29
Language Learning ISSN 0023-8333 Natural or Artificial: Is the Route of L2 Development Teachable? Xian Zhang a,b and James P. Lantolf c a Guangdong University of Foreign Studies, b Rice University, and c The Pennsylvania State University The current study was designed to assess the central claim of the Teachability Hypothesis (TH), a corollary of general Processability Theory (PT), which predicts instruction can- not alter posited universal, hierarchically organized psycholinguistic constraints behind PT’s developmental sequences. We employed an interventional design, which adhered to instructional procedures of Systemic Theoretical Instruction, and we taught four university learners at Stage 2 (subject-verb-object) Chinese topicalization for Stage 4 (object-first, e.g., Pizza t¯ a yˇ e ch¯ ı le, Pizza , ‘Pizza he also ate’). We be- lieve the findings show that, under the instructional conditions utilized in the study, the predictions of TH do not hold. We conclude it is possible to artificially construct a de- velopmental route different from the one predicted by natural developmental sequences, in agreement with the claims of Vygotsky’s developmental education. Keywords teachability hypothesis; processability theory; sociocultural theory; natural sequence; concept-based instruction Introduction Processability Theory (PT) was formulated by Pienemann (1998) and his asso- ciates (e.g., Pienemann, Di Biase, & Kawaguchi, 2005; Pienemann & Johnston, 1987) in order to explain the apparent fact that second language (L2) learners develop the cognitive ability to process certain features of L2s in accordance with universal hierarchically organized psycholinguistic constraints. Its basic Support for this project was provided to XZ by a Gil Watz dissertation fellow award from the Center for Language Acquisition at Penn State University, the National Social Science Fund of China (#12&ZD224) and the National Key Research Center for Linguistics and Applied Linguistics of Guangdong University of Foreign Studies. Correspondence concerning this article should be addressed to: James Lantolf, The Pennsylvania State University, Department of Applied Linguistics, 304 Sparks Building, University Park, PA 16802. E-mail: [email protected] Language Learning 65:1, March 2015, pp. 152–180 152 C 2015 Language Learning Research Club, University of Michigan DOI: 10.1111/lang.12094

Zhang Et Al-2015-Language Learning

Embed Size (px)

DESCRIPTION

Journal article

Citation preview

  • Language Learning ISSN 0023-8333

    Natural or Artificial: Is the Routeof L2 Development Teachable?

    Xian Zhanga,b and James P. LantolfcaGuangdong University of Foreign Studies, bRice University, and cThe Pennsylvania State

    University

    The current studywas designed to assess the central claim of the TeachabilityHypothesis(TH), a corollary of general Processability Theory (PT), which predicts instruction can-not alter posited universal, hierarchically organized psycholinguistic constraints behindPTs developmental sequences. We employed an interventional design, which adheredto instructional procedures of Systemic Theoretical Instruction, and we taught fouruniversity learners at Stage 2 (subject-verb-object) Chinese topicalization for Stage 4(object-first, e.g., Pizza ta ye ch le, Pizza , Pizza he also ate). We be-lieve the findings show that, under the instructional conditions utilized in the study, thepredictions of TH do not hold. We conclude it is possible to artificially construct a de-velopmental route different from the one predicted by natural developmental sequences,in agreement with the claims of Vygotskys developmental education.

    Keywords teachability hypothesis; processability theory; sociocultural theory; naturalsequence; concept-based instruction

    Introduction

    Processability Theory (PT) was formulated by Pienemann (1998) and his asso-ciates (e.g., Pienemann, Di Biase, &Kawaguchi, 2005; Pienemann & Johnston,1987) in order to explain the apparent fact that second language (L2) learnersdevelop the cognitive ability to process certain features of L2s in accordancewith universal hierarchically organized psycholinguistic constraints. Its basic

    Support for this project was provided to XZ by a Gil Watz dissertation fellow award from the

    Center for Language Acquisition at Penn State University, the National Social Science Fund of

    China (#12&ZD224) and theNational KeyResearch Center for Linguistics andApplied Linguistics

    of Guangdong University of Foreign Studies.

    Correspondence concerning this article should be addressed to: James Lantolf, The Pennsylvania

    State University, Department of Applied Linguistics, 304 Sparks Building, University Park, PA

    16802. E-mail: [email protected]

    Language Learning 65:1, March 2015, pp. 152180 152C 2015 Language Learning Research Club, University of MichiganDOI: 10.1111/lang.12094

  • Zhang and Lantolf Natural or Artificial

    logic is that L2 learners can produce and comprehend only those second lan-guage (L2) linguistic forms that the current state of the language processor canhandle (Pienemann, 2007, p. 137). Accordingly, the constraints limit learnercapacity to process linguistic information that is too far beyond their currentability. That is, learners at stage X in the processing hierarchy must proceedthrough stage X+1 before reaching stage X+2; that is, they cannot move fromX to X+2. Given that the constraints are assumed to be natural, they arepredicted to operate in all learning environments, including classrooms, andtherefore the Teachability Hypothesis (TH) corollary to the general theorypredicts that instruction cannot alter the stages postulated for the processinghierarchy. In this article we present the results of a study designed to assess thecentral claim of TH and which we believe shows that, under the instructionalconditions utilized in the study, the predictions of TH do not hold. Contrary tothe TH premised on Piaget (1950), and in agreement with the predictions ofVygotskys (e.g., 1978) developmental education, it is possible to artificiallyconstruct a developmental route different from the one predicted by naturaldevelopmental sequences.

    The article is organized as follows. We first briefly summarize general PTand examine TH as its corollary, followed by a review of four classroom studiesthat have challenged it in the past. Next, we review research carried out ontopicalization in L2Chinese, the instructional target of the research project to beconsidered here. Finally, we present the details and findings of the current study.

    The TH as a Corollary of PT

    The aspects of PT most relevant for the current study are associated withlexical mapping and the Topic Hypothesis (TOPH; discussed in a later section).Lexical mapping encompasses three independent levels of representation asproposed byBresnan (2001) for Lexical Functional Grammar (LFG). Argumentstructure describes who does what to whom and comprises argument roles suchas agent, experiencer, locative, and patient. Constituent structure consists ofuniversal components of sentences, such as verb phrase and noun phrase. Theorder within and between these components is language specific. For example,some languages favor prenominal while others prefer postnominal adjectiveposition (Pienemann, 2007). Functional structure includes the universal units,such as SUBJECT and OBJECT, that are related to the constituent structurein a language-specific way (Pienemann, 2007). Functional structure connectsargument structure and constituent structure.

    153 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    argument structure Agent theme locative

    functional structure SUBJ OBJ OBL

    NPsubj NPobj PP constituent structure

    John [threw] the ball into water

    Figure 1 Three levels of structure in LFG (Pienemann, 2007, p. 144, reprinted bypermission).

    The architecture of lexical mapping in LFG specifies the relationship be-tween argument structure and functional structure, which allows argument roles(argument structure) to be expressed by different grammatical forms (functionalstructure). For example, in Figure 1, each argument is mapped onto a gram-matical function: John as agent is mapped onto subject; Ball as theme ismapped onto object; Into Water as locative is mapped onto oblique. Based onlexical mapping theory, early PT research provided empirical support for thepredicted processing sequences that L2 learners of languages such as Englishand German should follow (details can be found in Pienemann, 1998).

    According to Pienemann (1987), PT and its corollary, TH, were inspiredby Piagets (1950) developmental theory:

    [T]he approach we have taken in the Predictive Framework of SLA and inthe Teachability Hypothesis was inspired by our admiration for JeanPiagets work on cognitive development. We adopted one concept inparticular from Piagets work, namely the implicational nature ofprocessing prerequisites for the operations possible at the different stagesof acquisition. (p. 92)

    For Piaget, and for Pienemann, instruction is subordinate to psychologicaldevelopment because stages of development determine what kind of knowledgean individual can understand and process at any give time. Pienemann (1989)argued that the acquisition process cannot be steered ormodeled just accordingto the requirements or precepts of formal instruction (p. 57). Consequently, ifinstruction is to promote L2 development it must take account of the processinghierarchy. If instruction aims at too high a stage (e.g., X+2 rather than X+1)a learner will not be able to cope with the complexities entailed in the higher

    Language Learning 65:1, March 2015, pp. 152180 154

  • Zhang and Lantolf Natural or Artificial

    stage. This is because the processing procedures that operate at each stage area prerequisite for the procedures that operate at the subsequent stage.

    As a corollary, TH makes two specific claims: (1) processing stages cannotbe skipped regardless of quality or quantity of instruction; (2) for instructionto be effective it must be aimed at the next immediate processing stage, X+1.However, it is important to understand that the theory does not claim that alearner must develop to the next stage even if instruction aims at this stage. AsPienemann (1987) makes clear, TH defines the possible range of influence ofexternal factors on the SL learning process; it does not imply that learning isguaranteed by the mechanisms internal to the learner (p. 92). What are calledvariable features (e.g., lexicon, prepositions, phrasal verbs in English) are notsubject to the same internal constraints assumed to operate in PT; therefore,instruction for these features need not be concerned with the learning barrier(Pienemann, 1989, p. 61) presented by processing constraints.

    Pienemann and his colleagues have conducted numerous observationalstudies designed to assess the validity of TH. The studies that provide supportfor TH include the following: Felix (1981), Pienemann (1984, 1989, 1998,2005), Boss (1996), Mansouri and Duffy (2005), Jansen (2008), Ellis (1989,2006), Y. Zhang (2001), Gao (2005), and Wang (2011). They have shown thatlearners must progress from one processing stage to the next; that learners can-not skip stages; that instruction is only effective if it aims at stage X+1 and notX+2 or higher; that the progress of learners can be plotted on an implicationalscale in a stepwise progression from least to most complex processing stage;and that if a learner can produce features at say Stage 4, he or she, implication-ally, can also produce features at the three preceding stages. The latter threestudies listed above are directly relevant for the current project and will bediscussed further below.

    At variance with Piaget (1950), Vygotsky (1986, p. 188) argued that theonly good instruction is that which leads rather than follows development. ForVygotsky, properly organized instruction does not wait for development to oc-cur. On the contrary, formal education is an intentionally organized activity thatsets in motion a variety of developmental processes not normally availablein the everyday world (Vygotsky, 1978, p. 90). Education not only influencescertain processes of development, but restructures all functions of behavior ina most essential manner (Vygotsky, 1997, p. 88).

    The TH allows us to compare the claims of Piaget (1950) and Vygotsky(1987) regarding the relationship between instruction and development, withspecific focus on the development of L2 processing ability.

    155 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    Challenges to the TH

    Four classroom studies have provided some evidence that challenges the centralclaims of TH. Two (Bonilla, 2012; Farley & McCollum, 2004) addressed theprediction that instruction can only be effective if it aims at the next immediatestage in the processing hierarchy (i.e., X+1). Two studies (Mackey, 1999; Spada& Lightbown, 1999) uncovered some evidence that stages can be skipped.

    The studies by Farley and McCollum (2004) and Bonilla (2012) used ex-plicit approaches to instruction on L2 Spanish. Farley and McCollam framedtheir research within Van Pattens (1996) Input Processing model, wherebylearners are provided with brief explicit information on a language featurefollowed by structured input activities designed to encourage learners to pro-cess for meaning texts containing exemplars of the relevant feature(s). Theexplanations and activities are sensitive to relevant processing strategies usedby learners, such as assign subject function to the first noun in a sentence orgive preference to lexical rather than grammatical items when processing forsemantic information.

    Farley and McCollum (2004) focused on a Stage 4 feature of Spanish,object-marker a, used to indicate animate direct object noun phrases (El gatomuerde al perro, the cat bites the dog/Al gato muerde el perro, the cat, thedog bites), and a Stage 5 feature, subjunctive mood in subordinate clauses (Elprofessor duda que los estudiantes salgan bien en el examen, the professordoubts that the students will do well on the exam). While some students intheir study were deemed ready for instruction (i.e., at X+1) with regard toobject-marker a, no student was judged ready for instruction with respect tosubjunctive. Statistical analysis of posttest scores showed no significant effectfor learner readiness; that is, following instruction some students unready forsubjunctive met the emergence criterion, and some students ready for object-marker a met the criterion but others failed to do so. It must be kept in mind,however, as we pointed out earlier, that TH does not require that anyone deemedready to progress must do so. The important finding of the study is that, whilenone of the students skipped stages, some of those that advanced to the nextprocessing level (i.e., X+1) did so on the basis of instruction that was beyondthe next level (X+2).

    Bonillas (2012) study focused on a number of different features of Spanishmorphology and syntax, including a Stage 3 feature, XP-adjunction (e.g., Endos semanas llega mi abuela, In two weeks arrives my grandmother, Mygrandmother is coming in two weeks); a Stage 4 feature, SV Inversion andclitic placement (e.g., El libro lo compro Roberto, the book it bought Robert,The book, Robert bought it); and a Stage 5 feature, use of subjunctive in

    Language Learning 65:1, March 2015, pp. 152180 156

  • Zhang and Lantolf Natural or Artificial

    subordinate clauses. Learners were given explicit explanations followed bygroup-work practice for each feature. Learner performance on post-instructionproduction tests confirmed PTs predictions that processing stages must emergein the predicted sequence. As with Farley and McCollumss (2004) study,however, Bonilla reported that instruction did not have to be targeted at thenext stage in the processing sequence in order to move development forward:aimed at next or at next + x stages was effective at increasing learnersproduction of the next stages as well as their current stage (p. 244).

    Spada and Lightbown (1999) used input flooding to teach English questionformation to grade six francophone children in Quebec. The input in the form ofquestion and answer (Q&A) activities implemented by the regular classroomteachers was aimed at Stages 4 (wh- + copula BE; yes/no questions withinversion) and 5 (wh- + aux second) in the processing hierarchy and wascarried out for 1 hour each day over a 2-week period. On the oral posttests,of the 79 students pretested at Stage 2 (subject-verb-object [SVO] order withquestion intonation), 23 progressed to Stage 3 (wh- fronting without inversion),54 remained at Stage 2, and 2 students skipped to Stage 4. Of the 39 studentsbeginning at Stage 3, 7 improved to Stage 4 and the remainder made no furtherprogress. The rest of the students, who began at higher stages, also showed noadditional progress. As with the previously discussed Spanish studies, severallearners progressed to the next processing stage even though input was two orthree stages higher than their initial stage of processing.Most relevant, however,is the fact that two students began at Stage 2 and skipped to Stage 4. Despitethis counterevidence, the researchers nevertheless concluded that the oralproduction findings are consistent with previous research documenting thatlearners progress through an acquisition sequence without skipping stages(p. 14). They also concluded that explicit teaching might be more effective thanimplicit instruction at ensuring learner progress (p. 14).

    Although Mackeys (1999) study on the effect of negotiated input on learn-ing English questions did not directly address TH, some of her data suggestthat stages might be skipped. In the study, learners who negotiated with nativespeakers of English in carrying out a series of tasks were compared to learn-ers who either only observed but did not participate in negotiated interactionsor who received prescripted input, which minimized the likelihood of nego-tiation. The learners who generated input through direct negotiation with thenative-speaking participants reached Stage 4 or 5 in the question-processing hi-erarchy, including several who were deemed developmentally unready prior toinstruction. This outcome lends further support to the findings of the previouslyreviewed studies that instruction need not be restricted to stage X+1. Mackey

    157 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    also considers the performance of an individual learner who was at Stage 3(wh- fronting) prior to treatment and at Stage 5 (do/aux-second) followingtreatment. The learner was exposed to several exemplars of Stage 5 questionsduring interaction. Mackey made no mention of the learner having producedStage 4 questions in any of the posttests or during treatment. It is conceivabletherefore that this learner provided additional evidence of stage skipping.

    To summarize the research considered in this section, the four studiesreviewed do not support the TH prediction that for instruction to be effective itmust be targeted at X+1. With regard to stage skipping, only one study, Spadaand Lightbown (1999), provides evidence that challenges this prediction, whileMackey (1999) presents data that is suggestive but not conclusive with regardto the prediction.

    The TOPH

    The TOPH, proposed by Pienemann et al. (2005), makes predictions derivedfrom the relationship between functional structures and constituent structures.The TOPH predicts that beginning learners of an L2 will not be able to differ-entiate between SUBJECT and other grammatical functions (such as TOPIC)in sentence-initial position (Pienemann, 2007; Pienemann et al., 2005). Thetopic (initial) position assumes the most prominent position in the grammaticalfunction hierarchy (Bresnan, 2001), which in canonical structures is usuallyoccupied by the grammatical SUBJECT (the agent in the argument structure).However, whenever speakers wish to profile different aspects of an event orscene other than the SUBJECT, this position may be filled by another el-ement such as OBJECT. The process through which this happens is calledtopicalization. When topicalization occurs, it triggers the differentiation of thegrammatical functions TOPIC and SUBJECT (Pienemann et al., 2005), whichresults in linguistic nonlinearity and is regarded as more costly in terms of pro-cessing effort, when compared to the canonical structure in which TOPIC andSUBJECT coincide. Thus, the production of a specific structure is constrainedby learners processing capacity (Pienemann et al., 2005). Based on this logic,PT makes the following prediction:

    In second language acquisition learners will initially not differentiatebetween SUB and TOP. The addition of an XP to a canonical string willtrigger a differentiation of TOP and SUBJ which first extends tonon-arguments and successively to arguments thus causing furtherstructural consequences. (Pienemann, 2005, p. 239)

    Language Learning 65:1, March 2015, pp. 152180 158

  • Zhang and Lantolf Natural or Artificial

    Stage 2: TOP = SUBJ: TOPsubj V(O) (SVO)

    e.g., Mary ate an apple. | | |

    TOP=SUBJ V OBJ

    Stage 3: TOP = ADJ(unct): TOPadj SV(O) (ADJ.+SVO)

    e.g., Yesterday Mary ate an apple. | | | |

    TOP=ADJ SUBJ V OBJ

    Stage 4: TOP = OBJ: TOPobj SV (OSV)

    e.g., Bob, I think, she will not forget. | | |

    TOP=OBJ SUBJ V

    Figure 2 Stages of topicalization in Chinese.

    TOPH predicts three stages in the mapping of functional structures ontoconstituent structures. Accordingly, L2 learners must pass through the hier-archical stages indicated in Figure 2 as they develop the ability to use andcomprehend topicalization in Chinese. We do not include Stage 1, lexical pro-cessing, because it is not relevant for our purposes.

    TOPH in L2 Chinese Research

    Three previous projects included a discussion of TOPH in L2 Chinese: Y.Zhang (2001, 2007), Gao (2005), and Wang (2011). Y. Zhang (2007) con-ducted a longitudinal study with three first language (L1) English speakers in abeginning-level Chinese course at an Australian university. Zhang engaged thelearners in a variety of tasks on a regular basis throughout the semester. Thetasks included elicited imitation, problem solving, role playing, picture-basedoral compositions, and storytelling. The main finding of the study was that theparticipants progressed from linear alignment of TOPIC and SUBJECT towardnonlinear alignment in accordance with TOPH. At the outset, the participantscould only produce Stage 2 SVO structures (see Figure 2). At a later point,Stage 3 ADJ+SVO structures emerged, and eventually the learners exhibitedthe ability to process Stage 4 OSV structures, thus confirming the sequencepredicted by PT for Chinese topicalization.

    159 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    Table 1 Instructional sequence in the textbook New Practical Chinese Reader used byY. Zhangs (2007) participants

    L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L13 L14

    3. OSV2. Adv-fronting

    + + + + + + + +

    1. SVO + + + + + + + + + + + + + +Note. L = lesson.

    Y. Zhang (2007) did not provide specific information on her instructionalprocedures. She did, however, mention the textbook used in the course, NewPractical Chinese Reader 1 (Book 1) (Xun, 2005). Table 1 summarizes thecontent of the book with regard to each of the relevant grammatical structurescovered in each lesson.

    Notice that the sequence of structures covered in the text coincides with twoof the three stages in the topicalization hierarchy: SVO and ADJ+SVO. It didnot include Stage 4, OSV; however, Y. Zhang provided instruction at this stagefollowing the two previous stages. Given that instruction appeared to followthe sequence indicated by the processing hierarchy, it is difficult to determinewhether or not Y. Zhangs study provided support for TH.

    Gao (2005) conducted a study to evaluate whether TOPH would also applyto learners of Chinese from different L1 backgrounds. In addition to L1 Englishspeakers, Gao included speakers of L1 Japanese (19) and L1 German (9) inher study. As with Y. Zhang (2007), Gao did not describe the instructionalprocedures implemented in her study. Nevertheless, she reported that all threegroups followed the predicted processing pattern: SVO ADJ+SVO OSV and that therefore the processing hierarchy for Chinese is unaffected bylearners L1.

    A third study conducted by Wang (2011) also evaluated TOPH in L2Chinese. She recruited eight learners of Chinese enrolled in a British uni-versity. The students reported heterogeneous experiences learning Chinese.Six had been to China prior to the study. Two had studied the language for1.5 months, while the remaining eight had studied the language for nearly ayear. Seven of the participants were L1 English speakers and one was an L1German speaker. Wangs general conclusion was that, despite the differencesin background, experience with the language, and the types of textbooks (noinstruction on stage X+1 structures), they all adhered to the sequence predictedby PT.

    Language Learning 65:1, March 2015, pp. 152180 160

  • Zhang and Lantolf Natural or Artificial

    The Present Study

    The current study was designed to answer the following two research questionregarding TH:

    1. Would instruction organized according to Vygotskian principles of devel-opmental education operationalized in Galperins (1970, 1992) SystemicTheoretical Instruction model induce learners to skip stages in the process-ing hierarchy postulated for Chinese L2 topicalization?

    2. If learners assessed to be at Stage 2 (SVO) received instruction targetingStage 4 (OSV), would they attain this stage of processing ability withoutfirst processing Stage 3 structures (ADJ+SVO), or would they move onlyto this stage (X+1) as predicted by PT?

    Method

    ParticipantsThe final participants were Leo, Alisa, John, and Amy (not their real names),all volunteers from a pool of L1 speakers of English enrolled in a universitybeginning-level Chinese course at a university in the United States.

    To identify qualified participants, X. Zhang (2014) (the first author) first ex-amined the textbooks and syllabi for the courses in which potential participantswere enrolled in order to confirm that Stage 3 and Stage 4 structures had notbeen taught. In addition, he conducted interviews with the Chinese instructorsin order to obtain information on the range of grammatical structures coveredat each level in the Chinese language program. It was determined that learn-ers at the beginning level would be suitable to take part in the study becausethe students as this level: (a) had studied some Chinese vocabulary, (b) hadencountered only the SVO structure in their classes, and (c) would not receiveinstruction on ADJ+SVO or OSV during the course of the study.

    Six students from the Chinese program volunteered for the study. On thepretest, one participant produced three ADJ+SVO sentences and an OSVsentence. She mentioned that she had encountered ADJ+SVO sentences inthe past and knew that time phrases can appear in initial position. Anotherparticipant produced an OSV sentence and she also mentioned that her Chineseteacher had briefly introduced the ADJ+SVO structure. These two participantsreceived instruction but are not included in the data considered here. Of theremaining four learners, all speakers of L1 English, none of them was ableto process structures at Stages 3 or 4 on the pretest. The students signed aconsent form prior to the start of the study and each received 60 dollars fortheir participation.

    161 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    Oral Production InstrumentsThree types of tasks were used in the study, all of which have been used inprevious research carried out on PT and TH. The tasks included an elicitedimitation task (EI), a question-and-answer session (Q&A), and an oral cartoondescription task (CD).

    Elicited ImitationSome may question EI as a valid instrument for assessing spontaneous perfor-mance. However, it has a long history in bilingual (e.g., Lambert & Tucker,1972; Radloff, 1991) and second language acquisition (e.g., Ellis, 2006; Erlam,2006; Flynn, 1987) research. It was also used in Y. Zhangs (2001, 2007) study.We believe it to be a legitimate instrument for tapping into a speakers linguisticcompetence.

    To minimize the possibility that participants might reproduce the targetedsentences from rote memory, the following procedures were followed whenconstructing the EI task. The sentences to be imitated were presented in sets ofthree. The total number of words in each set averaged 21, with a range of 16to 26 words per sentence. The targeted sentence was positioned in the middleof each three-sentence sequence, as recommended by Gallimore and Tharp(1981). Finally, before repeating the sentences, the participants were asked aquestion about content. An example of the procedure is given in (1):

    1. Sample EI procedure

    A. Researcher read the sentences in sequence

    ta zao sha`ng he le ka fei. .He morning drank coffee.

    Pizza ta ye ch le. [target structure]Pizza .Pizza he also ate.

    ta jn tian b jia`o gao x`ng .He today pretty happy.

    B. Researcher posed the question, Why do you think he is happy?C. Participant answers the question and then repeats the three sentences

    Language Learning 65:1, March 2015, pp. 152180 162

  • Zhang and Lantolf Natural or Artificial

    Three versions of the EI task were developed. The first was used in thepretest and the delayed posttest; the second was used in posttest 1, and the thirdin posttest 2. There were 72 sentences in the first version; 48 were SVO, 12were ADJ+SVO, and 12 were OSV. Both the second and the third versionscontained 60 sentences; 40 were SVO, 10 were ADJ+SVO, and 10 were OSV.

    As beginners, the participants had a limited vocabulary; therefore, the EItasks used a restricted set of lexical items. A set of words was selected from thefirst 11 chapters of the textbook used in the participants Chinese course. Thislist, together with their Pinyin and their English translation, was given to theparticipants to identify words that they recognized. Only words recognized bymost of the participants (no less than 70%) were used to construct the EI tests.

    The three versions of EIwere administered to five nativeChinese speakers toensure that the sentences were processable. None of the speakers had difficultieswith any of the sentences included in the EIs.

    Question&Answer SessionParticipants answered 5 to 10 questions in Chinese of the following type: Canyou tell me something about your country?, When and where did you eatlunch today?, What movie do you like most?, Can you describe it? Allquestions were read in English in order to avoid comprehension problems.

    Oral Cartoon D TaskFive 1-minute silent episodes from a Tom and Jerry cartoon were selected forthis task. Participants escription watched each episode twice before describingit in Chinese. If anyone failed to produce a sufficient number of utterances,the researcher asked questions or prompted the participant to say more about aparticular scene.

    Systemic Theoretical Instruction on Chinese Topicalization

    Teaching the target structures followed the procedures developed by Galperin(1970, 1992) known as Systemic Theoretical Instruction (STI). STI is a ped-agogical approach that integrates principles of developmental education asoutlined by Vygotsky (1987) in which instruction requires explicit presentationof systematic knowledge of concepts relevant to a particular academic sub-ject. In the present study, the key concept was Chinese topicalization. Whilein STI the concept is initially explained verbally (or in written form), learnersare also provided with a visual and/or material representation of the conceptas a way to avoid the tendency for learners to memorize verbal definitions

    163 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    without comprehension (Negueruela [2003] documents this tendency amongclassroom learners). Readers will find the visual materials employed in thepresent study to depict Chinese topicalization in the Supporting Informationonline. The visualization/materialization of a concept functions as a cognitivetool that mediates understanding, performance, and eventual internalization ofthe concept. In addition, learners are prompted to verbalize their understandingof the concept as well as their use of the specific features related to the conceptin concrete communicative activities. STI generally adheres to the followingsequence: systematic explanation of the concept visualizing/materializingthe concept in concrete communicative activities verbalizing the concept internalization (for details, see Lantolf & Poehner, 2014).

    The visualization/materialization phase of instruction is referred to as aSCOBA (Schema for the Orienting Basis of Action). The present study usedtwo different types of SCOBA. The first (shown in Figure S1 of the SupportingInformation online) depicts OBJECT topicalization and the second (shown inFigure S2) illustrates ADJUNCT topicalization. Both SCOBAs were animatedso that students were able to visualize the movement of the OBJECT rice inFigure S1a to sentence initial position in Figure S1b when that entity wasprofiled by a speaker. Similarly, the temporal adverb at 2 was moved fromFigure S2a to S2b (see Supporting Information online) to profile the temporalaspect of the event. A similar illustration was created for the locative adverbat home.

    After introducing the concept of topicalization and its manifestation inChinese, learners were asked to explain its function in English. All of the ini-tial explanations given by the learners appropriately described the pragmaticfunction of topicalization. The learners were then engaged in various practiceactivities, including sentence construction, gap filling, Q&A, translation, CD,and free talk. During this instructional phase learners were also provided withan activity designed to materialize the concept through manipulation of Cuise-naire rods (shown in Figure S3 in the Supporting Information online). The rods,originally developed for math instruction but introduced into language teach-ing by the Silent Way methodology (Gattegno, 1976), enabled the students tomaterially practice constructing topicalized sentences. The rods, which comein different colors and corresponding lengths, were used to represent differ-ent constituents of a Chinese sentence. The students rearranged the rods toindicate those constituents that are available for topicalization. In Figure S3,for example, the green rod represents the SUBJECT, the blue rod representsthe verb, and the yellow rod represents the OBJECT. Moving the yellow rod(representing OBJECT) to sentence-initial position showed that it had taken on

    Language Learning 65:1, March 2015, pp. 152180 164

  • Zhang and Lantolf Natural or Artificial

    topic status. While engaged in the various practice activities, the students wereable to manipulate the rods to mediate their language production. They werenot required to use the rods if they felt sufficiently confident in their ability toproduce correct sentences without material mediation.

    ProcedureAt Time 1 (T1), participants received the pretest, which consisted of an EI task,a Q&A task, and a CD task. (All subsequent posttests also comprised threesubcomponents but with different items in each case in order to avoid practiceeffects.) One to two days later, they received the first instruction session (lastingapproximately 1 hour) that exclusively taught Stage 4 OSV. One week later, atTime 2 (T2), the participants were given posttest 1 followed immediately byinstruction that focused on Stage 3 ADJ+SVO. The following week, at Time 3(T3), they were given posttest 2 immediately followed by the third instructionsession aimed at helping them practice the new grammatical structures. A littlemore than 1 month after posttest 2, at Time 4 (T4), the participants receivedthe delayed posttest. A final interview was then conducted to evaluate whetherthe participants understood the concept of topicalization and to discover theirattitude toward STI.

    Results

    The participants performance on each of the tests is presented in Table 2.We followed the criterion established by PT to determine processing stageattained by a learner. Accordingly, what matters is not accuracy of use in ahigh percentage of contexts, but first systematic use in obligatory contexts(Kessler & Pienemann, 2011, p. 94, italics in the original). The criterion foremergence requires use in at least four different contexts in order to avoid thesituation where learners use the relevant feature frequently but in the samecontext or where they might produce formulaic tokens without generalization.

    As observed in Table 2, the pretest (T1) showed that all four learners were atStage 2 prior to the start of instruction. On the first posttest (T2) administered1 week after instruction on Stage 4 OSV the learners met the criterion forprocessing ability at this stage. Only one participant, Amy, produced an instanceof Stage 3ADJ+SVO, and thiswas on theEI task, as explained below. Followinginstruction on ADJ+SVO, performance on the second posttest (T3) evidencedthe learners ability to use Stage 3 and Stage 4 structures. However, as we willexplain in the more detailed discussion of test performance below, there wassome variation in their ability to use both structures during the Q&A and CD

    165 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    Table 2 Summary of performance on pre- and posttests for all elicitation tasks

    Topic Stage T1 T2 T3 T4

    Leo 4 top = Obj. OSV 25/92 18/113 20/993 top = ADJ ADJ+SVO 17/113 12/992 top = Subj. SVO 72/72 67/92 78/113 67/99

    Alisa 4 top = Obj. OSV 17/69 14/91 23/1103 top = ADJ ADJ+SVO 20/91 19/1102 top = Subj. SVO 72/72 52/69 57/91 68/110

    John 4 top = Obj. OSV 19/91 12/95 23/1083 top = ADJ ADJ+SVO 14/95 24/1082 top = Subj. SVO 77/77 72/91 69/95 61/108

    Amy 4 top = Obj. OSV 25/86 17/113 26/1213 top = ADJ ADJ+SVO 1/86 19/113 16/1212 top = Subj. SVO 48/48 59/86 77/113 79/121

    Note. Denominator = possible contexts; Numerator = frequency of supplianceT1 = pretest; T2 = posttest 1; T3 = posttest 2; T4 = delayed posttest; = The relevant structure was missing despite the presence of appropriate contexts.

    tasks. This is also the case for the learners performance on the delayed posttestadministered 1 month after the final instruction session.

    In what follows, we consider learner performance on each of the tests andprovide examples of this performance.

    PretestOn the pretest (T1), administered 1 to 2 days prior to the first instruction session,all participants produced Stage 2 SVO sentences only. As illustrated in Excerpt1 below, John was unable to reproduce the target sentence (the EI procedure isdescribed above) in which the temporal adverb yesterday appeared in topicposition. In fact, not only did he omit the adverb, he also changed the subjectpronoun as well.

    Excerpt 1. EI on Pretest (Stage 3 ADJ+SVO)

    Target sentencezao sha`ng wo men da qiu. .Morning we play ball.In the morning we play ball.

    Language Learning 65:1, March 2015, pp. 152180 166

  • Zhang and Lantolf Natural or Artificial

    Johns imitation:

    nmen sh` . . . da qiu. . . . .You are . . . play ball.You play ball.

    The learners were also unsuccessful in their attempts to reproduce Stage4 OSV sentences. They frequently interpreted the preposed OBJECT as SUB-JECT, as Excerpt 2, also from Johns pretest, illustrates. Not only did he misin-terpret the argument status of Mr. Li, he inserted an OBJECT argument bookfrom one of the other sentences in the set.

    Excerpt 2. EI Pretest (Stage 4 OSV)

    Target sentencel lao sh wo x huan. .Mr. Li I like.I like Mr. Li.

    Johns imitation

    l lao sh xhuan y ben shu. .Mr. Li like a book.Mr. Li likes a book.

    We forgo examples of learner performance on the Q&A and CD tasksbecause they only produced SVO despite the presence of appropriate contextsfor Stages 3 and 4 structures. As predicted on the basis of the instructionalsyllabus and the teacher interviews, the pretest showed that the learners wereat Stage 2 in the processing hierarchy.

    Posttest 1: Following OSV InstructionOn posttest 1, administered at T2, 1 week after the first instruction session,learners were capable of producing OSV sentences but were unable to produceStage 3 ADJ+SVO structures (see Table 2). On the EI task, Leo reproduced4, Alisa 7, John 5, and Amy reproduced all 10 sentences. Amy managed tocorrectly reproduce 1 out of 10 ADJ+SVO sentences; however, this fails tomeet the PT criterion of emergence. In the interest of space we do not provideexamples of the learners EI performance on the posttests.

    167 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    On the Q&A task, a total of 9 contexts for OSV use were established by theresearchers questions. In response, the learners performance was as follows:Leo 6/9, Alisa 5/9, John 4/9, and Amy 6/9. Excerpt 3 provides an example ofan OSV utterance on the Q&A task produced by John.

    Excerpt 3. Posttest 1, Q&A (Stage 4 OSV)

    Q: Do you like Chinese food or American food?

    John: zhongguofa`n he meiguofa`n wo dou xhuan.

    .Chinese food and American food, I both like.I like Chinese food and American food.

    With regard to the CD, it was difficult to determine what counted as anappropriate context, given that topicalization is an optional feature of languageuse. In the EI, of course, context was determined by the model sentences, andon the Q&A, the researcher, based on the questions asked, had some ability togenerate contexts where topicalization was favored. In the CD, however, thelearners were free to topicalize or not, as determined by their communicativeintent. Consequently, it was difficult to determine the precise number of contextswhere topicalization was expected to occur. In reporting learner performanceon the CD, therefore, we provide the frequency of topicalized structures only.In each case, the utterances were not formulaic, nor did they repeat previousutterances. In the CD task, Leo produced 15 OSV sentences, Alisa 5, John 10,and Amy 9. Excerpt 4, taken from Amys and Leos respective performances,illustrates their ability to use OSV.

    Excerpt 4. Posttest 1, CD (Stage 4 OSV)

    Amy: lao shu mao da le yc`. .Mouse cat hit once.The cat hit the mouse once.

    Leo: niu nai Tom he le (hen)duo. Tom ().Milk Tom drank a lot.Tom drank a lot of milk.

    None of the participants produced the ADJ+SVO structure on either theQ&A or CD task on posttest 1.

    Language Learning 65:1, March 2015, pp. 152180 168

  • Zhang and Lantolf Natural or Artificial

    Posttest 2: Following ADJ+SVO InstructionOn posttest 2, administered at T3, 1 week after the second instruction session,which focused on Stage 3 ADJ+SVO, all of the participants were able toproduce both OSV and ADJ+SVO sentences (see Table 2). Of the 10 OSV and10 ADJ+SVO sentences on the EI task, Leo reproduced 6 of the former and9 of the latter; Alisa reproduced 7 OSV and all 10 ADJ+SVO sentences; Johnreproduced 6 of the 10 OSV and 9 of the 10 ADJ+SVO sentences; and Amysuccessfully repeated 5 of the OSV and 9 of the ADJ+SVO sentences.

    On the Q&A task, 13 possible contexts were available for each target struc-ture. Leo produced seven OSV and four ADJ+SVO utterances. Alisa producedthree OSV and six ADJ+SVO utterances. Alisa also produced four OSV sen-tences on the CD task (see below), which means she met PT criterion forboth stages. John produced four OSV utterances and one ADJ+SVO utterance;however, on the CD he produced four additional ADJ+SVO utterances, whichagain indicates that he met PT criterion for Stages 3 and 4. Amy produced sevenpossible OSV utterances and four ADJ+SVO utterances. Excerpt 5 illustrateslearner performance on the Q&A task from posttest 2.

    Excerpt 5. Posttest 2 (Q&A)

    ADJ+SVOResearcher: Where do you study?

    Amy: za`i xue xia`o wo mei tian xue x? ?At school I every day study.I study at school every day.

    OSVResearcher: What do you buy when you go shopping?

    Amy: dong x wo za`i shu dia`n mai.

    Stuff I in the bookstore bought.I bought some stuff in the bookstore.

    Researcher: Where do you study Chinese?

    Alisa: za`ixuexia`o wo fu`x le zhongwen. .At school I went over Chinese.I went over Chinese at school.

    169 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    za`i tushuguan wo ye xuex le zhongwen. .In the library I also studied ChineseI study Chinese in the library as well.

    Researcher: What else did you study in the library?

    Alisa: faguowen wo za`i tushuguan ye fu`x le . .French I in the library also reviewed.I also reviewed French in the Library.

    On the CD task, Leo produced five OSV sentences and four ADJ+SVOsentences. Alisa produced four OSV and two ADJ+SVO sentences. John pro-duced two OSV and four ADJ+SVO sentences. Finally Amy produced fiveOSV and six ADJ+SVO sentences. These results suggest that 1 week afterthe second instruction session the learners were capable of processing Stage 4OSV and Stage 3 ADJ+SVO structures. Examples from the CD task are givenin Excerpt 6.

    Excerpt 6. Posttest 2 (CD)

    Amy: za`i jia lao shu ka`n jia`n le niu nai. .At home mouse saw milk.The mouse saw some milk at home.

    niu nai mao jn tian za`i zhe` he le. .Milk cat today here drank.The cat drank some milk here.

    The Delayed PosttestOn the delayed posttest, administered at T4, approximately 1 month afterposttest 2 (the exact time depended on participant availability), the learnerscontinued to show the ability to produce ADJ+SVO and OSV structures. TheEI task included 12 OSV and 12 ADJ+SVO sentences. Leo accurately re-produced 11 of the former and 7 of the latter; Alisa reproduced 10 of 12for each structure; John reproduced 10 of the former and 9 of the latter; andAmy reproduced 8 of 12 in each case. On the Q&A task, Leo reproduced 5,Alisa 7, John 10, and Amy 9 of 10 possible OSV utterances in their respectiveresponses. Of the possible 10 ADJ+SVO responses on the same task, Leo

    Language Learning 65:1, March 2015, pp. 152180 170

  • Zhang and Lantolf Natural or Artificial

    produced 2, Alisa 6, John 10, and Amy 8 of the target structures. Finally, on theCD task, Leo produced four OSV sentences and three ADJ+SVO sentences;Alisa produced six OSV and three ADJ+SVO sentences; John produced twoOSV and two ADJ+SVO sentences. Amy produced nine OSV but did notproduce any ADJ+SVO sentences. Participant performance on the delayedposttest is illustrated in Excerpts 7 and 8.

    Excerpt 7. Delayed posttest (Q&A)

    OSVResearcher: Do you go to a bar?

    Alisa: jiu ba wo bu qu`. .Bar I not go.I dont go to bar.

    ADJ+SVOResearcher: When do you usually have dinner?

    John: ba dian he wu dian he jiu dian, wo chfa`n. , .8 oclock and 5 oclock and 9 oclock I have dinner.I have dinner at eight oclock, five oclock and nine oclock.

    Excerpt 8. Delayed posttest (CD task)

    ADJ+SVOLeo: za`i cao sha`ng ta men shu`jia`o.

    .On the grass they sleep.They slept on the grass.

    OSV

    Leo: gou de shou mao za`i gou jia xia` wu yao le. .Dogs hand cat in dogs house afternoon bit.The cat bit the dogs hand in the doghouse in the afternoon.

    171 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    Table 3 Chinese topicalization in current study

    TOPIC Hypothesis T1 T2 T3 T4

    4 top = obj + + +3 top = adj. + +2 top = subj. + + + +Note. T1 = pretest; T2 = posttest 1; T3 = posttest 2; T4 = delayed posttest.

    Discussion

    According to PT, L2 Chinese learners should not be able to process Stage 4OSV structures unless they are also able to process Stage 3 ADJ+SVO struc-tures. Because the sequence of stages is determined by cognitive processingconstraints, noncognitive factors such as teaching are assumed not to be ableto interfere with natural, internally determined processing stages. The currentstudy employed an interventional design which adhered to principles of de-velopmental education proposed by Vygotsky (e.g., 1978) and instructionalprocedures of STI formulated by Galperin (1970, 1992). It artificially con-structed a developmental route different from the one predicted by TOPH. Theconcept of topicalization and how it is specifically manifested in Chinese dis-course was taught using cognitive tools (i.e., SCOBAs), which visualized theconcept and provided learners with an effective and accessible understandingof the concept (see Supporting Information online). The learners ability toproduce Stage 4 and Stage 3 structures in that order was supported by materi-alized mediation in the form of Cuisenaire rods (see Figure S3). Consequently,the cognitive processes involved in learner development not only occurredinside of the head, but it was at the same time an embodied activity. Throughpractice with the rods, which we argue, had cognitive status (see Lantolf &Thorne, 2006), the learners not only came to understand how topicalizationoperates in Chinese, but they also appropriated, or internalized, the concept.Evidence for this comes from the delayed posttest. We believe that use of therods resulted in a strong trace in the learners long-term memory system. Inthe final interview, for example, one learner commented that, even though sheno longer needed overt support from the rods, she visualized them to helpher produce appropriate topicalized utterances. The rods also helped learnerscompensate for deficits in working memory (see X. Zhang, 2014), a key factorin successful learning (Williams, 2012).

    Tables 3 and 4 compare the general developmental pattern for Chi-nese topicalization uncovered in the current study and in Y. Zhang (2007).

    Language Learning 65:1, March 2015, pp. 152180 172

  • Zhang and Lantolf Natural or Artificial

    While the latter presents evidence that adheres to the predicted processingsequence, the former does not.

    Table 3 shows the general developmental pattern for Chinese topicalizationuncovered in the current study. It contradicts the findings reported by Y. Zhang(2007, p. 164). Her study turned up evidence that adheres to the predictedstepwise processing sequence. Clearly, ours does not.

    The Issue of ClassifiersAlthough the students had not received prior instruction on Stages 3 or 4in the topicalization hierarchy, they had been instructed on the use of someclassifiers, required in Chinese when nouns are counted or specified in someway (e.g., liang ge ren two persons.) According to Y. Zhang(2007), classifiers are Stage 3 structures. Lenzing (personal communication,March 24, 2014) has suggested that if our participants were able to produceclassifiers they could be at Stage 3 rather than Stage 2 prior to instruction. As itturned out, the participants were able to correctly use some classifiers during thepretest. For instance, one participant produced the following utterance duringthe pretest interview: wo you yi ge didi, I have a brother, with ge as theclassifier modifying didi (brother).

    Lenzings proposal is interesting and aligns with the argument made byPienemann et al. (2005) that for any given stage morphology is likely to emergebefore syntax and it may bootstrap syntax. However, two studies present evi-dence that calls into question the morphology-first argument. In a study of twolearners of English as an L2, Dyson (2009) uncovered some evidence to supportthe PT position; she also found a robust amount of counterevidence where infact syntactic features of a given stage emerged before morphological featuresat the same stage. She also reported examples where syntax and morphologyemerged simultaneously. Bonilla (2012) reported a similar circumstance forSpanish syntax and morphology. Assuming that classifiers are indeed Stage 3features, on the pretest in the present study the learners were unable to produceADJ+SVO sentences even though they were able to correctly use some clas-sifiers. Moreover, on posttest 1, they showed clear emergence of Stage 4 OSV,continued to produce classifiers, but still showed no evidence of ADJ+SVOemergence. In our view, the fact that the learners were able to process classifierson the pretest does not make a strong case in support of morphology bootstrap-ping syntax within the boundaries of the present study, because it would meanthat for some unexplainable reason the classifiers would have led the way forStage 4 but not Stage 3 syntax.

    173 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    The Issue of Memory StoreOne of the guest editors of this Special Issue (Ellis, personal communication,May 8, 2014) pointed out in his comments to an earlier version of this article thatif our instruction indeed fostered explicit knowledge of Chinese grammarknowledge stored in long-term declarative memory (see Paradis, 2009; Ullman,2005)the findings of the present study do not provide counterevidence to TH.This is because the hypothesis, as a corollary to PT, is assumed to hold only forimplicit knowledgeknowledge stored in long-term procedural memory. Ourresponse is twofold. First, given that we used the same spontaneous elicitationprocedures and emergence criteria utilized by Pienemann and his colleagues intheir research on PT and TH, and if indeed, as we believe, the knowledge thelearners internalized and used on the posttests was stored in declarativememoryas a result of explicit instruction, the elicitation procedures developed by Piene-mann and his colleagues are not sufficiently sensitive to distinguish between ac-cess to automatic proceduralized memory and accelerated declarative memory.Paradis (2009) argued that explicit declarative knowledge can, through appro-priate practice, be accessed with sufficient speed to support fluent spontaneousproduction. Ullman and his colleagues (Morgan-Short, Finger, Grey, &Ullman,2012) made a similar claim based on a comparative study of learners receivingexplicit and implicit instruction: similar proficiency levels, even at high levelsof proficiency, can be attained using quite different brain mechanisms and typesof processing (p. 14). If this is indeed correct, it would mean that behavioralcriteria assumed to hold between timed/unplanned and untimed/planned tasksmay not be sufficiently sensitive to distinguish between accelerated access todeclarative memory or automatic access to procedural memory. It might bethat the only way to confidently distinguish performances based on either ofthe memory stores must be neurological (e.g., derived from brain imagingtechniques such as event-related potentials [ERPs]).

    This leads us to our second response. Paradis (2009) and Ullman (2005)both argue that declarative knowledge cannot directly convert to proceduralknowledge regardless of amount of practice. This is because of the lack ofneural pathways connecting the neural substrates responsible for declarativememory (e.g., the lateral temporal lobe and the hippocampus) and the neuralnetworks that support the procedural memory system (e.g., the frontal lobe andthe basal ganglia). Both researchers also argue, however, that through sufficientpractice, which for Paradis (2009) requires extensive and intensive immersionexperiences, use of declarative knowledge can indirectly provide input forthe procedural memory store. We do not believe that the kind of experienceenvisioned by Paradis was available to the learners who participated in the

    Language Learning 65:1, March 2015, pp. 152180 174

  • Zhang and Lantolf Natural or Artificial

    present study. Had this been the case, the findings would not only challengeTH, they would also challenge general PT.

    In their ERP study of adult L2 classroom learners, Morgan-Short et al.(2012) report that learners receiving either explicit or implicit instruction in aninvented language, Brocanto2, reached a high level of proficiency, which wasmaintained following a 5-month time period without additional exposure to thelanguage. Several interesting and relevant findingswith respect to ERP emergedfrom the study. Immediately after instruction, the implicit group produced brainpatterns typical of native speakers of a language for procedural memory, whilethe explicit group emitted patterns normally supported by declarative memory.Following the 5-month time lag, however, the explicit group showed an increasein nativelike brain procedural memory patterns, although not to the same extentas the implicit group.At the same time, this group generated patterns consistentwith forgetting the underlying knowledge in declarative memory (p. 12). If it isthe case that declarative memory cannot directly convert to procedural memory,howmight this shift in ERPbe explained?One possibility suggested byMorgan-Short et al. (2012) is that, even during instruction, the explicit group might haveacquired some procedural grammatical knowledge that was obscured byreliance on declarative memory immediately after instruction (p. 12). In thetime lag between instruction and the second posttest, knowledge in the declar-ative memory system of the explicit group could have been forgotten while atthe same time knowledge in the procedural store could have consolidated, thusaccounting for the increase in nativelike ERP patterns in this group (p. 12).

    Is it possible that something similar occurred in participants in the currentstudy? This is a difficult question to answer at this point. For one thing, thetime lag between the posttest and delayed posttest was 1 month and not 5,which may not have been sufficient time for knowledge to consolidate in pro-cedural memory. For another thing, we did not measure ERP patterns and, asMorgan-Short et al. (2012, p. 13) pointed out, behavioral patterns do not nec-essarily reflect one or the other underlying neurological system. If indeed thelearners had been able to consolidate any knowledge entering their proceduralmemory store during the 1-month span between instruction and the delayedposttest, their performance on the delayed posttest would present a challengenot only to TH but to PT as well.

    Conclusion

    The data considered in this article provide evidence that stages in the processinghierarchy for topicalization in Chinese can be directly taught without regard

    175 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    for the processing sequence predicted by general PT. Thus, the predictionsof the TH corollary may not hold when instruction is organized according toVygotskys (e.g., 1978) principles of developmental education. These princi-ples include overt explanation of the relevant concept with focus on its meaning,visualization of the concept in the form of a nonverbal SCOBA, materializationof the concept (Cuisenaire rods) that enabled participants to manually manip-ulate sentences illustrating the concept, and practice activities accompanied byparticipant verbalization of their understanding of the concept.

    While the evidence presented in this article presents a challenge to TH andto general PT, we caution that it is one study of one feature assumed to entailprocessing stages. Additional research must be carried out on other featuresassociated with the general theory, including, for example, English questionformation and negation and German word order.

    An important issue raised in the study is the nature of learner knowledge thatresulted from developmental instruction. The tasks used to assess processingability, especially the Q&A session and the CD, have generally been assumedto entail spontaneous performance typical of implicit knowledge subserved byprocedural memory. We suggested, however, that it is conceivable that par-ticipant performance on the tasks might very well reflect explicit knowledgesubserved by declarative memory and accessed, as Paradis (2009) proposed,with sufficient speed to allow spontaneous performance to occur. The implica-tions of all of this for PT and for elicitation tasks used in general SLA researchremain to be worked out.

    Final revised version accepted 18 September 2014

    Postscript

    Space does not permit us to respond here in detail to Pienemanns (2015) re-action to our study. We will provide a more detailed response in a future issueof the journal. Nevertheless, we would like to note at this point that we do notagree with Pienemanns claim that elicited imitation tasks do not provide accessto underlying linguistic knowledge and that they only generate formulaic lan-guage. They have a long history in second language acquisition and assessmentresearch as a valid means of tapping into participant knowledge of a language,provided they are structured in an appropriate way that avoids rote memoriza-tion. We would also like to point out that in addition to the aggregated datapresented in Table 2, we indeed provide data on the performance of each learnerfor the Q&A and CD tasks from the pre- and posttests included in the study.

    Language Learning 65:1, March 2015, pp. 152180 176

  • Zhang and Lantolf Natural or Artificial

    References

    Bonilla, C. (2012). Testing processability theory in L2 Spanish: Can readiness ormarkedness predict development? Unpublished doctoral dissertation, University ofPittsburgh.

    Boss, B. (1996). German grammar for beginners: The teachability hypothesis and itsrelevance to the classroom. In C. Arbones Sola`, J. Rolin-Ianziti, & R. Sussex (Eds.),Whos afraid of teaching grammar (pp. 93103). Queensland, Australia: Universityof Queensland Press.

    Bresnan, J. (2001). Lexical-functional syntax. Malden, MA: Blackwell.Dyson, B. (2009). Processability theory and the role of morphology in English as a

    second language development: A longitudinal study. Second Language Research,25, 355376.

    Ellis, R. (1989). Are classroom and naturalistic acquisition the same? A study of theclassroom acquisition of German word order rules. Studies in Second LanguageAcquisition, 11, 308328.

    Ellis, R. (2006). Modelling learning difficulty and second language proficiency: Thedifferential contributions of implicit and explicit knowledge. Applied Linguistics,27, 431463.

    Erlam, R. (2006). Elicited imitation as a measure of L2 implicit knowledge: Anempirical validation study. Applied Linguistics, 27, 464491.

    Farley, A., & McCollam, K. (2004). Learner readiness and L2 production in Spanish:Processability theory on trial. Estudios de Lingustica Aplicada, 40, 4769.

    Felix, S. W. (1981). The effect of formal instruction on second language acquisition.Language Learning, 31, 87112.

    Flynn, S. (1987). A parameter-setting model of L2 acquisition: Experimental studies inanaphora. Dortrecht, Netherlands: Reidel.

    Gallimore, R., & Tharp, R. (1981). The interpretation of elicited imitation in astandardized context. Language Learning, 31, 369392.

    Galperin, P. I. (1970). An experimental study in the formation of mental actions. In E.Stones (Ed.), Reading in educational psychology: Learning and teaching (pp.142154). New York: Barnes and Noble.

    Galperin, P. I. (1992). Stage-by-stage formation as a method of psychologicalinvestigation. Journal of Russian and Eastern European Psychology, 30, 6080.

    Gao, X. D. (2005). Noun phrase morphemes and topic development in L2 MandarinChinese: A processability perspective. Unpublished doctoral dissertation,University of Wellington, Victoria, New Zealand.

    Gattegno, C. (1976). The common sense of teaching foreign languages. New York:Educational Solutions.

    Jansen, L. (2008). Acquisition of German word order in tutored learners: Across-sectional study in a wider theoretical context. Language Learning, 58,185231.

    177 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    Kessler, J-U., & Pienemann, M. (2011). Research methodology. How do we knowabout developmental schedules? In M. Pienemann & J-U. Kessler (Eds.), Studyingprocessability (pp. 8496). Amsterdam: John Benjamins.

    Lambert, W. E., & Tucker, R. G. (1972). Bilingual education of children: The St.Lambert experiment. Rowley, MA: Newbury House.

    Lantolf, J. P., & Poehner, M. E. (2014). Sociocultural theory and the pedagogicalimperative in L2 education: Vygotskian praxis and the theory/research divide. NewYork: Routledge.

    Lantolf, J. P., & Thorne, S. L. (2006). Sociocultural theory and the genesis of secondlanguage development. Oxford, UK: Oxford University Press.

    Mackey, A. (1999). Input, interaction and second language development: An empiricalstudy of question formation in ESL. Studies in Second Language Acquisition, 21,557587.

    Mansouri, F., & Duffy, L. (2005). The pedagogic effectiveness of developmentalreadiness in ESL grammar instruction. Australian Review of Applied Linguistics,28, 8199.

    Morgan-Short, K., Finger, I., Grey, S., & Ullman, M. T. (2012). Second languageprocessing shows increased native-like neural responses after months of noexposure. PLoS One, 7, e32974.

    Negueruela, E. A. (2003). A sociocultural approach to the teaching-learning of secondlanguages: Systemic-theoretical instruction and L2 development. Unpublisheddoctoral dissertation, The Pennsylvania State University, University Park, PA.

    Paradis, M. (2009). Declarative and procedural determinants of second languages.Amsterdam: John Benjamins.

    Piaget, J. (1950). The psychology of intelligence. London: Routledge and Kegan Paul.Pienemann, M. (1984). Psychological constraints on the teachability of languages.

    Studies in Second Language Acquisition, 6, 186214.Pienemann, M. (1987). Determining the influence of instruction on L2 speech

    processing. Australian Review of Applied Linguistics, 10, 83113.Pienemann, M. (1989). Is language teachable? Psycholinguistic experiments and

    hypotheses. Applied Linguistics, 10, 5279.Pienemann, M. (1998). Language processing and second language development:

    Processability theory. Amsterdam: John Benjamins.Pienemann, M. (2005). Cross-linguistic aspects of processability theory. Amsterdam:

    John Benjamins.Pienemann, M. (2007). Processability theory. In B. VanPatten & J. Williams (Eds.),

    Theories in second language acquisition (pp. 137154). Mahwah, NJ: Erlbaum.Pienemann, M., Di Biase, B., & Kawaguchi, S. (2005). Extending processability

    theory. In M. Pienemann (Ed.), Cross-linguistic aspects of processability theory(pp. 199252). Amsterdam: John Benjamins.

    Language Learning 65:1, March 2015, pp. 152180 178

  • Zhang and Lantolf Natural or Artificial

    Pienemann, M., & Johnston, M. (1987). Factors influencing the development oflanguage proficiency. In D. Nunan (Ed.), Applying second language acquisitionresearch (pp. 45141). Adelaide, Australia: National Curriculum Resource Centre.

    Radloff, C. F. (1991). Sentence repetition testing for studies of community bilingualism.Arlington: Summer Institute of Linguistics and the University of Texas at Arlington.

    Spada, N., & Lightbown, P. M. (1999). Instruction, first language influence, anddevelopmental readiness in second language acquisition. Modern LanguageJournal, 83, 122.

    Ullman, M. T. (2005). A cognitive neuroscience perspective on second languageacquisition: The declarative/procedural model. In C. Sanz (Ed.),Mind and contextin adult second language acquisition: Methods, theory, and practice (pp. 141178).Washington, DC: Georgetown University Press.

    Van Patten, B. (1996). Input processing and grammar instruction in second languageacquisition. Northwood, NJ: Ablex.

    Vygotsky, L. S. (1978).Mind in society: The development of higher psychologicalprocesses. Cambridge, MA: Harvard University Press.

    Vygotsky, L. S. (1986). Thought and language. Cambridge, MA: MIT Press.Vygotsky, L. S. (1987). The collected works of L. S. Vygotsky: Volume 1. New York:

    Plenum.Vygotsky, L. S. (1997). The collected works of L. S. Vygotsky. Volume 3: Problems of

    the theory and history of psychology. New York: Plenum.Wang, X. (2011). Grammatical development among Chinese L2 learners: From a

    processability account. Unpublished doctoral dissertation, Newcastle University.New Castle Upon Tyne, UK.

    Williams, J. N. (2012). Working memory and SLA. In S. M. Gass & A. Mackey (Eds.),The Routledge handbook of second language acquisition (pp. 427441). New York:Routledge.

    Xun, L. (2005). New practical Chinese reader (Book 1). Beijing, China: BeijingLanguage and Culture University Press.

    Zhang, X. (2014). The teachability hypothesis and concept-based instruction:Topicalization in Chinese as a second language. Unpublished doctoral dissertation,The Pennsylvania State University. University Park, PA.

    Zhang, Y. (2001). Second language acquisition of Chinese grammatical morphemes: Aprocessability perspective. Unpublished doctoral dissertation, The AustralianNational University, Canberra, Australia.

    Zhang, Y. (2007). Testing the topic hypothesis: The L2 acquisition of Chinese syntax.In F. Mansouri (Ed.), Second language acquisition research: Theory-constructionand testing (pp. 145171). Newcastle, UK: Cambridge Scholars Publishing.

    179 Language Learning 65:1, March 2015, pp. 152180

  • Zhang and Lantolf Natural or Artificial

    Supporting Information

    Additional Supporting Information may be found in the online version of thisarticle at the publishers website:

    Figure S1. SCOBA for object topicalization.Figure S2. SCOBA for adjunct topicalization.Figure S3. Cuisenaire rods and Chinese topicalization.

    Language Learning 65:1, March 2015, pp. 152180 180