EDITORS: Nick C. Ellis and Diane Larsen-Freeman TITLE: Language as a Complex Adaptive System SERIES TITLE: Best of Language Learning Series PUBLISHER: Wiley-Blackwell YEAR: 2010
Julie Bruch, Dept. of Languages, Mesa State College
This special 60th anniversary issue of the journal ''Language Learning'' contains ten papers from the 2008 conference on Language as a Complex Adaptive System. These papers will be of interest to researchers in the fields of Complex Adaptive Systems, Connectionism, Emergentism, Cognitive Linguistics, and Dynamic System Theory. They represent a unified and substantive set of innovative perspectives that explore and help more fully explain language learning, processing, and variability. The research concepts and methodology in this collection, while wide-ranging, are successfully presented as an accessible and cohesive thread that ties scientific and humanistic approaches into a well integrated whole. Together, these papers reveal an exciting paradigm shift that is in stark contrast to the traditional Generativist, Universal Grammar, and Minimalist agendas, which largely ignored the influence of context and the innate variability of language. This collection takes a huge step toward further moving the Emergentist agenda into the mainstream.
The basic assumption of the book is that social interaction as well as general human cognitive processing abilities are fundamental to language acquisition and to the dynamics of ongoing language development and change over time. The position paper labels language in this view as a complex adaptive system (CAS) and claims that the CAS perspective unifies recent theoretical perspectives from cognitive and sociolinguistics which challenge the traditional view of language as a fixed system of symbols.
The papers here are based on a usage-based theory which sees language as fundamentally involving form-meaning constructions. The usage-based perspective is also used for quantitative modeling of language processing and emerging language change within both individual users and whole linguistic communities. The position paper presents seven characteristics of language as a CAS, many of which debunk formerly held generativist characterizations. These characteristics emphasize language as inherently: 1) socially formed, represented, and constrained, 2) largely variable, mutable, and emergent, and 3) dependent on human cognitive mechanisms.
''A Usage-Based Account of Constituency and Reanalysis'' by Clay Beckner and Joan Bybee: This paper describes how corpus data (from COCA) reveals constituent structure as a natural outflow of general cognitive categorization, chunking, and memory processes rather than as part of any innate or universal linguistic competence. It also points out how usage trends can result in constituent reanalysis and changes in distribution over time. This contradicts the generativist view that reanalysis of structural forms in language happens predominantly through incomplete first language acquisition and therefore occurs between successive generations. The study focuses on changing constituent boundaries of the complex phrasal prepositions ''in spite of'' and on an analysis of its syntacto-semantic functions. It details the reanalysis of ''in spite of'' from its original lexical function to its current grammaticalized constituent function with its own peculiar meanings, structural behaviors, and distributions. The paper ends with a well-supported defense of this analysis of the change and a rebuttal of possible generativist-based objections to it. The data provided support the modern usage of ''in spite of'' as a reanalyzed, formulaic, single-constituent chunk. Finally, the paper underlines the perspective that such constituent structure is emergent and adaptive, and therefore, is a prime example of complex interacting factors.
''The Speech Community in Evolutionary Language Dynamics'' by Richard A. Blythe and William A. Croft: This paper reports on the use of mathematically-based language analysis for modeling language change as a result of speaker interaction over time. This work is grounded in the notion that all past and current interaction of multiple language users, together with sociocultural, cognitive, and environmental factors, are directly related to the evolution of language forms. One key goal of the model developed by these authors is to describe and explain how language variants are selected and replicated by speakers. They used modeling from the field of statistical physics to formulate a statistically-based ''utterance selection model'' based on probabilities of occurrence of variants in interaction as speakers engage in linguistic communication. It is clear from the examples the authors provide that such a model is capable of explaining and predicting the extinction, selection and replication, and variability of specific features of language over time. One interesting application of this model discussed in the paper is the work done on the New Zealand dialect of English, which emerged from the various dialects of immigrants in the mid 1800s. This paper points to intriguing possibilities in the use of mathematical modeling for explaining language as a CAS.
''Linking Rule Acquisition in Novel Phrasal Constructions'' by Jeremy K. Boyd, Erin A. Gottschalk, and Adele E. Goldberg: This third paper discusses two psycho-linguistic experiments designed to assess adults' ability to learn novel syntactic-semantic pairings (linking rules) after simple exposure to input. The authors point out that one of the greatest weaknesses of the constructionist model is the paucity of testing of the model, so these experiments attempted to address this weakness. In the first experiment, adults were exposed to novel form-meaning linked constructions (through voiced-over movie clips) and then tested for both learning and retention. The authors report that the novel linking rules were learned, but that the rules were not retained to a significant degree after a lapse of one week. The second experiment required participants to describe the novel events they saw in the movies using the paired mapping constructions they learned, a test of production as well as of comprehension. Here too, the authors report that learning was significantly evident, and they conclude that the ability of adults to quickly and easily learn novel linking rules involving unknown syntax-semantics constructions is evidence for the constructionist model of language acquisition which favors cognitive-based learning processes over innateness.
''Constructing a Second Language: Analyses and Computational Simulations of the Emergence of Linguistic Constructions from Usage'' by Nick C. Ellis with Diane Larsen-Freeman: This fourth paper looks at second language acquisition of three types of English verb-argument constructions (verb plus locative, verb-object-locative, and ditransitive verbs). The research here examines inductive learning based on usage, corpus linguistics based analyses of second language learner usage, frequency effects of construction islands, and computational modeling of these factors. The authors show in this paper that social co-adaptation and cognitive processing actively follow from and then feed into language usage in a manner that facilitates language learning, and at the same time, reinforces the mapping of human socio-cognition onto the structure of language. The common thread of the findings here is that form, function, meaning, context, and usage are all inextricably linked. The paper reports on corpus data work done by Ellis and Ferreira-Junior which shows that order of acquisition of verb-argument constructions for second language learners is affected by usage factors such as input frequency distributions, semantic prototypicality of verbs, and strength of association within verb islands, much like the order of acquisition for children. The authors conclude that learner input is biased toward frequently occurring prototypical forms, thereby increasing learnability. Later in the paper, the authors report on two connectionist computer simulations that used learning algorithms to help describe how verb-argument constructions emerge in language. For all three verb-argument construction types, the model was able to learn higher frequency, higher prototypicality verbs first, and then used that learning to bolster learning of successive verbs. It was also able to use cues from the surrounding words. The second simulation removed the layer of semantic bootstrapping, and results showed similar but slower learning. The authors point out that this work empirically supports the idea that verb-argument constructions, and by extension, other linguistic constructions, emerge based on category learning and social co-adaptation through usage. They suggest that natural language systems themselves evolve through co-adaptive usage into easily acquired patterns.
''A Usage-Based Approach to Recursion in Sentence Processing'' by Morten H. Christiansen and Maryellen C. MacDonald: This paper theorizes that recursivity in linguistic constructions is learned and finite rather than part of potentially infinitely nested grammar structures as suggested in traditional models. The paper reports on results from simulation modeling and on four behavioral experiments. These authors simulated acquisition of linguistic constructions by training a ''Simple Recurrent Network'' to recognize and process constituents, in the form of various types of right, left, and center-branching recursion. This indicated to the authors that there is a close interaction between ''intrinsic architectural constraints'' and input received. The four human experiments asked participants to judge and rate grammaticality of sentences containing varying degrees of right, left, and center-branching recursion. The previous simulations predicted that with increased depth of recursion, human ratings of acceptability would decrease. Actual results reported here confirmed that modeling predictions and human processing matched. The authors conclude that the close match between the connectionist computational models and human grammaticality judgments challenges the Minimalist view of recursivity as being bound only by external memory limitations. They close by observing that this work strongly supports the view that the ability to use recursion in language is acquired gradually over time through usage experiences, and therefore, that it is variable across users and languages, making it an unlikely candidate for innateness or language evolution.
''Evolution of Brain and Language'' by P. Thomas Schoenemann: This paper assesses the evidence for changes in human cognition and language being co-evolutionary and argues that such changes are characteristic of complex adaptive systems. The author points out some important correlates of increasing brain size and the development of language, namely that brains developed for learning ability, interactive sociability and social complexity, conscious awareness, and areas of specialized conceptual processing. The author compares key areas of the human brain with their homologous areas in other primates and suggests that as higher level language processing (together with other types of higher level cognition) evolved in humans, these areas became more developed and interconnected (''biased evolutionary changes''). He indicates that this points to the inevitable evolution of semantic processing and processing of complex syntax. The author concludes that overall increases in brain size allowed for elaboration of specialized language processing areas of the human brain and for corresponding increases in cognitive ability.
''Complex Adaptive Systems and the Origins of Adaptive Structure: What Experiments Can Tell Us'' by Hannah Cornish, Monica Tamariz, and Simon Kirby: This paper reports on how an unstructured artificial language became increasingly regularized and learnable through the very process of being learned by various ''generations.'' This iterative learning showed that language evolves as an adaptive system. In the experiments outlined here, participants learned an artificial language as it was paired with pictures. They were then asked to describe pictures using the language they had learned. Their descriptions were used as the training material for the next set of learners, whose descriptions continued the iterative learning in a chain carried on for ten ''generations'' of learners. The authors indicate that the results coincided closely with results of computational models of iterated learning in which languages themselves adapt, through repeated transmission, becoming increasingly learnable and stable with progressively fewer variants and more regularization. Each generation of learners unknowingly contributed to the understanding of the succeeding generation by adapting, regularizing, and selecting certain forms that would be more likely to continue to be transmitted. The authors state that this showed that language systems themselves generate emerging structure as they are learned, used, and transmitted. Finally, the authors suggest that this work has explanatory relevance for the original evolution of language as well as for the dynamics of continued transmission of language and language change.
''Meaning in the Making: Meaning Potential Emerging From Acts of Meaning'' by Christian M. I. M. Matthiessen: The paper focuses on the construction of meaning occurring over the lifetime of individual language users. Similar to the other papers in this volume, the author invokes the idea of language as a continuous loop which simultaneously expresses, potentiates, and evolves meaning, leading to constant learning. The author provides quantified examples of various types of discourse, illustrating how certain aspects of the language system contribute to overall probability of learning and continued usage of those aspects. He shows how these types of ''probabilistic profiles'' interact with varying registers in various contexts leading to further adaptive variation in the system itself. Significantly, the author suggests that these characteristics of adaptability and variability are precisely the conditions requisite to language learnability, especially as learning occurs over time. He details the human phases of ''learning how to mean,'' emphasizing that acquisition of additional register repertoires is a key component of continued learning. The author reiterates the central point that individuals construct meanings throughout life by expanding their understanding of registers and that these personalized meaning potentials collectively create the larger language system.
''Individual Differences: Interplay of Learner Characteristics and Learning Environment'' by Zoltán Dörnyei: This paper attempts a firmer conceptualization of the dynamic interplay between learner, language, and environment, specifically in second language acquisition, focusing on the highly variable nature of each factor. The author criticizes traditional conceptualizations of second language acquisition that monolithically distinguish single aspects such as motivation or aptitude, arguing that it is essential to consider the highly fluctuating nature of these aspects as they relate to both temporal and contextual variables. He insists that motivation, aptitude, learning styles, and other aspects are in constant interplay with each other and with the changing contexts and environments of language learning. He discusses the importance of carrying out quantifiable research on the dynamic interrelationships among language, cognition, social interaction, and environment. The author maintains that unless all these factors are studied as an integrated and dynamic complex, analyses will fall into over-simplistic or even false characterizations. In the last section of the paper, the author tries to provide direction for overcoming the challenges posed by empirically studying language acquisition as a dynamic enterprise. Many of the papers in this book are already attempting to meet the challenge of innovating more adequate experimental design in the face of the adaptive systems paradigm shift as suggested by this author.
''If Language Is a Complex Adaptive System, What Is Language Assessment?'' by Robert J. Mislevy and Chengbin Yin: The last paper encourages a reconceptualization of language assessment based on the emerging view of language as a complex adaptive system. The authors, using a connectionist paradigm, suggest that rather than testing discrete language traits, assessments should consider interactions among people and situations. They demonstrate the application of evidentiary argument structure to formulations of language assessment structure and consider how various assessment configurations might be reconceptualized in the interactionist approach. The authors criticize the behavioral approach of many traditional decontextualized test configurations, claiming that such tests do not adequately predict performance in real usage situations. They encourage a more realistic approach to testing, that of using communicative situations to gain more valid observations of learner features such as appropriateness and effectiveness of language use within a context as well as fluency and accuracy of use. The authors point out that factors such as individual profiles of exposure to the language items being tested and previous experience with task demands can affect test results as much as actual language proficiency. They argue that tests that target specific structures are bound to under-represent dimensions of language that are required in real situations and suggest that tests that simulate real usage situations are more fully able to provide evidence about students' real-world interactive capabilities in language.
Readers of this book are certain to gain a great sense of increased understanding, not only of the workings of language but also of current research innovations within the Emergentist paradigm. All ten of the papers are clearly written so that those with little previous exposure to this type of work will be easily engaged and be able to follow the evidence and arguments presented. At the same time, there is enough technical detail to provoke thought among those who are already immersed in connectionist endeavors. The research described here is careful and convincing, yet many of the papers also point out ways in which future research can be refined, so there is a wealth of ideas for those hoping to carry out similar studies.
The goal of the editors of this book, to present a ''path-breaking'' perspective on how to understand language processing and learning, has been amply met. The chapter contributors do not neglect to emphasize the problems inherent in dealing with the complexity of linguistic forms, language usage, cognition, and change and variability. Yet, they clearly point out multiple avenues for approaching such complex systems in ways that are both valid and viable.
One strength of the book is the continuity of perspective present throughout the various papers. The papers are unified in their recognition of the same problems, such as accounting for variability and change, but they represent a wide range of fields (from sociolinguistics to anthropology to psycholinguistics to evolution) so that together they are able to build a solid front composed of mutually supporting research. Before going into the specifics of their particular project, the authors of each paper provide a short discussion of the problems and questions involved in re-imagining language from the connectionist viewpoint; however, this reiteration is done from such a variety of perspectives that it does not become redundant.
The only minor inconsistency in the book is that the first nine papers are largely research-based and address the larger problem of conceptualizing language processing and acquisition while the last paper jumps to a more practical question, that of language testing. While this topic is certainly relevant to the paradigm shift suggested in the book, it might have been a better fit to place the last paper in a collection of work that is more praxis oriented.
While the emergentist paradigm has been developing for at least the last twenty years, it has yet to gain full acceptance in the mainstream, so this book is bound to spark controversy with its clear rejection of traditional paradigms. Even so, or perhaps because of this, the book is an insightful contribution to those who are interested in exploring the connections between cognition, language, and society. It has certainly broken ground for research to come.
ABOUT THE REVIEWER
ABOUT THE REVIEWER:
Julie Bruch received her Ph.D. in Linguistics from the University of
Kansas. She currently teaches Linguistic Diversity, History of English,
Structure of English, and Beginning Japanese at Mesa State College in
Colorado. Her principle research interests are culture and language and
language change and diversity.