Basic vs. derived context

The level of difficulty of going to a phonemic split for L2 from an existing allophonic variation in L1 has a correlation with various contexts. In a thorough examination of this issue, Eckman et al. (2003) state that whenever there is a conflict between L1 and L2 in the above manner (i.e. two sounds are in contrast in L2 but are the allophones of a single phoneme in L1), the target language phonemic contrast will be acquired first in basic (tautomorphemic) contexts, then in derived (heteromorphemic) contexts. Going back to the Portuguese–English mismatch regarding the /t/ – /tʃ/ contrast of English, the expected rendition of target English words tea and betting in the earlier stages of learning will be [tʃi] and [bεtʃɪŋ] respectively, which fails in regard to the target contrast. Eckman et al. describe this as “stage I” (“no contrast”), where the native language allophonic rule applies in both the basic context (i.e. tauto-morphemic situation where the affected sound /t/ becoming [tʃ], and the relevant environment – following vowel /i/ – are in the same morpheme), and in a derived context (in heteromorphemic context). Eckman et al. predict the next stage (“stage II”) will show a partial contrast in the way the native rule applies only in the derived (heteromorphemic) context (i.e. the sound affected, /t/ becoming [tʃ] before high vowels taking place in betting realized as [bεtʃɪŋ], while tea is realized as [ti] and not as *[tʃi]). The last stage (“stage III”) is the one where the target contrast is acquired in both the basic and the derived context (i.e. in both tautomorphemic and heteromorphemic contexts). Thus, the following implicational relationship holds: if a target pattern is acquired in a heteromorphemic context, it implies that the same is acquired in a tautomorphemic context, but not vice versa.

This excludes a situation where the learner is successful in a derived context (e.g. betting pronounced as [bεtɪŋ]) but will fail in the basic context (e.g. tea pronounced as [tʃi]). This hypothesis receives support from studies where the acquisition patterns reflect such an order (i.e. learning is earlier in basic contexts than in derived contexts).