In the previous chapters, I have discussed ways in which language attrition may affect the lexicon of a language. I have pointed out that the lexicon of the average speaker consists of thousands of items, changes quickly even in predominantly monolingual communities, and often overlaps a great deal between languages. It is therefore not surprising that we often find indications of change, both due to CLI and to attrition, in the lexicon of attriters.

The situation is quite different when it comes to those things that we normally consider part of linguistic structure: phonology and phonetics, morphology and syntax. Here, we are dealing with much smaller sets of items. For example, most languages have only a few dozen phonemes (phonemic inventories vary between as few as fifteen and as many as eighty items according to Crystal, 1980). Similarly, most languages have a rather limited number of inflectional endings, although there are some exceptional cases where this set is much larger (some notable examples are discussed by Mithun, 1998 and Senft, 1996). That means that while a speaker may only hear or use a certain word a few times over many years, and may even know words which he or she has never actually produced him- or herself, the sounds, grammatical morphemes and sentence patterns present in a language will usually all be heard (or read) and produced multiple times in any interaction.

In chapter 1 I introduced two native speakers of German who had both grown up using their first language exclusively in their daily lives until age 13, when they moved to an English-speaking environment. Nearly six decades later, one of them – Albert L. – retained a fairly fluent command of German, while the other – Gertrud U. – had great difficulties expressing herself in that language.

This discrepancy raises the question of what attrition is and where it begins. Is there a clear and perceptible boundary in linguistic skills which marks the distinction between an attriter and a non-attriter? Can we define it in terms of time – can we, for example, say that it sets in after ten years? Obviously not, or not exclusively: Gertrud U. and Albert L. had both been living in an English-speaking environment for approximately the same length of time.

Cast your mind back to some of the differences between the excerpt from the narratives of Gertrud U. and Albert L. In part I above, I pointed out that they were different from each other on a range of features (see Table 7.1).

When you see this kind of difference between two speakers who have lived in emigration for more or less the same number of years, the first intuition about how these differences came about is likely to be that Albert L. had probably had more opportunities to ‘practise’ his German over the years than Gertrud U.

It is an apparently reasonable and natural assumption that speakers who continue to use a language remain fluent in it while speakers who don't forget it. Human knowledge and memory need to be maintained to prevent them from deteriorating: the more often knowledge is called upon, the easier it becomes to recall. If, on the other hand, a certain memory is not activated for a long time, it will eventually become difficult to retrieve.

When Barbara Köpke and I first became interested in the study of language attrition, we felt that one of the main problems of the field was the fact that there was no consistent methodology available for attrition studies. It seemed that everyone who had ever looked at attrition had invented their own approach, tests and questionnaires. This not only seemed like a huge waste of time but also contributed to the fact that little overall progress was made in the common understanding of how attrition takes place: no two findings could be compared to each other, and so nothing could ever be generalized beyond the scope of an individual investigation. We therefore decided to target our collective effort at providing a methodological framework that people would be able to adapt to different languages and different contexts.

The data collection instruments and experiments presented in this section are the outcome of these efforts. They were largely developed between 2003 and 2009 in the European Graduate Network on Language Attrition, a group of young researchers and Ph.D. and M.A. students working on language attrition in a variety of languages and settings. This group met on a regular basis and during those meetings we selected various tests and evaluated their potential for application in language attrition research. This means that for all of the tasks described here, previous findings from attrition in a range of first languages are available.

The transcription and tagging of free speech, as discussed in the previous section, is by far the most difficult and time-consuming way of coding linguistic data. Data elicited by controlled experiments or tests, on the other hand, can usually be coded relatively quickly once you have set up your database. During my own fieldwork I was usually able to enter all of the data (the sociolinguistic questionnaire, two Verbal Fluency Tasks, five C-Tests in each language, two Wug-tests and two Can-Do Scales) into my database while the next participant was completing the written tasks.

Before you start setting up this database, however, there are some important considerations and decisions to make. These mainly concern the following issues:

1. Types of data

2. Missing data

3. How to report and visualize your results

TYPES OF DATA

The first thing that you will be taught in any statistics course is that numbers are not always numbers: the mathematical and statistical operations you are allowed to perform crucially depend on how the numbers within the scale that you are applying relate to each other.

Chapter 9 acquainted you with the independent variables or predictors which play a role in the process of L1 attrition. These are the external factors which may impact on the processes of language attrition that we observe among bilingual migrants. In chapter 10, aspects of dependent variables will be discussed. These are those elements in an experiment which the researcher attempts to elicit, and which we expect will change or have changed as a result of differences in the predictors. Your dependent variables are therefore those linguistic features for which you expect to find differences between the experimental and the control population as a result of the process of attrition (their vocabulary or specific fields of it, their grammar, etc.). Again, you will make life a great deal easier for yourself if you try to be as specific as possible as to which linguistic features, and which aspects of these features, you are planning to investigate.