The morphological continuum of grammaticization has been considered to be unidirectional, in which a lexical item undergoing the grammaticization process would become cognitively and phonologically bound to the root morpheme for which it specifies its inflection, but not vice-versa (Givan 1971, Lehmann 1986, Heine et al 1991). The trend is for these lexical items to lose their phonological and morphological weight along with their semantic integrity and become fused to the root, eventually losing their original lexical meaning, while acquiring a more grammatical meaning. Consider, for example, the Spanish future tense suffixes that have evolved from an independent verb of
possession, now cognate with the present tense forms of the verb haber (Givan 1971). The independent verb forms now used to express future tense have fully grammaticized through a process of semantic bleaching and phonological reduction to the extent that native speakers cannot easily predict their lexical origin:
yo he 'I have' yo ser-e 'I will be'
tu has 'you have' tu ser-as 'you will be'
el ha 'he has' el ser-a 'he will be,' etc.
As the majority of historical processes documented do in fact follow this pattern of development, grammaticization theory has been considered a unidirectional process. This paper will introduce data from three languages, llokano, Hebrew, and Hungarian, that will counter the notion that independent lexemes may not develop from phonologically bound affixes, and will review two similar cases of lexeme genesis in Irish (Bybee et ai, in press) and Estonian (Campbell 1991).