Academic literature on the topic 'Type/token ratios'

AbstractAs reflected by the distinction between types and tokens, frequency is standardly conceived of in a two-dimensional fashion. This conception draws on the linguistic hierarchy in that token frequency is calculated at a higher structural level than type frequency. The fact that the linguistic hierarchy consists of more than two levels raises the possibility of investigating frequency from a multiplanar angle. This approach is illustrated with a three-level analysis of derivational affixes in English, German, and Dutch. The lowest level represents the number of affix types, the intermediate level the number of words of which these affixes are a part, and the highest level the textual frequency of these affixed words. The three languages exhibit significant variation. While English shows a consistent increase, German shows a consistent decrease in the suffix-to-prefix ratios from the lower to the higher levels. Dutch is inconsistent in displaying an increase from the lowest to the intermediate level but a decrease from the intermediate to the highest level. A good part of the explanation for this difference lies in the disparate role of prefixing in the three languages. The multiplanar approach provides novel perspectives on a number of issues such as the locus of linguistic change. Preliminary evidence suggests that the loss of a lower-level unit may be accompanied by “collateral damage” at the higher level.

Creoles (here including expanded pidgins) are commonly viewed as being more analytic than their lexifiers and other languages in terms of grammatical marking. The purpose of the study reported in this article was to examine the validity of this view by measuring the frequency of analytic (and synthetic) markers in corpora of two different English-lexified creoles — Tok Pisin and Hawai‘i Creole — and comparing the quantitative results with those for other language varieties. To measure token frequency, 1,000 randomly selected words in each creole corpus were tagged with regard to word class, and categorized as being analytic, synthetic, both analytic and synthetic, or purely lexical. On this basis, an Analyticity Index and a Syntheticity Index were calculated. These were first compared to indices for other languages and then to L1 varieties of English (e.g. standard British and American English and British dialects) and L2 varieties (e.g. Singapore English and Hong Kong English). Type frequency was determined by the size of the inventories of analytic and synthetic markers used in the corpora, and similar comparisons were made. The results show that in terms of both token and type frequency of grammatical markers, the creoles are not more analytic than the other varieties. However, they are significantly less synthetic, resulting in much higher ratios of analytic to synthetic marking. An explanation for this finding relates to the particular strategy for grammatical expansion used by individuals when the creoles were developing.

ABSTRACTThis study employed multiple regression analysis to examine the relationship between global writing quality (holistic scores) and lower level analytic measures of writing, with a focus on cohesive indices. The subjects were 9-year-old English-speaking children who participated in either a story free-writing condition or a story rewriting condition. The results showed that both cohesive indices and lower level writing measures (type-token ratios, mean length of utterances in morphemes, composition length, etc.) each accounted for a significant amount of the variance in holistic scores. The story rewriting procedure proved to facilitate the children's writing processes and, hence, resulted in higher quality writing (in terms of both global writing quality and text cohesion) than the story free-writing condition.

This paper presents an investigation of lexical first language (L1) attrition, asking how a decrease in lexical accessibility manifests itself in long-term residents in a second language (L2) environment. We question the measures typically used in attrition studies (formal tasks and type–token ratios) and argue for an in-depth analysis of free spoken data, including factors such as lexical frequency and distributional measures. The study is based on controlled, elicited and free data from two populations of attriters of L1 German (L2 Dutch and English) and a control population (n = 53 in each group). Group comparisons and a Discriminant Analysis show that lexical diversity, sophistication and the distribution of items across the text in free speech are better predictors of group membership than formal tasks or elicited narratives. Extralinguistic factors, such as frequency of exposure and use or length of residence, have no predictive power for our results.