Music samples Emergentism theory Deb Roy, part 1 Connectionism Gephi, The machine that spoke for itself Deb Roy part 2
The EG metaphor for language is a blueprint. Blueprints do not specify every brick and every nail, but with a blueprint a builder is enabled to constitute an artifact by combining structural outline with a set of conventions and outlines and experience, the meaning of which is determined through consultations of various kinds. .\\ Language is a code because its forms stand for invisible meanings that exist independently of it. The task is one of encoding and decoding. Language is also a conduit through which ideas formed in the head of one speaker are encoded and transmitted to the head of another speaker, where they are decoded and reconstituted as the same ideas. Grammar is a byproduct of communication Use drives form-meaning Use based on what speakers have heard, current context, and history. Utterances are repeated Grammar is prefabricated and repeated in local variations Grammar lives in lexicon Grammar/lexical items reliant on context ALL discourse is specific to individual and context Grammar is bottom-up Grammar is innate Form drives language Acquisition by innate knowledge-checking off grammar items as you learn them Grammar/lexical items are accessible at all times Meanings are fixed until deployed in conversation Grammar is processed top-down
We say things that have been said before. Our speech is a vast collection of hand me downs reaching back to the beginning of language, the aggregation of changes and adjustments that are made to this inheritance on each occasion and use result in constant erosion and replacement of the sediment we call grammar.
Utterances are novel. Language is not fixed. Therefore grammar must be studied as it occurs naturally - in context with all of its imperfections
Through “chunking” Bottom-up and top-down processes interact. Linguistic Chunks: learning groups of letters, words, phrases that collocate and often and help the speaker predict what’s coming up in the discourse. More repetition = stronger association
Memory and analysis create and restructure your neural associative network. Gephi and Machine video This means that the task of learning a language must be reconceived as expanding a repetoire of communicative contexts. New contexts and new occasions of negotiation occur constantly thus language learning never ends.
<ul><li>Is grammar the chicken or is it the egg? </li></ul><ul><li>Does an innate grammar make conversation possible or does conversation make a “grammar” possible? </li></ul>APG vs. EG
<ul><li>How do children acquire language given that the input they receive is both imperfect and incomplete? </li></ul><ul><ul><li>Innate knowledge fills the gaps (i.e., we fill in the details of an already existent pattern). </li></ul></ul><ul><ul><li>Common possession of rules is a prerequisite for understanding. </li></ul></ul><ul><ul><li>Without the grammatical egg, you have no linguistic chicken. </li></ul></ul>The Egg Hypotheggsis (APG)
<ul><li>Grammar is not the source of understanding and communication, but a byproduct of it. </li></ul><ul><li>Structure or regularity comes out of discourse and is shaped by discourse in an ongoing process. </li></ul><ul><li>Grammar is simply the name given to observed repetitions in discourse. </li></ul>The Chicken Theory (EG) note not: (egg)
<ul><li>Simple learning mechanisms operating within and across human systems for perception, motor action and cognition, while exposed to data in a communicatively rich human social environment navigated by an organism eager to exploit the functionality of language suffice to drive the emergence of complex language developments. (Ellis, 2003). </li></ul>How does linguistic chicken come about?
<ul><li>The acquisition of grammar is the piecemeal learning of many thousands of constructions and the frequency-biased abstraction of regularities within in them. </li></ul>Usage-based perspective
Linguistics (Hopper, 1998) <ul><li>Oral and written communication </li></ul><ul><li>Snapshots of utterances out of context deprive the analysis of important data (temporal change) </li></ul>~Don’t go in that little hole. ~Just think of me as a used mount. ~It goes much quicker when I’m on the boss. ~It’s really important that Mildweasel gets that sheep off . (Excerpts of dialogue from an online game, which uses a large amount of its own emergent terminology)
Connectionism… <ul><li>“ provides computational tools for exploring conditions under which emergent properties arise”. (Ellis, 1998, p.645) </li></ul><ul><li>asserts that learning language is not different from other skills. </li></ul><ul><li>sees the brain as a recorder of statistics. </li></ul><ul><li>claims pattern analysis comes after accumulating enough data. </li></ul>
How associative learning works <ul><li>Memory </li></ul><ul><li>Analysis </li></ul><ul><li>Prediction </li></ul>
Memory <ul><li>Working memory, Long-term memory and chunking </li></ul><ul><li>Normal working memory capacity is 7 +/- 2 units </li></ul><ul><li>Organization and repetition facilitate memorization </li></ul><ul><li>Words vs. Syntax </li></ul>
Analysis This is your brain on connectionism. Any questions?
Prediction <ul><li>Comprehension relies on knowing which patterns to expect in which contexts </li></ul>
Prediction <ul><li>Idiomaticity: Token vs. Type Frequency </li></ul><ul><li>Native competency relies on mastery of idiomatic structures, not just grammar </li></ul>
Discussion <ul><li>1. Do you think grammar is the egg or the chicken? If it’s the chicken, how do we cross the road of acquisition? </li></ul><ul><li>2. Does connectionism, using computers to mimic human learning processes, work as a model of acquisition? </li></ul><ul><li>3. What implications do you see for SLA from a constructionist/emergentist view of grammar? Role of input, output, authentic materials, individual differences, etc? </li></ul>