Kiparsky, Paul, and Menn, Lise. (1977). On the acquisition of phonology. In John Macnamara (Ed.), Perspectives in Neurolinguistics and psycholinguistics. New York, NY: Academic Press. pp. 47-78.

Kiparsky and Menn (hereafter KM) present a theoretical argument for children as active discoverers of grammar, building structural representations based on evidence from the ambient language. In the process, KM propose a dual lexicon. The split includes one path between phonetic and phonological forms (i.e., some phonological processes map acoustic forms to the underlying phonological representations that link related words) and another path between incoming phonetic forms and the phonetic output that children create.

The chapter begins with “The Learning of the Phonetic Repertoire”, a discussion of the two major proposals for child phonology that existed in 1977. The first was Roman Jakobson’s, who proposes that phonology develops according to a universal system of contrasts, and contrasts are learned by children in the order of most to least universal. For example, children should contrast /d/ and /g/ before they contrast /d/ and /b/ (pp. 48-49). The problem with Jakobson’s approach is that it says nothing about the order in which the sounds themselves will be acquired. Furthermore, the absence of a contrast may indicate that children are intentionally, or selectively, avoiding a particular sound, but Jakobson says nothing about this or why sound evasion should happen. Therefore, KM consider Jakobson’s theory to be difficult to falsify.

Stampe’s theory is specific about when sounds will be acquired, but makes a distinction between phonological rules and phonological processes. Rules are the grammatical means by which speakers convert from phonological to phonetic word forms, such as the flapping or homorganic nasal cluster rules. Processes, on the other hand, are innate rule-like conversions that explain the kinds of errors that children make. For example, children produce voiced word-final stops without voicing (/d/ –> [t]/__#) because of a devoicing process. Speakers of languages like English, which do voice final stops, must overcome these processes.

KM describe several problems with this view. First, it appears that Stampe’s theory requires children to learn phonological rules in the same order as they would unlearn phonological processes. This is an empirical but unstudied question.* Second, KM find no reason to assume that adult speakers maintain rules on the one hand and processes on the other (i.e., German speakers do not appear to be stuck in a word-final devoicing process, and regardless, they must still learn the allomorphy that relates allomorphs with voiced and voiceless final stops).

KM also criticize both Jakobson and Stampe as being overly deterministic and not allowing for the kind of variability inherent to child language learners. As evidence, they point to the fact that children break up consonant clusters in a variety of ways, and to the fact that children often produce phonological idioms, words that are produced more accurately than the phonological processes apparently at work in their language would predict. In sum, KM state that we need a new model of phonological development. However, they do not focus much on the development of sounds or sound contrasts. Instead, they focus on the fact that children’s production abilities lag behind their perceptual abilities.

KM propose the dual lexicon to account for a distinction between cognitive grammar learning and articulatory sound implementation. Children may learn the cognitive grammar at whatever pace (KM describe it as going on over many years, although I think that today’s infant literature would generally contradict that**), but the development of a productive sound repertoire is separate from the cognitive grammar. Thus, we have two lexicons.

The second part of the book, “The Learning of Morphophonemics”, is somewhat orthogonal to the dual lexicon proposal, so I do not discuss it.

Here, I identify what I think are outstanding issues in the paper, some of which will be addressed in future posts. First, is the dual lexicon meant to be only a description of the grammar, or is it also a processing model? In other words, when formulating a message, does a child start with the phonological grammar, which is translated into a phonetic form, which is then translated into the child’s pronunciation? KM suggest that, in fact, their may be yet another step, in which physical limitations act on the message, as would be the case for a lisp. Second, KM propose that children do not have allomorphy. Is this really true? It seems to me that children could be learning meanings and linking related word forms at a fairly early age. However, I’m not familiar with the literature on this topic. Third, the logical dependency of the dual lexicon and KM’s proposal of the child as “language discoverer”*** view is not clear to me.