A construction may be grammaticalized from a free configuration at the text level down to a word form at the morphological level. Such a construction typically comprises a lexical item which functions as a pivotal element in being reduced to a grammatical formative. The general grammaticalization scale displayed here visualizes the various phases that such an element runs through from its first recruitment from the lexicon to its final complete dissolution.

General grammaticalization scale
degree of grammaticalizationlowhigh
status of
grammaticalized element
wordformativeaffixinner
modification
0

Not every historical case of grammaticalization enters this scale at the start and leaves it at its end. It is entirely possible and even frequent that grammaticalization affects an element that is already a grammatical formative and only converts it into an affix. Historical cases of grammaticalization which represent the entire gamut are hard to come by. This is due to the external circumstance that grammaticalization takes its time and that few languages in the world have a sufficiently long documented history for a single element to run through all the stages.

Also, it is not necessary that a grammaticalized element undergo each of the phases in turn. In particular, the stage of the inner modification is mostly skipped. On the one hand, an affix may be directly reduced to zero. This is observable in the conjugation of today's Brazilian Portuguese. The following table shows the development of the present-tense conjugation of the verb 'come' from Latin to Brazilian Portuguese.

Conjugation of vir ‘come’ in present tense
stage
category  ╲
LatinEur. Port.Bras. Port.
singular1veni-ovenhovenho > vem
2veni-svensvem
3veni-tvemvem
plural1veni-musvimosvimos > vem
2veni-tisvindesvem
3veni-untvêmvem

It must be conceded that this is an extreme example. Other verbs such as matar ‘kill’ would preserve a few more conjugation endings, for phonological reasons. Also, the endings in the second person do not disappear by erosion but because the pronoun used for address is grammatically a third person form.

On the other hand, the transition of an affix to an inner modification is not very frequent and much less observable in historical times. Little is known about the genesis of inner phonological modifications such as apophony.

There are at least some data concerning the development of infixes out of external affixes. Some Philippine languages show an alternation between an affixal and an infixal marking of verbal voice. Proto-Indo-European had a suffix -n which formed an imperfective present form in on conjugation class. It appears as a suffix in Sanskrit, but as an infix in Latin. In the Mayan languages, the conversion of an original suffix into an infix may be reconstructed by genetic comparison. The suffix appears, inter alia, as a submorphemic component of a passive participle in -bil, shared at least among the Yucatecan, Tzeltalan and Ch’olan branches (cf. Lehmann 2020). A passive morpheme -b (of unknown provenience) can therefore safely be reconstructed for some prehistorical stage of Yucatecan. It survives as such in the Yucatecan branch. In Mopán and Itzá, it remains a suffix (). In Yucatec and Lacandón, it develops an infixal allomorph ().

.k=utz'on-b-ol
ItzáIPFV=A.3shoot-PASS-INCMPL
it is shot(Hofling 1991:33)
.k=uts’o<’>n-ol
YucIPFV=A.3shoot<PASS>-INCMPL
it is shot