The past tense of the weak verbs of the first conjugation class is formed by attaching one of the suffixes -de/-də/ or -te/-tə/ to the stem (see paradigm of class I). The choice between the two suffixes depends on the (underlying) voice specification of the stem-final segment. If the latter is voiced, the suffix -de is selected, if it is voiceless, -te is. See the examples in (1):

The feature [voice] is distinctive within obstruents, whereas sonorant consonants are redundantly [+voice]. Since it is only obstruents that can be specified as [-voice], the suffix -te is only attached after stems ending in an obstruent. Obstruent sequences must agree in voicing, hence they consist of either voiced or voiceless segments (see onset: sequences of two obstruents and word-final sequences of two obstruents), with which the distribution of the two past tense suffixes is in line.

The distribution of /-də/ and /-tə/ can be accounted for by means of the distributional statement in (2):

As is clear, the above statement considers the context of /-tə/ as the special case and that of /-də/ as the 'elsewhere case'. This approach − with the two independent underlying suffixes /-də/ and /-tə/ − obviates the necessity to invoke progressive voice assimilation in deriving the surface forms of the past tense of verbs of the first weak conjugation class.

However, it meets with several objections. In the first place, the suffixes /-də/ and /-tə/ have a complementary distribution. This is usually taken to be indicative of the fact that one suffix is the basic one, from which the other one is a variant. If this line of reasoning is pursued, one cannot but consider /-də/ as the underlying representation. First, it has a wider scope than /-tə/. Second, it also follows verb stems ending in a vowel. If /-tə/ were the underlying suffix, the stem-final vowel would have to induce the change of suffix-initial /t/ to [d]. In Regressive Voice Assimilation, only voiced plosives can enforce the voicing of preceding plosives (see regressive voice assimilation: type 1 and regressive voice assimilation: type 2). Vowels are the least consonant-like kind of segments, so it is highly unlikely for them to have this voicing capacity with respect to following plosives in Progressive Voice Assimilation.

In the second place, an approach with both underlying /-də/ and /-tə/ leaves out of account a) that both suffixes are almost identical and b) that the difference in their distribution is not arbitrary from a phonological point of view. The latter can be the case within a distributional approach like the one in (2), for there is no principled phonological reason why the suffix /-tə/ could not follow a stem ending in a voiced non-obstruent (a vowel, a glide, or a sonorant consonant). Therefore, an analysis with the underlying suffix /-də/ is called for.

Combining a verb stem ending in a voiceless obstruent and /-də/, as in /bak+də/baked and /paf+də/puffed, leads to a voicing conflict ‒ /bakdə/ and /pafdə/ ‒, which is resolved by the devoicing of suffixal /d/: [baktə] and [paftə].