Problem solving in a foreign language Lena Heine

Sách dành cho các bạn sinh viên đại học và thạc sĩ chuyên ngành Tiếng Anh nghiên cứu về phương pháp giảng dạy, và đi sâu vào giải quyết các vấn đề trong quá trình học ngoại ngữ. Bao gồm 11 chương và và các nội dung tương ứng

Studies on Language Acquisition 41

Editor

Peter Jordens

De Gruyter Mouton

e-ISBN 978-3-11-022446-7

ISSN 1861-4248

Library of Congress Cataloging-in-Publication Data

Heine, Lena.

Problem solving in a foreign language / by Lena Heine.

p cm ⫺ (Studies on language acquisition ; 41)

Includes bibliographical references and index.

ISBN 978-3-11-022445-0 (alk paper)

1 Second language acquisition ⫺ Study and teaching 2

Lan-guage and lanLan-guages ⫺ Study and teaching 3 Language and

cul-ture 4 Sociolinguistics I Title.

P118.2H45 2010

418.0071 ⫺dc22

2010003181

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

쑔 2010 Walter de Gruyter GmbH & Co KG, Berlin/New York

Printing: Hubert & Co GmbH & Co KG, Göttingen

⬁ Printed on acid-free paper

Printed in Germany

www.degruyter.com

For Doris and Werner Heine who make things grow and sail the sea

Acknowledgements

This book is based on my PhD thesis “Kognitive Prozesse bilingualer Lerner bei der fremdsprachlichen Verarbeitung von Fachinhalten” (2007, Osnabrück University), which was partly funded by the German Research Foundation within the project “Fachlernen und (Fremd-)Sprachlichkeit: aufgabenbasierte Kognition, Kooperation, Kommunikation”

I would like to thank H Johannes Vollmer for all his support and for taking me ‘on board’, Debbie Coetzee-Lachmann for the many stimulating discussions and the enjoyable working hours spent together, and Verena Barbosa Duarte, Kristin Möller and Randi Troschke for their help with the coding of the think-aloud data

A special thanks goes to Alex Bergs for his support under the final stages

2 Linguistic knowledge in multilingual speakers 10

1 Task-solving as a complex problem solving activity 57

Chapter 6 Task design and task analysis 68

1 The importance of the research design 68

2 Thinking aloud and language of thought 84

3 Individual differences in cognitive processing 89

5 Think-aloud data as protocols of problem solving activites 94

2 Change of focus between the problem spaces 107

3 Different phases – different mental activities 114

1 Language as a catalytic converter: depth of processing 119

2 L1-L2 processing differences in language production 128

6 Differences in semantic processing of text information? 156

7 Critical thoughts on the process data 159

1 Limitations of the think-aloud method 162

5 Is thinking aloud a valid elicitation method? 173

Chapter 11 Results and discussion 177

5 Implications for the design of learning contexts 185

Figures

Fig 1: Cognitive processes on a continuum of descriptive levels 26 Fig 2: The model of text composition by Flower and Hayes (1981) 44 Fig 3: Bereiter and Scardamalia’s (1987) model of knowledge telling 46 Fig 4: Bereiter and Scardamalia’s (1987) model of knowledge

Tables

Table 1: Framework for the description of complex problem solving

tasks

32 Table 2: Task solving phases of the conceptual task solution 61 Table 3: Task solving phases of the linguistic-rhetorical task solution 63

Table 4: Cognitive Task Analysis of Task 2 (knowledge structures and

cognitive processes)

78

Table 5: Problem solving phases in the conceptual and the

linguistic-rhetorical problem spaces

102

Table 6: Cases of deeper semantic processing, caused by activity in the

linguistic-rhetoric problem space

129

Table 7: Deeper semantic processing of the CLIL learners, caused by

ac-tivity in the linguistic problem space

147

Table 8: CLIL learners’ mispronunciations when reading 152 Table 9: Subject-specific scores in the answers of Task 6 156 Table 10: Contribution of individual data types to the study 182

Chapter 1

Introduction

1 Bilingual education and CLIL

Language is the most elaborated semiotic system we use, and it is cably linked with the meaning that it conveys In all linguistic communica-tion, conceptual content is either constructed on the basis of linguistic structure – namely when utterances are decoded, and a mental representa-tion of the message is constructed –, or it is merged into such a structure –

inextri-in speech production or text composition

This book is about the effects of foreign language use on the cognitive processes of meaning-focused problem solving It is an attempt to under-stand more clearly what happens when humans deal with non-linguistic tasks in a foreign language setting Thus, I will try to find answers to the following two main questions:

have a focus on meaning, but not on form?

solved in an L2 dominated environment, compared to an L1 text?

con-The question whether the use of a foreign language has any effects on problem solving, is particularly relevant for contexts in which an L2 is used

as a working language in educational settings, such as immersion programs

or Content and Language Integrated Learning classrooms Indeed, the

main reason why I decided to undertake this research was that I wanted to

acquire a better understanding of the characteristics of Content and guage Integrated Learning (CLIL) I will therefore provide some back-

Lan-ground information for this educational approach and the state of research

in the field

In CLIL classrooms, often also called Bilingual Education, a foreign

language is used as the medium of instruction and communication in a ited number of content subjects in an otherwise L1 dominated context In Europe, it has become a very popular form of teaching, and we find it in a wide variety of formats CLIL differs from immersion programmes, known first and foremost from Canada (Genesee 1987; Harley, Allen et al 1990; Swain 1998; Swain and Lapkin 1995), in that the dominance of the L2 is smaller and generally only covers one to three school subjects (often sci-ence subjects such as geography, history, or biology) Still, differences in individual programmes are large, even within the same country This makes general statements somewhat difficult, but certain aspects can be generalized (see Dalton-Puffer and Smit 2007 for a summary of the histori-cal development of the discipline, and Garcia 2009 for an overview with an international perspective)

lim-Historically, CLIL has in many European contexts emerged as an native to the traditional language classrooms, in which not the linguistic structure is in the centre of the syllabus, but its communicative use in authentic learning situations Systematic comparisons between foreign lan-guage proficiency have provided evidence that learners in CLIL classes outperform learners who only participate in the traditional language class-rooms (Bredenbröker 2000; DIPF 2006; Helbig 2001; for an overview, see also Vollmer 2000; Wode 1995; Zydatiß 2005, but see Sylven 2004 and Lim Falk 2008) As a rationale for this superior effect on language acquisi-tion, the Comprehensible Input Hypothesis by Krashen (1984; 1985) or, as

alter-an alternative approach, the Output Hypothesis (Swain 1985) have been considered However, only general foreign language proficiency has been measured in the studies mentioned above, while the subject-specific as-pects of language use was not in focus

Therefore, it is not surprising that predominantly research on language acquisition has dealt with CLIL, while its character as a subject classroom setting has largely been neglected This stands in contrast to the fact that in CLIL curricula, linguistic structure and function is only seldom the explicit

topic of instruction, in spite of the comprehensive term Content and guage Integrated Learning; language learning is assumed to take place

Lan-more or less automatically, and CLIL classrooms are generally regarded by

teachers and syllabus developers as subject classrooms with a focus on meaning (Long 1991, Long and Robinson 1998), not language classrooms;

introducing and practising aspects of linguistic structure is generally garded as a domain of the foreign language classrooms

re-Mirrored in CLIL curricula and the scientific discussion is the following understanding: It is assumed that language learning and content learning are regarded as separate processes, and subject-specific content learning occurs more or less independently from language In contrast to this, we find in any meaning-focused classroom that language plays an important role, even if it is not the centre of attention: Language serves as the domi-nant means of communication, and the medium in which the subject con-tent is presented and addressed; through language conceptual relations are discussed and negotiated in classroom interactions; subject-specific knowl-edge is built up by reading textbook chapters, or through discussions with fellow students in group work Furthermore, language is the predominant semiotic system used in testing learning outcome: In typical achievement tests, students are required to demonstrate learning outcome by putting their knowledge, skills and competences into words, either written or spo-ken Even in testing formats, where no oral or written texts have to be pro-duced, language has to be dealt with at least in a receptive way, namely in decoding the task prompts or test questions

Until now only few studies have focused on the aspects of content learning in CLIL settings and the role that language plays therein So, al-though several studies have revealed that L2 proficiency is improved in

CLIL settings even if it is not in focus, we still do not know what effects

the use of the L2 as a working language has on the processing of subject content One major reason for this is a methodological obstacle: It is much more difficult to measure levels of subject-specific competence than lin-guistic proficiency, because carefully validated tests for general linguistic proficiency have long been in use, but reliable tests for subject-specific competences are still lacking Thus in the few studies we find about sub-ject-specific knowledge and its use in typical problem solving situations (Badertscher 2004; Gajo and Serra 2002; Golay 2005; Koch 2005; Koch and Bünder 2006; Stohler 2006), no competence tests have been used up to now; instead pre-post test designs that test only a selection of more or less randomly chosen subject knowledge have been preferred That is why they cannot claim to give an overall measure of the central subject-specific competences Within the scope of these tests the studies show that CLIL learners achieve a similar subject-specific proficiency as the control group

of learners taught in traditional monolingual environments

First attempts to measure subject-specific competence have been made

by Bonnet (2004) for the field of chemistry and Vollmer (2006a; 2006b,

2007, see also Passon 2007) for geography Vollmer uses a competence

model developed by the German Geographical Society (Deutsche sellschaft für Geographie 2006) The results are difficult to compare, and Vollmer’s study has not been finalized yet, but here too it seems obvious that the CLIL learners are in their subject-specific competences on a com-parable level to learners from traditional classes

Ge-These results are particularly interesting with regard to what has been addressed as the ‘CLIL paradox’ (De Florio-Hansen 2003; Thürmann 2000; Zydatiß 2002) The puzzle addressed here is that CLIL learners have

to acquire a range of the categories and skills of the subject, but are at the same time lacking the linguistic competence that would allow them to infer these categories from context So, subject learning and language learning have to take place at the same time, despite the fact that one is the precon-dition for the other When we compare this paradox to the outcomes of the studies mentioned above – why are CLIL learners apparently not hindered

by the use of the L2 as a working language? How do they compensate for the difficulties they should experience in the foreign language setting? This interrelationship of the two sides of CLIL learning, subject side and language side, needs to be clearly described, and the two sides put into relation to each other to form a scientific basis for the development of cur-ricula, teaching material and suggestions for their implementation

2 Objectives

Although the focus of this book is the relationship between (foreign) guage and the construction of conceptual meaning relationships, its aim is not to contribute to the discussion on linguistic relativity, e.g how speakers

lan-of different languages conceptualise the world differently through the ticular linguistic microstructures they are using

par-What I will focus on instead is whether it is possible to see an effect of the use of a foreign language as a working language in processes of task solving The study thus approaches the intimate relationship between lan-

guage and thought from the macro-perspective of focus It asks how phases

of problem solving in which a person’s focus is on linguistic information influence processes of meaning construction So, not subconscious micro-processes of grammatical or lexical processing are at the centre of this study, but potentially conscious phases in a student’s attempt to find a solu-tion to a problem

It follows that the aims of this book are basically twofold, because in der to understand which role a foreign language plays in content-focused set-tings, two things are necessary: Firstly, we need a theoretically well-based model of how language and problem solving are interrelated in cognitive processing in general I will suggest such a model, which integrates the so-cio-cultural embedding of the cognitive processes

or-Secondly, assumptions and statements about the effects of the foreign language on the task solving processes should be based on empirical evi-dence I will therefore provide an analysis of empirical data that has been elicited, transcribed and categorized especially for this purpose The data corpus consists of a set of three kinds of data: 1 think-aloud protocols pro-duced by subjects solving meaning-focused tasks in a foreign language set-ting, 2 the written texts they composed as an answer for the elicitation tasks and 3 interview data collected after the task-solving sessions While the think-aloud corpus can be regarded as the main data set that has been elicited in order to gain insight into the online-processing of language and conceptual content, the written answers and the interview data serve as complements that allow a methodological triangulation The following sec-tion gives a more detailed outline of the content of the book

3 Structure

The book contains eleven main chapters After this introductory chapter I will account for the theoretical view underlying the empirical data elicita-tion of the study

Chapter 2 starts with a short overview over definitions of linguistic and non-linguistic knowledge from different theoretical perspectives and the re-lationship between language and thought

Chapter 3 continues with a discussion on the processing of information in the context of tasks that can be defined as problems (in other words, situations

in which a goal state cannot be reached automatically, but which require the application of knowledge and cognitive activity of some sort) I will discuss different degrees of abstractness in cognitive processes, and distinguish be-tween problem solving processes with a focus on language on the one hand and with a focus on content on the other

In Chapter 4 I will give a short overview of psycholinguistic approaches

in order to account for the question of what happens when linguistic

infor-mation is processed Here, I briefly discuss processes of speech production and reception, and elaborate on models of text composition and reading Chapter 5 integrates these basic theoretical assumptions into a model of task-based content and language processing This model accounts for the assumption that different content-based and linguistic activities can be identified during the macro-process of problem solving, and that the focus can alternate between them In addition to the process dimension, the model also covers the aspect of learner characteristics and that of the con-text and the task setting, which account for macro-theoretical assumptions

of the individual construction of a mental representation and the necessity

of embedding any analysis of cognitive processes into their social contexts

In Chapter 6, the basic framework for the elicitation of the empirical data is provided The elicitation tasks are firstly described and a way for analysis presented On the one hand, the problem solving tasks show a clear focus on content in that non-verbal sources of information such as climate graphs and geographical maps had to be interpreted by the subjects

On the other hand, even linguistic information had to be processed in order

to solve the tasks: Part of the information was provided in linguistic form (task assignment and information texts) that had to be decoded Further-more, the subjects had to formulate a written answer in which they pre-sented their findings One group of subjects had to work with all linguistic material in their L1 German, while another group had to process all linguis-tic input in their L2 English, and had to write their answers in the L2 as well This kind of task design was chosen on the basis of the following rea-soning: If interactional effects between content processing and linguistic processing are to be investigated, both kinds of processes take place when such a task is solved, and should thus be accessible in process data

After depicting how the elicitation tasks are designed in order to trigger the mental activities to be investigated, I will in Chapter 7 discuss possibili-ties for making these mental processes observable What is interesting here

is that most of the time language use is an automatic process that people only seldom reflect on Normally, we do not notice the words and gram-matical constructions we use, but, as it were, we immediately decode their underlying meaning structures This point has methodological implications for this study in that it has led to the decision to use think-aloud data as the main data source In the remainder of Chapter 7, the think-aloud method is outlined and discussed in terms of both its advantages and restrictions, and the suitability of think-aloud data as a source for investigating mental proc-esses in problem solving activities is explained

In Chapter 8, the process categories from the model introduced in ter 5 are exemplified, using think-aloud data examples from the study With their help it is possible to separate the process data into content-focused and form-focused processing activities In order to give a reliable interpretation, a cognitive task analysis is shown to be necessary in order to identify different phases of the problem solving process The focused in-stances in the data analysis are thus instances where attention on any lin-guistic form triggers a deeper processing of meaning relationships, espe-cially when an L2 is used as a working language Thus, no linguistic construction is in particular focus, although it turns out that many problem solving activities are caused by knowledge gaps on the lexical or phrasal level

Chap-The results of the data analyses are presented in Chapter 9 Here, the empirical findings are discussed and illustrated with examples from the data This results in the formulation of a number of hypotheses concerning the role of language in problem solving activity, with a special focus on foreign language use

Owing to the fact that before and during the data elicitation, the aloud method led to many questions that could not be answered by the exist-ing literature, part of the data elicitation was conducted in order to address these problems Chapter 10 is dedicated to the presentation of these results concerning the validity of the method Here, evidence is provided for the in-dividual and shifting character of mental processing and the need for psycho-linguistic modelling to respond to this

think-By presenting the hypotheses that have been generated, Chapter 11 summarizes the findings, gives an outline of recommendations for didactic situations as well as perspectives for future research The chapter concludes with final remarks concerning qualitative and quantitative research designs and hypothesis testing

Chapter 2

Language and thinking

1 Linguistic knowledge

The interrelation between language and non-linguistic knowledge has been

of interest in a wide range of fields, ranging from logic and philosophy of language over cognitive psychology to linguistic semantics, and is of cru-cial interest if we want to find out about possible impacts of language on meaning-focused problem solving activities

In this chapter I will present the theoretical view underlying the design for the empirical study Because my aim is to study the impact of an L2 on learners’ mental processes in problem solving situations, it is necessary to specify the basic assumptions about linguistic and non-linguistic knowl-edge

It is not the space here for any detailed accounts of theoretical views; thorough overviews can be found in specialized monographs, such as Edmondson and Burquest (1998) or Herdina and Jessner (2002) for the modelling of linguistic knowledge, or Löbner (2002) or Saeed (2003) for linguistic semantic theories, to name just a few Instead, I will give a rough overview of how the variety of research strands conceptualise linguistic knowledge, and the relationship between language and thought in general and in problem solving situations in particular This will form the basis for the analysis for the empirical data, presented in Chapter 9

Since the cognitive shift in the late 1950s the main task of linguistics has been to model linguistic knowledge In the modular linguistic theories in the Chomskyan tradition that have developed to become the dominant view since then, linguistic knowledge is assumed to be based on innate, specific and independent cognitive mechanisms that differ from those of other con-ceptual knowledge representations Although embedded into the general cognitive apparatus, and in principle dependent on a theory of cognition, linguistic knowledge is modelled separately from general cognition, de-scriptions are made module- internally Therefore, results from psychologi-cal studies on perception and cognition are generally not integrated into linguistic model building

The terminology used in the neurosciences or cognitive psychology, such

as ‘perception’, ‘activation’, or ‘memory’, is seldom used in traditional guistic approaches, and general cognitive structures and functions play only a subordinate role, if any (but cf Marantz 2005, who argues that gen-erative linguistic theories are compatible with results from the neuro-sciences) The same holds for sociopsychological concepts such as ‘expec-tations’ and ‘assumptions’, ‘affects’ or ‘context’ Rather, linguistic processing is described in its own terminology, such as ‘mental lexicon’ and ‘rules’ for the generation of grammatically correct utterances, without asking detailed questions about the cognitive-psychological reality of these abstract concepts, nor the sociocultural embedding of language use Within these traditional theories, linguistic competence is regarded as a composite of several autonomous subsystems in which different fields of knowledge are stored, each subsystem controlling a specific level of lin-guistic representation In most models, we find a separation into phonological, morphosyntactic, lexical and semantic information (for an overview, see e.g., Aitchison 2003), with varying assumptions about which entities are stored as a whole and what is generated on the basis of rule knowledge (for an overview of different models of morphosyntactic proc-essing, see Haspelmath 2002)

lin-What is interesting for us here is the question which structures should

be regarded as linguistic knowledge, and which ones have to be thought of

as being part of non-linguistic modules The answer is rather ward for knowledge of phonological or grammatical rules, which can clearly be counted as linguistic knowledge Yet, a distinction becomes more difficult when it comes to the meaning part of language

straightfor-Generally, linguistic knowledge is assumed to contain a semantic part in the mental lexicon, where the meaning of linguistic entities, such as words,

is stored It contains information about what is permanent of a word’s core meaning, or its sense (e.g., the general, isolated meaning of a word like

‘shoe laces’, Evans and Green 2006: 208) Semantic knowledge is trasted with encyclopaedic, or conceptual, knowledge, which is assumed to

con-be stored outside the linguistic module, and contains knowledge of the world (e.g., where to buy shoe laces or how to tie them, Evans and Green 2006: 208) Encyclopaedic knowledge is by traditional linguists regarded

as belonging to the field of cognitive psychology, rather than linguistics, and is not integrated into the modelling of linguistic core disciplines This traditional view has been challenged, however, as I will explain in the fol-lowing section

2 Linguistic knowledge in multilingual speakers

After this short overview of general linguistic theories, the question mains how foreign language acquisition research attempts to model the or-ganization of linguistic knowledge

re-Foreign language research has suggested different approaches (for overviews, see e.g Appel and Muysken 1987; Baker and Prys Jones 1998; Bhatia and Ritchie 2008; Herdina and Jessner 2002; Kroll and de Groot 2005) Because at least two language systems are in focus here, a bigger grain size is chosen for the theoretical view This often leads to the unfa-vourable situation that models are restricted to an internal theorizing of this aspect, but fail to embed it into a larger framework Compare for Herdina and Jessner (2002), who criticise

that a large number of theories currently discussed in research on language acquisition and multilingualism lack an overall theoretical foundation

In the following, I will only present theories that do address the issue of knowledge representation

Older theories describe multilingual knowledge as subordinate partments of the language module, and with that tie in with modular theo-ries of grammar Here, the models by Weinreich (1953) and Ervin and Os-good (1954) used to have a highly influential status A central concept in their assumptions is the differentiation between compound and coordinate bilingualism A compound bilingual speaker has learned two languages in the same cultural context, while a coordinate bilingual learner has formed each language system in a different cultural surrounding In the first case, it

com-is assumed that both languages form a common mental representation, so that words from the different language systems relate to the same concep-tual representation In the second case, on the other hand, two separate rep-resentations are formed, in which each language’s lexicon is stored sepa-rately and linked to separate conceptualizations

If the proficiency for both languages is different in coordinate guages, which is the norm in foreign language acquisition, Weinreich sug-gests that the concepts from the language with the higher proficiency can influence the interpretation of word meaning of the other language (= in-fluence from L1 to L2) Generally, it is assumed that is more probable that more interferences occur between compound than between coordinate lan-guage systems

lan-In order to test these theoretical assumptions, studies in error analysis and contrastive research designs were conducted Their goal was to inves-tigate on a structural and behaviourist background which stages of lan-guage development occur and which transfer and interference phenome be-tween language systems can be found (see Edmondson and House 2000; Herdina and Jessner 2002) The theoretical assumption behind this is that the learner language develops successively from an incomplete to a native speaker status Furthermore, an influence of the L1 on the L2 is assumed in the way that identical structures in the L2 should be easy to acquire, while different structures should be difficult, and result in linguistic structures that deviate from the target language structure This contrastive hypothesis has met harsh critique:

Learner language is typically instable and in many respects independent from the L1, so that the individual and constructive character of language learning with its embedding into learner motivations and other factors needs to be integrated in explanations Most learners do never reach an L2 proficiency comparable to their L1, but tend to reach a standstill on certain stages (Corder 1971; Nemser 1971; Selinker 1972)

Besides that, the degree of difficulty cannot be defined by the degree of difference between linguistic structures, because this value would remain constant (Edmondson and House 2000) This means that English learners

of German would have the same difficulties as German learners of English, but that is obviously not the case: Any second language teacher will attest that the German verb position in subordinate clauses usually poses big problems for native speakers of English, while the corresponding syntax of English is only seldom problematic for German learners of English Like-wise, the contrastive hypothesis cannot explain why learners of the same L1 do experience idiosyncratic difficulties when learning the same L2 And finally, early research (Osgood 1949) already indicates that a lack

of contrast can cause errors in the L2 system as well So, even when there are only minor differences between L1 and L2, more problems can arise than when the differences are very obvious

Another highly influential early study for theories on L2 representations

is Peal and Lambert (1962) In a large-scale investigation, it could show that bilingual Canadian children (English-French) were not only superior to

a French-speaking control group in verbal tests in both languages, but even

in nonverbal tests of cognitive ability This study is highly important for the following reasons: Before it, L1 influence on the L2 was only assumed

to lead to negative transfer phenomena Now it could not only be shown

that even the L2 had an effect back on the L1, but even on other cognitive areas These results could be confirmed in further studies (Baker and Prys Jones 1998)

These results provide evidence against the modular hypothesis, because

a modular view cannot account for interactions between the linguistic module and other cognitive modules

Another theoretical model that explicitly addresses the representation of different language systems in multilinguals is suggested by Cummins (1979; 1984; 1991) In part, it can provide explanations for the Peal and Lambert study

According to Cummins, linguistic knowledge can be subdivided into different types of knowledge: Basic interpersonal communication skills (BICS) und Cognitive/Academic Language Proficiency (CALP) BICS covers linguistic elements such as pronunciation, lexicon, and grammar, which enable the speaker to communicate CALP, on the other hand, refers

to the ability to use language as a cognitive tool in order to construct and structure abtract thoughts and master metalinguistic requirements For in-stance, a general text decoding ability is assumed that transfers general de-coding-related strategy knowledge to new L2 knowledge This means that a proficient L1 reader will also be a proficient L2 reader (of course this is de-termined by general proficiency) Likewise, abilities that have been ac-quired in L2 contexts are accessible for L1 usage as well

Another model is linked to this interdependence hypothesis which lenges the traditional view of multilingualism as multiple monolingualism (see Baker 1996) Within the framework of multiple monolingualism, lan-guage acquisition is regarded as a reduplicated number of processes and cognitive resources, so the abilities for each language are separate (Cum-mins calls this SUP for Separate Underlying Proficiency) This assumption underlies the old view that multilingualism is actually restricting cognitive capacities: The more cognitive resources are bound by one language, the less capacities are left over for the other languages According to this logic, the optimal case is monolingualism

chal-As a contrasting theoretical view, Cummins suggests a Common lying Proficiency (CUP) Multilinguals use a common pool that is underly-ing all their languages systems This is conformous with the fact that in-crease of proficiency in one language can have even a positive effect on the other language systems

Under-The studies by Peal and Lambert (1962), and concurrently the pendence hypothesis by Cummins (1979; 1984; 1991) have provided im-

interde-portant impulses for more recent models of multilingual language edge In particular, we can state that the potential for interaction between cognitive and language-specific knowledge is a well-accepted concept by now Furthermore, the constructivist assumption that there are individual differences in how the linguistic subsystems are build up and interact with each other (cf Riemer 1997) Language systems are highly dynamic, so that a once reached proficiency stage can be lost again Newer models make use of system theory and chaos theory, and the basic assumptions of Cognitive Linguistics have been adapted for foreign language research (ibid., also Achad and Niemeier 2004) This has implications which lead to

knowl-a renknowl-aissknowl-ance of topics from eknowl-arlier reseknowl-arch under new focus:

Still the question is unanswered whether multilingual speakers store knowledge in one cumulative or several separate mental lexica In this con-text, the Subset model by Paradis (1987) takes an intermediate position, and might have the biggest explanatory power It starts out from a single mental lexicon of all languages, while each linguistic code forms an inter-nal subsystem within it, either because these elements are used in combina-tion so often (Raupach 1994) or because of structural features such as the phonological-prosodical system

As this short overview shows, it seems not justified to assume separate cognitive modules, in which the linguistic module is subdivided further into different linguistic codes Instead, the interlinking and interdepen-dency of all knowledge ares has to be stressed, and conceptual meaning is not bound to single linguistic systems

3 Language and meaning

The relationship between linguistic structure and underlying conceptual representations, that has been addressed already, is the traditional field of linguistic semantics I will therefore give a short summary over different theories here that might be relevant for the theoretical basis of the present study

Traditional linguistic semantics assumes formal, clear analytical ries Within the formal approach of feature semantics (Katz 1972; Wierz-bicka 1996), word meaning is modelled with the help of a finite number of semantic primitives (the concept BACHELOR would show the features +HUMAN, +MALE, +ADULT, -MARRIED) In order to identify these primitives, a strictly structuralist analysis is used: The semantic fields

catego-within the mental lexicon are structured according to lexical relations, so that the single elements of a field come to stand in semantic relationships

of homonymy, antonymy, hyponymy etc to each other Besides that, the single element is internally structured in that it is distinguished from other elements by its distinctive features What follows from this is that word meaning can be described as being built up of smaller entities All other as-sociations that are not triggered by the semantic primitives are assumed to belong to encyclopaedic world knowledge

The criticism of models of feature semantics starts out from the vation that humans do not distinguish phenomena in the world as strictly distinguishable on the basis of a number of features Feature semantics cannot account for the fact that people often experience difficultiy in sort-ing phenomena into clear categories; hedging expressions like ‘this thing is

obser-a kind of X’ or ‘this thing resembles obser-an X’ hint obser-at this fobser-act, obser-as well obser-as obser-a range of experiments, e.g the cups and mugs experiment by Labov (1973) and a range of experiments by Rosch (1975, 1977, 1999) This led to the formulation of prototype semantics (Rosch 1975, 1977, 1999) Prototype semantics investigates how humans construct meaning through category building, and describes this system of categories The experimental evi-dence suggests that humans perceive the world as structured, which is mir-rored in their cognitive categories, and not the number of necessary distinc-tive features determines the meaning of a phenomenon, but the features that make it to a more or less good representative of a category Thus, a

while a pug or a chihuahua are untypical representatives This view grates concepts of gradience and fuzziness (Aarts 2007)

inte-For linguistics these findings set the stage for the integration of bouring disciplines, particularly cognitive psychology, because they made

neigh-it obvious that for linguistic structures the same basic mechanisms apply as for other forms of cognition It is crucial here that prototypes cannot be ex-plained without the integration of world knowledge, because what is con-sidered to be a good or a bad member of a category is often to a high de-gree culture specific and grounded in personal experience; so the strict distinction between linguistic knowledge and encyclopaedic knowledge is suspended

A similar approach is taken in frame semantics (Fillmore 1975, 1977, 1982; Fillmore and Atkins 1992) and script theory (Schank and Abelson 1977), which introduce basic ideas from newer branches of Cognitive Lin-guistics Here, it could be shown that humans store prototypical and sche-

matic situations (frames) or typical scenarios (scripts) on the basis of their individual experiences, which guide processes of categorisation, expecta-tions, and actions The meaning of single words can only be understood in the context of these frames and scripts, because the connected field of ex-perience reveals different aspects of lexeme meaning So, the question which features of a concept become meaningful can only be answered by embedding them into larger relational schemes; denotation and connotation

of a word fall into one

Evans and Green (2006: 161) illustrate this fact with the following amples:

ex-(a) The child is safe

(b) The beach is safe

(c) The shovel is safe

In these examples, the word safe shows a range of different meanings,

which can only be inferred from context Whether the subject of either example sentence is perceived as being potentially endangered, or on the contrary as a potential source of danger, depends on a language user’s world knowledge about children, beaches and shovels:

[T]he conventional meaning associated with a particular word is just a

‘prompt’ for the process of meaning construction: the ‘selection’ of an

ap-propriate interpretation against the context of the utterance [ibid, bold face

in original|

So, single lexemes trigger processes of conceptual meaning tion, but do actually contain less fixed meaning than is traditionally as-sumed They are the key that opens up the path through the conceptual con-tent that is attached to a particular concept, of which different features can

construc-be foregrounded in different contexts

4 Language and the format of thought

An important issue that needs to be taken up in the discussion of the tionship between language and conceptual processing is the actual format

rela-of thoughts If it should turn out that thoughts are dependent on a linguistic format, then language would be insolubly interconnected with processes of thinking, and thought would not be possible without language, so that lan-

guage might in fact be the actual cause for our thinking abilities If, on the other hand, thought is possible without language, linguistic and conceptual knowledge can be assumed to be separable on a cognitive level, and this has consequences for the setup of the empirical study presented here We find a range of different theories about the format of thinking and the inter-relation between language and conceptual content, which I summarize in the following

4.1 Thinking for speaking and linguistic relativity

The view that language and thought are insolubly interconnected has a long tradition Going back to von Humboldt (ca 1919), it has been assumed that the native language is the actual format we think in, so that the language

we speak is the format that codes our thoughts

Following from this is that language contains a theory of how speakers mentally represent the world, so that each natural language can be seen as a system for the conceptualisation of reality Because certain conceptual structures are linguistically conventionalised in a given language, certain ways of mental representation are conventionalised as well Cultural con-ventions are thus mirrored in linguistic structures and patterns, and their repeated use reinforces the conventions in return

As a result it has been hypothesised that speakers of different languages must think differently, because languages with different grammatical struc-tures and semantic word boundaries serve as different classificatory sys-tems that control a person’s perception of the world

A closely related implication to this view is known as ‘linguistic ity’, traditionally linked to the work of Edward Sapir and Benjamin Lee Whorf (Sapir 1949; Whorf 1956) According to the strong version of this view, which is often called ‘linguistic determinism’, language does not only provide a framework for classification of thoughts, but actually im-poses restrictions on possible ways of thinking and reasoning, so that we are only able to see reality through our linguistic filter

relativ-This view equates thought with language and has met strong criticism, mainly for two reasons: On the one hand, there is evidence of thinking without language (cf the results obtained by Barsalou et al 2003, see next section), on the other hand there is the fact that linguistic structure never fully comprises the semantic richness of a message (see section above)

Linguistic determinism can today be said to be largely discredited, on the grounds that it is seems unlikely that linguistic categories embody the only kind of influence on our thought processes; we all share the experi-ence that we can create both new words and new concepts Nevertheless, the basic idea that the linguistic structure we use has an impact on the way

we conceptualise has met renewed interest in Cognitive Linguistics, both under the names ‘linguistic relativity’ and ‘thinking for speaking’ (Slobin 1987; 1996; 2000; Boroditsky 2001, Li and Gleitmann 2002)

Still, linguistic relativism has been criticised harshly (e.g., Pinker 1995) One of the main flaws addressed is its methodological circularity – Slobin (1987; 1996; 2000) for instance only uses linguistic data to infer results about non-linguistic thinking What he can show is only that grammar and thought correlate (which indeed does not come as a surprise when verbal data are used as the source of information), but fails to prove that language really has an impact on cognition Pinker parodies this reasoning in the fol-lowing circular argumentation:

[They] speak differently so they must think differently How do we know that they think differently? Just listen to the way they speak! (Pinker 2007: 50)

Nevertheless, in studies that have taken regard to this methodological issue and used non-linguistic data, it could be shown that different lan-guages embody different conceptual classifications of the world, for in-stance in that semantic contrasts expressed by grammar lead to differences

in perspective and foregrounding of certain aspects of concepts (Lucy 1992; see also overview in Pavlenko 2005), while here too, it has to be concluded that linguistic structure is not the only factor that influences thought

4.2 Sociocultural theory

Because it also addresses the relationship between thinking and language, and has become a widely used theoretical framework, I want to mention

Vygotsky’s socio-cultural theory of mind (Vygotsky 1978, 1986) in this

context It has a strong developmental focus and tries to explain how dren develop their cognitive abilities In that, it stresses the social aspect of thinking and perception, and describes the role that language as a semiotic tool plays in the internalisation of concepts and thus the conceptual and at the same time social development of humans

chil-Sociocultural theory can generally be seen in the tradition of the Whorf hypothesis (Vygotsky 1986: ch 1) but puts the focus on some spe-cial concepts One central assumption of this theory is that

Sapir-[e]very function in the child’s cultural development appears twice: first on the social level, and later, on the individual level: first, between people (in- terpsychological), and then inside the child (intrapsychological) (Vygotsky 1978: 57)

According to this assumption, higher thinking processes are originally derived from structures and functions of social interaction; these are even-tually internalised in that they manifest themselves in internal mental struc-tures

Because of its origin in social and cultural structures, these mental structures mirror the socioculture in which the individual is raised Special

emphasis is laid upon the Zone of Proximal Development: This concept

re-fers to the internalisation of culture (which equals with the establishment of consciousness and learning processes) by means of semiotic mediation, and assumes that the process of internalisation optimally takes place when cog-nitively more mature individuals interact with immature ones (e.g., in teacher – student relations) This relationship provides situations in which the expert leads the novice to insights that he or she would not have been able to reach alone Language plays a central role in this mediation process:

It is the most elaborated semiotic system that humans use, and it is a vital tool in social interaction In that, it provides the interface that allows inter-action between an individual’s mind and the outside world, it is the cause for higher order thinking, and stays indelibly interconnected with the men-tal structures that it helped to create

Vygotsky does not assume that language is necessary in order to think

in general (e.g., he assumes a prelinguistic phase in the development of children, in which nevertheless thinking takes place), but that a close con-nection emerges in the run of the ontogenetic development of an individ-ual In his considerations about word meaning Vygotsky characterises the character of meaning as basically procedural and changing, ‘a continual movement back and forth from thought to word and from word to thought […] Thought is not merely expressed in words; it comes into existence through them’ (Vygotsky 1978: 57)

In this context, Vygotsky uses the concept of inner speech which he

de-fines as a kind of condensed and abbreviated speech the speaker does not

utter but produces for himself in the interaction with the world Inner

speech and external speech are assumed to be two different types of

lan-guage because they have different functions: External lanlan-guage is believed

to be an externalisation of thoughts, while inner speech is a translation of

words into thoughts So, Vygotsky assumes that thought and speech unite

into verbal thought (Vygotsky 1986: ch 1), and that inner speech is closely

connected to it

Sociocultural theory does not so much answer questions of how tures of thought and cognitive processes can be modelled, which I will fo-cus on in more detail further on Still, it can be combined with approaches from cognitive science without any greater friction (Frawley 1997) I re-gard this theory as a supplement to the cognitive models I will describe in the next section, and particularly appreciate the stress it lays on the social context in which the individual stands

struc-Even when the focus is not explicitly on linguistic structure, the fact that according to this theory abstract thoughts in particular are dependent

on language might have an impact on the conceptualisation of the different foci in teaching situations and shall be kept in mind here

4.3 The format of thought

The issue of the format of thinking and the role language plays therein has been the cause of a long debate in cognitive psychology, known as the ‘im-agery debate’ (overview in Kosslyn 1994; Tye 1991)

Generally, there have been two contradictory positions: One, nantly represented by Fodor (1995), states that thoughts are coded in a

domi-‘language of thought’, which he calls mentalese Although mentalese is a

language-like structure that makes use of propositional entities, it is not the same as the speaker’s mother tongue: It is an abstract format in which propositions are combined on the basis of a mental syntax So, verbatim word order of a specific language is not encoded, but sense relationships The English and German sentences ‘Bill ate 20 cookies’ and ‘Bill aß 20 Kekse’ are represented by the same single proposition with the following elements:

Time – PAST

Relation – EAT

Object – COOKIES

Agent – BILL

The same proposition is contained in the sentence ‘20 cookies were eaten by Bill’ and its translation into German, French, or any other lan-guage

According to this theory, thoughts are represented in this abstract positional code, and have to be translated into the language actually spoken

pro-by the thinking subject This theory stresses the innateness of general nitive principles, assumes that human thinking is based on the linking of conceptual entities and is not dependent on any natural language system in particular

cog-Contrasting this, the dual coding theory by Paivio (1971; 1986) takes an intermediate position between the two extremes ‘thinking is possible with-out language’ and ‘thinking is only possible because of language’ It as-sumes that there are different cognitive codes for verbal and non-verbal in-formation and gives equal weight to both kinds of processing Paivio assumes that both subsystems process information simultaneously but have different functions: Verbal coding takes place when a person is confronted with verbal information, e.g., when reading, or listening to speech If the input is visual, then the representation is coded in a visual-spatial represen-tation Contrary to the verbal coding system, the visual-spacial system is not sequential but can process different dimensions of information, such as size or colour, at the same time Both subsystems can be activated inde-pendently, but depending on the semantic content of the input, the other system can be activated in a crosstalk-like process So, for example, when a person reads a description of a house, the linguistic subsystem is activated, but simultaneously a mental image of a house is formed by the visual-spacial subsystem By this, dual coding of information can take place Evidence for visual-spacial coding stems furthermore from research on mental models (Bucciarelli and Johnson-Laird 2005; Craik and Lockhart 1972; Garnham and Oakhill 1996; Gentner and Stevens 1983; Johnson-Laird 1983; Oakhill and Garnham 1996) which proves that humans build mental representations in an analogue, visual-spatial, mode These repre-sentations structurally resemble the structure of the perceived phenomenon Mental model theory can explain the fact that humans are able to predict features and effects of complex entities: Because we experience phenom-ena as systems, we can mentally simulate and modify them (e.g by mental rotation) which in turn enables us to make predictions and hypothetical as-sumptions, e.g about cause-effect relationships Inferences and analogies play an important role here Representations of this kind are not assumed to consist of discrete elements, as for example the language of thought hy-

pothesis by Fodor would assume, but form a structural and functional tity that is holistic in nature

en-Recent models of enactive theories of thought activity (Barsalou 1999, 2003; Barsalou et al 2003) follow up the basic assumptions of the mental model approach in that they model thinking in an analogue format Enac-tive models move away from the theory that knowledge is stored in dis-crete symbolic entities and stress the constructive and procedural character

of thinking Here it is assumed that conceptual processing is based on dality-specific systems of auditory, tactile, olfactory and visual-spacial processing, so not only is a dual coding assumed, but a multitude of sen-sory formats Information is not transferred into amodal symbols, but mo-dality-specific states are stored in memory in their modal form (for empiri-cal evidence see Barsalou et al 2003) Although this theory is not concerned with linguistic representations whatsoever but models concep-tual formats from a cognitive perspective only, the evidence supports the view that there is conceptual content that is not connected with language

mo-5 Conclusions and summary

As this overview indicates, the relationship between language and thought has been modelled in various ways and with different purposes

As Prototype Theory suggests, the modular-linguistic approach with its distinction between objectively describable semantic knowledge and sub-jective encyclopaedic knowledge cannot account for a range of experimen-tal results It seems to make more sense to explicitly integrate the construc-tive character of any kind of knowledge into a model of linguistic knowledge A range of experimental results supports the assumption that there is no general distinction between linguistic meaning and encyclopae-dic meaning It is thus not necessary to use specific criteria for describing semantic knowledge, as is indeed assumed in newer linguistic theories that integrate concepts from cognitive psychology into linguistic modelling, such as Construction Grammar or Cognitive Linguistics (Goldberg 2006; Evans and Green 2006; Geeraerts 2006) These branches do not describe linguistic knowledge as being in need of a distinct language-specific for-malism Instead, they stress that conceptual structure and processes of categorisation are sufficient to describe linguistic meaning

Furthermore, research from cognitive psychology suggests different mats for thought, but none of the dominant theories assumes natural lan-

for-guage to play a nesessary role in the coding of thought Empirical evidence makes it likely that there are a range of modes of mental representations, among them visual, auditory, and linguistic ones, the latter being connected with the linguistic language systems of which the speaker possesses knowl-edge (Barsalou et al 2003; Ericsson and Simon 1993)

Although the actual format of thinking seems not to be bound to guage, language can be assumed to play an important role in building up ab-stract knowledge categories, in the ontological development into a mature social being, and although people using different languages are in principle able to think any thought, the association of thoughts with specific lexical and grammatical structures might lead to differences in conceptualisation and perspective

lan-So, semantic concepts in the mental lexicon have their origin in ual and social experiences Between encyclopaedic and linguistic informa-tion stored in a speaker’s mind, a close interrelation can be assumed, because linguistic concepts belong to our experience of reality and have a big impact

individ-on the individ-ontogenesis of each individual’s system of cognitive cindivid-oncepts The sult is a closely intertwined relationship between cognitive and linguistic concepts Conceptual structure does not necessarily form a part of linguistic structure; language, on the other hand, serves as a highly sophisticated semi-otic tool for the encoding of meaning, thus for thinking

re-An immediate consequence from this is highly interesting for the gation of how linguistic structure influences conceptual content: When we use lexical items and syntactic constructions in order to express a thought,

investi-we automatically feature the conceptual structure that the lexicon and mar of that particular language encodes; by this, certain aspects are put in the background, while others are pushed into the focus of attention

gram-On this basis a distinctive definition of linguistic and conceptual edge on pure functional grounds seems useful I will therefore define linguis-tic knowledge as knowledge underlying the semiotic system of language This is knowledge that enables humans to formulate thoughts in language and to interact with others verbally It refers to those elements of knowledge that contain knowledge about speech sounds, lexemes and the possible ways

knowl-of connecting them to phrases and sentences, but even about stylistic and rhetorical rules, about different discourse forms and metalinguistics knowl-edge Contrasting to this, the definition of conceptual knowledge I will use here is the following: Conceptual knowledge refers to any kind of knowledge structure that is necessary in order to construct meaning, and which is not concerned with the semiotic system of language

lan-is thinking? How can we grasp what happens on a mental level when we

plan and focus our attention on different mental concepts?

When we try to describe processes of thinking in everyday language, we can choose from an overwhelming list of vocabulary: We consider, re-member, learn, plan, compare, know, recognise, imagine, calculate, trans-late, analyse, evaluate, solve problems – a list that can easily be continued (cf Fortescue 2001) But which mental processes are actually described by these verbs? What is, cognitively speaking, the difference between them?

Is it possible to define them distinctively at all?

Some further thought reveals that not every expression seems suitable for a cognitive distinction between processes of thought Let’s take a look

at the verb ‘understand’: On the one hand, it can refer to the result of the mental act in which information already stored in memory is retrieved; formulated in a more abstract way, it refers to processes of reconstruction and activation On the other hand, ‘understand’ can refer to processes in which new knowledge elements are added or new connections between concepts are established A paraphrase for this process would be ‘learning’ This example suggests two notions: Firstly, that there are different lev-els of description that can help define what we actually mean when we speak of mental processes; secondly, when we analyse what verbs of men-tal processes actually depict, we evoke a theoretical framework about what cognition is and how it works

All recent theories of cognition, including theories of language tion (Grotjahn 1997), regard human behaviour as complex, that means as the result of a range of simpler processes (see e.g Palmer and Kimchi 1986 and Massaro and Cowan 1993; for overviews in the area of language ac-quisition, see Grotjahn 1997 and Robinson 1995) Mental events are de-scribed functionally as events in which information is processed Informa-tion, according to Massaro and Cowan (1993: 386), is the mental

acquisi-representation that a person constructs, either by individual interpretation

of external data, or by activating existent cognitive structures from ory Information as this internal mental representation is to be understood within a constructivist framework, which means that it is a result of a sub-jective and individual interpretation process; it cannot directly be linked to any objective reality, and can differ from person to person Thus informa-tion can be understood as a specific state of the cognitive apparatus

mem-Within this view, every process can be regarded as a three-step module that consists of the following components (Palmer and Kimchi 1986: 40):

1 the information that forms the starting point

2 the operation that is performed on the initial information and which modifies it

3 the resulting information as a result of the processing event

Each of these three processes provides information that is necessary for the next one This, however, does not imply that mental processing can only be modelled serially; parallel processing is also possible Neverthe-less, within each basic module, the order is logically obligatory

Starting from here any processing event can in principle be described as the sum of sub-processes which makes it possible to describe it on different levels of abstraction

In cognitive psychology, two different kinds of theoretical description

of cognitive processes are usual: One uses the physiognomy, the mental

‘hardware’, as the reference point for the description; the other is cerned with the ‘software’ that runs on it, that is which functions the proc-esses have in a computational system (Palmer and Kimchi 1986: 49f), without taking the physiological basis into account

con-On the level of a physical description with the highest degree of detail, cognitive processes are described as neurological activities The mental state in question corresponds here to the interplay between different bio-chemical processes which consist in their most basic elements of coordi-nated activity or inactivity of neurons, or in Palmer and Kimchi’s (1986: 42) words:

Such descriptions are ultimately reducible, at least in theory, to quantal penings among countless quarks, or whatever the most microscopic level of physical reality happens to be

hap-Theories at this end of the continuum use terminology such as tion’, ‘stimulus’, ‘activating’, ‘inhibiting’, ‘connecting’, ‘disconnecting’ (cf Bruner 1966, Dörner 1999: 61; Rumelhart and Norman 1978) More dominating in cognitive science are, however, theories about the ‘cognitive software program’:

‘percep-There is, in addition, a strategy for dealing with the conditions under which to stop decomposing, and that is simply not to worry about it Many

IP [information processing, L.H.] psychologists, perhaps even most of them, are quite satisfied to work at a level that is well above any ultimate

‘primitives’ and leave theorizing at such a ‘low’ level to other theorists (Palmer and Kimchi 1986: 49)

These theories use more complex and more functional descriptions and employ terminology that coincides with everyday language use Mostly, terminology like ‘thinking’, ‘problem solving’, ‘language comprehension’,

‘judging’, ‘drawing conclusions’, ‘learning’ or ‘memory activation’ is used (e.g., Anderson 2000; Eysenck and Keane 2000; Solso 2005) Cognitive processes here are defined independently of their physiological basis Be-cause of the functional description, there is often no reason for cognitive psychologists to draw clear lines between them; therefore, terminology can overlap

From this follows that a dichotomic view of process descriptions on the material basis on the one side and functional descriptions on the other does not seem useful; rather, the information processing paradigm that I will use here suggests that both positions should be considered to be the end points

on a continuum in which complexity and abstractness increase tively

cumula-This view is illustrated in Figure 1 The processes indicated above the axis illustrate exemplary taxonomies, with an increasing degree of com-plexity from the left to the right The processes indicated here range from biochemical descriptions on an extreme micro level of detail (to the left) to

a more functional description that does not take the material basis of tive processes into account at all (to the right)

cogni-Note that the processes to the left can be regarded as cumulative ments of the processes to their right At the same time, even within the tax-onomies there can be different degrees of cognitive complexity:

ele-So, the process of ‘comparing’ is not always cognitively the same, ing on the complexity of the objects of comparison, and how it is anchored

depend-in the knowledge store of the subject

Figure 1 Cognitive processes on a continuum of descriptive levels

As Figure 1 shows, functional and neurological standpoints can be linked smoothly if we take the theoretical view that every cognitive activity can ultimatively be described in terms of neurological microprocesses Both are compatible with each other

Note that in such a decomposition of operations, the sub-components are not only quantitatively simpler, but show a different quality (Palmer and Kimchi 1986): In a decomposition of the process ‘search in memory’, for instance, the following range could result: a ‘perception of external stimulus’, b ‘comparison of perceived information with content in mem-ory’, c ‘decision that congruence is the case’ and d ‘conditional control (if congruence, give out result, otherwise go on searching)’ None of these sub-components bears any resemblance with the overall process ‘search in memory’ – it is a result of the combination of them Speaking of a ‘search’

is therefore actually only reasonable if we take a higher level of ness

abstract-Thinking can be defined as situation-specific activation and tion of different semantic relationships between pieces of information in memory According to this, the solving of meaning-focused tasks can be described as a sequence of cognitive processes, which lead to the estab-lishment of meaningful relations between knowledge structures

construc-In order to develop the analytical tools for the present study, I now need

to decide at which level of abstractness mental processes can be analysed

Perceive Learning Judging Problem Solving

Basic cognitive

processes as

neu-rological events

Basic cognitive processes as functional entities

Neurons

firing

Activating Inhibiting Linking Dissolving

Comparing Concatenating Abridging Transform

Functional process clusters that characterise specific performance

with enough illustrative and explanatory power In order to do this, I will turn back to the actual processes to be investigated

2 Task solving as problem solving

On the basis of the theoretical grounds established in the above paragraphs,

it is now possible to develop a clear understanding of how the mental ity can be captured Because I want to investigate how learners mentally deal with content-focused activities while using a foreign language, I will use the concept of task to capture all activities that a learner can go through

activ-in these situations This concept is embedded activ-into the approach of problem solving, a well-established branch of applied cognitive psychology Prob-lem solving research investigates how humans solve complex tasks for which they do not have any immediate solutions, and this provides a suit-able framework for the kind of activities I will focus on here

In order to classify a situation as a problem, three components are essary: a a beginning state that is perceived as being in some way unsatis-factory, b a desired goal state and c a lack of knowledge of how to imme-diately change a into b From this follows that automatised activities are not subsumed under problem solving activities, because for them, only the feature of goal directedness would suffice

nec-On this basis, the process of problem solving can be subdivided into two macro processes, namely a process in which the problem is identified, and a search process

Before the solving activity can be started, a mental representation of the problem has to be created This means that the starting point needs to be perceived by the individual as being unsatisfactory, and a mental represen-tation of goal state has to be constructed

Processes of problem solving show central features that make them compatible with the information processing approach I have presented in the above: Problem solving is goal-directed, the problem is divided into subordinate goals, and operations are used that transform the actual prob-lem state in a step-by-step manner on the way to the desired final goal state So, in the process of being solved, a problem takes different shapes The total of these states is called problem space It can be characterised by the sum of all possible states and operations that can be applied in order to reach the goal state Thus the process of problem solving can be regarded

as a search in the problem space by which subordinate problems are