Conditioned consonant changes

There are four sets of major consonant changes in Eastern Bantu that need to be distinguished — two result from a following element, two from a preceding one.

The major consonant changes in Eastern Bantu:
  1. before a tense vowel *i, *u (Bantu Spirantization): [+ATR] vowels can be inside morphemes or immediately follow.  
  2. before a glide (Glide Spirantization): glide formation in roots like *dɪ́-a ‘eat’, *pɪ́à ‘new’ or before causative suffixes. These changes are sometimes included under Bantu Spirantization, but best to distinguish them because they can have different outcomes.
  3. after a nasal (Post-nasal Effects): The basic Bantu syllable structure (CV) does not usually create C-C clusters other than nasal combinations.  After nasals, voiced consonants usually become pre-nasalized, but unvoiced consonants can also have other results.
  4. after a glide (Preceding-I/Y Effects): high vowels can lead to glide formation after other vowels.

All four changes were major in that they affected many consonants in many languages and show evidence that they were common to some early or intermediate branches. Additionally, specific consonant lenitions or more limited effects like the palatalization of velars before front vowels are seen in certain languages (e.g. the class 7 prefix *kɪ- > Shonic and Chewa chi-, or Nguni si-).  Altogether these changes could lead to a large number of reflexes of a single proto-phoneme, as seen in Zezuru:


 Reflexes of *p in Zezuru 
  1.     unconditioned *p > p       -pá ‘give’ < *pá (BLR 2344)
  2.     *pi-C > svi-C                    -svìka ‘arrive’ < *pìk (BLR 2570)
  3.     *pɪ-V > tsv-V                  -tsvá ‘be on fire’ < *pɪ́-a (BLR 2491)
  4.     *pu-C > fu, pfu-C           -fúra ‘blow gently’, -pfùtsa ‘blow (fire)’ < *púd (BLR 2672)
  5.     *pu-V > f-V                      -fána ‘resemble’ < *pú-an (BLR 2670)
  6.     *N-p > mh                        mhàrá 9 ‘impala’ < *pàdá (BLR 2355)
  7.    *Y-p > b                            bángá 5 ‘knife’ < pángà (BLR 2399)
To which one can add later Shonic voicing of class 5 for (c):
8.     *pɪ-V + voicing > dzv          dzvá ‘new’ cl. 5 < *pɪ́à ‘new’ (BLR 2495)
 

Inevitably, some of these reflexes overlapped or merged with changes to other phonemes, so a full diachronic phonology would need to track all those paths and establish a relative chronology as well as arguments for branching into subgroups. Certain changes are shared by various languages: (1) lenition and rhotacism of unvoiced stops (S20-60, P30), (2) special Preceding-I reflexes before back vowels (S20-40), (3) *NC > C (S30, P30), (4) various reflexes of Bantu Spirantization. All these changes need more detailed study in combination. 

For example, the distinctive consonant change *t > r is seen widely in South Bantu (everywhere except the Nyanjic, Shonic and Nguni groups) and has been considered a possible isogloss for subgrouping. But changes like *i-t > th (Gitonga), ʈh (Makhuwa) or spirants/affricates (S20-30-50) make it likely that Preceding-I Effects happened before the completion of *t > r. For our purposes, it is sufficient to consider briefly what other changes might have affected the input or output of Preceding-I Effects.  Likewise, some Preceding-I reflexes of unvoiced stops overlap with reflexes of *s and *ti, so a relative chronology might be constructed. For example, it appears that Proto-Nguni *i-k > *s and then *s > hl[ɬ].

The four major conditions above form two pairs of alternatives and in principle over time a consonant could be affected by two changes, i.e. by one of the preceding conditions (nasal or front vowel) and by one of the following conditions (tense vowel or glide). So, any stem could potentially have derivational or inflectional forms affected by various combinations of these conditions (as well as the unconditioned form of the stem). From this panoply, we can predict that certain languages will engage in levelling to reduce the number of allomorphs. In fact, since multiple morphemes create Bantu Spirantization, e.g. agent suffix *i and perfect *ide, verbal stems can even vary in their participation in a single conditioned change.

The two effects caused by preceding sounds (post-nasal and PIE) eventually merge or overlap in some languages, e.g. aspiration in Chewa (*mp > mph, *i-p > ph), as was noticed by Eiselen (1923/24) and Bourquin (1932). For Makhuwa, Schadeberg (1999: 385) proposes “that the strengthening influences exercised by high closed vowels and by prenasalization are not unrelated”, and later adds (386) that “the strong reflexes kh and ph are obviously conditioned by a preceding rather than following *i. They are equal to the reflexes of the prenasalized consonants *nk and *mp.” Creissels (1999) notes that the effects in Tswana of preceding and following glides are the same.

Bourquin, W. (1932–1933). Entstehung von nasalen durch den einfluss von i im Bantu. Zeitschrift für Eingeborenen‑Sprachen, 23, 195–202.
Eiselen, W. (1924). Die Veränderung der Konsonanten durch ein vorhergehendes i in den Bantusprachen. Zeitschrift für Eingeborenen‑
Sprachen, 14(2), 81–153.
Schadeberg, T. C. (1999). Katupha’s Law in Makhuwa. In J.‑M. Hombert, & L. M. Hyman (Eds.), Bantu historical linguistics: Theoretical
and empirical perspectives (pp. 79–394). CSLI (Center for the Study of Language and Information) Publications.