Here is a short description of lenition, gradation, and mutation for the Conlang FAQ. Let me know what else I can do to make it clear.

The terms lenition, gradation, and mutation can refer to processes which apply to vowels or consonants; the definitions below refer to processes affecting consonants. Also, gradation can also be applied to roots or stems as a whole; Estonian has gradation which affects the prosodic make-up of the whole stem and not just to consonant or vowel alternations.

1)  Lenition is the phonetic result of reducing the amount of effort expended in the production of a segment. Lenition involves alternations among non-distinctive elements. American English flapping and Spanish spirantization are good examples of lenition.

2)  Gradation is an alternation among consonants which has a fairly transparent phonological basis, but which is normally triggered by morphology. Finnish has gradation: consonants alternate upon suffixation when that suffixation produces a closed unaccented syllable. Shoshoni also has gradation; a short description is given below.

3)  Mutation is an alternation among consonants which marks some morphological category. It usually is not phonologically transparent; that is, it often isn't obvious why the consonants alternate the way they do given the phonological environment they find themselves in (although in some cases the historical source may be recoverable). Celtic languages have mutation, as do many languages of western Africa. A short description of Fula mutation is given below.

The particular mutations that a language exhibits will be bound by the segmental inventory, since mutation typically involves alternations among distinctive elements. So if a language has /p/ and /w/ but no /f/ or /v/, then a spirantizing mutation would likely alternate /p/ and /w/ rather than some non-distinctive segment like /f/ or /v/.

I think that these are pretty standard definitions of these terms. Other uses of these terms are generally confined to the study of a particular language or language family, as is the case for the term 'lenition' in Celtic studies.

Gradation in Shoshoni

In Shoshoni, a stem-final nasal assimilates to a following voiceless stop or nasal. A voiceless stop following a nasal is voiced. This is one of a series of gradation patterns in this language; the others alternate a voiceless stop with a voiced fricative, a voiceless fricative, and a geminate voiceless stop. As an example, consider the following forms ([kw], [gw], [Gw], [xw] are labio-velar obstruents, [N] is a velar nasal, [e] is a high central unrounded vowel, [E] is a voiceless version of [e]).

        kwasu  'shirt'
        eNgwasu  'your (2s) shirt'
        neGwasu  'my shirt'
        tawExwasu  'our (dual incl) shirt'
        tuukkwasu  'soldier' (lit: 'black shirt')

In each case, the particular grade of consonant is determined by the final element of the preceding morpheme. That is, any voiceless stop or nasal will be geminated following _tuu_ 'black', &c.

Mutation in Fula

Nouns in Fula are divided among 25 possible classes in such a way that each noun stem belongs to a singular class, a plural class, and up to five diminutive and augmentative classes. Each class is marked by a prefix and an initial consonant mutation. The alternating segments are:

        b / mb / w
        d / nd / r
        j / nj / y
        g / Ng / y~w (/y/ if the following vowel is front, /w/ otherwise)

So for example, the noun stem /baa/ 'monkey' appears in classes 3, 6, and 11 as:

        (3)  baa-Ngel
        (6)  mbaa-kon
        (11) waa-ndu

There are two kinds of exceptions to this 3-way mutation pattern. The first concerns alternations involving a voiceless stop; these alternations don't show a prenasalized version:

        p / p / f
        S / S / s (/S/ acts like a stop in mutation)
        k / k / h

so /payan/ 'cooking pot' appears in the same classes as:

        (3)  payan-Ngel
        (6)  payaN-kon
        (11) fayan-nde

This is explainable as the avoidance of a nasal/voiceless stop cluster. This tendency is apparent in many languages of the world (see my other post on this).

The second kind of exception are stems which have a voiced stop which either fails to show mutation, or which only shows partial mutation. So /beebe/ 'deaf mute' is invariant across noun classes:

        (3)  beebe-yel
        (6)  beebe-hon
        (2)  beebe-'en

and /bukka/ 'shelter' is only partially variable, missing a continuant-initial variant:

        (3)  bukka-yel
        (6)  mbukka-hon
        (11) bukkaa-ru

While these exceptional patterns are not generalizable to mutation systems elsewhere, all mutation systems will show exceptional behavior of some sort. Lenition and gradation are more regular.

Dirk Elzinga                                               [log in to unmask]

"It is important not to let one's aesthetics interfere with the appreciation of
fact." - Stephen Anderson