Nouns-first, verbs-first and computationally-easier first: a preliminary design to test the order of acquisition

Avcu, Enes
The primary accounts for early lexical differences can be broken down into two distinct theoretical positions that either defend early noun acquisition or provide evidence that challenges this account. This work is trying to bring a computational perspective to the problem of early lexical acquisition of words. It is a preliminary investigation to see if the underlying mechanism relates to computational complexity by which short, frequent and unambiguous words are supposed to be acquired first; and long, ambiguous or infrequent words (including nouns) are predicted not to be acquired early. Our database consists a longitudinal data of three Turkish children between 8 months and 36 months. We conducted three analyses to test this; (1) in frequency analysis, we compared the type token ratios and the number of types and tokens of nouns and verbs in both child directed speech and child speech; (2) in ambiguity analysis, we examined the role of social and attentional cues on word learning; and (3) in phonological analysis, we measured the effect of word length on learning of words. Results revealed that most frequent words did not prove any noun predominance; in place of this, the usage rates of verbs was close to nouns and sometimes much more than that. Furthermore, caretakers did not have any bias to nouns or object names on the contrary there was a preference to verbs at some parts. The high rate of verbs in child speech also challenged the noun-first view. Ambiguity analysis showed that social and attentional cues in the natural language’s context were important factors for word learning. v Therefore, disambiguated words in the context were the most frequent words in child speech. Phonological complexity analysis indicated that word length affected the infant’s ability of word learning. Thus, short words were more advantageous when compared to long words.