of the 10th prefix, the concord or corresponding pronoun persists long after the prefix has fallen out of use as a definite article. Thus, though it is absent as a plural prefix for nouns in the Swahili of Zanzibar, it reappears in the concord. For instance: - ñombe hizi zangu - Cows these mine (These cows are mine), although ñombe has ceased to be ziñombe in the plural, the Zi- particle reappears in hizi and zangu. In fact, the persistence of this concord, which exists in almost every known Bantu language in connexion with the 10th prefix, shows that prefix to have been in universal use at one time. The 11th prefix -Lu- seems to be descended from an older form, Ndu-. Its commonest type is Lu-, but it sometimes loses the L and becomes U-, and in the more archaic dialects is usually pronounced Du- or Ru-. It is also Nu- in one or two languages. The 12th prefix (Tu-), always used in a diminutive sense, disappears in many of these languages. Where met with it is generally Tu- or To-, but sometimes the initial T becomes R (Ru-, Ro-) or L (Lu-, Lo-) or even Y (Yo-), the concord following the fortunes of the prefix. The 13th prefix (Ka-) is sometimes confused with the 7th (Ki) and merged into it and vice versa.

Ka- very often takes the 8th prefix as a plural, more commonly the 12th, sometimes the 14th. This prefix (Ka-) entirely disappears in the north-western section of the Bantu languages. Bleek thought that it persisted in the attenuated form of E- so characteristic of the Cameroon and northern Congo languages, but later investigations show this E- to be a reduction of Ki- (Ke-) the 7th prefix. The 14th prefix Bu- is very persistent, but frequently loses its initial letter B, which is either softened into V or W, or disappears altogether, the prefix becoming U- or O- or Ow-. Sometimes this prefix becomes palatized into By- or even Tš- (C-). The concord follows suit. The 15th prefix, Ku-, occasionally loses its initial K or softens into Hu or χυ or strengthens into Gu. Its concord under these circumstances sometimes remains in the form of Ku-. The 16th, Pa-, prefix is one of the most puzzling in its distribution and its phonetic changes. A very large number of the Bantu languages in the north, east and west have a dislike to the consonant P, which they frequently transmute into an aspirate (H), or soften into V, W, or F, or simply drop out.

There is too much evidence in favour of this prefix having been originally Pa- or Mpa-pa to enable us to give it any other form in reconstructing the Bantu mother-tongue. Yet in the most archaic Bantu dialects to the north of the Victoria Nyanza it is nowhere found in the form of Pa-. It is either Ha- (and Ha- changes eastward into Sa-!) or Wa-.[15] But for its existence in this shape in the language of Uganda one might almost be led to think that the 16th locative prefix began as Ha-, and by some process without a parallel changed in the east and south to the form of Pa-. There are, however, a good many place names in the northern part of the Uganda protectorate, in the region now occupied by Nilotic negroes, which begin with Pa-. These place names would seem to be of ancient Bantu origin in a land from which the Bantu negroes were subsequently driven by Nilotic invaders from the north. They may be relics therefore of a time before the Pa- prefix of those regions had changed to the modern form of Ha-. In S.W. and N.W. Cameroon the initial p of the 16th prefix reappears in two or three dialects; but elsewhere in North-West Bantu Africa and in the whole basin of the Congo, except the extreme south and south-east, the form Pa- is never met with; it is Va-, Wa-, Ha-, Fa-, or A-. In the Secuana group of dialects it is Fa- or Ha-; in the Luyi language of Barotseland it assumes the very rare form of Ba-, while the first prefix is weakened to A-.