Lade Inhalt...

Switching Costs and Language Proficiency in Bilinguals and Trilinguals

Hausarbeit 2013 15 Seiten

Anglistik - Linguistik


Table of Contents

1 Introduction
1.1 General Introduction
1.2 Key Concepts and Methods: Inhibitory Control and Stroop Test

2 Analysis of Data
2.1 Meuter et al. (1999): Bilingual Language Switching in Naming: Asymmetrical Costs of Language Selection
2.1.1 Assumptions and Predictions
2.1.2 Participants, Materials, and Procedure
2.1.3 Results and Main Findings
2.2 Marian et al. (2012): Multilingual Stroop performance: Effects of trilingualism and proficiency on inhibitory control
2.2.1 Assumptions and Predictions
2.2.2 Participants, Materials, and Procedure
2.2.3 Results and Main Findings
2.3 Comparative Analysis of Both Studies

3 Conclusion


List of figures

Figure 1: Overall RT (in milliseconds) on nonswitch and switch trials

Figure 2: Effects of practice in language switching

Figure 3: Number of disfluencies on colour-word congruent vs. incongruent trials

Figure 4: Performance on stimulus/response language matched vs. mismatched trials

List of abbreviations

Abbildung in dieser Leseprobe nicht enthalten

1 Introduction

1.1 General Introduction

When we communicate, we have a goal in mind and have to choose the words that best suit that goal. For people speaking more than one language, this is even more challenging as they usually have a much greater vocabulary to choose from than monolinguals and have several words for the same object. Therefore, one of the most discussed questions in bilingual speech production has been lexical access, i.e., how bilinguals manage to select the appropriate word. Amongst others, much research has focused on two questions that are of particular interest for us here: first, what is the relationship between a bilinguals two languages, i.e., what effect does the constant switching between these languages have (known as “switching costs”) (Meuter & Allport, 1999; Bialystok, Craik, Klein, & Viswanathan, 2004); second, what is the role of language proficiency in this relationship (Meuter & Allport, 1999; Costa & Santesteban, 2004), since most bilinguals will have different proficiencies in their two languages.

Yet, as our world becomes more and more globalised, the number of people who speak more than two languages increases, thus making research into trilingualism an attractive enterprise in order to see what the effects of three languages in one mind might be. In recent years, there has been an increase in studies investigating, amongst other aspects, cross-linguistic influences on third language acquisition (Molnár, 2008; Tremblay, 2006) or word recognition in trilinguals (Lemhöfer, Dijkstra, & Michel, 2004). However, in general, such research is still relatively scant. Trilingualism is often seen as merely a special case of bilingualism and most of the findings for bilingualism have been regarded as extendable to trilingualism, often not even making a distinction between bilingualism, trilingualism, and multilingualism in general (Clyne, 1997; Hoffmann, 2001). Some studies have tried to determine whether trilinguals have special competences that bilinguals have not, yet without conclusive results (Edwards & Dewaele, 2007; Hoffmann, 2001). Yet, it would be very interesting to know whether or not trilingualism is really just ‘bilingualism plus one’, i.e., the findings regarding lexical access for bilinguals extend to trilinguals.

This paper will therefore investigate whether the findings for switching costs for bilinguals’ two languages extend to trilingualism, and to what extent these costs are influenced by language proficiency. In a first step, some general concepts and methods relevant for the papers discussed here will be reviewed. In chapter 2, two studies dealing with the issue of switching costs in bilinguals and trilinguals, respectively, will be presented and compared. In the final chapter, I will try to answer my research questions, but also to touch on alternative accounts for the findings presented here.

1.2 Key Concepts and Methods: Inhibitory Control and Stroop Test

It is still not certain how exactly the lexical retrieval in bilinguals works. One possible explanation could be given by the Inhibitory Control model (Green, 1998), which states that lexical competitors need to be suppressed on the cognitive level, the allocation of resources to this process causing latencies and inaccuracies in response (=inhibitory costs). This model is widely accepted and many studies, amongst them the two studies that will be discussed here, used it as a tool to explore the issue of language switching.

To understand this model, it is necessary to outline the problem it tries to explain, i.e., lexical access. Obviously, most words contain more than one syllable, though many words share a certain amount of the same syllables, e.g., mis- can be part of the word mistake, misty, missed, etc. In order to know which word is meant in each particular utterance, there is a constant process of narrowing down the competing options when producing and comprehending language, excluding more and more options while additional syllables are uttered (Marian, Blumenfeld, Mizrahi, Kania, & Cordes, 2012). For multilinguals, this process becomes more complicated since research strongly suggests that all their languages are co-activated, i.e. lexical access is non-selective (Lemhöfer, Dijkstra, & Michel, 2004; Costa & Santesteban, 2004) and hence there are even more possibilities as to which word is meant. For instance, ang- could not only result in English angel or angina, but also in German angeln or Angeber, as well as French ange or anguille, to name but a few. Hence, there must be a process in place to suppress the incorrect words and select only the correct one.

For that reason, much research has concentrated on how bilinguals still manage to select the correct word. Research on how executive functions are controlled in general (Norman & Shallice, 1980) suggests the ‘Supervisory Attentional System’ (SAS) as a possible model, saying that for each stimulus a certain action or thought scheme becomes activated, depending on whether that impulse is rare or frequent. Based on this assumption and adapted to linguistics, Green (1998) proposed the Inhibitory Control model (ICM) in order to model how bilinguals solve the word selection issue mentioned above. This model states that during production of speech a communicative goal is set and according to the frequentness of that goal, the task schemes are automatically activated (frequently-performed) or task activation depends more on the SAS (rarely-performed) (Marian, Blumenfeld, Mizrahi, Kania, & Cordes, 2012). Each language is represented by a language scheme and usage of one language needs inhibition of the non-target language scheme.

This is where the Stroop test relevant for the two studies presented here comes into play. It is a test employed to measure inhibitory control and has been used for many studies on monolinguals (J.G. & Gilpin, 1978; Warren & Marsh, 1978) as well as bilinguals (Costa, Albareda, & Santesteban, Assessing the presence of lexical competition across languages, 2008; Zied, et al., 2004). Participants are asked to name the ink colour of coloured words with the printed word either matching the ink colour (congruent condition) or not (incongruent condition) (Stroop, 1935). Usually, responses are slower and less accurate for incongruent in comparison to congruent trials, which is commonly referred to as “Stroop effect”. The explanation given before could explain the classic Stroop effect: the colour-naming scheme and the reading scheme compete with each other, a conflict that needs to be resolved by the SAS, thus draining resources from other processes and resulting in slower reaction times and less accuracy for incongruent trials. In a cross-language trial, according to the ICM, the SAS needs even more resources to resolve the lexical conflict between the different available languages, hence causing the between-language Stroop effect in addition to the within-language Stroop effect (Marian, Blumenfeld, Mizrahi, Kania, & Cordes, 2012). This also means that both within-language and between-language inhibition can act independently, though they rely on the same mechanisms (Marian, Blumenfeld, Mizrahi, Kania, & Cordes, 2012).

2 Analysis of Data

2.1 Meuter et al. (1999): Bilingual Language Switching in Naming: Asymmetrical Costs of Language Selection

2.1.1 Assumptions and Predictions

The aim of the study was to find out how bilinguals switch between their languages and what the effects of that switching are. Their assumption was that the processes bilinguals employ for selecting their language are similar to processes “responsible for the control tasks set in other monolingual and/or nonlanguage task domains” (Meuter & Allport, 1999, p. 25). A competing assumption is that any language/linguistic processes (including language selection) are unique to human beings and, therefore, are separate from any other task control processes (Meuter & Allport, 1999). One goal of the study was to refute the later assumption. Research into these non-language processes had shown that task switching costs are determined by the task preceding the switch, not the task that is switched into (Allport, Styles, & Hsieh, 1994) (Meiran, 1996). Apparently, switching from a dominant task into a weaker task needs more suppression than vice versa (“Task Set Intertia hypothesis”) (Meuter & Allport, 1999, p. 27).

The study tried to transfer these findings to bilingualism and predicted that there would be asymmetrical switching costs between L1 and L2: when switching from a stronger (usually L1) to a weaker (usually L2) language, the costs would be larger when switching from L2 into L1 than from L1 into L2. The degree of the switching costs is predicted to depend on the relative proficiency of the two languages: the more equal their proficiency is, the smaller the switching costs should be and vice versa (“Relative Strength hypothesis”) (Meuter & Allport, 1999, p. 28).

2.1.2 Participants, Materials, and Procedure

The participant group consisted of 11 women and 5 men, aged between 23 and 44 years. All participants were bilinguals, spoke English as either first or second language, and judged their second language skills as reasonable. An effort was made to select different European languages in order to avoid systematic errors in certain language combinations; the various second languages were Portuguese, Spanish, Italian, German, and French.

The task was performed twice and consisted in reading aloud and as fast as possible a list of 20 numerals in each of the participants’ languages. The numerals ranged from 1 to 9 and were given in sequences, their length ranging unpredictably from 5 to 14 items; the total number of sequences was 200, resulting in approx. 2000 trials per participant. Participants were presented with all of these 200 lists, though their order varied randomly. Each numeral was displayed on a VGA screen in either a blue or a yellow rectangle. The colour of the rectangle indicated in which language the participants were supposed to answer; a colour switch therefore resulted in a language switch, the chance for that being 0.3, with the number of switches per sequence ranging from 0 to 4. The display of the numerals was activated through a voice key and the participants had up to three test runs with practice lists to get accustomed to using it correctly. A computer recorded the latency between stimulus onset and response.

2.1.3 Results and Main Findings

Three factors were important for the analysis: first, Response Language (L1, L2); second, “Run Length: the number of consecutive naming responses consistently in one language immediately preceding each switch or nonswitch trial” (Meuter & Allport, 1999, p. 30), ranging between 1 and 13; third, “Ordinal Position: for switch trials, the first, second, or third occurrence of a switch of response language in a list, and for nonswitch trials, those occurring (a) before the occurrence of the first switch, (b) after the first switch but before the second switch, and (c) after the second switch within a list” (Meuter & Allport, 1999, p. 31). Figure 1 shows the results of the RT calculations and yields a clear pattern for each of the switch positions. First, switch trials are slower than nonswitch trials. Second, if a switch occurs, RTs are faster in L1; if no switch occurs, RTs are faster in L2. Third, with each successive switch, RTs are inclined to increase. Based on these data, the mean switching costs were calculated and yielded 143ms for L2 to L1 to 85ms for L1 to L2 (language switching cost is measured in the difference between the RTs of non-switch trials and the RTs of switch trials). These asymmetrical switching costs clearly show that the switching costs from a weaker language (L2) to a stronger language (L1) are higher than vice versa, thus confirming the study’s Relative Strength hypothesis. As another important finding, language proficiency seems to play a significant role in switching costs (Figure 2), as the difference in RT decreases when proficiencies are more equal.

Abbildung in dieser Leseprobe nicht enthalten

Figure 1: From “Bilingual Language Switching in Naming: Asymmetrical Costs of Language Selection” by R.F.I. Meuter, and A. Allport, 1999. Journal of Memory and Language 40, p. 31.

Abbildung in dieser Leseprobe nicht enthalten

Figure 2: From “Bilingual Language Switching in Naming: Asymmetrical Costs of Language Selection” by R.F.I. Meuter, and A. Allport, 1999. Journal of Memory and Language 40, p. 34.



ISBN (eBook)
ISBN (Buch)
674 KB
Institution / Hochschule
Eberhard-Karls-Universität Tübingen
switching costs language proficiency bilinguals trilinguals




Titel: Switching Costs and Language Proficiency in Bilinguals and Trilinguals