1
Kristin Kersten
Profiling child ESL acquisition: Practical and methodological issues
Kristin Kersten
Abstract
Immersion programs have been claimed to be the most effective educational programs for the acquisition of a second language. This study focusses at ESL data from an immersion elementary school in Germany, which are analyzed within the framework of Processability Theory (PT, Pienemann 1998, 2005) and subsequently compared to PT data from naturalistic L2 acquisition. The paper puts a special focus on methodological issues of data analysis, especially with regard to coding decisions resulting from the form-function interface of linguistic structures. A fine-grained analytical grid is suggested, which is mainly based on the work of Pienemann (1998) and Pallotti (2003, 2007). The results indicate that, after four years of immersion schooling in a monolingual German environment, the participants in the program reached the final two stages of the processing hierarchy suggested by PT (stages 5 and 6) in L2 English, and are thus comparable to children learning English as a second language in a naturalistic context. With regard to data elicitation, it could be demonstrated that the profiling procedures suggested by PT can also be applied, with some limitations, to data sources not directly related to the PT framework.
1. Introduction
This paper focusses on data from an immersion (IM) elementary school in Kiel, Germany, in which monolingual German children aged 6-11 are instructed in English in almost all parts of the curriculum. The immersion method has been called the most successful educational program for second[1] language acquisition in schools (Genesee 1987, Wesche 2002, Wode 2004, cf. also results from a comparative study by Pienemann et al. 2006[2]).In order to shed more light on the effectiveness of such programs, semi-spontaneous oral narratives from a four-year longitudinal study with four children of the Kiel immersion school, originally elicited in the framework of temporality and narrative structure (e.g. Berman & Slobin 1994, Dietrich et al. 1995, for an overview on temporal semantics cf. Bardovi-Harlig 2000), are profiled with the help of acquisitional stages for L2 English as suggested by the profiling hierarchy within the framework of Processability Theory (PT, Pienemann 1998, cf. also online information by Pienemann et al. on Rapid Profile[3]). Developed by Manfred Pienemann in his influential 1998 book, PT is based on the premises of Levelt's (1989) model on language production and relies mainly on Kempen Hoenkamp's (1987) Incremental Procedural Grammarand Kaplan Bresnan's (1982)Lexical-Functional Grammar (for a concise introduction to the theory see Pienemann 2005, chapter 1). A special emphasis of the discussion in this paper lies on the criteria used for data coding.
In the first section of this paper, the program, the data elicitation procedure,and the research questions will be introduced. The next part will describe the structures the analysis focusses on, and discuss other theoretical and methodological issues relevant to the coding of the data, especially with regard to the form-function interface in the acquisition of grammatical structures. Afterwards, the operationalization of criteria used for the analysis will be discussed in detail. The following two sections will present the results of the analysis and relate them to attainment levels of naturalistic L2 learners of English (Pienemann & Mackey 1993).
2. The study
2.1 The immersion project
The data presented here were collected in a bilingual elementary school, the Claus-Rixen Grundschule in Kiel, Germany. The school is one of several educational institutions,ranging from preschool to secondary levels (ages 3-18), which are monitored under the supervision of Henning Wode at KielUniversity(see Kersten 2005, Wode 2001for more information on the Kiel Bilingual Project).
The elementary school incorporates a partial immersion program. Bilingual classes (grades 1-4) mainly composed of monolingual German children from a middle class background are taught in English in all content matter except for German language arts. This amounts to approximately 70% of L2 input throughout the curriculum.
2.2 Data elicitation procedure
Thepresent analysis is part of a larger study which was designed to elicit guided, semi-spontaneous oral narratives in the framework of the acquisition of temporality based on a comparative research study by Berman & Slobin (1994). The data are currently being analyzed with respect to the acquisition of narrative structures (e.g. Möller 2006) and verbal morphology in the framework of temporal semantics (Kersten 2007, 2008). Data was elicited longitudinally over four years in the children’s L1 and L2 at the end of each grade (grades 1-4), and cross-sectionally with an L1 English-speaking comparison group[4] from an elementary school in White Bear Lake, Minneapolis, in the USA, grades 1-4.
Subjects: The analysis of this study was carried out on the samples of four subjects of the L2 data set, i.e. subjects 03, 06, 07, and 08. For each of the children, one test is available at each grade level. All subjectsare female and started learning the L2 at age 6, with the exception of child 06, who had prior experience with the L2 in a bilingual preschool.
Method:A picture story was used for data elicitation.[5]During each elicitation, the narration was delivered twice in the L2, first with a German-speaking interviewer whom the children were able to ask for vocabulary (Test A), and subsequentlywith an interviewer who was known to the children as a native speaker of English (Test B). During the second test, no questions were permitted.Bothversionsprovided data for this study. The grade 4 data also include a short interview on personal stories preceding the narration. The tests were audio- and videotaped and subsequently transcribed. The samples average about 7:50 min of recording.
Coding conventions: Direct repetitions within the narrations (e.g. where is my frog, where is my frog), repetitions of interviewers' utterances (up to 3 turns), hesitations and self-corrections, and uninterpretable elements such as tokens with unclear endingswere excluded from the analysis.
2.3 Research questions
The developmental stages suggested by PT are used to evaluate the effectiveness of language acquisition in an IM educational program. As the data were collected in a different theoretical framework, the first step of the analysis will focus on the applicability of PT to this specific set of child data. Only after having established the coding criteria for the analysis can the results be compared to naturalistic L2 acquisition. Thus, the study will focus on the following research questions:
1. Do the data confirm the stages predicted by PT?
2. What are the operational criteria for an application of PT to this data set?
3.Are the results of L2A in an immersion school as indicated by the PT stages comparable to results achieved by learners from a naturalistic learning context?
3. Theoretical and methodological considerations
The operational methodology applied here is based on Pienemann (1998) and includes criteriasuggested byPallotti(2003, 2007),[6] who, in his recent work, sought to operationalize the coding criteria presented in Pienemann (1998) for his application of PT to Italian. The following section introduces the selection of structures focussed on in the analysis, anddiscusses some problematic issueswhich have emerged in the process of data interpretation.
3.1 The structures
The classification of linguistic structures into six developmental stages of ESL identified by PT is based on Pienemann (1998), Rapid Profile, and on Pienemann & Johnston (1987) and will not be repeated here. In order to use PT as a profiling measure for data not elicited in the PT framework,
6 / cancel inv
5 / 3.sg –s
4 / (wh-)copula inv
yes-no inv
part-verb
3 / do-front
topical
wh-front
adv-front / aux+ing
aux+en
poss pro/det
obj.pro
2 / SVO
neg+V / past reg
past irreg
IL-ing
plural –s
1 / single wds
Table 1: PT-Structures
it is important to identify those structures which occur with a relatively high frequency and thus increase data density.
Verbal inflections are likely candidates since they appear in almost every clause of the narrations. Table 1 gives an overview of the syntactic and morphological structures coded for analysis. Irregular past was classified together with regular past and IL (interlanguage) -ing as stage 2 since no exchange of phrasal information is involved. The two auxiliary constructions not present in Rapid Profile were classified as stage 3 because of the unification of values between auxiliary and the lexical verb, which is an instance of phrasal agreement (Pienemann 1998:175). Note that this does not require the target-like use of the auxiliary.The target-like agreement of auxiliary and subject is not relevant here; otherwise the structure would have to be placed at stage 5. To illustrate this with examples from the data:
Child 06.1:
and the dog falling down off the window(IL-ing, stage 2)
and the boy are looking to a tree (aux+ing, stage 3)
Since there were too few occurrences of do-front and aux 2ndin the sample they were excluded from the analysis.
3.2 Form-function interface and the emergence criterion:
some problem cases[7]
Research in SLAhas repeatedly shown that a learner's IL-system is a system in its own right with its individual rules, which do not necessarily need to correspond to those of the target language (Pienemann 1998: 138f, 160f). Especially in the beginning stages of acquisition, investigations into the distribution of linguistic features run the risk of confusing the emergence of interlanguage rules with random variation and chunks (Housen 1995, Perdue 1993, Pienemann 1998 chapter 4, 117ff).Such confusion can be avoided when researchers explicitly define "what forms and what functions are considered evidence for the emergence of a certain structure" (Pallotti 2003:1).
The emergence criterion is a concept crucial for data analysis within the framework of PT. For an analysis of learner language, PT proposes a methodological operation which is based on the criterion of the first emergence of morhosyntactic elements in the language produced by a learner. More specifically, Pienemann characterizes emergence as the "first systematic use of a structure" in question (1984:191). This refers to the moment "at which certain operations can, in principle, be carried out" (1998:138).
In recent years, Pallotti has further developedthe definition of the term emergence(2003, 2007).In his 2007 article, he seeks to operationalize the different components of emergence as quoted above to validify data analyses based on this criterion. The relation of form vs. function of a structure becomes especially important in two aspects of his definition, i.e. the notions of the productive[8] and the systematic use of grammatical morphemes:
The learner may in fact be supplying the morpheme randomly, with no clear function, in free allomorphic variation. In this case, one would not say that a systematic form-function association has emerged, but rather that the learner is still experimenting with a phonological form. A criterion must specify a way of differentiating such cases from systematic uses. ... Only when a form begins to be used with a specific, selective function can one conclude that a rule has emerged. This point is also made by Pienemann (1998: 126).(Pallotti 2007:371f)
It has to be pointed out that the term functionshould refer solely to the grammatical funtion of a linguistic element as opposed to a conceptual or semantic function in a specific context. To give an illustration of this distinction: if the V-s inflection is used by a learner in many different contexts, most of which do not refer to the grammatical function of 3rd pers. sg., V-s cannot be interpreted as having emerged in terms of a systematic use of the structure. The application of V-s rather seems to be used randomly in the data. However, this example only relates to the suppliance of the grammatical function of 3rd pers. -s. The temporal (i.e. conceptual) function of present tense which the -s inflections carries as well should not be taken into consideration in the analysis, as PT does not make any claims about conceptualization but only about the processing load in terms of syntactic and morphological grammar formulation. This distinction will be discussed in more detail below (section 3.2.3).
While the description of morpho-syntactic forms seems to be a quite straightforward operation, the description of their various functions thus seems a somewhat more complicated matter.It is therefore necessary to carefully operationalize the criteria which lead to the conclusion that a structure has emerged in an interlanguage system with respect to the first systematic occurrence (Pallotti 2007). To account for these criteria, Table 2 shows a distributional analysis which takes into account the variables context, target-like types, and target-like tokens of a given element, as well as its under- and over-suppliance (i.e. the null-hypothesis). Table 2 is an example for two structures which have emerged in an interlanguage system, i.e. 3rd sg. -s and plural -s (subject 06,grade 1):
Structure / Contexts/target-liketypes/target-like tokens
3.sg -s
3.sg s
-s 3.sg
others –s / 59/11/21
38
0
10/2/10 / – under-suppliance
– over-suppliance
pl -s
pl s
-s pl / 11/5/11
0
0 / – under-suppliance
– over-suppliance
without
Table 2: Example of a distributional analysis (Child 06.1)
To illustrate this distribution with the example of 3rd pers. sg. -s: In 59 contexts of 3rd pers. sg., the child uses V-s with 11 types (i.e. with different lexical verbs) and 21 tokens (i.e. total occurrences independent of the verb type). 38 contexts of 3rdpers. sg. are used without the -s inflection (i.e. as base form V-ø), which is counted as under-suppliance, but V-s is never used in a context other that the 3rd pers. sg., thus there is no instance of over-suppliancein this example. For this reason, the structure was assigned the status emerged in the analysis (cf. chapter 4).
While in Table 2 the form-function interface seems unproblematic, difficulties may arise in the interpretation of corpora like the one at hand with regard to other kinds of functions. The following sections will discuss the notions of grammatical vs. conceptual function and their relevance to a PT analysis with reference to the interpretation of reported speech vs. relative clauses as an example of the former notion, as well as to maturational and lexico-semantic influences on learner language as an example of the latter.
3.2.1 Conceptual vs. grammatical function
The preceding sectionhas hinted at certain difficulties in coming to terms with the form-function relationship which has to be taken into consideration when assigning the emergence criterion to interlanguage data. The term function is, in itself, not without problems, though. As already indicated earlier, it is necessary to tease apart different functional aspects of a linguistic element.
Taking again the function of 3rdpers. sg.-s as an example, two different aspects can be distinguished: On the one hand, the inflection denotes subject-verb agreement. This specific exchange of inter-phrasal grammatical information pertains to the S-procedure in the Formulator of Pienemann's model (1998, 2005, based on Levelt's 1989 model of language production). On the other hand, the inflection marks present tense.In other words, one can differentiate between a grammatical and a semanticfunction (Tarone 1988) of the linguistic element. In the example of 3rdpers. sg. -s the grammatical function would be inter-phrasal agreement, and the conceptual or semantic function relates the respective event within a temporal sequence. Thus, in the tense/aspect system of a language it is the second, the semantic function, which comes into play (cf. also Huddleston 1993:80f).
In Levelt's model of speech production (1989:9) which underlies PT, time reference is generated in the Conceptualizer (cf. also Pienemann 1998:76), whereas the diacritic feature for tense marking is part of the lemma information stored in the Lexicon. PT's predictions on the processing hierarchy focusses solely on the functioning of the Formulator (Pienemann 1998:74). It pertains fully to syntactic structures, whose functions are often described as "grammatical relations" (Huddleston 1993:7): there is no conceptual difference in saying Turn off the radio or Turn the radio off; nor in Where are you? vs. *Where you are?Thus, in the processing of syntactic structures, it is the grammatical function which indicates that a specific procedure is at work. Still, when it comes to data coding, not all syntactic structures in learner language are easy to interpret in this framework. An example will be discussed in section 3.2.2. In section 3.2.3, a more detailed discussion of form and function with reference to time and morphological tense-marking follows, to which the predictions of PT refer only in part.
3.2.2 Cancel inversion
In his characterization of cancel inversion, Pienemann states that “word order phenomena observed in direct questions do not apply in the context of indirect questions” (1998:170). He gives the example I wonder whether/why/where (he had lunch yesterday). An example for this structure from this study is found in the data of child 08.3:
08.3
... shouted: "Where is my frog?"
... shouted where the frog is.
The introductory verb shout is not quite target-like, but the opposition of both phrases points to a cancel inversion in the context of an indirect question in the second phrase. But consider the following example:
06.2
... looks where the bees are.
... to see where is the frog.
According to Greenbaum & Quirk (1990:298f), only cognitive verbs or verbs describing a mental activity introduce indirect questions. However, it seems obvious that, in 06.2, the child overgeneralizes the inversion rule from indirect contexts to relative contexts introduced by look and see. It is apparent in the data that the difference between indirect questions and other forms of subordination is not as clear-cut as theory would have it. The same wh-elements that introduce interrogative clauses can as well introduce relative clauses (Greenbaum & Quirk 1990:309, 367f). And Huddleston (1993:396) points out that the same fronting mechanisms are at work in both kinds of clauses. What adds to the confusion in the latter example of 06.2 is the fact that the antecedent is missing, i.e. the element which the relative pronoun/adverbial refers to. Huddleston (1993:396) calls these structures fused relative constructions. He points out that they even partially overlap with interrogative constructions (as e.g. in he spent what they gave him vs. she told me what they gave him, p. 404).
Without going into too much detail, it obviously seems rather unlikely that the learners distinguish between relative and interrogative wh-elements in their interlanguage. It is much more probable to assume that once a lexical item is annotated for a certain grammatical function, in this case for inversion, it is initially overgeneralized to all contexts, and only later in development becomes differentiated with respect to different grammatical functions.
This might be different, however, in contexts of relative clauses with antecedent. There are many instances of such full relative clauses especially in the data of the third and fourth year. The last section raises the question of whether these elements, too, should be classified as contexts for cancel inversion. Indeed, some instances of the corpus suggest that inversion seems to be an issue for the children even in full relative contexts: