Input, Interaction, Feedback, Evaluation Second Language Acquisition and Multimedia Environments
SIXTH CONFERENCE Université de Technologie of Use of New de Compiègne Technologies in Foreign Language 24 - 26 March 2005 Teaching
Nick Ellis Professor of Psychology Research Scientist, English Language Institute University of Michigan University of Michigan [email protected] InterfaceInterfaceInterfaces in SLA ands CALL: DYNAMIC interactionsinteractions
SLACALL CMCSLA
ofof
Languageexplicit and Media implicit and languagelanguage knowledge Learners
2 The Interface question Interactions of
Acquisition Learning Naturalistic Instructed Implicit learning Explicit learning Skilled performance Metalinguistic knowledge Procedural knowledge Declarative knowledge Usage-based acq’n Focus on form Communicative Grammar instruction
3 Strong interface position Focus on forms Traditional FL instruction Grammar-translation methods: explicit bias, formal explanation of L2 rules, a deductive approach to learning 1960s - explicit instruction of grammatical rules in the Cognitive Code Method, ‘a modified, up-to-date translation theory’ (Carroll, 1966): perception and awareness of L2 rules precede their use
But L2 learners who can tell more about a language than a native speaker, yet whose technical knowledge of grammar leaves them in the lurch in conversation tell us
4 Krashen’s Input Hypothesis a Non-interface position adults can both subconsciously acquire languages and consciously learn about language, however, (i) subconscious acquisition dominates in L2 performance; (ii) learning cannot be converted into acquisition (iii) conscious learning can be used only as a Monitor, an editor to correct output after it has been initiated by the acquired system
But (1) all of the times input fails to become intake suggests SL Acquisition isn’t enough… and (2) the evidence on effects of instruction shows there is interface
5 What is the nature of the interface?
• However separate and unalike they are, implicit and explicit knowledge do interact • How does explicit language instruction affect implicit learning?
• Their interface is dynamic: It happens transiently during conscious processing, but the influence upon implicit cognition endures thereafter
6 Broad Plan Dissociating Implicit and Explicit knowledge and processing Language leaning is implicit learning Usage-based perspectives hold that the acquisition of language is the piecemeal learning of many thousands of constructions and the frequency-biased abstraction of regularities within them. Rational L1 acquisition? SLA Irrational? When frequency fails to drive learning as a result of X- linguistic transfer . Implicit learning limitations . Learned Attention in Associative Learning and Automatization . When language learning is explicit learning. FonF. . The importance of noticing in the initial registration of a pattern recognition unit. Conscious processing . The role for explicit instruction . Interface of implicit and explicit learning in SLA
7 Dissociations
MEMORY:
• Explicit memory: where recall involves a conscious process of remembering a prior episodic experience
• Implicit memory: where there is facilitation of the processing of a stimulus as a function of a recent encounter with an identical or related stimulus but where the subject at no point has to consciously recall the prior event
8 Neuropsychological evidence: The dissociation between Explicit and Implicit memory in Anterograde Amnesia •H.M. (Milner et al 1968): –became amnesic in 1953, at the age of 27, following bilateral removal of his medial- temporal lobe, including the anterior two- thirds of the hippocampus, the posthippocampal gyrus, and the amygdala. 9 10 HM could not learn new concepts or new vocabulary. Indeed he could not form any new explicit memories at all:
The following are excerpts from a description of H.M.'s condition 14 years after the onset of his amnesia: He still fails to recognize people who are close neighbours or family friends but who got to know him only after the operation . . . although he gives his date of birth unhesitatingly and accurately, he always underestimates his own age and can only make wild guesses as to the date . . . During . . . nights at the Clinical Research Center, the patient rang for the night nurse, asking her, with many apologies, if she would tell him where he was and how he came to be there. He clearly realized that he was in a hospital but seemed unable to reconstruct any of the events of the previous day. On another occasion he remarked 'Every day is alone by itself, whatever enjoyment I've had, and whatever sorrow I've had.' Our own impression is that many events fade for him long before the day is over. He often volunteers stereotyped descriptions of his own state, by saying that it is 'like waking from a dream'.
Profound anterograde amnesia (inability to consolidate new explicit memories) resulting from bilateral hippocampal damage, yet…
11 Preserved implicit memory HM still good at –Perceptual learning - mirror reading –Perceptual learning - lexical repetition priming –Perceptual learning: visual object recognition –Perceptual learning - nonword reading –Motor learning - pursuit- rotor tracking –Classical conditioning
12 • The hippocampal system • The neocortical systems subserves rapid EXPLICIT subserve IMPLICIT memory: one-off learning, memory • the establishment of new • the tuning of associative conjunctions of arbitrarily systems to reflect repeated different elements into a patterns of local activity and unitized representation to generalize from them, • the learning of separate discrete episodic memories • generalizations rather than where we do not want an episodic memory. average, an abstraction, or a gist: • To operate efficiently in the • There is benefit in being world we need to be able to able to keep some records identify general patterns by straight, complete, and abstracting from instances - distinct. to classify and categorize.
13 Dissociations
LEARNING: Implicit and explicit learning are quite different styles of learning, varying, in the degree to which acquisition is driven by conscious beliefs, as well as in the extent to which they give rise to explicit verbalisable knowledge
• Implicit learning: acquisition of knowledge about the underlying structure of a complex stimulus environment by a process which takes place naturally, simply and without conscious operations.
• Explicit learning: a more conscious operation where the individual attends to particular aspects of the stimulus array and generates and tests hypotheses in a search for structure.
• Both modes of learning apply to differing extents in all learning situations.
14 Experimental evidence: Repetition Priming • Word processing is faster if you have processed that word recently • This occurs whether you explicitly remember the prior encounter or not, and thus it demonstrates implicit memory - the tuning of the lexicon by experience • In a lexical decision task subjects have to rapidly decide whether a particular letter string is a word (e.g. watch) or not (e.g. wetch); priming is reflected by a decreased latency in the making of a lexical decision on the second presentation of a letter
string relative to the first: 15 Priming Studies Of Implicit Memory Systems Repetition Priming to Measure Implicit Memory Effects
Lexical Decision Task
Subject's response Stimulus items presented hours or Letter string presented even days previously Keypress reaction time
t0
857ms nufce N
table 744ms child train Y nurse 723ms nurse Y fire 754ms television horse Y
.. 717ms table Y ...
Lexical decision responses are made faster for stimuli which they have seen before, 16 (even though the subject may have no conscious memory of having done so). Priming Studies Of Implicit Learning Systems
• Marcel (1970s) - you get this priming even if the first processing of the prime is so fast, so automatic, or so masked to be conscious
17 Dissociations
The NCC - Unconscious and conscious brain processing. Dozens of studies:
•Without consciousness, optical input only activates local visual cortex, where the stimulus is taken apart and then reconstructed as objects and events
•Conscious stimuli, on the other hand, activate not just the back of the brain (where visual cortex is located), but parietal, frontal, and medial-temporal regions, including the hippocampus and amygdala. It is as if conscious contents are globally distributed throughout many brain regions
18 Conscious perception activates widespread brain regions Dehaene et al 2001, Nature Neuroscience
• Words like “LION” and “NOTE” are shown for 20 ms. • In the unconscious case they are preceded and followed by pattern masks. In the visible case they are followed by black fields, making the words conscious. • Below, recognition tests show that there is no memory for the masked words, while memory for the conscious words is quite good. This is to check whether masking is effective. • This experiment might therefore answer the question: What is the difference between conscious and unconscious visual words, in terms of their cortical activity? 19 Results of Dehaene’s (2001) experiment: Conscious vs. Unconscious visual words
• Masked words activate visual cortex. • The same words, when they are conscious, also activated much wider regions in parietal and frontal cortex. Conscious Unconscious
20 Similar results have now been found in more than 10 studies for vision
21 Dehaene et al 2001, Nature Neuroscience. • Explicit Memory Implicit Memory
• Explicit Learning Implicit Learning
• Conscious Processing Unconscious Processing widespread, synchronized, unitary localized, parallel, competition resolution of ambiguity in a statistical ensemble
23 IMPLICIT LANGUAGE LEARNING
Implicit language learning from usage - tallying, tuning, competition, local neocortical tuning, connectionist learning
• We learn language while using language. When things go right, when routine communication flows easily, this time on task tunes our skills without us giving much thought to the learning process • Implicit learning of language occurs during fluent comprehension and production.
24 Question, What does wZn mean? That's wZn small step for man, one giant leap for mankind To love wZn self is the beginning of a lifelong romance Alice in wZn derland wZn the battle, lost the war The human brain is a wZn derful thing. It starts working the moment you are born, and never stops until you stand up to speak in public wZnce upon a time... things were ever ambiguous…
25
“Perception is of definite and probable things” (James, 1890) There is a lot more to the perception of language than meets the eye, or ear. A percept is a complex state of consciousness in which antecedent sensation is supplemented by consequent ideas which are closely combined to it by association. Apperception. Where the sensation is associated with more than one reality, unconscious processes weigh the odds, and we perceive the most probable thing: "all brain- processes are such as give rise to what we may call FIGURED consciousness" (James, 1890, p. 82).
27 Probabilistic language processing
Frequency (& relative frequency) effects across all levels of language representation
28 Constraint-satisfaction accounts of adult language comprehension - Syntactic ambiguity resolution geometric element? airplane? tool? The_plane left for_the…. direction? past tense leave - active voice? past tense leave - passive voice?
1) The plane left for the East Coast 2) The plane left for the reporter was missing 3) The note left for the reporter was missing 29 Psycholinguistic experiments show these ambiguities are resolved by rapidly exploiting a variety of probabilistic constraints derived from previous experience first-order frequency information:- plane (vehicle) much more frequent that other meanings left used more frequently in active than passive voice & combinatorial constraints:- sentence 3 is more easy to comprehend as a reduced relative clause than sentence 2 because it is more plausible for a note to be left than to leave 30 “Adults have a vast amount of statistical information about the behaviour of lexical items in their language… Comprehenders know the relative frequencies with which individual verbs appear in different tenses, in active vs. passive structures, and in intransitive vs. transitive structures, the typical kinds of subjects and objects that a verb takes, and many other such facts. This information is acquired through experience with input that exhibits these distributional properties…. This information is not some idiosyncratic fact in the lexicon isolated from "core" grammatical information; rather, it is relevant at all stages of lexical, syntactic and discourse comprehension.” (MacDonald et al, 1994)
31 Comprehension: "a comprehender tends to perceive the most probable analysis of a new utterance on the basis of frequencies of previously perceived utterance analyses" (Bod, 1998, p. 8) Grammatical constructions that a person has performed before are preferred to analyses that must be newly constructed (Bock, 1986)
Production: The greater the working memory demands of the task, the greater the need to rely on formulae - "it is easier for us to look something up than to compute it" Bresnan (1981). "a language user tends to produce the most probable utterance for a given meaning on the basis of frequencies of utterance-representations" (Bod, 1998, p. 3). 32 Frequency (& relative frequency) effects across all levels of language representation bigram frequencies in word identification (Broadbent) phonotactic knowledge in speech segmentation (Aslin) spelling-to-sound correspondences in reading (Coltheart) cohort effects in spoken word recognition (Marslen Wilson) friends/enemies effects in morphology, spelling, reading competition model (MacWhinney) syntactic productivity depends on type frequency (Bybee) constraint satisfaction in sentence processing (Seidenberg) essential to statistically augment NLP grammars (Jurafsky) And the rest….. Psycholinguistics is the body of evidence demonstrating that language acquisition is usage-based and probabilistic Ellis, N. C. (2002). Frequency effects in language acquisition: A review with implications for theories of implicit and explicit language acquisition. (Target article) Studies in Second Language Acquisition, 24, 143-188. 33 Implicit Language Learning
We are very well tuned to our language input at all levels But the on-line conscious experiences of language learning involve language understanding rather than counting So, language language is usage-based implicit learning, the gradual strengthening of associations between co-occurring elements of the language and their functional interpretations.
34 Connectionism, Competition, and the Emergence of Constructions
35 The Emergence question Connectionist explorations of whether we can we get from Usage to Structure
Implicit learning ExplicitRule-like learning language processing Acquisition ReadLearning slopendash aloud Tom had a wug. We gave him another. So he Skilled performance hadMetalinguistic two …. knowledge Procedural knowledge Declarative * Mary Tom knowledge splugged. Usage-based acq’n Focus√ Mary on formsplugged Tom. `Twas brillig, and the slithy toves Did gyre Naturalistic andInstructed gimble in the wabe Communicative Grammar instruction
36 Language acquisition as contingency learning The learner as an intuitive statistician
Language processing ability emerges from the conspiracy of the memories of all of the utterances we have experienced ‘Rules’ emerge from the collaborations of patterns in exemplars ‘System’ is a set of frequency-biased statistical abstractions, ‘tuned’ by experience
Ellis, N. C. (1998). Emergentism, connectionism and language learning. Language Learning, 48, 631-664.
37 38 39 But what about when frequency fails to drive learning?
Input | Input | Intake Input | Input | Intake Input |
40 A founding observation of applied linguistics
although learners are surrounded by language, not all of it “goes in” Corder’s distinction between input, the available target language, and intake, that subset of input that actually goes in and which the learner utilizes in some way
41 Klein & Perdue: The Basic variety ESF project, 40 adult learners of different L1s and L2s picking up the language of their social environment by everyday communication A special language form we called the Basic Variety. All learners, independent of source language and target language, developed and used it; about one-third of them also fossilised at this level—that is, they learned more words, but they did not complexify their utterances in other respects, particularly in morphology or syntax.
42 Klein (1998) The Basic Variety is highly efficient for communicative purposes. by far most lexical items correspond to nouns, verbs and adverbs; closed-class items, in particular determiners, subordinating elements, and prepositions, are rare, if present at all. there is no functional morphology whatsoever: no tense, no aspect, no mood, no agreement, no casemarking, no gender assignment
43 Why these apparent Irrationalities in SLA?
Insights from Associative Learning theory
Learned Attention in Animal and and Second Language Learning:
Pigeons & Pidgins
44 Learning depends upon Attention
45 Selective attention for SLA
“the concept of attention is necessary in order to understand virtually every aspect of second language acquisition (SLA), including the development of interlanguages over time, variation within IL at particular points in time, the development of L2 fluency, the role of individual differences such as motivation, aptitude, and learning strategies in L2 learning, and the ways in which interaction, negotiation for meaning, and all forms of instruction contribute to language learning” (Schmidt, 2001).
46 Attention depends upon the physical salience of the CS (stimulus) and its outcome
47 Rescorla & Wagner (1972) model of conditioning expressing the capacity any cue (CS) has to become associated with an outcome (US) at any given time dV = ab(L - V) V the associative strength of the US to the CS dV the change in this strength which occurs on each trial of conditioning a the salience of the CS b the salience of the US L the amount of processing given to a completely unpredicted US. • Learning depends on the salience of the cue • Learning depends on the importance of the outcome (interpretation) • The more a CS is associated with a US the less additional association the US can induce (power law of learning). • The more predicted the US (from context and other cues), the less the additional association on this trial 48 Grammatical particles tend to be psychophysically non-Salient -> lenition / erosion in production, lack of perceptability, language change
49 Salience Salience and Production (Bybee, Zipf) frequent morphemes are more susceptible to erosive lenition processes which cause loss and overlap of gestures grammatical morphemes tend to be shorter and phonologically fused with surrounding material Salience and Perception (Bates) When grammatical function words are clipped out of connected speech and presented in isolation, adult native speakers can recognize them no more than 40% to 50% of the time (Herron & Bates, 1997)
50 and Salience depends upon Learning History and Learned Attention: Overshadowing and Blocking
51 Overshadowing and Salience Salience for the birds •Two cues, C1 and C2, are always presented together during training and jointly predict an outcome.
-> the strength of conditioning to each cue depends on their relative intensity.
If C1 is a dim light and C2 a bright light then the learned response to the bright light is very strong while the dim light alone produces little or no reaction
•C2 has been overshadowed by the more salient C1 (Kamin, 1969). 52 Blocking is automatically learned inattention
. What we attend to is determined by our prior experience
. Salience is a psychological not just a physical property
53 Blocking in SLA L2 Temporal expression interlanguage development Meisel, Schumann 1. No explicit temporal reference 2. Reference established exclusively by adverbials (soon, now…) serialization (presention in order of occurrence) calendric reference (May 12, Monday) 3. Introduction of verbal morphology …increasingly systematic use Bardovi-Harlig (2000, pp. 414-415) “The expression of temporality exhibits a sequence from pragmatic to lexical to grammatical devices” “Whereas all learners apparently achieve the pragmatic and lexical stages of development, fewer learners achieve the morphological stage of development.
54 Blocking
Second language learners already know about temporal adverbs and narrative strategies for serialization, these strategies are effective in the communication of temporality
the high salience of these means of expression leads L2 learners to attend to them and to ignore the phonologically reduced tense-markings
55 Learned attention becomes Automatized as a result of Implicit Learning from Usage
56 Control Stroop Stroop The Stroop Effect Condition Condition Condition
Name the ink colours *** RED ROUGE from top to bottom as ***** BLACK NOIR quickly as you **** GREEN VERT ***** BLUE BLEU can ***** YELLOW JAUNE ****** GREEN VERT **** RED ROUGE ***** BLUE BLEU *** BLACK NOIR RED ROUGE *****
Familiar words are coherent perceptual units, so immediately and automatically available that a literate person is unable
not to recognize them 57 Automaticity
• Automatization is the acme of L1 learning in that it represents the implicit language system being so optimally tuned that it will ballistically provide the appropriate interpretation of a cue while conscious thought is occupied elsewhere
58 But Automatization drives Transfer and Interference too
e.g. intonation patterns
59 Iambic English:
Trochaic French: The d'Antin Manuscript
Noyé, l’ami, dans tout, sa lippe, Après d’alarmants sauts, l’équipe, En duvet deuil beffroi évêque… Apprête alore ma salle de teck.[1]
[1] Lament: “Scornful of life, the friend was drowned After alarming leaps by the clique. In downy morning the bishops tower… Prepare then my room of teak.” A room of teak is obviously a coffin. The first line, however, could mean that, unlike Shelley, our nameless friend did not drown, but drank himself to death -- a much more common and unromantic end. Un petit d'un petit [1] The d'Antin S'étonne aux Halles [2] Manuscript Un petit d'un petit Ah! degrés te fallent [3] Indolent qui ne sort cesse [4] Indolent qui ne se mène [5] Qu'importe un petit d'un petit Tout Gai de Reguennes. [6] [1] The inevitable result of a child marriage. [2] The subject of this epigrammatic poem is obviously from the provinces, since a native Parisian would take this famous old market for granted. [3] Since this personage bears no titles, we are led to believe that the poet writes of one of those unfortunate idiot-children than in olden days existed as a living skeleton in their family's closet. I am inclined to believe, however, that this is a fine piece of misdirection and that the poet is actualy writing of some famous political prisoner, or the illegitimate offspring of some noble house. The Man in the Iron Mask, perhaps? [4,5] Another misdirection. Obviously it was not laziness that prevented this person's going out and taking himself places. [6] He was obviously prevented from fulfilling his destiny, since his is compared to Gai de Reguennes. This was a young squire (to one of his uncles, a Gaillard of Normandy) who died at the tender age of twelve of a surfeit of Saracen arrows before the walls of Acre in 1191. Automaticity • Automatization is the acme of L1 learning … • But if that same form occurs in the L2 but cues something else, then • Automatization is the bane of SLA
• There is no way that implicit learning mechanisms can cope with a new interpretation, as they are automatized to do different.
63 What then to do w/r facilitating SLA? Exploit explicit learning and instruction
64 EXPLICIT LANGUAGE LEARNING
• When communication breaks down, we try hard to negotiate meaning, and we learn a lot about linguistic construction in the process
• Explicit learning of language occurs in our conscious efforts to negotiate meaning and construct communication. FonF- noticing, attending, following the attentional guidance of our interlocutors, scaffolding
•It also occurs during explicit instruction of different types, and during conscious analysis of language form and its functions WM - analogizing, problem solving guided by declarative metalinguistic knowledge FonFs
65 These are some of the targets / mechanisms of the effectiveness of explicit L2 instruction Norris & Ortega (2000). Effectiveness of L2 Instruction: A research synthesis and quantitative meta-analysis. Language Learning, 50:3. Meta-analysis of 1980-1998 findings from experimental and quasi-experimental investigations into the effectiveness of L2 instruction. Comparisons of average effect sizes from 49 unique sample studies reporting sufficient data indicated that focused L2 instruction results in large target-oriented gains that explicit types of instruction are more effective than implicit types that Focus on Form and Focus on Forms interventions result in equivalent and large effects that the effectiveness of L2 instruction is durable
66 Baars on Consciousness & Learning “The more novelty we encounter, the more conscious involvement is needed for successful learning and problem-solving” The NCC involve a coalition of forebrain neurons implicated in working memory and planning, interconnected via widespread cortico-cortico and cortico-thalamic feedback loops with sets of neurons in sensory and motor regions that code for particular features. Long distance fibres, gamma synchronization etc.
67 What is human consciousness for?
To solve problems that existing modules can’t - mismatch-driven To step outside the here-and-now. Scrutiny and manipulation of material made contiguous in Working Memory Noticing & focus of attention Uniting & solving the binding problem Consolidation of unitized representation. Formation of a pattern-recognition unit …a lesson in Quinean…
68 “Gavagai”
69 Quine (1960)
Referential Indeterminacy: Single words cannot be paired with experiences, since they confront experience in clusters
Why translate 'gavagai' as 'rabbit’, why not translate it as, say, 'undetached rabbit-part'?
For any experience which makes the use of 'rabbit' appropriate would also make that of 'undetached rabbit- part' appropriate
W.V.O. Quine (1960). Word and Object, MIT Press, Cambridge Mass.
70 Consciousness for focussed binding
“Hippocampus” 71 having changed what learners focus on in their processing, this focus lives on, changing what their implicit learning mechanisms subsequently tune
72 Scott Kelso on Dynamic processes In the human brain, information (as a marginally coupled, phase-locked state) is created and destroyed in the metastable regime of the coordination dynamics, where tendencies for apartness and togetherness, individual and collective, segregation and integration, phase synchrony and phase scattering coexist. New information is created because the system operates in a special régime where the slightest nudge will put it into a new coordinated state. In this way, the (essentially nonlinear) coordination dynamics creates new, informationally meaningful coordination states that can be stabilized over time. The stability of information over time is guaranteed by the coupling between component parts and processes and constitutes a dynamic kind of (nonhereditary) memory.
73 Some stability, and some slight nudges…
74 THE EFFECTS OF SET UPON PERCEPTION
The d'Antin Manuscript
Reine, reine, gueux éveille.
Gomme à gaine, en horreur, taie. [1]
[1] "Queen, Queen, arouse the rabble
Who use their girdles, Horrors, as pillow slips."
The d'Antin Manuscript was published in 1967 by Courtlandt H. K. Van Rooten by Angus & Robertson (U.K.) Ltd., 16 Ship Street, Brighton, Sussex. It was edited and annotated by Luis d'Antin Van Rooten Once you learn to perceive an object in that way, that's the way you perceive it (or that's the way you first perceive it), and those are the features whose strengths are incremented on each subsequent processing episode. Exaggerated input / FonF to get you first to notice in a different way. Once consolidated, it is this new way of seeing whose strengths are incremented on each subsequent processing episode.
Review: Ellis, N. C. (in press, 2005). At the interface: How explicit knowledge affects implicit language learning. Studies in Second Language Acquisition
78 Baars on Consciousness & Learning Implicit learning occurs largely within modality and involves the priming or chunking of representations or routines within a module; it is the means of tuning our zombie agents, the menagerie of specialized sensori-motor processors that carry out routine operations in the absence of direct conscious sensation or control . conscious processing is spread wide over the brain and unifies otherwise disparate areas in a synchronized focus of activity. Conscious activity affords much more scope for focused long-range association and influence than does implicit learning. It brings about a whole new level of potential associations. 79 Consciousness is the interface
Consciousness is the publicity organ of the brain. It is a facility for accessing, disseminating, and exchanging information and for exercising global coordination and control
“Paying attention—becoming conscious of some material—seems to be the sovereign remedy for learning anything, applicable to many very different kinds of information. It is the universal solvent of the mind”. (Baars)
80 In conclusion
Separable representational systems In dynamic interaction Just the slightest nudge will do it…
81 Eau la quille ne cole Old King Cole Oise à mer est haulte de soles Was a merry old soul Aîné marié au sol, vas-y! merry old soul was he École vorace paille face He called for his pipe, face Pain école vorace boule he called for his bowl, En école vorace And he called for his fille de loterie. fiddlers three Et vérifie d’allure, Each fiddler ah! des fidèles face had a fiddle, En avarie faille ne fille te l’a dit very fine fiddle had he; Et puis, tu lui dis, tu lui dis, vingt-deux filles de loure. Oh! d’hère, se nom soeur erre there’s none so rare Ascain compère As can compare Huit qui ne collent With King Cole and his et ne se fient de loterie. fiddlers three
Learning mechanisms Associative Learning & implicit frequency gathering, the chunking of conjunctions that are within the scope of implicit learning Noticing and explicit learning may be needed for laying down the recognition unit Consciousness to counteract Automatization and Interference from L1, both content-full (the representations), and attention-focussing - where you look for cues in language, thinking for speaking Complementary learning and memory systems: Cortex Hippocampus & WM Implicit memory Explicit memory Connectionist tuning Unitized episodes 84 CALL & Explicit learning
Hippocampus & WM Explicit memory Unitized episodes
Noticing and explicit learning may be needed for laying down the recognition unit Consciousness to counteract Automatization and Interference from L1, both content-full (the representations), and attentional biases - where you look for cues in language, thinking for speaking
85 CALL & Explicit learning Norris & Ortega (2000): focused L2 instruction results in large target-oriented gains explicit types of instruction are more effective than implicit types Focus on Form and Focus on Forms interventions result in equivalent and large effects Rescorla & Wagner: dV = ab(L - V) • Learning depends on the salience of the cue • Learning depends on the importance of the outcome (interpretation) • BLOCKING:The more a CS is associated with a US the less additional association the US can induce • The more predicted the US (from other cues in context), the less the additional association on this trial -> • Interaction of Imp/Exp FonF/FonFS and salience / importance / degree of overshadowing / ΔP. The lower the salience, the less the functional significance, the greater the blocking, the less reliable the cue- interpretation contingency, the more the recast has to be explicitly focussed 86 CALL & Explicit learning Transfer appropriate processing - that slightest nudge has to come to mind when appropriate Levels of processing / Depth / Elaboration Context embeddedness Increasing Ratio Review of practice and repetition Retrieval Practice effects …….
87 CALL & Implicit learning Cortex Implicit memory Connectionist tuning Associative Learning & implicit frequency gathering, the chunking of conjunctions that are within the scope of implicit learning Corpus linguistics Representative written / spoken samples of authentic language for tallying and probabilistic tuning from Usage EAP ESP specialist corpora AWL AFL Illustrative range of the options for ambiguity (probability ordered). Corpus-driven testing Shortcuts from psycholinguistic analyses - WordNet Visual
Thesaurus - and from corpus analyses - Collocations 88 Pierre Bastien & Lukas Simoni Interface or bust // Your busted interface?
Luria , having separately analyzed the workings of the three principal functional units of the brain (the unit for regulating tone or waking, the unit for obtaining, processing, and storing information, and the unit for programming, regulating, and verifying mental activity), emphasized that it would be a mistake to imagine that each of these units carry out their activity independently: Each form of conscious activity is always a complex functional system and takes place through the combined working of all three brain units, each of which makes its own contribution … all three principal functional brain units work concertedly, and it is only by studying their interactions when each unit makes its own specific contribution, that an insight can be obtained into the nature of the cerebral mechanisms of mental activity. (pp. 99-101, italics in original) William James (1890) "The particulars of the distribution of consciousness, so far as we know them, point to its being efficacious. ... It seems an organ, superadded to other organs which maintain the animal in the struggle for existence; and the presumption of course is that it helps him in some way in the struggle ..."
90 Global Workspace theory, functions of consciousness for me, the New Psycholinguistics of SLA is an Associative-Cognitive CREED
Construction-based (probabilistic) Rational (the best model of usage, the input to date) Exemplar-driven (friends and enemies) Emergent (prototypical patterns, productive schema, tuning to L1, learned attention) Dialectic (explicit/ implicit) neuroscience of noticing “The novelties of one generation are only the resuscitated fashions of the generation before last” (George Bernard Shaw) 92
At the interface: Dynamic interactions
face vase face Nick Ellis Professor of Psychology Research Scientist, English Language Institute University of Michigan of University of Michigan [email protected]
explicit and implicit language knowledge
94 OISE / University of Toronto MLC colloquium series in Applied Linguistics, February 25, 2005 This paper considers how implicit and explicit knowledge are dissociable but cooperative. It reviews various psychological and neurobiological processes by which explicit knowledge of form-meaning associations impacts upon implicit language learning. The interface is dynamic: it happens transiently during conscious processing, but the influence upon implicit cognition endures thereafter. The primary conscious involvement in SLA is the explicit learning involved in the initial registration of pattern recognizers for constructions that are then tuned and integrated into the system by implicit learning during subsequent input processing. Neural systems in the prefrontal cortex involved in working memory provide attentional selection, perceptual integration, and the unification of consciousness. Neural systems in the hippocampus then bind these disparate cortical representations into unitary episodic representations. These are the mechanisms by which Schmidt’s ‘noticing’ helps solve Quine’s problem of referential indeterminacy. Explicit memories can also guide the conscious building of novel linguistic utterances using processes of analogy. Formulae, slot-and-frame patterns, drills, and declarative pedagogical grammar rules all contribute to the conscious creation of utterances whose subsequent usage promotes implicit learning and proceduralization. Flawed output can prompt focused feedback by way of recasts that present learners with psycholinguistic data ready for explicit analysis. Other processes of acquisition from output include differentiation, analysis, and preemption. These processes of conscious construction in working memory underpin relationships between individual differences in working memory capacities and language learning aptitude.
Ellis, N. C. (in press, 2005). At the interface: Dynamic interactions of explicit and implicit language knowledge. Studies in Second Language Acquisition Ellis, N. C. (2002). Frequency effects in language acquisition: A review with implications for theories of implicit and explicit language acquisition. (Target article) Studies in Second Language Acquisition, 24, 143-188. Ellis, N. C. (2002). Reflections on frequency effects in language acquisition: A response to commentaries. Studies in Second Language Acquisition, 24, 297-339.
95 ESL failure to acquire third person singular -s (1) Conditioning depends upon contingency, not mere temporal pairing
96 Contingency, not temporal association Pavlovian Classical conditioning involves a cue (a to-be-conditioned stimulus, CS, for example, a bell) being temporally paired with an outcome (an unconditioned stimulus, US, for example, food), with after several such pairings the animal emitting a conditioned response (CR, salivation) on encountering the cue alone. temporal pairing of the CS and the US causes learning? However, Rescorla (1968) showed that if one removed the contingency between the CS and the US, preserving the temporal pairing between CS and US but adding additional trials where the US appeared on its own, then animals did not develop a conditioned response to the CS. This result was a milestone in the development of learning theory because it implied that it was contingency, not temporal pairing, that generated conditioned responding.
97 A contingency table showing the four possible combinations of events showing the presence or absence of a target Cue and an Outcome
Outcome No Outcome Cue a b No cue c d a, b, c, d represent frequencies, so, for example, a is the frequency of conjunctions of the cue and the outcome, and c is the number of times the outcome occurred without the cue. The one-way dependency statistic ΔP (Allan, 1980) ΔP = P(O/C) – P(O/-C) = a/(a+b) - c/(c+d) 98 ΔP
ΔP is the probability of the outcome given the cue (P(O/C) minus the probability of the outcome in the absence of the cue (P(O/-C)
30 years of associative learning research into human sensitivity to the contingency between cues and outcomes :
Shanks (1995) The Psychology of Associative Learning chapter 2 when given sufficient exposure to a relationship, people’s judgments match quite closely the contingency specified by normative ΔP theory
99 ΔP and language Normative ΔP theory applies in language acquisition too
it’s the contingency between the language cue and it’s functional interpretation that’s important (sic)
MacWhinney’s competition model and connectionist learning, And, for the case of 3rd person -s… MacWhinney, B. (2001). The competition model: The input, the context, and the brain. In P. Robinson (Ed.), Cognition and second language instruction (pp. 69-90). New York: Cambridge University Press. Goldschneider, J. M., & DeKeyser, R. (2001). Explaining the "natural order of L2 morpheme acquisition" in English: A meta-analysis of multiple determinants. Language Learning, 51, 1-50. 100 Cues vicious hits serious redress cats it’s eats hurts Thomas its bites Nick’s
Interpretations Noun Verb possNoun Adverb 3rd Verbperson possNoun presentVerb poss Article 3rd3rd person personVerb Adjective present3rd person presentpresent Noun Pronoun pluralNoun Pronoun3rd person poss pluralNoun Pronoun3rd person 3rd personposs plural poss 101 Cues vicious hits serious redress cats Homophony Plural -s it’s eats third person singular present –s possessive -shurt … s Thomas its bites Nick’s ΔP = P(O/C) – P(O/-C) Outcome No Outcome Cue a b No cue c d Interpretations= a/(a+b) - c/(c+d) Noun Verb possNoun Adverb 3rd Verbperson possNoun presentVerb poss Article 3rd3rd person personVerb Adjective present3rd personallomorphy [s, z , z ] presentpresent Noun Pronoun pluralNoun Pronoun3rd person poss pluralNoun Pronoun3rd person 3rd personposs plural poss 102 Range of other basic associative learning factors causing limitations of SLA: Interference Latent Inhibition Perceptual learning
103 Independent parallel evolution of role of attention in Psychology and SLA
.criticisms of the prescriptions of the original behaviorist CAH of transfer mark the development of the field of SLA as an independent discipline .the rejected spouse has been growing in maturity too, behaviorism matured into cognitivism, and modern accounts of learning theory have attention, perception, competition, and rationality at their very heart .On reflection, then, it might be said that SLA and the psychology of learning divorced too early, or if not, at least that there are better grounds than both parties might have once have believed for their reconciliation now.
104 •An understanding of associative learning theory illuminates both the rationality of fluency and the apparent irrationalities of fragile SLA and fossilization
•For these various reasons, implicit learning in SLA fails to produce an L1-like end-state.
•Ellis, N. C. (submitted 2004). Associative learning of first and second languages.
105
Rational Analysis and Optimal Processing Things that were likely yesterday tend too to be likely in the same context today: (1) Something which has been frequently required in the past is likely to be required now (2) Something which has been recently required is likely to be required now (3) Something which has been often required in this particular context is likely to be required now
Rational Psycholinguistics the last half century’s work in psycholinguistics similarly demonstrates that fluent language users are optimal processors of words
107 Word production and recognition FREQUENCY and the POWER LAW OF PRACTICE
Naming latency Learning Curve
Number of exposures Memory strength/ Proficiency/ Power law of Practice Skill/ Accuracy
Number of exposures
Error rate
Number of exposures
108 Word production and recognition RECENCY
The probability of a word occurring p. word in, say, speech to children (from the occurring CHILDES database), or the New York Times, or the electronic mail a time since last person receives), is predicted by its occurrence recency of occurrence --> there is a power (=log-log linear) Forgetting Curve function relating probability of a word occurring in the headline in the New York Times on day n to how Memory strength long it has been since the word previously occurred. Retention interval
109 CONTEXT A Rational Analysis of Collocations On top of the effects of time-since-last-occurrence and occurrence-frequency on the probability a word will occur in the environment, there there are effects of immediate context: HORSE-CA__?, PROFOUND-IGN____? ONCE-UPON_A_TI__? ALICE IN WO______? vs. CA__? IGN____? TI__? WO______?
110 a) intros b) b) 1:35 to 2:45 you talk c) c) 2:45 to 3:00 --take general questions from the audience, particularly the non-studentswho won't be back after the break d) d) 3:00 to 3:15 break e) e) 3:15 to 4:00question and answer period with the speaker with Sharon and me chiming ina bit too. f) All sessions will be held in 4-414. Brain -
Complementary Explicit and Implicit Memory Systems
112
.
. Author Method Unconscious Conscious . Kjaer et al (2001) Subliminal vs. supraliminal Activity peaks in … PLUS parietal visual words. (PET.) ventral visual word and prefrontal areas only. cortex.
Beck et al (2001) Change blindness vs. Ventral visual regions. … PLUS parietal Turatto et al (2002) change detection. and Right (fMRI and ERPs) prrefrontal. Vuilleumier et al Seen and unseen faces in Ventral visual regions … PLUS parietal (2001) parietal neglect and prefrontal. (fMRI and ERP)
Driver et al (2001) Extinguished vs. conscious Ventral visual regions. … PLUS parietal stimuli in unilateral neglect. and prefrontal areas (fMRI and ERP). (of the intact left hemisphere).
115
Baars 9 functions of Consciousness 1. Definition and Context-setting. By relating global input to its contexts, the system underlying consciousness acts to define the input and remove ambiguities. Conscious global messages can also evoke contexts, which then constrain later conscious experiences. 2. Adaptation and Learning. Conscious experience is needed to represent and adapt to novel and significant events. 3. Editing, Flagging, and Debugging. Unconscious processors can monitor any conscious content, edit it, and try to change it, if it is consciously "flagged" as an error. 4. Recruiting and Control Functions. Conscious goals can recruit subgoals and motor systems, to organize and carry out mental and physical actions.
117 Baars 9 functions of Consciousness
5. Prioritizing and Access-Control. Attentional mechanisms exercise conscious and unconscious control over what will become conscious. By relating some particular conscious content to deeper goals, we can raise its access priority, making it conscious more ofen and increasing the chances of successful adaptation to it. 6. Decision-making or Executive Function. When automatic systems cannot routinely resolve some choice-point, making it conscious helps recruit unconscious knowledge sources to help make the proper decision. In the case of indecision,we can make a goal conscious to allow widespread recruitmentof conscious and unconscious "votes" for or against it.
118 Baars 9 functions of Consciousness 7. Analogy-forming Function. Unconscious systems can search for a partial match between their contents and a globally displayed (conscious) message. This is especially important in representing new information, when no close models of the input are available. 8. Metacognitive and Self-monitoring Function. Through conscious imagery and inner speech we can reflect upon and control our own conscious and unconscious functioning. 9. Autoprogramming and Self-maintenance Function. The deeper layers of context can be considered as a "self- system", which works to maintain maximum stability in the face of changing inner and outer conditions. Conscious experience provides information for the self-system to use in its task of maintaining stability. By "replaying" desirable goals, it can recruit processors able to produce solutions and thereby reprogram the system itself.
119 Analogy from the visual domain
(a) (b) (c) (d) (e)
120 ’s possession Tom’s ball His dog Mary’s crayon Hers Its downfall (it’s a gas) Dad’s car My pen
Word order seems the more reliable cue And anyway, there’s no clear sense to be had of -s And, as native undergraduates attest, none at all for that_’!
121 Learning Period 1 Learning Period 2 Test Phase Mean Judgment A predicts X LearnedAC predict selective X attention:C 31 B noBlocking relation X BDof predict Associative X D Conditioning77 Blocking: a cue that is experienced in a compound along with a known strong predictor is blocked from being seen as predictive of the outcome. Chapman & Robbins (1990) - participants track & predict changes in a fictitious stock market. First period of Thethe learning prior phase learning •whenever stock A rose in price (cue A), the market rose as well (outcome X). •The rise or not of stock B during this period had no effect on the market. •Thus A wasof a good cue predictor ofA outcome ‘ blockedand B was not. ’ •Second period of the learning phase •on some trials stocks A and C rose together and the market increased, on other trials stocks Bthe and D rose acquisition together and the market again rose. of •The number of cases of AC cue combination and BD cue combination in period 2 were the same, so they were equally good predictors of market growth. •Final testing phase cue C. •learners were asked to rate on a scale from –100 to +100 (perfect predictor of market rising) how well each stock predicted a change in the market. 122 The IDIOM principle
"a language user has available to him or her a large number of semi-preconstructed phrases that constitute single choices, even though they might appear to be analysable into segments. To some extent this may reflect the recurrence of similar situations in human affairs; it may illustrate a natural tendency to economy of effort; or it may be motivated in part by the exigencies of real-time conversation. However it arises, it has been relegated to an inferior position in most current linguistics, because it does not fit the open-choice model" (Sinclair, 1991, p. 110)
123 Psycholinguistics
Construction frequency and processing,
but what about
Construction frequency and acquisition?
124