When learnability and universals disagree: The case of phonological repetition

In the past few decades, much research on the connection between language acquisition and linguistics has been driven by the notion that human language is shaped by learning biases that favour certain types of structures over others. A broad prediction that follows from this idea is that linguistic patterns that are systematically preferred in human languages should be consistent with what is more readily learned, and vice versa. This prediction has generated a large body of findings indicating, for example, that phonological or syntactic structures that are typologically unmarked in natural languages emerge earlier in children’s production or are preferred by adult participants in artificial language experiments compared to structures that are typologically marked.

In this talk, I examine a case where this generalization fails, and discuss its implications for this general strategy in exploring the link between language acquisition and linguistic theory. The particular case relates to phonological repetitions within a lexical unit, as exemplified by the repetition of the same syllable in words like cocoa (/koko/). There is typological, statistical and psycholinguistic evidence that, except when it is employed morphologically (that is, as reduplication marking plurality, iterativity, intensity etc.), languages tend to avoid such repetitions (Berent, Bat-El, Brentari, Dupuis, and Vaknin-Nusbaum 2016, Monaghan and Zuidema 2015, Pozdniakov and Segerer 2007). Yet, there are learning theories that argue that human perception and memory should prefer string-internal repetitions, including those in linguistic structures (Endress, Nespor, and Mehler 2009). This claim is not only consistent with neonates’ neurological responses to auditory stimuli with repetition (Gervain, Macagno, Cogoi, Peña, and Mehler 2008; Gervain, Gerent, and Werker 2012) and the preponderance of sound repetitions in infant-directed vocabulary (Ferguson 1964, 1977), but also with experimental evidence that infants are more adept at segmenting and learning words consisting of identical, rather than non-identical, syllables (Ota and Skarabela 2016, in press). These findings indicate that, when it comes to phonological repetitions, there are diametrically opposite tendencies between what is preferred by learners and what is preferred by language systems.

What are the implications of this observation? First, it shows that the putative link between learning and language universals may not be as straightforward as we might think, and a more nuanced understanding of the relationship is in order. Second, the presence of general linguistic patterns in conflict with learning biases suggests that some factors that shape human languages can trump learnability pressures. Lastly, it raises the possibility that, although having served us extremely well, the general strategy of focusing on consistencies between acquisition and language universals may also have blinded us from potentially illuminating dynamics between learning and linguistic patterns that we have not paid sufficient attention to. I will address these issues by drawing from the insights and methodology that have been developed in the study of language evolution.