In lieu of an abstract, here is a brief excerpt of the content:

33 3 ■Does Language Zipf Right Along? Investigating Robustness in the Latent Structure of Usage and Acquisition Nick c. ELLis University of Michigan MAtthEw BRook o’DoNNELL University of Michigan UtE RöMER Georgia State University ■ EACH OF US AS language learners has had different language experiences, but somehow, we have converged on the same general language system. From diverse and often noisy samples, we end up with similar linguistic competence. How so? Do language form, language meaning, and language usage come together across scales to promote robust induction by means of statistical learning over limited samples? The research described here outlines an approach to this question with regard to English verb-argument constructions (VACs), their grammatical form, semantics, lexical constituency, and distribution patterns. Measurement and analysis of large corpora of language usage identifies Zipfian scale-free patterns in VAC type-token frequency and in the structure of their semantic networks. Using methods from cognitive linguistics , corpus linguistics, learning theory, complex systems, and network science, we explore how these latent structures of usage might promote the emergence of linguistic constructions in first and second language acquisition. Literature Review We seek an understanding of robust language acquisition. As a child, you engaged your parents and friends by talking about shared interests and using words and phrases that came to mind; this is how you learned language. None of the authors of this paper was privy to this system, but somehow, we have all converged on a similarenough English to be able to communicate. How did this happen? We take a two-pronged approach to this question. First, we consider the psychology of learning as applied to linguistic constructions. This is generally the approach to language acquisition as pursued within usage-based linguistics, cognitive linguistics , construction grammar, child language research, second language acquisition 34 Nick C. Ellis, Matthew Brook O’Donnell, and Ute Römer (SLA), and psycholinguistics (Bybee 2010; Goldberg 2006; Robinson and Ellis 2008; Tomasello 2003). These views share the assumption that language acquisition is similar to the rest of cognition in that we learn language like we learn anything else. Of course, human cognition is not simple; there is much to the psychology of learning. The problem-space of language—mapping thoughts to serial sequences of sound—is particularly special, but it is parsimonious to assume that language is subject to the same learning mechanisms and cognitive constraints as the rest of our experience. Second, for the factors that promote robust acquisition, we look to research within emergentism, dynamic systems theory, and complex adaptive systems (CAS). CAS are characterized by their robustness to different kinds of perturbations, by their scale-free properties, and by their structures emerging from the interactions of agents and components at many levels (Holland 1995; Page 2009), as shown by the recent explorations of Language as a Complex Adaptive System (Beckner et al. 2009; Ellis 1998; Ellis and Larsen-Freeman 2006, 2009b; Larsen-Freeman 1997; MacWhinney 1999; Solé et al. 2005). A significant discovery in the early cognitive analysis of language involved how basic objects in natural categories underpinned the robust acquisition of nouns. Rosch et al. (1976) showed how basic categories—those that carry the most information in clustering the things of the world—are those whose members possess significant numbers of attributes in common, are visually imageable with similar shapes, and have similar associated motor programs. Basic natural categories are organized around prototypes. These prototype exemplars are most typical of the category, similar to many other category members, and not similar to members of other categories. People categorize prototype exemplars (like robin as bird) faster than those with less common features or feature combinations like geese or penguins (Rosch and Mervis 1975; Rosch et al. 1976). Basic categories are also those that are the most codable (faster naming), most coded, and most necessary in language (highly frequent in usage). Children acquire basic-category terms like dog, bird, hammer, and apple earlier than they do their superordinates animal, tool, and fruit, or subordinates collie, wren, ball-peen hammer, and Granny Smith. It is reliable visual and motor perceptual experience along with frequent and highly contingent labeling that makes these nouns reliably and robustly learnable, despite individual children experiencing different types of dogs and birds. Cognitive linguistics, particularly construction grammar, has since extended these ideas to language as a whole. Nouns typically relate to the things of the world, but because language has emerged to describe our experiences, whole sentences are used to describe the...

Share