Effect_of_Tech_on_SLA

Views:
 
Category: Education
     
 

Presentation Description

CALL IS Academic Session at TESOL NYC, April 4, 2008

Comments

By: ElizabethHS (79 month(s) ago)

This presentation for the CALL IS Academic Session at TESOL NYC, April 4, 2008, reviews in general terms what is known about second language acquisition and discusses the implications of Landauer and Dumais' theory of Latent Semantic Analysis, a general learning mechanism that resolves Plato's problem of "excessive knowledge." LSA uses computer-based algebraic matrixes that in effect simulate the human brain's neural processing of inferred inductions. The presentation examines some of the reasons computers may be an optimal means to learn a language and explores some of the difficulties in CALL research in an open-ended media-rich environment.

Presentation Transcript

The Effect of Technology on Second Language Acquisition(and vice versa) : 

Elizabeth Hanson-Smith Professor emeritus, Calif State Univ, Sacramento, USA Computers for Education Command Performance Language Institute http://webpages.csus.edu/~hansonsm ehansonsmi@yahoo.com The Effect of Technology on Second Language Acquisition(and vice versa)

Topics- What are the problems facing CALL researchers?- What is SLA?- Are there any solutions? : 

Topics- What are the problems facing CALL researchers?- What is SLA?- Are there any solutions?

de Lisle's summary of Clarke’s (1994) “Media Will Never Influence Learning”:Instructional Methods differ from Instructional Media : 

Methods are fundamental ways of working to bring about a change in the learner's cognitive processes Media are the means by which the method is delivered to the learner Media attributes are the delivery methods which a particular medium offers (e.g., zooming in video) de Lisle's summary of Clarke’s (1994) “Media Will Never Influence Learning”:Instructional Methods differ from Instructional Media

"…there is no single media attribute that serves a unique cognitive effect for some learning task, …[so] the attributes must be proxies for someother variables that are instrumental in learninggains" (Clarke, 1994, p. 22). : 

"…there is no single media attribute that serves a unique cognitive effect for some learning task, …[so] the attributes must be proxies for someother variables that are instrumental in learninggains" (Clarke, 1994, p. 22).

Media are like the delivery truck…we are grateful to UPS, but thecontent of the box isn't theirs. : 

Media are like the delivery truck…we are grateful to UPS, but thecontent of the box isn't theirs.

De Lisle's summary of Clarke: : 

De Lisle's summary of Clarke: The learning effects demonstrated are due [not to the delivery method, but] to superior Instructional methods being built into Computer Based learning situations.

In brief, my claim is that media research is atriumph of enthusiasm over substantive examination of structural processes in learning and instruction. (Clarke, 1994, p. 27) : 

In brief, my claim is that media research is atriumph of enthusiasm over substantive examination of structural processes in learning and instruction. (Clarke, 1994, p. 27)

What is SLA? - Definition(s)Krashen's Input Model : 

What is SLA? - Definition(s)Krashen's Input Model Graph constructed at Gliffy.com

The above is admittedly a very simple version, based on Stevick's illustration (1980, p. 270). Krashen by 1981 has a much more elaborated version (see especially p. 101) that takes into account the impact of learning on acquisition, though clearly he never strayed from the idea that input is primary to the process of SLA. : 

The above is admittedly a very simple version, based on Stevick's illustration (1980, p. 270). Krashen by 1981 has a much more elaborated version (see especially p. 101) that takes into account the impact of learning on acquisition, though clearly he never strayed from the idea that input is primary to the process of SLA.

What is SLA? - Definition(s)An Interactionist Model : 

What is SLA? - Definition(s)An Interactionist Model Based on Chapelle, 1998, Fig. 2, p. 23 (see also, Swain, 1985). Graph constructed at Gliffy.com

What is SLA? - Definition(s)Stevick's Levertov Machine : 

What is SLA? - Definition(s)Stevick's Levertov Machine Graph constructed at Gliffy.com

The "Rheostat" in Stevick's model turns up attention and hence acquisition and output capabilities. Social forces are important: How learners feel about school and learning, and also the reactions of others during communication. What we acquire and what we learn interact with each other. Reactions to our output aid both acquisition and learning. (Stevick, 1980, pp. 270-279.) : 

The "Rheostat" in Stevick's model turns up attention and hence acquisition and output capabilities. Social forces are important: How learners feel about school and learning, and also the reactions of others during communication. What we acquire and what we learn interact with each other. Reactions to our output aid both acquisition and learning. (Stevick, 1980, pp. 270-279.)

What is SLA? - Plato's ProblemorHow do we we acquire so much knowledge on the basis of so little information? : 

What is SLA? - Plato's ProblemorHow do we we acquire so much knowledge on the basis of so little information?

Landauer & Dumais (2004): "A typical American seventh grader knows the meaning of 10-15 words today that she didn't know yesterday" (p. 2). Most of these words must have been acquired through reading because the majority of English words are used only in print, and she has already acquired the full complement of oral vocabulary; however, she has acquired less than one word through direct instruction since yesterday.About one word for every twenty paragraphs read in a school text goes from wrong to right on a daily vocabulary test. Yet the typical seventh grader would have read fewer than 50 paragraphs since yesterday. How did she acquire these words she didn't encounter? : 

Landauer & Dumais (2004): "A typical American seventh grader knows the meaning of 10-15 words today that she didn't know yesterday" (p. 2). Most of these words must have been acquired through reading because the majority of English words are used only in print, and she has already acquired the full complement of oral vocabulary; however, she has acquired less than one word through direct instruction since yesterday.About one word for every twenty paragraphs read in a school text goes from wrong to right on a daily vocabulary test. Yet the typical seventh grader would have read fewer than 50 paragraphs since yesterday. How did she acquire these words she didn't encounter?

Plato's solution for this "mystery of excessive knowledge" is that people are born with most of their knowledge and need only hints or contemplation to retrieve it. : 

Plato's solution for this "mystery of excessive knowledge" is that people are born with most of their knowledge and need only hints or contemplation to retrieve it.

What is SLA?Latent Semantic Analysis (LSA) : 

What is SLA?Latent Semantic Analysis (LSA) A general theory of "acquired similarity and knowledge representation" (Landauer & Dumais (2004): p. 1). Landauer and Dumais used the Grolier Encyclopedia, an electronic text for young students, to analyze 30,473 articles with a mean text length of 151 words. (Some articles were only one sentence long, e.g., "Constantinople; see Istanbul," and longer articles were cut off at 2,000 characters.) The words were placed into a matrix, each column representing an article, and each row one of 60,768 word-types that appeared in at least 2 articles. Each cell contained the frequency with which a word appeared in that article:

Table based on Landauer & Dumais, Figure 2, p. 13 : 

Table based on Landauer & Dumais, Figure 2, p. 13

Latent Semantic Analysis 2 : 

Latent Semantic Analysis 2 The computer is then able to determine logrithmically what words (i.e., strings of characters) appear in what contexts (i.e., with other words/strings), but even more importantly, what words might appear in similar contexts. When given the synonym portion of the TOEFL, the machine approximated the average scores of EFL applicants to U.S. colleges. The model got 64.4% correct, and the students got 64.5% (Landauer & Dumais, p. 14).

This is all without the computer understanding the words tested semantically, and without being able to use grammar/syntax cues."…A substantial portion of the information needed to answer common vocabulary test questions can be inferred from the contextual statistics of usage alone" (Landauer & Dumais, 2004, pp. 2-3) : 

This is all without the computer understanding the words tested semantically, and without being able to use grammar/syntax cues."…A substantial portion of the information needed to answer common vocabulary test questions can be inferred from the contextual statistics of usage alone" (Landauer & Dumais, 2004, pp. 2-3)

The machine acquired knowledge about synonymity "from the kinds of experience on which a human relies" (Landauer & Dumais, p. 15). That is, the hundreds of billions of neural networks in the brain can exploit both indirect inference and co-occurrence relations, both within and beyond a particular text. : 

The machine acquired knowledge about synonymity "from the kinds of experience on which a human relies" (Landauer & Dumais, p. 15). That is, the hundreds of billions of neural networks in the brain can exploit both indirect inference and co-occurrence relations, both within and beyond a particular text.

Several observations by Landauer & Dumais: : 

Several observations by Landauer & Dumais: ...weak local constraints [e.g., in the search for synonymity] can combine to produce strong inductive effects (p. 6) - ...the effects of constraints [which may be very weak in themselves] may emerge only in very large naturally generated ensembles. In other words, experiments with miniature or concocted subsets of language may not be sufficient to reveal or assess the forces that hold conceptual knowledge together. (p. 6)

(To learn a language, you need masses of input data as encountered naturally by people.) : 

(To learn a language, you need masses of input data as encountered naturally by people.) - Knowledge comes not just from an immediate stimulus or direct experience with something, e.g., encountering a word, but "with everything else ever experienced." (p. 11)

What is SLA? - LSA 4The LSA model accounts for other aspects of human knowledge: : 

What is SLA? - LSA 4The LSA model accounts for other aspects of human knowledge: First- through fifth-graders learn about 10 words a day, despite learning only about one word a week by direct instruction. (p. 17)

This is possible because: : 

This is possible because: A late grade school child will have read about 3.8 million words. The direct learning effect is calculated at .0007 words per word encountered X 70 (the approximate number of words in a paragraph). The indirect effect is .15 words per word encountered. An average student reads 50 paragraphs daily, amounting to 10 new words/day learned. (p. 22)

What is SLA? - LSA 5 : 

What is SLA? - LSA 5 Landauer & Dumais (2004): Expert readers "get more" out of what they read. There is "inductive power inherent in the possession of large bodies of old knowledge." (p. 31) By the end of secondary school, a knowledge of 100,000 words is probably a low estimate.

The expert reading her 2 millionth paragraph has a .56 probablility of correct extrapolation when encountering a new term in, say, an academic journal, while the novice encountering only his second sample of a similar context has a .14 probability of correct meaning: he is about 1/4 as able to read for meaning.There is "inductive power inherent in the possession of large bodies of old knowledge." (p. 31) : 

The expert reading her 2 millionth paragraph has a .56 probablility of correct extrapolation when encountering a new term in, say, an academic journal, while the novice encountering only his second sample of a similar context has a .14 probability of correct meaning: he is about 1/4 as able to read for meaning.There is "inductive power inherent in the possession of large bodies of old knowledge." (p. 31)

Is the medium the message? Back in 1994, Clarke told us no, it was not. If you drill and grill on the computer, it's no different than doing it in the classroom. You can click for animation or zoom for details, but these are delivery, not content or methods. - Only superior methods can produce better learning. Therefore, - CALL researchers are in fact investigating methods. : 

Is the medium the message? Back in 1994, Clarke told us no, it was not. If you drill and grill on the computer, it's no different than doing it in the classroom. You can click for animation or zoom for details, but these are delivery, not content or methods. - Only superior methods can produce better learning. Therefore, - CALL researchers are in fact investigating methods.

Does CALL provide us with superior methods of delivering instruction?Or superior quantities of natural language?Or something else? : 

Does CALL provide us with superior methods of delivering instruction?Or superior quantities of natural language?Or something else?

A 4-year-old, when asked what she was fishing for behind the TV, replied "the mouse." (Carr, 2008, C7)Another 4-year-old asked to see the movie she had just viewed on broadcast TV at the babysitters' house. When told it wasn't on TV just then, she asked, "Is it broken?" (Carr, 2008, C7) : 

A 4-year-old, when asked what she was fishing for behind the TV, replied "the mouse." (Carr, 2008, C7)Another 4-year-old asked to see the movie she had just viewed on broadcast TV at the babysitters' house. When told it wasn't on TV just then, she asked, "Is it broken?" (Carr, 2008, C7)

- You point, you click, you don't wait. - You don't have to allow others to select what you read or listen to. : 

- You point, you click, you don't wait. - You don't have to allow others to select what you read or listen to.

So possibly both method (in the largest sense) and input--and the methods of input--as well as motivational quality are all superior in a CALL environment. : 

So possibly both method (in the largest sense) and input--and the methods of input--as well as motivational quality are all superior in a CALL environment.

Putting together all these sources: * "Methods" may in fact be seen as less important than the delivery of input--and how input is delivered. * Masses of input are more important than direct instruction. (Are we back to Krashen??) * Authentic content is more important for indirect learning than prepared texts with narrow foci. : 

Putting together all these sources: * "Methods" may in fact be seen as less important than the delivery of input--and how input is delivered. * Masses of input are more important than direct instruction. (Are we back to Krashen??) * Authentic content is more important for indirect learning than prepared texts with narrow foci.

* Human readers can readily disambiguate terms through local context, using their hundreds of billions of parallel computational elements (Landauer & Dumais, p. 32); hence, concordancers should be important tools in language input. * Since humans are exposed to spoken language as well as print, the newer forms of oral Web interaction, such as VoIP should become increasingly important for communication beyond classmates and teacher. : 

* Human readers can readily disambiguate terms through local context, using their hundreds of billions of parallel computational elements (Landauer & Dumais, p. 32); hence, concordancers should be important tools in language input. * Since humans are exposed to spoken language as well as print, the newer forms of oral Web interaction, such as VoIP should become increasingly important for communication beyond classmates and teacher.

* Since, in language, learners can produce the same events that they perceive--and receive feedback on their approximations (remembering Stevick's Levertov machine)--opportunities for communicative production are very important to expanding the knowledge base and neural networks. * Some degree of autonomy in the selection of media and learning goals appears to be useful both for motivation and in making inferences about content and linguistic structures. : 

* Since, in language, learners can produce the same events that they perceive--and receive feedback on their approximations (remembering Stevick's Levertov machine)--opportunities for communicative production are very important to expanding the knowledge base and neural networks. * Some degree of autonomy in the selection of media and learning goals appears to be useful both for motivation and in making inferences about content and linguistic structures.

The media-rich, Internet-enhanced computer-based environment offers multiple opportunities to expand input and interaction far beyond schoolroom walls.Additionally, it offers the potential for learner autonomy and motivation. : 

The media-rich, Internet-enhanced computer-based environment offers multiple opportunities to expand input and interaction far beyond schoolroom walls.Additionally, it offers the potential for learner autonomy and motivation.

Taking into account these ideas, what can we conclude about the proper functions of CALL --and by extension what should be the scope of CALL research? : 

Taking into account these ideas, what can we conclude about the proper functions of CALL --and by extension what should be the scope of CALL research?

* Content-based learning is important--it provides the contextual clues that allow inferred induction to develop * Authentic, extensive materials are important--they provide the large masses of information necessary for inferential learning * Learners need opportunities to direct their own learning : 

* Content-based learning is important--it provides the contextual clues that allow inferred induction to develop * Authentic, extensive materials are important--they provide the large masses of information necessary for inferential learning * Learners need opportunities to direct their own learning

* Opportunities for creative output and interaction are important, e.g., in extensive project-based learning involving an authentic audience--they allow learners to experiment and refine hypotheses about what they are learning(see CALL Environments, Egbert & Hanson-Smith, eds., 2007, particularly Chapter 1) : 

* Opportunities for creative output and interaction are important, e.g., in extensive project-based learning involving an authentic audience--they allow learners to experiment and refine hypotheses about what they are learning(see CALL Environments, Egbert & Hanson-Smith, eds., 2007, particularly Chapter 1)

Of lesser importance in SLA and hence in researching CALL: * Learning of discrete items, e.g., parts of speech in decontextualized sentences * Tests judging rote learning, e.g., lists of vocabulary items : 

Of lesser importance in SLA and hence in researching CALL: * Learning of discrete items, e.g., parts of speech in decontextualized sentences * Tests judging rote learning, e.g., lists of vocabulary items

A significant problem is how to control the variables when dealing with autonomous learner choices, massive amounts of authentic materials (as in extensive reading and long-term projects), and multiple media resources for both receptive and expressive communications--and often all of these at the same time. : 

A significant problem is how to control the variables when dealing with autonomous learner choices, massive amounts of authentic materials (as in extensive reading and long-term projects), and multiple media resources for both receptive and expressive communications--and often all of these at the same time.

ReferencesBerry, M. W. (1992). Large scale singular value computations. International Journal of Supercomputer Applications, 6(1), 13-49.Carr, D. You want, you click (no waiting). 2008. New York Times, Business Day, 31 March, pp. C1 + C7.Chapelle, C. A. (1998). Multimedia CALL: Lessons to be learned from research on instructed SLA. Language Learning & Technology, 2(1), 22-34. Available at http://llt.msu.edu/vol2num1/article1/index.html.Clarke, R. E. (1994). Media will never influence learning. Educational Technology Research and Development, 42(2), 21-30. : 

ReferencesBerry, M. W. (1992). Large scale singular value computations. International Journal of Supercomputer Applications, 6(1), 13-49.Carr, D. You want, you click (no waiting). 2008. New York Times, Business Day, 31 March, pp. C1 + C7.Chapelle, C. A. (1998). Multimedia CALL: Lessons to be learned from research on instructed SLA. Language Learning & Technology, 2(1), 22-34. Available at http://llt.msu.edu/vol2num1/article1/index.html.Clarke, R. E. (1994). Media will never influence learning. Educational Technology Research and Development, 42(2), 21-30.

References, cont.de Lisle, P. (n.d.) Summary of Clarke (1994): "Media will never influence Learning" (M.Ed. Project). Available athttp://hagar.up.ac.za/catts/learner/peterdl/ClarkeKozma.htm.Egbert, J. & Hanson-Smith, E., eds. (2007). CALL Environments: Research, practice, and critical issues. Alexandria, VA: TESOL.Krashen, S. (1981). Second language acquisition and second language learning. New York: Pergamon Press.Landauer, T. K., & Dumais, S. T. (2004). A solution to Plato's problem: The Latent Semantic Analysis theory of acquisition, induction and representation of knowledge. Available at http://lsa.colorado.edu/papers/plato/plato.annote.html. : 

References, cont.de Lisle, P. (n.d.) Summary of Clarke (1994): "Media will never influence Learning" (M.Ed. Project). Available athttp://hagar.up.ac.za/catts/learner/peterdl/ClarkeKozma.htm.Egbert, J. & Hanson-Smith, E., eds. (2007). CALL Environments: Research, practice, and critical issues. Alexandria, VA: TESOL.Krashen, S. (1981). Second language acquisition and second language learning. New York: Pergamon Press.Landauer, T. K., & Dumais, S. T. (2004). A solution to Plato's problem: The Latent Semantic Analysis theory of acquisition, induction and representation of knowledge. Available at http://lsa.colorado.edu/papers/plato/plato.annote.html.

References, cont.Stevick, E. W. 1980. The Monitor Model and the Levertov Machine. In E. Stevick, ed., Teaching languages: A way and ways, pp. 267-282. Rowley, MA: Newbury House.Swain, M. (1985). Communicative competence: some roles of comprehensible input and comprehensible output in its development. In S. M. Gass & C. G. Madden, eds., Input in second language acquisition, pp. 235-253). Rowley, MA: Newbury House. : 

References, cont.Stevick, E. W. 1980. The Monitor Model and the Levertov Machine. In E. Stevick, ed., Teaching languages: A way and ways, pp. 267-282. Rowley, MA: Newbury House.Swain, M. (1985). Communicative competence: some roles of comprehensible input and comprehensible output in its development. In S. M. Gass & C. G. Madden, eds., Input in second language acquisition, pp. 235-253). Rowley, MA: Newbury House.

*Notes:A free software version of the program used to analyze large matrices by Landauer and Dumais is available from http://www.netlib.org/svdpack/index.html (see Berry 1992). University-affiliated researchers may obtain a research-only license and complete software to replicate Landauer & Dumais' work in LSA by contacting Susan Dumais at the Information Sciences Research Bellcore, Morristown, NJ 07960. : 

*Notes:A free software version of the program used to analyze large matrices by Landauer and Dumais is available from http://www.netlib.org/svdpack/index.html (see Berry 1992). University-affiliated researchers may obtain a research-only license and complete software to replicate Landauer & Dumais' work in LSA by contacting Susan Dumais at the Information Sciences Research Bellcore, Morristown, NJ 07960.

An outline of this paper may be viewed athttp://tesol-tech-sla.wikispaces.com/Links to my other papers and work in software and Webware design are found at my homepage:http://webpages.csus.edu/~hansonsm : 

An outline of this paper may be viewed athttp://tesol-tech-sla.wikispaces.com/Links to my other papers and work in software and Webware design are found at my homepage:http://webpages.csus.edu/~hansonsm

authorStream Live Help