Sage Journals: Discover world-class research

Abstract

Aims and objectives:

Bilingual speakers are rarely equally proficient in both languages, making the determination of language dominance a major challenge. For a regional language like Low German, an additional challenge is the absence of a standard variety as a reference. To assess language dominance in both standardized and non-standardized languages, examining oral fluency provides a promising, non-invasive source of information. In this paper, we investigate several variables of speed and breakdown fluency to assess how well they reflect variations in language dominance among bilingual speakers of Low and High German, while also examining the role of the speaking task.

Design:

We recruited 95 bilingual speakers of Low and High German. The participants ranged in age from 15 to 88 years (47 female and 48 male). They completed four different tasks in each language (narrative task, picture-story task, direction-giving task, and reading task).

Data and analysis:

We analyzed seven fluency variables: articulation rate, speech rate, phonation/time ratio, mean length of runs, mean length of silent pauses, duration of silent pauses per minute, and number of silent pauses per minute. Generalized linear mixed models were fitted for these variables with language, dominance score, and gender as fixed effects and age as a covariate.

Findings:

Six of the seven fluency variables were found to be sensitive to variations in language dominance. Mean length of runs, duration of silent pauses per minute, and number of silent pauses per minute yielded the most consistent results across speaking tasks.

Originality:

This study links fluency measures with language dominance in a demographically changing group of speakers of an endangered regional language.

Implications:

We conclude that variables of speed and breakdown fluency are sensitive to variation in language dominance, even for two closely related languages such as Low and High German.

Keywords

Age dominance score fluency High German language dominance Low German

Introduction

Language dominance

Bilingual speakers are rarely completely balanced in their two languages. More often one language is more dominant than the other. Language dominance refers to the relative proficiency level in two languages as well as to the input and use of both languages (Montrul, 2016; Treffers-Daller, 2019). Speakers tend to be more dominant in one of the languages depending on factors such as age of acquisition and frequency of use (Birdsong, 2014; Flege et al., 2002; Grosjean, 1982; Olsson & Sullivan, 2005). An imbalance in the level of dominance between two languages may involve differences in language processing and production that are reflected in differences in fluency, complexity, and accuracy (Birdsong, 2014). These differences may be due to the increased cognitive load of more difficult tasks, especially when speaking the less-familiar language. Grosjean (1980) reports a slower speech rate, longer pauses, shorter mean length of runs and a lower phonation/time ratio for more complex tasks. Daller et al. (2010) found a strong correlation between speaking rate (words per second) and language dominance, with participants speaking more slowly in the less-dominant language. Language dominance is a dynamic construct that varies according to different speaking situations and circumstances and may also change over time (Olsson & Sullivan, 2005). This inherent variability makes assessing language dominance a challenging task.

Fluency

In second language acquisition research, the term “fluency” is commonly used to refer to specific temporal features of spoken language, including speech rate, pausing behavior, and repair phenomena. More broadly, fluency can be interpreted as an overall indicator of language proficiency (Lennon, 1990; Lintunen et al., 2020; Segalowitz, 2010). We use the term in the narrow sense by examining the relationship between language dominance and temporal variables of speech.

Variation in fluency can be described in three dimensions: speed fluency, breakdown fluency, and repair fluency (Skehan, 2003; Tavakoli & Skehan, 2005). Speed fluency refers to how quickly speech is delivered. It is usually assessed by the articulation rate, which is measured by the number of words, syllables, or phonemes per time interval excluding pauses, and by the speech rate, which includes pauses (Cucchiarini et al., 2000, 2002; De Jong et al., 2013, 2015; Di Silvio et al., 2016; Guz, 2015). Breakdown fluency involves silent and filled pauses that interrupt the flow of speech (De Jong et al., 2013, 2015; Di Silvio et al., 2016; Guz, 2015; Housen et al., 2012). Repair fluency is evident in corrections, false starts, and repetitions that occur during speaking (Lahmann et al., 2017). The phonation/time ratio is a particularly sensitive indicator of breakdown fluency, highlighting differences between L1 and L2 speech. Speech rate as well as mean length of runs are complex measures which combine speed and breakdown fluency (cf., e.g., Kormos & Dénes, 2004; Lennon, 1990; Tavakoli et al., 2020).

Furthermore, the type and complexity of speech tasks have been reported to have a significant influence on fluency (cf., e.g., Tavakoli & Wright, 2020). The effects of task complexity are usually explored by modifying task design and implementation features. Notable effects on speed and breakdown fluency have been found when changing the number of elements to process, the interconnectedness of those elements, the amount of preparation time, and the familiarity with the task through practice or repetition (De Jong et al., 2012; Ellis, 2005; Foster & Skehan, 1996; Hunter, 2017; Kovač & Vickov, 2018; Lambert et al., 2017; Levkina & Wright, 2012; Mehnert, 1998; Ortega, 1999; Tavakoli & Skehan, 2005).

Fluency in bilinguals can vary widely, depending on several factors, including age of acquisition, context of acquisition, frequency of use, language proficiency, language processing strategies, and the individual’s motivation and attitudes toward each language (Grosjean, 2013). Bilinguals who have acquired both languages early in life and had consistent exposure to both languages tend to be more fluent in both languages than those who acquired one language later in life (Bialystok, 1999; Daller et al., 2010; Flege et al., 2002). Bilinguals who have a high level of proficiency in both languages are generally more fluent in both languages than those who are less proficient in one of their languages. In addition, the way in which bilinguals process and use their two languages can affect fluency. For example, bilinguals who have developed language processing strategies that enable them to switch easily between the two languages tend to be more fluent in both languages than those who have not developed such strategies (Green, 1998).

To summarize, fluency in bilinguals is a complex and multi-faceted concept that is influenced by a range of factors, including the age of acquisition, the context of acquisition, the frequency of use, and the individual’s motivation and attitudes toward each language. Bilinguals who have developed high proficiency in both languages and who have developed effective language processing strategies tend to be more fluent in both languages.

Low German and High German

Low German is an endangered language spoken in the northern part of Germany with more than 2.5 million active speakers. Furthermore, the closely related Plautdietsch is spoken in linguistic enclaves in the United States, Mexico, Bolivia, Brazil, Paraguay, Russia, Kazakhstan, among others (Louden, 2020). In Germany, Low German is recognized as a regional language and protected by the European Charta for Regional or Minority Languages since 1999 (Council of Europe, 2021).

Low and High German are two closely related languages, with (High) German being the everyday language throughout Germany. High German is highly standardized in grammar, spelling, and pronunciation, however, variation in standard-near registers can be observed in Germany and the neighboring countries with German as official language. In contrast, Low German is a non-standardized language with strong dialectal variation in grammar, vocabulary, and pronunciation. It differs from High German mainly in that it has not undergone the Second Germanic Consonant Shift and that it has a less-complex case system. The proportion of North Germans who consider themselves to have a good or very good command of Low German decreased from 35% in 1984 (Stellmacher, 1987) to 15% in the latest surveys of 2007 (Möller, 2008) and 2015 (Adler et al., 2016), with the highest proportion of proficient speakers being those aged older than 80. Nowadays, Low German is no longer acquired monolingually, but only alongside High German. The use of Low German is largely confined to the private sphere, while High German prevails in public life, particularly in the education system and in more formal contexts. The demographic change in language proficiency suggests a shift in dominance from older to younger bilingual speakers of Low and High German. Among the older speakers, a dominance of Low German or a balanced bilingualism can be expected, while among younger speakers, High German is likely to be the dominant language.

Research question

As Low German is a language without a codified standard variety, established procedures for the standardized assessment of language proficiency based on the grammar, vocabulary, or pronunciation are not available. Against this background, the present paper addresses the question of how well commonly used fluency variables reflect variations in language dominance. For the purposes of the present study, we disregard variables of repair fluency and concentrate on speed and breakdown fluency variables, which can be captured automatically, thus allowing for the analysis of large corpora.

In line with previous research (e.g., Daller et al., 2010; Tavakoli & Wright, 2020), we anticipate that both speed and breakdown fluency decrease in the less-dominant language. In particular, we expect to find a lower articulation rate, a lower speech rate, a lower phonation/time ratio, a shorter mean length of runs, a longer mean length of silent pauses, a longer duration of silent pauses, and a lower number of silent pauses in the less-dominant language. As the surveys mentioned in the previous section suggest, High German dominates over Low German among young speakers, while older speakers can be expected to be more balanced bilinguals. Therefore, we expect a strong correlation between speaker age and language dominance which is why we include age as a co-variate in our statistic models. We also include gender as a factor for exploratory purposes, although we have no specific expectations about the possible effects of gender on the relationship between language dominance and fluency.

Since the type of speech task can also have an influence on the cognitive load associated with language use, we include four different task types for exploratory purposes: a narrative task, a picture-story task, a direction-giving task, and a reading task, the last three at different levels of difficulty.

Method

Subjects

We recruited 95 participants (47 female and 48 male), aged between 15 and 88 years at the time of recording. All participants grew up in the municipality of Krummhörn in East Frisia, in the northwest of the federal state of Lower Saxony in Germany, and were native speakers of East Frisian Low German (LG) and standard High German (HG). The municipality of Krummhörn is renowned for having one of the highest numbers of LG speakers (Strybny, 2009) and the local dialect is documented in detail in a study by Reershemius (2004). Four of the participants had to be excluded from the investigation due to technical errors while recording. Six additional participants were excluded because it was later discovered that they did not meet the requirement of having been raised in Krummhörn. The remaining 85 participants (43 female and 42 male) were between 15 and 88 years of age. All participants volunteered to take part in this study and had given their written informed consent prior to the recording. They were compensated for their time with a small amount of money. The general study design was approved by the ethics committee of the University of Oldenburg.

Dominance score

All participants completed a questionnaire containing a sociodemographic part as well as questions about their age of language acquisition, language use, and self-attributed language proficiency for LG as well as HG. Methods for assessing language dominance include self-report questionnaires (e.g., Flege et al., 2002; Poarch et al., 2019), oral proficiency tests (Daller et al., 2010), cognitive tasks, or a combination thereof (e.g., Olsson & Sullivan, 2005). An example for a self-report questionnaire is the Bilingual Language Profile (BLP) by Birdsong et al. (2012) which includes four components: language history, language use, language proficiency, and language attitudes. Each answer option is linked to a numeric value which is used to calculate a language dominance score. Marian et al. (2007) developed the Language Experience and Proficiency Questionnaire (LEAP-Q) based on standardized language tests and self-reported measures to get individual language profiles for bilingual and multilingual speakers.

The questionnaire used in this study consisted of 55 questions, which resulted from a combination and adaptation of Adler et al. (2016), Birdsong et al. (2012) and Poarch et al. (2019) (see Supplemental Appendix). According to their self-assessment, all speakers scored at least a 4 in both languages on a 7-point scale of language competence, where 1 stands for gar nicht gut (“not at all”) and 7 for sehr gut (“very well”), with an overall average of 6.2 for LG and 6.6 for HG. Based on their answers to the questions on age of acquisition, frequency of use, and language proficiency,¹ we calculated separate language scores for LG and HG for each participant. For statistical analysis, a dominance score was calculated by subtracting each individual’s HG score from their LG score, resulting in positive values for LG-dominant speakers and negative values for HG-dominant speakers.

Tasks

The tasks were organized in two blocks, one block per language. The order of the languages was counter-balanced across participants. In each block, the participants were asked to complete different versions of the same tasks to avoid repetition effects between the language blocks. We prepared two different versions of each task for both LG and HG to avoid version-specific effects, and ensured an even distribution of versions across recordings, taking into account participants’ age and gender. All LG tasks and instructions had previously been translated from HG by a local speaker. We trained a group of bilingual Krummhörn residents to conduct the experiment and interview the speakers, with technical assistance from the first and second author. This group consisted of eight people who were divided into four groups of two. They were female, aged between 60 and 65 years, and had grown up and lived in Krummhörn all their lives. Each interview was led by one of these four teams. Before each recording, they decided who would give instructions in which language, and then they spoke only the chosen language throughout the recording.²

The participants completed four tasks in each language, with two different levels of difficulty for the last three: a narrative task (“= Monologue”), a picture-story task (= “Story”), a direction-giving task (= “Route”), and a reading task (= “Reading”). In the first task (Monologue), our participants were asked questions to elicit 2–3 minutes of monologue-like spontaneous speech. The questions included topics such as their daily lives, hobbies, interests, life in Krummhörn, and attitudes toward the local dialect. The remaining three tasks, each with two difficulty levels, were recorded in descending order: “level 2” (higher difficulty) first, “level 1” (lower difficulty) second. In the second task (Story), participants were asked to retell a story depicted in a comic strip consisting of six pictures. The comic strips were taken from the “Father and Son” series by Plauen (2016). At the higher difficulty level, they had to start their story immediately. At the lower level, they had 30 seconds to prepare. In the third task (Route), the participants were asked to perform a map task. The recording assistant described a scenario in which she had to visit two different places in Pewsum (the largest village in Krummhörn) and needed help choosing a route to her destination. The participants were given a map of Pewsum with an outlined route. At the higher difficulty level, the map did not include any landmarks or street names. At the lower level, the map included landmarks and street names. In the fourth task (Reading), the participants were asked to read aloud a fable by Aesop (either “The Northwind and the Sun” or “The Raven and the Fox”). At the higher difficulty level, they had to start reading immediately without any preparation time. At the lower level, they read the same story a second time.

Recording procedure and acoustic analysis

All speech samples were recorded with a portable digital recorder (Tascam DR-100 MKIII) and a head-mounted microphone (DPA 4066).³ The recordings were digitized at a sampling rate of 48 kHz and with 24 bits/sample quantization and were later downsampled to 16 kHz for further analysis. The recordings took place in the speaker’s homes. Our aim was to build a comfortable and natural atmosphere for our participants. The rooms were carefully checked for background noises.

The recordings were cleaned up by excluding all instruction parts and analyzed using the syllable nuclei-script by De Jong et al. (2021) in Praat (Boersma & Weenink, 2020). To evaluate the accuracy of the script, 16 recordings were manually annotated. These recordings were equally distributed over language, age, gender, and task type and otherwise chosen randomly. We calculated the differences between the number of syllables counted by the script and by hand and calculated the error rate per 100 syllables. This resulted in an average error rate of 12.96 per 100 syllables. Although slightly higher than typical inter-rater discrepancies, the error rate shows no systematic bias or correlation with our variables and is therefore not considered a threat to the validity of our findings. To find the optimal setting for silent pauses, we tested various script settings using this procedure and found that the default settings with a minimum pause length of 200 ms yielded the best results. Fragmented intonation phrases, word repetitions, and filled pauses were not excluded. However, they only occurred relatively rarely in our corpus.

For the purpose of comparison, we examined seven common variables of speed and breakdown fluency, although they are not entirely independent of one another: articulation rate, speech rate, phonation/time ratio, mean length of runs, mean length of silent pauses, duration of silent pauses per minute, and number of silent pauses per minute (cf., e.g., Cucchiarini et al., 2002; Kormos & Dénes, 2004). All variables were either obtained directly from the script or derived from its output.

Articulation rate was calculated by dividing the total number of syllables by the total speaking time (excluding silent pauses) and is expressed in syllables per second. We also measured articulation rate in phones per second for a subset of 20 speakers evenly distributed across gender and age to account for possible differences in the syllable structure between LG and HG. Since we got very similar results for both measures, we chose the syllable as the unit of measurement to increase comparability with other studies. Speech rate was obtained by dividing the total number of syllables by total speaking time (including silent pauses). Phonation/time ratio is the total speaking time excluding silent pauses divided by total speaking time including silent pause time, which was multiplied by 100 to get a percentage value. Mean length of runs is the mean number of syllables between two silent pauses, which was calculated by dividing the total number of syllables by the total number of runs. Mean length of silent pauses is the total silent pause time divided by the number of silent pauses. Duration of silent pauses per minute is the total duration of silent pauses divided by the total speaking time including silent pauses divided by 60. Finally, the number of silent pauses per minute is the number of silent pauses divided by the total speaking time including silent pauses divided by 60.

Statistical analysis

As we were interested in the individual sensitivity of each fluency variable, we fitted separate generalized linear mixed models using the glmmTMB package (Magnusson et al., 2023) in R (R-Core-Team, 2022) for Articulation Rate, Speech Rate, Phonation/Time Ratio, Mean Length of Runs, Mean Length of Silent Pauses, Duration of Silent Pauses per Minute and Number of Silent Pauses per Minute as dependent variables. As fixed effects, we included Language (LG vs. HG) and Dominance Score as within-subject factors. In addition, we included the between-subject factor Gender (male vs. female) as a control variable and Age as a co-variate. We included random intercepts and by-subject random slopes for Language, Order (order of both languages in each recording), and Difficulty⁴ (for all tasks except for Monologue). We fitted the linear mixed models with either a Gaussian or gamma distribution, depending on which best fitted the residuals of our dependent variables.

For each dependent variable, we reduced the full model step by step to select the model with the best fit using the AIC (Akaike information criterion) score as the selection criterion. Based on Matuschek et al. (2017), the models were reduced using the following procedure: First we excluded the intercept for Order; second, we excluded the intercepts for Order and Difficulty; third, we changed the by-subject slopes for Language to an intercept for Subject with the inclusion of the intercept for Difficulty and Order; forth, we kept the intercept for Subject and Difficulty but excluded Order; and finally, we only kept the intercept for Subject. For the Story, Route, and Reading tasks, we were left with six models to choose from for each dependent variable. Due to the missing difference in difficulty in the Monologue, we only had four models per dependent variable for this task. The p-values were calculated using Likelihood Ratio Tests for each model. The level of significance was corrected for multiple testing with Bonferroni correction, multiplying the threshold of p = .05 by six for Story, Route and Reading and by four for Monologue.

Post hoc analyses of the interactions were performed to examine the difference between slopes for both levels of Language using the emtrends function of the emmeans package in R (Lenth, 2023; R-Core-Team, 2022). All figures were created with the tidyverse packages (Wickham, 2023) in R.

As indicated in the Research question section, we expect lower Articulation Rate, Speech Rate, Mean Length of Runs, and Phonation/Time Ratio in the less-dominant language, as well as higher Mean Length of Silent Pauses, Duration of Silent Pauses per Minute, and Number of Silent Pauses per Minute. However, we also expect that these differences will be greater the more dominant the speakers are in HG, which should be reflected in an interaction between the Dominance Score and Language. To test these predictions, we only report main effects of Language and interactions involving Language and Dominance Score.

Results

Before examining the relationship between language dominance and fluency, we will first analyze the relationship between the speakers’ age and their language scores for LG and HG, which were used to calculate Dominance Score.

Age and language dominance

Figure 1 shows the language scores for LG and HG as a function of Age. Younger speakers have a lower score for LG than for HG. These values converge with age, and the regression lines cross around the age of 75, indicating a tendency toward HG dominance below this age and LG dominance above.⁵ The statistical analysis revealed a significant main effect for Language (χ² = 83.40, p < .001), a significant interaction between Age and Language (χ² = 38.06, p < .001) and no interaction with Gender. The analysis of contrast for two languages revealed a significant difference between the slopes of the regression lines for LG and HG, Estimate = 1.259, SD = 0.202, t-ratio = 6.234, p < .0001.

Figure 1.

Regression lines for individual global language scores for LG and HG and AGE as a predictor variable.

Fluency variables

Figure 2 shows regression lines for LG and HG for the seven fluency variables analyzed, broken down by task. All graphs, except those showing results for the Reading task and for most tasks of the dependent variable Mean Length of Silent Pauses, show crossing regression lines for LG and HG, with lower values for LG (black lines) than for HG (blue lines) in the HG-dominant speakers for Articulation Rate, Speech Rate, Phonation/Time Ratio, Mean Length of Runs, and with higher values for LG in the HG-dominant speakers for Duration of Silent Pauses per Minute and Number of Silent Pauses per Minute, a pattern that tends to reverse as the dominance scores increase. For the Reading task, Figure 2 shows lower values for LG than for HG for the first four variables and lower values for HG than for LG for the remaining three variables, with little or no variation across the range of Dominance Score. For Mean Length of Silent Pauses, the differences between LG and HG in the Monologue, Story, and Route tasks are small overall.

Figure 2.

Regression lines for Articulation Rate, Speech Rate, Phonation/Time Ratio, Mean Length of Runs, Mean Length of Silent Pauses, Duration of Silent Pauses per Minute, and Number of Silent Pauses per Minute with DOMINANCE SCORE as a predictor variable.

Table 1 lists main effects of Language and interactions involving Language and Dominance Score. It shows a significant main effect of Language for all variables in the Reading Task. For Mean Length of Runs and Number of Silent Pauses per Minute, a significant interaction between Language and Dominance Scores was found in all three remaining tasks. For Duration of Silent Pauses per Minute, a significant interaction was found for both the Story and Route task. For Articulation Rate, Speech Rate, and Phonation/Time Ratio, a significant interaction effect was found only for the Story task. No significant interactions involving Gender were observed.

Table 1.

Assessing model fit (χ²) with Dominance Score as predictor variable.

	Monologue	Story	Route	Reading
Articulation rate
Language	4.21	0.24	0.85	383.22***
Dominance: Language	2.76	12.99**	2.91	0.27
Dominance: Language: Gender	0.45	1.70	1.13	2.25
Speech rate
Language	0.79	0.03	0.27	411.68***
Dominance: Language	4.68	11.94**	5.73	1.94
Dominance: Language: Gender	2.08	2.98	0.74	1.30
Phonation/Time ratio
Language	0.18	0.07	0.09	129.29***
Dominance: Language	4.95	7.40*	6.87	0.73
Dominance: Language: Gender	2.05	2.66	0.44	0.18
Mean length of runs
Language	3.13	1.97	3.21	196.28***
Dominance: Language	10.62**	10.96**	11.99**	1.96
Dominance: Language: Gender	3.42	1.79	0.07	0.70
Mean length of silent pauses
Language	1.64	2.63	1.68	48.40***
Dominance: Language	0.45	0.11	0.48	0.03
Dominance: Language: Gender	1.23	0.02	0.77	0.01
Duration of silent pauses per minute
Language	0.01	0.07	0.06	137.08***
Dominance: Language	4.71	7.40*	7.12*	2.04
Dominance: Language: Gender	1.72	2.66	0.53	0.30
Number of silent pauses per minute
Language	3.86	5.81	5.44	91.17***
Dominance: Language	11.79**	10.05**	11.45**	2.53
Dominance: Language: Gender	2.42	1.62	0.01	0.38

Corrected p levels for monologue: * < .0125, ** < .0025, *** < .00025; corrected p levels for Story, Route, and Reading: * < .0083, ** < .0016, *** < .00016.

The contrast analyses in Table 2 demonstrate for the interaction effects a significant difference between the slopes of the regression lines for LG and HG, suggesting that for all fluency variables except Mean Length of Silent Pauses, the lower the dominance score of the bilingual speakers, the less fluent they were in LG compared to HG. Due to the correction factor, contrasts for Phonation/Time Ratio and Duration of Silent Pauses per Minute do not reach the 5% level of significance. Nevertheless, they do show a tendency in the direction that was expected.

Table 2.

Contrasts between LG and HG slopes for interactions involving Language and Dominance Score.

Dependent variable	Task	Contrast	Estimate	SE	z/t-ratio^a	p
Articulation rate	Story	LG—HG	-0.0022	0.0007	-3.3560	.0009***
Speech rate	Story	LG—HG	-0.0037	0.0011	-3.2180	.0014**
Phonation/Time ratio	Story	LG—HG	-0.0524	0.0210	-2.4980	.0130
Mean length of runs	Monologue	LG—HG	-0.0027	0.0009	-2.9070	.0037*
	Story	LG—HG	-0.0027	0.0009	-3.0610	.0022*
	Route	LG—HG	-0.0021	0.0006	-3.4070	.0007***
Duration of silent pauses per min.	Story	LG—HG	0.0315	0.0126	2.4980	.0130
	Route	LG—HG	0.0015	0.0006	2.5550	.0106
Number of silent pauses per min.	Monologue	LG—HG	0.0405	0.0129	3.1480	.0020**
	Story	LG—HG	0.0372	0.0127	2.9320	.0036*
	Route	LG—HG	0.0304	0.0091	3.3540	.0009***

Selected distributions for analysis: Gaussian distribution: Articulation Rate, Speech Rate Phonation/Time Ratio, and Number of Silent Pauses per Minute; gamma distribution: Mean Length of Runs and Duration of Silent Pauses per Minute.

Corrected p levels for Monologue: * < .0125, ** < .0025, *** < .00025; for Story, Route, and Reading: * < .0083, ** < .0016, *** < .00016.

Discussion

This study aimed to investigate how well individual fluency variables reflect variations in language dominance among balanced and non-balanced bilingual speakers of LG and HG. In line with previous studies, we expected to find a decrease in speed and breakdown fluency in the less-dominant language. We also expected that the more balanced the speakers were, the smaller the difference between the two languages. In addition, we expected language dominance to vary with age, with younger speakers more likely to be dominant in HG and dominance becoming more balanced with increasing age.

We observed a strong correlation between language dominance and age, which is consistent with the ongoing dominance shift reported for the LG speaking population in Germany by Möller (2008) and Adler et al. (2016). To assess speed fluency, we examined articulation rate and speech rate. While the first variable reflects the speed of articulatory movements, the latter is a complex variable that combines speed and breakdown fluency by including pauses. In line with previous research, the participants showed both a lower articulation rate and a lower speech rate in the less-dominant language, but only in the Story task (Daller et al., 2010; Di Silvio et al., 2016; Guz, 2015). We conclude that both variables of speed fluency might be useful predictors of language dominance. However, since both variables show significant interactions between Language and Dominance Score in only one of the four task types, they might not be the most reliable predictors when not combined with other fluency variables.

For breakdown fluency, almost all measures show significant interactions between Language and Dominance Score at least for the storytelling task. All interactions align with our expectations and previous studies (Daller et al., 2010; De Jong et al., 2015; Guz, 2015; Housen et al., 2012). Mean Length of Silent Pauses is the only variable examined that does show no significant interactions. Duration of Silent Pauses per Minute and Number of Silent Pauses per Minute are among the most promising predictors. Mean Length of Silent Pauses is the only variable tested that does not show any significant results. We conclude that the length of individual pauses is not as relevant as the proportion of pausing time to speaking time in determining language dominance. However, Phonation/Time Ratio, which reflects the proportion of speaking time without silent pauses to total speaking time, turns out to be sensitive to variation in language dominance only in the Story task. Mean Length of Runs also shows consistent results across all task types except Reading, such as Number of Silent Pauses per Minute to which it is related. No effect of Gender was found for any of the variables. In general, it appears that the most reliable indicator of language dominance in our data is the division of speech into smaller parts by pauses. Further research will have to show whether this is also the case if filled pauses and interruptions due to repair phenomena are accounted for in the analysis.

Since the use of a non-dominant language is usually associated with increased cognitive load, the variation found in this study can be interpreted as the effect of different levels of cognitive load when using the two languages. However, we observed a relatively large influence of the task type on the fluency variables, which we did not expect to this extent. While the Story task seems to show very consistent results, the Monologue and Route task seem less predictable in their outcome. While only two variables show significant results for the Monologue task (Mean Length of Runs and Number of Silent Pauses per Minute), for the Route task, we found three significant interactions. It is also striking that none of the fluency variables varied depending on language dominance in the reading task. Instead, we found only a main effect of Language. This finding reflects a pattern typical of L2 learners who tend to show significantly lower reading proficiency in the L2. In our data, this tendency persists across all ages and dominance score ranges. This finding confirms the common observation that LG is primarily a spoken language, and that most LG speakers have low literacy skills.

Varying the difficulty level in the Story, Route, and Reading task led to no or inconclusive results. The effects were not reported in the previous section as the results raise the suspicion that the manipulations were either not large enough or not appropriate to be effective in inducing systematic variation in cognitive load (cf. Frank et al., 2023 for a similar conclusion regarding the analysis of vowel spaces). We therefore included task difficulty as a random factor rather than as a predictor variable in our regression models. The unexpected outcome of the manipulation of difficulty levels may also be due to the way our tasks were approached. In case of the Route task, for example, where participants were given maps labeled with street names and names of landmarks in the easier version of the task, many participants chose to reduce their vocabulary output to a minimum by only naming the street names along the indicated route, resulting in very short utterances. On the other hand, in the more difficult version, participants used a wide range of different ways to solve the task. Some participants simply said “left” and “right” without actually describing the route, while others either recalled landmarks from memory or told the interviewer who lived in each house they were passing during their route description. For future studies, we recommend placing greater emphasis on the premise that the interviewer must be able to reach the intended destination on the base of a careful route description. Otherwise, the necessary cognitive load varies significantly in both languages, making it difficult to relate observed variations in fluency to variations in language dominance. In the case of the Story task, unexpected effects of manipulations of the difficulty level arose from the fact that in the easier version of the task, in which additional preparation time was granted, participants may have used the extra time for planning the overall structure of the narrative or for planning individual utterances at the stages of conceptualization, grammatical encoding, and morpho-phonological encoding in the sense of Levelt’s model of speech production (Levelt, 1999). This strategy may have increased the complexity of the narrative or of individual utterances rather than the fluency of speech. Such a behavior would be consistent with Skehan’s Limited Attentional Capacity model (Skehan, 2015; Skehan & Foster, 2001), according to which cognitive resources can be variably allocated to different aspects of performance (see also Peters et al., 2023, 2024).

In summary, this study shows that most of the included variables of speed and breakdown fluency reflect variation in language dominance for Low and High German, two closely related languages that are very similar in terms of grammar, vocabulary, and pronunciation, making them particularly suitable for the comparative study of fluency in bilingual speakers. In our study, Mean Length of Runs, which reflects aspects of both speed and breakdown fluency, and the breakdown fluency variables Duration of Silent Pauses per Minute and Number of Silent Pauses per Minute yielded the most consistent results across speaking tasks, with the exception of the Reading task, which showed only a main effect of Language. In contrast, the Story task, which required participants to retell a story depicted in a comic strip, was found to be suitable for demonstrating the sensitivity of all fluency variables, except Mean Length of Silent Pauses, to variations in language dominance. No systematic effects of our manipulations of task difficulty were found and no effects of Gender. The present study is a first step in assessing the impact of a lack of proficiency on individual speed and breakdown fluency variables in a non-standardized regional language that is no longer acquired monolingually. For a more complete picture, it will be necessary to include variables of repair fluency as well as measures of accuracy and complexity of the speech output, which will allow for a better understanding of the variable effects found for speech tasks that differ by task type and task difficulty.

Supplemental Material

sj-docx-1-ijb-10.1177_13670069251353090 – Supplemental material for On the relationship between language dominance and fluency in bilingual speakers of Low and High German

Supplemental material, sj-docx-1-ijb-10.1177_13670069251353090 for On the relationship between language dominance and fluency in bilingual speakers of Low and High German by Tio Rohloff, Marina Frank and Jörg Peters in International Journal of Bilingualism

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by the German Research Foundation (DFG), grant number PE 793/3-1. We would like to thank Heinz Richter for his support in translating the instructions and tasks into the regional dialect of Low German spoken in Krummhörn as well as the Low German instructors and participants for making this research possible.

ORCID iDs

Tio Rohloff

Marina Frank

Jörg Peters

Supplemental material

Supplemental material for this article is available online.

Notes

References

Adler

Ehlers

Goltz

Kleene

Plewina

(2016). Status und Gebrauch des Niederdeutschen 2016. Erste Ergebnisse einer repräsentativen Erhebung. Institut für Deutsche Sprache.

Bialystok

(1999). Cognitive complexity and attentional control in the bilingual mind. Child Development, 70(3), 636–644.

Birdsong

(2014). Dominace and age in bilingualism. Applied Linguistics, 35(4), 374–392.

Birdsong

Gertken

L. M.

Amengual

(2012). Bilingual language profile: An easy-to-use instrument to assess bilingualism. COERLL.

Boersma

Weenink

(2020). Praat: Doing phonetics by computer (Version 6.1.30). Praat.

Council of Europe. (2021). European Charter for regional or minority languages: Collected texts.

Cucchiarini

Strik

Boves

(2000). Quantitative assessment of second language learners’ fluency by means of automatic speech recognition technology. Acoustical Society of America, 107(2), 989–999. https://doi.org/10.1121/1.428279

Cucchiarini

Strik

Boves

(2002). Quantitative assessment of second language learners’ fluency: Comparisons between read and spontaneous speech. Acoustical Society of America, 111(6), 2862–2873. https://doi.org/10.1121/1.1471894

Daller

M. H.

Yıldız

De Jong

N. H.

Kan

Başbaği

(2010). Language dominance in Turkish–German bilinguals: Methodological aspects of measurements in structurally different languages. International Journal of Bilingualism, 15(2), 215–236. https://doi.org/10.1177/1367006910381197

10.

De Jong

N. H.

Groenhout

Rob

Hulstijn

J. H

. (2015). Second language fluency: Speaking style or proficiency? Correcting measures of second language fluency for first language behavior. Applied Psycholinguistics, 36(2), 223–243. https://doi.org/10.1017/S0142716413000210

11.

De Jong

N. H.

Pacilly

Heeren

. (2021). PRAAT scripts to measure speed fluency and breakdown fluency in speech automatically. Assessment in Education: Principles, Policy & Practice, 28(4), 456–476. https://doi.org/10.1080/0969594X.2021.1951162

12.

De Jong

N. H.

Steinel

M. P.

Florijn

. (2012). The effect of task complexity on functional adequacy, fluency and lexical diversity in speaking performances of native and non-native speakers. In Housen

Vedder

Kuiken

(Eds.), Dimensions of L2 performance and proficiency: Complexity, accuracy and fluency in SLA (pp. 121–142). John Benjamins Publishing Company.

13.

De Jong

N. H.

Steinel

M. P.

Florijn

Schoonen

Hulstijn

J. H

. (2013). Linguistic skills and speaking fluency in a second language. Applied Psycholinguistics, 34, 893–916. https://doi.org/10.1017/S0142716412000069

14.

Di Silvio

Donovan

Diao

. (2016). The development of L2 fluency during study abroad: A cross-language study. The Modern Language Journal, 100(3), 610–624. https://doi.org/10.1111/modl.12343

15.

Ellis

(2005). Planning and task-based performance. Theory and research. In Ellis

(Ed.), Planning and task performance in a second language (pp. 3–34). John Benjamins Publishing Company.

16.

Flege

J. E.

Mackay

I. R. A.

Piske

(2002). Assessing bilingual dominance. Applied Psycholinguistics, 23, 567–598.

17.

Foster

Skehan

(1996). The influence of planning and task type on second language performance. Studies in Second Language Acquisition, 18, 299–323.

18.

Frank

Rohloff

Peters

(2023). Effects of cognitive load on vowel production in bilingual speakers of High and Low German. ICPhS.

19.

Green

D. W.

(1998). Mental cotrol of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1, 67–81.

20.

Grosjean

(1980). Temporal variables within and between languages. In Dechert

H. W.

Raupach

(Eds.), Towards a cross-linguistic assessment of speech production (pp. 39–53). Lang.

21.

Grosjean

(1982). Life with two languages: An introduction to bilingualism. Harvard University Press.

22.

Grosjean

(2013). Bilingual and monolingual language modes. In Chapelle

C. A.

(Ed.), The encyclopedia of applied linguistics (pp. 1–9). Blackwell Publishing. https://doi.org/10.1002/9781405198431.wbeal0090

23.

Guz

(2015). Establishing the Fluency Gap between native and non-native-speech. Research in Language, 13(3), 230–247. https://doi.org/10.1515/rela-2015-0021

24.

Housen

Kuiken

Vedder

(Eds.) (2012). Dimensions of L2 performance and proficiency: Complexity, accuracy and fluency in SLA (Vol. 32). John Benjamins Publishing Company.

25.

Hunter

A.-M.

(2017). Fluency development in the ESL classroom: The impact of immediate task repetition and procedural repetition on learners’ oral fluency. St Mary’s University, Twickenham.

26.

Kormos

Dénes

(2004). Exploring measures and perceptions of fluency in the speech of second language learners. System, 32, 145–164. https://doi.org/10.1016/j.system.2004.01.001

27.

Kovač

Vickov

(2018). The impact of immediate task repetition on breakdown fluency. Govor, 35, 139–160.

28.

Lahmann

Steinkrauss

Schmid

M. S.

(2017). Speed, breakdown, and repair: An investigation of fluency in long-term second-language speakers of English. International Journal of Bilingualism, 21(2), 228–242. https://doi.org/10.1177/1367006915613162

29.

Lambert

Kormos

Minn

(2017). Task repetition and second language speech processing. Studies in Second Language Acquisition, 39, 167–196.

30.

Lennon

(1990). Investigating fluency in EFL: A quantitative approach. Language Learning, 40(3), 387–417.

31.

Lenth

R. V.

(2023). Estimated marginal means, aka least-squares means (Version 1.8.8). https://cran.r-project.org/web/packages/emmeans/emmeans.pdf

32.

Levelt

W. J. M.

(1999). Producing spoken language: A blueprint of the speaker. In Brown

C. M.

Hagoort

(Eds.), The neurocognition of language (pp. 83–122). Oxford University Press.

33.

Levkina

Wright

(2012). The effects of cognitive task complexity on L2 oral production. In Housen

Kuiken

Vedder

(Eds.), Dimensions of L2 performance and proficiency: Complexity, accuracy, and fluency in SLA (pp. 171–198). John Benjamins Publishing Company.

34.

Lintunen

Mutta

Peltonen

(2020). Defining fluency in L2 learning and use. In Lintunen

Mutta

Peltonen

(Eds.), Fluency in L2 learning and use (pp. 1–15). Multilingual Matters.

35.

Louden

M. L.

(2020). Minority Germanic languages. In Putnam

M. T.

Page

B. R.

(Eds.), The Cambridge handbook of Germanic linguistics (pp. 807–832). Cambridge University Press.

36.

Magnusson

Skaug

Nielsen

Berg

van Bentham

(2023). Generalized linear mixed models using template model builder (Version 1.1.7). https://github.com/glmmTMB/glmmTMB

37.

Marian

Blumenfeld

H. K.

Kaushanskaya

(2007). The Language Experience and Proficiency Questionnaire (LEAP-Q): Assessing language profiles in bilinguals and multilinguals. Journal of Speech, Language, and Hearing Research, 50, 940–967.

38.

Matuschek

Kliegl

Vasishth

Baayen

Bates

(2017). Balancing Type I error and power in linear mixed models. Journal of Memory and Language, 94, 305–315.

39.

Mehnert

(1998). The effects of different lengths of time for planning on second language performance. Studies in Second Language Acquisition, 20, 83–108.

40.

Möller

(2008). Plattdeutsch im 21. Jahrhundert. Bestandsaufnahme und Perspektiven (Vol. 34). Schuster.

41.

Montrul

(2016). Dominance and proficiency in early and late bilingualism. In Silva-Corvalán

Treffers-Daller

(Eds.), Language Dominance in bilinguals: Issues of measurement and operationalization (pp. 15–35). Cambridge University Press.

42.

Olsson

Å.

Sullivan

K. P. H

. (2005). Provoking dominance shift in a bilingual Swedish-American English 4-year-old child. In Cohen

McAlister

K. T.

Rolstad

MacSwan

(Eds.), Proceedings of the 4th international symposium on bilingualism (pp. 1750–1764). Cascadilla Press.

43.

Ortega

(1999). Planning and focus on form in L2 oral performance. Studies in Second Language Acquisition, 21, 109–148.

44.

Peters

Frank

Rohloff

(2023). Effects of cognitive load on vocal fold vibratory patterns in bilingual speakers of low and high german. Journal of Voice. Advance online publication. https://doi.org/10.1016/j.jvoice.2023.09.016

45.

Peters

Frank

Rohloff

(2024, July 2–5). Effects of task type and task difficulty on oral fluency in native and non-native speech [Conference session]. Speech Prosody, Leiden, The Netherlands.

46.

Plauen

E. O.

(2016). Vater und Sohn. Sämtliche Bildgeschichten (7th ed.). Nikol.

47.

Poarch

G. J.

Vanhove

Berthele

(2019). The effect of bidialectalism on executive function. International Journal of Bilingualism, 23(2), 612–628.

48.

R-Core-Team. (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing. www.R-project.org/

49.

Reershemius

(2004). Niederdeutsch in Ostfriesland. Zwischen Sprachkontakt, Sprachveränderung und Sprachwechsel. Franz Steiner Verlag.

50.

Segalowitz

(2010). Cognitive bases of second language fluency. Routledge.

51.

Skehan

(2003). Task-based instruction. Language Teaching, 36(1), 1–14. https://doi.org/10.1017/S026144480200188X

52.

Skehan

(2015). Limited attention capacity and cognition: Two hypotheses regarding second language performance on tasks. In Bygate

(Ed.), Domains and directions in the development of TBLT (pp. 123–156). John Benjamins Publishing Company.

53.

Skehan

Foster

(2001). Cognition and tasks. In Robinson

Robinson

P. J.

(Eds.), Cognition and second language instruction (pp. 183–205). Cambridge University Press. https://doi.org/10.1017/CBO9781139524780.009

54.

Stellmacher

(1987). Wer spricht Platt? Zur Lage des Niederdeutschen heute. Eine kurze Bestandsaufnahme. Schuster.

55.

Strybny

(2009). Plattdeutsche Sprachlandschaften in Ostfriesland. Ostfriesische Landschaft.

56.

Tavakoli

Nakatsuhara

Hunter

A.-M.

(2020). Aspects of fluency across assessed levels of speaking proficiency. The Modern Language Journal, 104(1), 169–191. https://doi.org/https://doi.org/10.1111/modl.12620

57.

Tavakoli

Skehan

(2005). Strategic planning, task structure and performance testing. In Ellis

(Ed.), Planning and task performance in a second language (pp. 239–273). John Benjamins Publishing Company.

58.

Tavakoli

Wright

(2020). Second language speech fluency: From research to practice. Cambridge University Press.

59.

Treffers-Daller

(2019). What defines language dominance in bilinguals. Annual Review of Linguistics, 5, 375–393.

60.

Wickham

(2023). Easily install and load the “tidyverse” (Version 2.0.0). https://tidyverse.tidyverse.org

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB