Reconciling Inter- and Intra-Individual Variation in L2 Socio-Pragmatic Development: Intensifier Variation in Spoken German

Abstract

This study is the first to scrutinize the rates of, and the lexical diversity in, adjective intensification in second language (L2) German. We additionally attend to the issue concerning whether sociodemographic variables (i.e., length of residence, age, and gender) and individual learner differences (i.e., L2 proficiency, intensity of exposure to the L2, and L2 socioaffect) can predict (a) the inter-individual variation in syntactic adjective intensification, and (b) the observed intra-individual variation based on a weighted measure of intensifier lexical diversity. We analyzed spoken data collected via virtual reality (VR) elicitation tasks from 40 learners of L2 German (first language [L1] English). We found that learners engaged in adjective intensification at similar rates as those reported in the literature, despite some cases of overshooting the target; learners also preferred markers of intensification consistent with the lexical choices of L1 German speakers. Sociodemographic variables did not predict different rates of adjective intensification; rather, individual learner differences such as those relating to L2 proficiency and L2 exposure correlated with more target-like use of intensifiers, though the correlations were weak. The diversity in adjective intensification was also only marginally related to demographic factors and individual learner differences. Our findings suggest that L2 learners indeed engage in similar intensification practices, as do L1 speakers; however, systematically predicting more ‘successful’ adoption of target-like sociopragmatic norms among L2 learners remains challenging.

1. Introduction

As second language (L2) learners advance to higher proficiency levels, they may find it increasingly challenging to select and appropriately use various forms across linguistic and social contexts (Geeslin, 2011), especially when those forms have similar but not identical meanings (Canale & Swain, 1980). This more nuanced body of knowledge is assumed to develop in relation to the degree and quality of meaningful exposure L2 learners have to authentic target-language input (e.g., Bardovi-Harlig & Bastos, 2011; Gautier & Chevrot, 2015; Wirtz & Pfenninger, 2024). That said, the target-like use of certain linguistic variables poses both structural and pragmatic issues for L2 learners, one instance being intensifiers. For one, this is because intensifiers are structurally confined to certain syntactic structures and semantic relationships; for another, because their pragmatic function—i.e., scaling a quality upward or downward from some assumed norm in order to heighten or weaken communicated ideas (e.g., Baños, 2013; Czerwionka & Olson, 2020)—requires learners to acquire sensibility for subtle social and contextual cues and develop “competence for variability” (Kanwit et al, 2018, p. 458), which can be challenging especially in L2 learning. Following in the steps of the abundant work on intensification patterns in English, a recent wave of variationist-inspired studies has explored intensifier variation in German (e.g., A. Mora & Bülow, 2013; Stratton, 2020; Stratton & Beaman, 2025; Pheiff, in press). This cumulative body of work demonstrates that intensification, specifically of adjectives, is a form of lexical-pragmatic variation which is subject to rather rapid change within and across individuals. Indeed, according to Ito and Tagliamonte (2003, p. 257), “[t]he most rapid and the most interesting semantic developments in linguistic change […] occur with intensifiers.” Furthermore, sociolinguists generally agree that intensifier variation is socially constrained, for example, in relation to age and gender (e.g., Ito & Tagliamonte, 2003; Stratton, 2020). Since L2 learners are known to acquire patterns of variation reflective of salient social categories such as gender (e.g., Adamson & Regan, 1991; Major, 2004) and age (e.g., Meyerhoff & Schleef, 2013), a question one might ask is to what extent L2 learners of German acquire patterns of intensification similar to those found in the L2 community. In this article, we explore this issue. Based on spoken data from 40 learners of L2 German (first language [L1] English), we address (1) at what rates L2 learners engage in syntactic adjective intensification, (2) whether sociodemographic variables (i.e., length of residence, age, and gender) and individual learner differences (i.e., L2 proficiency, intensity of exposure to the L2, and socioaffective factors) constrain rates of adjective intensification among L2 learners of German, and (3) to what extent the lexical diversity in adjective intensification may also relate to such sociodemographic variables and individual learner differences. The latter point in particular aims to chart new territory in the study of intra-individual variation among L2 learners, positioning these analyses as pioneering contributions on the issue of how diverse L2 learners’ intensification repertoires are and which factors influence this form of linguistic diversity.

2. Background

2.1. Intensification

Intensifiers are defined as linguistic devices which scale quality upward or downward from some assumed norm (e.g., Tagliamonte, 2008; D’Arcy, 2015). Scaling a quality upward via intensification involves the use of amplifiers (e.g., sehr gut ‘very good’), while scaling a quality downward involves the use of downtowners (e.g., das ist ein bisschen komisch ‘that is a bit weird). Intensifiers can modify numerous phrase categories, such as adverb, nominal, and prepositional phrases (Androutsopoulos, 1998; Claridge et al., 2024). According to Bäcklund (1973, p. 279) and Androutsopoulos (1998, pp. 457–458), however, the most frequent function of an intensifier is to modify adjectives. Following in the variationist sociolinguistic tradition (e.g., Ito & Tagliamonte, 2003; Stratton, 2020), the present study examines syntactic adjective intensification and adds to these findings by focusing on intensifier variation among learners of L2 German.

The use of intensifiers represents a way for speakers to impress, persuade, praise, and generally influence the interlocutor’s reception of a message (Partington, 1993, p. 178). Because of this, it comes as no surprise that intensifiers can index social meaning and can be socially constrained by factors such as gender and age (e.g., Ito & Tagliamonte, 2003; Tagliamonte, 2008; Barnfield & Buchstaller, 2010; D’Arcy, 2015; Fuchs, 2017; Stratton, 2020; Stratton & Sundquist, 2022). For example, younger speakers have been found to use intensifiers more frequently than older speakers (Ito & Tagliamonte, 2003; Barnfield & Buchstaller, 2010; Stratton, 2020), and women register higher rates of intensification compared to men (Fuchs, 2017; Stratton, 2020; Stratton & Sundquist, 2022). Stratton and Beaman (2025, p. 112) propose that these patterns may be a manifestation of younger individuals attempting to “use intensification to exaggerate qualities in an effort to have their voices heard in an adult-dominated world” and women may employ intensifiers more frequently as a means to “reinforce their opinions in a male-dominated world.”

In addition to being socially constrained, rates of intensification vary in relation to the syntactic function of the adjective modified. Androutsopoulos (1998), for instance, maintains that predicatively used adjectives most often undergo intensification, followed by attributive adjectives and adverbials. This effect has been verified on multiple occasions; in Swabian German and Swiss German, for instance, predicative adjectives are more amenable to intensification than attributive adjectives (Pheiff, in press; Stratton & Beaman, 2025). In fact, this trend appears to hold cross-linguistically and has been attested for varieties of English (e.g., Ito & Tagliamonte, 2003) and Norwegian (Stratton & Sundquist, 2022) as well.

Studies on contemporary varieties of German (e.g., Androutsopoulos, 1998; Christen, 2003; A. Mora & Bülow, 2013; Pheiff, in press; Stratton, 2020) have begun mapping the terrain on the rates of intensification, and the lexical variation therein. In his study of German youth language, Androutsopoulos (1998) investigated the intensification of adjectives, nouns, verbs, and prepositional phrases, and he found that, among others, intensifiers such as absolut ‘absolutely’, echt ‘really’, total ‘totally’, voll ‘fully’, völlig ‘completely’, extrem ‘extremely’, and recht ‘quite’ represent central intensifying particles of youth language. Others, such as A. Mora and Bülow (2013) in their corpus of two photo novellas, Stratton (2020) employing a corpus of spoken German (Foschungs- und Lehrkorpus, FOLK), Stratton and Beaman (2025) in their longitudinal corpus of Swabian German, and Pheiff (in press) drawing on the Swiss SMS Corpus, have similarly identified high frequencies of the German intensifiers so ‘so’, echt ‘really’, total ‘totally’, voll ‘fully’, ganz ‘completely’, and sehr ‘very’. Stratton (2020) emphasizes that amplifiers occur with a greater token and type frequency than downtoners (67% vs. 33%), and Van Olmen (2023) echoed this finding, maintaining that amplification is preferred over downtoning in four West Germanic languages (Afrikaans, Dutch, English, and German). In general, this diversity in potential adjective intensifiers leads us to believe that there is likely a rather pronounced degree of intra-individual variation in intensification among L1 speakers, though analyses assessing the degree of this intra-individual variation have not featured in previous research. We address this issue by asking to what extent L2 learners acquire intra-individual variation in adjective intensification.

While studies of intensifier variation in different varieties of German are quickly coming to light, the issue has remained overlooked in the context of Austrian German varieties. Our findings on the rates of and lexical manifestations of L2 adjective intensification can therefore not be directly compared to Austrian speakers of L1 German. That said, previous research on intensifier variation in German shows remarkable consistency between speakers of intended standard German in Germany analyzed in Stratton (2020) and vernacular varieties such as Swabian German (Stratton & Beaman, 2025) and Swiss German (Pheiff, in press), both in terms of the general rate of adjective intensification and the lexical manifestations thereof. Because of this, it seems feasible that at least the rates of adjective intensification among Austrian speakers could fall within a similar range (though, as of now, this remains speculative). Accordingly, we draw on these and related studies as a point of reference when assessing the patterns of L2 acquisition of intensifier variation.

2.2. Constraints on L2 Sociopragmatic Variation

Much like patterns of linguistic variation observed among L1 speakers, learner language also varies in systematic ways (Selinker, 1972). Traditionally, the field of second language acquisition (SLA) has taken native-like performance as a critical yardstick in assessing L2 attainment. This line of inquiry focuses on non-native-like errors in L2 learning, characterized by deviations from target-language norms that are not typically found in the language use of L1 speakers. Since the late 20th century (e.g., Adamson & Regan, 1991), however, a wave of research at the intersection between variationist sociolinguistics and SLA has shifted the object of study from non-native-like errors to a focus on “learner’s ability to use native-speaker sociolinguistic and sociopragmatic variants” (Howard et al., 2013, p. 341). This is because such an ability allows learners to go “[a]bove and beyond knowledge of linguistic structures alone” and “enables humans to bond with others: identifying with others, accommodating to their speech, indicating empathy and solidarity” (Regan, 2010, p. 22)—behavior that is essential in everyday social life. Developmental variation in learner speech with a distinction between target-like and non-target-like forms has been referred to as Type I variation (Rehner, 2002) or learning-related variation (Durham, 2014) whereas the “alternation between forms that are each used by native speakers of the target language” (Rehner, 2002, p. 15) refers to Type II variation (Rehner, 2002) or target-based variation (Durham, 2014). The latter form, which “entails developing competence for variability” (Kanwit et al., 2018, p. 456), informs the current study on adjective intensification.

The growing body of research on L2 intensifier variation has explored intensifier frequency, collocational knowledge and use involving intensifiers, and L2 learners’ preference for different intensifiers (e.g., Edmonds & Gudmestad, 2014; Kanwit et al., 2018; Lorenz, 1998; Pérez-Paredes & Díez-Bedmar, 2012; Philip, 2007; Recski, 2004). It has been frequently attested (e.g., Lorenz, 1998; Philip, 2007; Recski, 2004) that L2 learners, compared to L1 speakers, tend to demonstrate higher rates of intensification, pointing towards what Czerwionka and Olson (2020) characterize as an ‘overuse’ of intensifiers. Others, however, have identified rates consistent with L1 norms. Edmonds and Gudmestad (2014), for example, studied amplifier-adjective collocations among learners of L2 English and concluded that, with increasing L2 proficiency, learners behaved more similarly to L1 speakers, and Kanwit et al.’s (2018) exploration of the alternation between muy [very] and bien [very] in L2 Spanish revealed that L2 speakers adjusted to the intensifier norms of the L2 community during the course of study abroad. Davydova (2024), too, found that her group of advanced German learners of English, especially those with prolonged exposure to naturalistic settings, follow the trend set out by their L1 peers. For example, her L2 sample exhibited rates of intensification similar to those reported for L1 speech communities and adhered to L1-like linguistic constraints such as favoring certain intensifier forms in predicative versus attributive contexts. What is more, Davydova (2024) and Czerwionka and Olson (2020) have found that the L2 learners preferred markers of intensification consistent with the lexical choices of L1 speakers, indicating that L2 learners can acquire a similar degree of intra-individual variation in intensification patterns. Broadly speaking, the current findings would suggest that intensification rates, and also the diversity in intensifiers, increase with increasing proficiency and exposure to target-language input, either mirroring or even exceeding typical rates observed among L1 speakers.

These findings on the effects of proficiency and target-language exposure on intensifier variation are hardly surprising, however. It is well established in variationist SLA that the acquisition of sociolinguistic and sociopragmatic norms is sensitive to differences in proficiency, with advanced learners showing a propensity towards increased engagement with the social nuances of language (e.g., Adamson & Regan, 1991; George, 2014; Howard et al., 2013; Regan et al., 2009). In a related vein, the quantity and quality of naturalistic exposure to authentic target-language input, as well as the diversity in the types and depth of L2 learners’ social interactions with L1 speakers, represent indispensable steppingstones in achieving high L2 performance, also in terms of sociolinguistic and sociopragmatic behavior (Bardovi-Harlig & Bastos, 2011; Gautier & Chevrot, 2015; George, 2014; Howard et al., 2013; Regan et al., 2009; Wirtz & Pfenninger, 2024).

Additionally, a range of affective variables have been linked to more target-like patterns of stylistic and regional variation (e.g., George, 2014). Interest in the target-language community and the linguistic diversity therein and an individual’s anxiety towards speaking the L2, for example, have been found to affect learners’ acquisition of vernacular features in the L2 (e.g., Ender, 2022; Wirtz, 2025; Wirtz & Pfenninger, 2024) and may thus impact the degree of sociopragmatic accommodation to the local intensifier frequency norm in the ambient input. What is more, since intensifier variation is also a socially motivated phenomenon (see the Section 2.1), the degree of external social peer pressure to acquire the L2 and the sociopragmatic norms of the target-language community may affect L2 intensifier frequency, much like such an affective variable is known to affect L2 sociolinguistic development (e.g., Wirtz & Pfenninger, 2024) as well as L2 performance more generally (e.g., Geeslin & Long, 2014).

2.3. This Study

This study approaches L2 intensifier variation from two main perspectives: (1) investigating the relative contribution of individual factors that affect syntactic adjective intensification in L2 German (inter-individual variation); and (2) seeking to gain insights into whether we can systematically explain the diversity in intensifiers L2 speakers employ (intra-individual variation). The following research questions (RQs) and corresponding hypotheses (Hs) were formulated:

• RQ1: What rates of usage characterize L2 learners’ syntactic intensification of adjectives?
  • H1: Because it has frequently been found that L2 speakers engage in intensifier variation at higher rates than those reported for L1 communities (e.g., Lorenz, 1998; Philip, 2007; Recski, 2004; Czerwionka & Olson, 2020), we expect our sample to demonstrate intensification rates at the higher end of what is reported for L1 German.
• RQ2: To what extent do rates of adjective intensification correlate with sociodemographic variables (i.e., length of residence, age, and gender) and individual learner differences (i.e., L2 language proficiency, intensity of exposure to the L2, and L2 socioaffect)?
  • H2: According to various variationist SLA scholars (e.g., Davydova, 2024; Czerwionka & Olson, 2020; Kanwit et al., 2018), target-like patterns of intensification can be expected with prolonged contact with the L1 community, which is why we hypothesize that length of residence may emerge as a significant constraint. Similarly, variationist sociolinguistics work emphasizes that social variables, namely age (Ito & Tagliamonte, 2003; Barnfield & Buchstaller, 2010; Stratton, 2020) and gender (Fuchs, 2017; Stratton, 2020; Stratton & Sundquist, 2022), can constrain adjective intensification; since L2 learners are known to acquire patterns of variation reflective of salient social categories (e.g., Adamson & Regan, 1991; Major, 2004; Meyerhoff & Schleef, 2013), we expect age and gender to influence L2 intensifier variation. Similarly, because it is assumed that increasing L2 proficiency and exposure to the target language (e.g., Davydova, 2024; Kanwit et al., 2018), as well as affective variables (e.g., Czerwionka & Olson, 2020; Ender, 2022; Wirtz, 2025; Wirtz & Pfenninger, 2024), relate to L2 adoption of sociolinguistic and sociopragmatic norms, we expect intensification rates to vary in relation to an individual’s L2 German proficiency, their intensity of exposure to the target language, and their affective orientation towards the L2.
• RQ3: To what extent does the weighted diversity in adjective intensification (i.e., a measure of intra-individual variation) relate to sociodemographic variables and individual learner differences?
  • H3: In line with previous work on adverb–adjective collocations and the lexical diversity in intensification (e.g., Edmonds & Gudmestad, 2014; Kanwit et al., 2018; Lorenz, 1998; Pérez-Paredes & Díez-Bedmar, 2012; Philip, 2007; Recski, 2004), we hypothesize that the diversity in adjective intensification and thus the degree of intra-individual variation demonstrated by the respective L2 learner may relate to differences in sociodemographic variables and individual learner differences, also for the reasons mentioned in H2.

Note that, as will be detailed in Section 3.4, we also coded for syntactic function of the adjectival head, as this has been shown to be a strong constraint on L1 adjective intensifier rates (see, e.g., Stratton, 2020), and L2 learners are known to be sensitive to this linguistic constraint (e.g., Davydova, 2024). We thus included syntactic function in the analyses pertaining to RQs 1 and 2; while it functions by and large as an extraneous variables (i.e., variables that are not at the forefront of the investigation, but that can potentially affect the dependent variable), we do discuss the results pertaining to syntactic function when it aids in explaining the effects of the individual differences presented in RQ2.

The data for this study stem from a larger research project in which the goal was to investigate L2 learners’ acquisition of sociolinguistic variation in the Austro-Bavarian naturalistic learning context. The previous findings from this project have provided insights into how L2 learners integrate standard German and Austro-Bavarian vernacular varieties into their linguistic repertoire (see, e.g., Wirtz & Pfenninger, 2024; Wirtz et al., 2024; Wirtz, 2025). The current study adds to this scope of research by homing in on a specific type of sociopragmatic variable, namely adjective intensification.

3. Materials and Methods

The experimental materials, data, and analysis scripts can be found on the Open Science Framework (OSF): https://osf.io/tk2mu/.

3.1. Participants

We recruited a total of 40 individuals with German as a L2 living in Austria (22 men, 17 women, 1 gender-diverse individual). Thirty-eight speakers were born in L1 English-speaking countries (British Isles: 12; USA: 20; New Zealand: 2; Canada: 2; South Africa: 1; Australia: 1); two participants reported being born in Japan and Peru, respectively, but were raised predominately in the USA after birth. Participants were recruited via fliers and word-of-mouth, and each individual gave their written informed consent before participation. After successful completion of the full test battery (see Wirtz, 2025), participants were financially compensated 20 euros. The L2 learners were in early to mid-adulthood (M = 30.52 years, SD = 8.10, range = 20–57) and had been living in Austria for 4.01 years on average (SD = 3.31 years, range = 0.00–13.83, where 0 indicates a participant who arrived within one month of the time of the study). In terms of occupations, 15 participants classified themselves as some form of student, 4 participants reported being a lecturer of English in some capacity, and the remaining participants pursued diverse careers, including a mobile hydraulics technician, an engineer, an artist, a human resources specialist, a manager, to name a few (for additional information on the participant sample, see the biodata file on OSF).

3.2. Virtual Reality (VR) Experiment

The spoken data for this study were collected via a ‘virtual reality oral dialogue construction task’, which lasted approximately 15 minutes per participant. Peeters (2019) maintains that virtual reality (VR) instruments are especially useful since they allow us to reconcile the ecological validity of interviews and conversation analysis and the experimental control of psycholinguistic designs. VR also does justice to Ushioda’s (2015) call from the field of SLA to carefully delineate and empirically capture what is meant by context. In other words, VR designs constrain the environmental-external factors to those observable in virtual space. What is more, previous work has demonstrated that interactions occurring in virtual space elicit emotional responses similar to those in real-life settings (Moustafa & Steed, 2018), and typical patterns of accommodation as observed in the broader speech community have been elicited via VR (Gijssels et al., 2016; Wirtz, 2022). We therefore argue that VR represents a suitable tool to access L2 learners’ sociolinguistic repertoires in an ecologically valid, and also experimentally controlled, way.

Participants were subjected to two sets of VR oral dialogue construction tasks. Each set involved an interaction with a VR interlocutor who spoke standard German (24 years old, woman) and a VR interlocutor who spoke an Austro-Bavarian dialect variety (27 years old, woman), both of whom grew up in the Bavarian-speaking dialect region of Austria (specifically, Upper Austria). Much like traditional Labovian-style sociolinguistic interviews used to facilitate data collection in variationist designs, the VR tasks were informal in nature (VR set 1: running into a friend outside and VR set 2: meeting a friend for coffee in her apartment). Each VR set began with explicit contextual information presented verbally in English (i.e., participants’ L1 so as not to prime the standard German variety in the instructions) in order to ensure that the participants clearly understood the social-situational backdrop. During the VR tasks, the respective VR interlocutor (participants only interacted with one at a time) asked the participant a series of conversational questions (e.g., about when/where to use Austrian dialects, whether the participant spoke a dialect in their L1, if the participant was planning to stay in Austria, about problems in navigating dialects in Austria, etc.). Each VR interlocutor asked a total of three questions per VR set, resulting in a total of six questions per VR set (for more detailed descriptions of this task, see Wirtz, 2022; Wirtz & Pfenninger, 2024; Wirtz et al., 2024; Wirtz, 2025). Notably, many of the conversational questions targeted meta-linguistic knowledge, which we acknowledge may have influenced L2 learners’ use of vernacularity. The inclusion of such questions in the VR was guided by considerations of research efficiency, as they also served to elicit qualitative, introspective data on participants’ language use (see, e.g., Wirtz et al., 2024).

Note that, while rates of adjective intensification did not systematically differ in relation to VR interlocutor (β = 0.20 ± 0.23, p = 0.38), the diversity in adjective intensification (see Section 3.5) did (β = 0.12 ± 0.06, p = 0.05). We will thus consider L2 learners’ diversity in adjective intensification both at the global level (i.e., combined with both VR interlocutors) and with each VR interlocutor individually, but we neglect this variable when inspecting L2 learners’ rates of adjective intensification.

3.3. Sociodemographic Variables and Individual Learner Differences

The overarching project (see Wirtz, 2025) addressed the issue of how individual learner differences affect the acquisition of standard German and vernacular varieties. Data were therefore collected on the following variables:

• L2 proficiency: Proficiency was measured via two tasks: (1) A translation task, in which participants translated two sentences from standard German into the Austro-Bavarian dialect variety, and another where they translated two sentences from the Austro-Bavarian dialect variety into standard German (for more details, see Wirtz, 2025; Wirtz & Fischlhammer, 2024). Participants could achieve 21 points in each direction (translation into standard German: M = 12.18, SD = 4.51, range = 4–21; translation into the dialect variety: M = 6.92, SD = 4.97, range = 0–20). (2) Self-reports of L2 learners’ proficiency in standard German and dialect, respectively, with respect to reading, writing, listening, and speaking on 100-point slider scales. The four items assessing proficiency in standard German and dialect were averaged such that participants could achieve a score of 100 for each variety (self-reported proficiency in standard German: M = 59.98, SD = 21.17, range = 16–100; self-reported proficiency in the Austro-Bavarian dialect: M = 24.12, SD = 21.11, range = 0–78.75). The self-reported measures and translation scores in both varieties were correlated (standard German: r_s = 0.36, CI = [0.04, 0.63]; dialect: r_s = 0.68, CI = [0.40, 0.86]), which is why the scores were z-scored and subsequently aggregated to operationalize standard German and dialect proficiency, respectively.
• L2 exposure: Intensity of exposure was measured via an adapted version of the Bilingual Language Profile (BLP; Birdsong et al., 2012), namely the Multilingual Language Profile (MLP; see Wirtz & Pfenninger, 2023). The MLP assessed participants’ self-reported language history, use, and contact with respect to standard German and Austro-Bavarian dialect. To operationalize varietal exposure (i.e., exposure to standard German and Austro-Bavarian dialect), the scores from the aforementioned three modules were extracted and weighted in accordance with the instructions for use of the BLP, such that individuals could achieve a total of 54.5 points per module. After aggregation of the three modules to operationalize varietal exposure, participants could achieve a maximum varietal exposure score of 163.5 per variety (intensity of exposure to standard German: M = 33.48, SD = 24.33, range = 4.9–103.01; intensity of exposure to the Austro-Bavarian dialect variety: M = 24.09, SD = 22.80, range = 0–92.11).
• L2 socioaffect: Investigating varietal socioaffect, i.e., L2 learners’ affective orientation towards standard German and the Austro-Bavarian dialect variety, involved exploring a combination of six components, each of which was measured via two-item scales (for more details, see Wirtz, 2025; Wirtz & Pfenninger, 2024): Interest in dialect, interest in standard German, peer encouragement to learn standard German, peer encouragement to learn dialect, anxiety when speaking standard German, and anxiety when speaking dialect. Participants were presented with statement-type, closed-ended items that participants judged on a slider scale from 0 (strongly disagree) to 100 (strongly agree). To operationalize these six socioaffective scales, we averaged the item responses for each scale. Note that the scores for L2 anxiety were reverse coded so that all socioaffective variables can be interpreted consistently, with higher values indicating more favorable conditions (i.e., higher L2 anxiety measures indicate less anxiety in speaking the respective variety).

For our current study on syntactic adjective intensification, the distinction between standard German and Austro-Bavarian dialect is less relevant, as intensifier variation is evident both in spoken standard language varieties (e.g., Stratton, 2020) as well as in vernacular varieties (e.g., Pheiff, in press; Stratton & Beaman, 2025). In order to attain global scores for learners’ L2 proficiency, intensity of exposure to the L2, and their affective orientation towards the target-language (i.e., L2 socioaffect), we averaged the measures for standard German and the dialect variety for each of the three aforementioned domains of individual learner differences. This resulted in three variables: Global L2 proficiency (i.e., the average of L2 learners’ proficiency in standard German and dialect, based on the translation task and self-reports for reading, writing, listening, and speaking), intensity of exposure to the L2 (i.e., the average of learners’ exposure to standard German and the Austro-Bavarian dialect variety based on their scores from the MLP), and L2 socioaffect (i.e., averages of L2 learners’ interest in, peer encouragement to learn, and anxiety when speaking standard German and the Austro-Bavarian dialect variety). We argue that considering both standard German and vernacular varieties in our operationalization of L2 proficiency, exposure, and socioaffect aligns with recent calls in (variationist) SLA to move beyond assessing individual learner differences based on a single, more or less homogeneous standard language norm (e.g., Ender, 2022; Ruck & Shafer, 2020). Instead, it is essential to recognize and account for the diverse usage contexts learners encounter, especially in naturalistic learning environments.

3.4. Speech Coding and Annotation

The speech data collected were first segmented into clausal units using ELAN (2021) and were then manually transcribed quasi-orthographically (see Wirtz, 2025 for details), resulting in a corpus of 3282 clausal units of speech (20,781 words). The data were manually annotated to extract all instances of adjectives within the relevant context of variation. In circumscribing the variable context to intensifiable adjectives, and also for reasons of comparability with related studies of intensifier variation, we excluded adjectives from analysis for the following reasons:

• The adjective was not intensifiable, such as non-gradable adjectives (e.g., berufstätig ‘employed’) or classifiers (e.g., friesische Insel ‘Frisian Islands’) (Ito & Tagliamonte, 2003, p. 264; Stratton & Beaman, 2025, p. 116).
• The adjective was produced in the comparative or superlative form, as such tokens can block intensification (Ito & Tagliamonte, 2003).
• The adjective phrase was negated, since negation can affect the “semantic-pragmatic thrust” (D’Arcy, 2015, p. 459).
• The adjective was proceeded by a focus particle such as sogar ‘even’ (Stratton, 2020, p. 195), or an adjective followed wie ‘how’ (Stratton & Sundquist, 2022, p. 395).

After circumscribing the variable context, each intensifiable adjective was coded based on the presence or absence of an intensifier. At the same time, we coded for syntactic function, that is whether the adjective phrase was in a predicative (e.g., das Auto ist sehr schnell ‘the car is very fast’), attributive (e.g., das sehr schnelle Auto ‘the very fast car’), or adverbial function (e.g., das Auto fährt sehr schnell ‘the car drives very quickly’).

3.5. Data Analysis

To answer RQ1 regarding the rates of usage of adjective intensifiers among L2 learners of German, we present descriptive data based on relative frequencies.

RQ2 asks to what extent sociodemographic variables (i.e., length of residence, age, and gender) and individual learner differences (i.e., L2 proficiency, intensity of exposure to the L2, and socioaffect) correlate with variation in adjective intensification (i.e., the dependent variable). Due to data sparsity by virtue of the comparatively small corpus for the present study (see Section 4.1) and, at the same time, our consideration of a total of six predictor variables, our first step was to identify which variables plausibly constrain rates of intensification. To this end, we turn to classification methods, which are well-suited to model unbalanced datasets and identify the smallest, best-possible, and non-redundant set of constraints that predict the outcome variable. Random forests are one such method. These evaluate a set of randomly permuted conditional inference trees and compute importance measures (Tagliamonte & Baayen, 2012). Following recent strides in quantitative variationist methods (e.g., Beaman, 2025; Beaman et al., under review), we employ a random forest classification method with the Boruta algorithm as a wrapper, which expands the dataset by making copies of the original predictors and reduces its dimensionality by eliminating unimportant predictor variables, the goal being to simplify the model. The Boruta algorithm creates shadow predictors (i.e., copies of the original predictors) and randomly shuffles them so that their distribution remains the same. The original predictors are subsequently compared to the best-performing shadow predictors to determine which ones are most important. Predictors that outperform the best shadow predictors are confirmed; predictors that under-perform are rejected from the subsequent iterations; predictors that can neither be confirmed nor rejected are classified as tentative. Subsequently, we submit the Boruta-confirmed predictors to binomial mixed-effects regression analysis, which aims to elucidate the respective predictive strength of each predictor variable whilst controlling for the effects of the others and accounting for idiosyncratic variance in the data via random effects. In these models, we included individual speaker (40 levels) and adjective lemma (79 levels) as random intercepts in order to mitigate speaker- and adjective-related variance. Figure A1 in the Appendix A visualizes the random intercepts for the null model (i.e., a model only including the random effects), which shows the structural variability associated with these two variables. The fixed-effects structure will be elucidated in the forthcoming sections based on the results of the Boruta classification method.

In order to address RQ3, which questions whether sociodemographic variables and individual learner differences correlate with the diversity in adjective intensification, we adapted the Variation Intensity (VI) measure from Wirtz et al. (in press). VI is calculated as the sum of the relative frequencies of each intensifier used by a single L2 learner (i.e., relative to the other intensifiers used by an individual L2 learner), multiplied by one minus the same value (relative frequency) for that intensifier, resulting in a measure bounded between 0 and 1. VI thus represents a single-score measure of intra-individual variability in the use of linguistic variants, tailored in our case to capture the weighted diversity in adjective intensification by each participant, i.e., the degree of intensifier variation demonstrated by each learner (see the Appendix A for two examples).

The VI measure offers a more nuanced assessment of lexical diversity than the more traditional Type-Token Ratio (TTR). Unlike TTR, which simply captures the proportion of unique intensifiers (types) to total intensifier occurrences (tokens), VI incorporates the relative frequency of each intensifier, providing a weighted measure of variation. This approach accounts for both the richness (number of different intensifiers) and evenness (balance in their distribution) of usage. If one intensifier is used overwhelmingly more than others, TTR might still suggest high diversity as long as there are multiple unique intensifiers. The VI measure, however, penalizes dominance—in other words, if one intensifier is used disproportionately often, VI will be lower, reflecting a lower true diversity in intensifier variation.

As opposed to RQ2, which includes binary data that are most fruitfully modelled via logistic regressions (variable selection was determined via the Boruta algorithm so as not to overload the models), RQ3 includes a single score for participant, and in this case simpler correlational analyses were preferred.

4. Results

4.1. RQ1: Rates of Usage

We recorded a total of 442 tokens of intensifiable adjectives among the 40 speakers of L2 German, of which 193 were intensified (produced by 37 speakers). This means that the overall intensification rate of adjectives is 44%, which is generally on the higher end of what has been found for L1 German speakers (e.g., Stratton, 2020) as well as for L1 speakers of English (e.g., Ito & Tagliamonte, 2003; Tagliamonte, 2008).

We observed 14 lexical manifestations of intensifiers (note that vernacular forms were coded as the standard variant—e.g., a bissl was coded as ein bisschen ‘a bit’). Figure 1 descriptively illustrates how adjective intensifiers rank in terms of frequency among L2 learners. Not dissimilar to previous analyses of adjective intensification in L1 German (e.g., Stratton, 2020; Stratton & Beaman, 2025; Christen, 2003; Pheiff, in press), intensifiers such as sehr ‘very’ and ganz ‘completely’ rank among the most frequently occurring intensifiers, alongside ziemlich ‘rather, quite’ and ein bisschen ‘a bit’. Also not dissimilar to previous work is that the individual lexical items are very unevenly distributed, which is generally in line with Zipf’s law, i.e., that the most common word occurs approximately twice as often as the next common one, etc.

In line with the findings from Stratton (2020), among others, amplification (i.e., scaling a quality upward) was preferred to downtoning (i.e., scaling a quality downward). Excluding the adjective ganz ‘completely’, as its functional status is often ambiguous (see Stratton, 2020, p. 200), 70% (n = 113) of the intensifiers were employed in an amplifying function. Example (1) provides several excerpts from the most frequent amplifiers in the corpus:

(1)	a.	Das Leben hier ist sehr gut. (07_t_41_c)
		‘Life here is really good.’
	b.	Es war so schwierig am Anfang. (10_m_43_c)
		‘It was so difficult at the beginning.’
	c.	Salzburg ist eine super schöne Stadt. (05_k_36_c)
		‘Salzburg is a super pretty city.’
	d.	Das ist wirklich eine voi andere Sprache. (16_a_25_c)
		‘That is really an entirely different language’

Among the 14 lexical manifestations, three were consistently used as downtoners (30%, n = 49), specifically ein bisschen ‘a bit’, ziemlich ‘rather, quite’, and relativ ‘relatively’, as Example (2) demonstrates:

(2)	a.	Es ist a bissl schwer. (11_t_31_c)
		‘It’s a bit difficult.’
	b.	Dialekt ist ein bisschen schwierig zu verstehen. (17_n_28_c)
		‘Dialect is a bit difficult to understand.’
	c.	Die meisten Leute in den USA sprechen ziemlich gut Englisch. (07_t_41_c)
		‘Most people in the USA speak English pretty well.’
	d.	I find des eigentlich relativ blöd teilweise. (15_p_30_c)
		‘I actually find that to be relatively stupid sometimes.’

Turning to rates of adjective intensification in relation to syntactic function, we find 248 instances of intensifiable adjectives in predicative function (133, i.e., 54% were intensified), followed by 153 instances in attributive function (32, i.e., 21% were intensified), and finally 41 instances in adverbial function (28, i.e., 67% were intensified). In order to determine whether intensification varied systematically in relation to syntactic function, we computed a binomial mixed-effects regression model considering syntactic position as the sole predictor and individual learners and adjective lemma as random intercepts. As Figure 2 shows, intensifier expressions of adjective phrases in attributive function were predicted to be the lowest. There was a modest difference in the rates of intensification of adjective phrases in predicative and adverbial function (β = −0.81 ± 0.43, p = 0.06), though this difference was likely artificially inflated by the low count of adjectives in adverbial function. Given this, and also to reduce model complexity, we collapsed intensifier expressions for these two categories into a single factor level (see also Pheiff, in press for a similar approach).

4.2. RQ2: INTER-Individual Variation

In order to address the issue of inter-individual variation (i.e., variables predicting intensifier variation), we first attempted to narrow down the pool of potentially influential variables by conducting random forests with the Boruta algorithm (Kursa & Rudnicki, 2010; for sociolinguistic application, see, e.g., Beaman, 2024; Beaman et al., under review), which classifies predictors as either important, unimportant, or undecided in terms of their predictive force on the outcome variable (i.e., intensifier variation). The Boruta algorithm suggests the three individual learner differences—as opposed to the three sociodemographic variables—as being the strongest constraints on intensifier variation, in addition to the syntactic function of the adjectival head (for visual inspection see Figure 3). These three variables (in interaction with syntactic function, i.e., attributive versus predicative/adverbial), consequently, were subjected to regression analysis in order to test their predictive power in interaction with syntactic position.

When integrating the three individual learner differences into a binomial mixed-effects regression model and thus accounting for potential participant- and adjective lemma-related idiosyncrasy via random effects, we found only marginal effects of global L2 proficiency on differences in adjective intensification (see Figure 4), with more proficient learners tending to increasingly intensify in the context of attributive adjectives (β = 0.56 ± 0.42, CI = [−0.26, 1.37], p = 0.18) and those used predicatively or adverbially (β = 0.46 ± 0.27, CI = [−0.06, 0.98], p = 0.09). We also found a marginal negative effect of L2 socioaffect on the intensification of adjectives in predictive or adverbial function (β = −0.27 ± 0.17, CI = [−0.59, 0.05], p = 0.10). Notable effect sizes were identified for the effect of intensity of L2 exposure on intensification of attributive adjectives (β = 0.45 ± 0.34, CI = [−0.21, 1.11], p = 0.18). That said, all of the observed effect sizes were generally weak, and the correlations non-significant, indicating that these relationships need to be treated with caution. What is more, given the small size of the corpus, we did not further reduce statistical power by computing separate models for amplifiers and downtowners; we therefore acknowledge that the current results must be interpreted with special care given that some scholars (e.g., D’Arcy, 2015; Stratton & Beaman, 2025) maintain that amplifiers and downtoners may not always be functionally comparable.

The highest—albeit non-significant—effect size was observed for the influence of global L2 proficiency on increasing rates of usage of intensifiers in the context of attributive adjectives. To illustrate this trend, Example (3) provides several excerpts from the corpus as produced by L2 learners with above-average levels of proficiency (i.e., above the average in this sample of learners), contrasted with similar excerpts in Example (4) from L2 learners with below-average L2 proficiency.

(3)	a.	Man redet Dialekt, wenn man weiß, dass mit ziemli guter Wahrscheinlichkeit die andere Person aus der gleichen Gegend kommt. (05_k_36_c)
		‘One speaks dialect if one knows that the other person, with a pretty high probability, is from the same region.’
	b.	In südamerikanische Staaten […] haben [sie] teilweise relativ starken Dialekte (15_p_30_c)
		‘In southern American states [they] sometimes have relatively strong dialects.’
	c.	Da spricht man eigentlich sehr klares Englisch. (13_e_21_c)
		‘There one speaks pretty clear English.’

(4)	a.	Ja ein paar Leute sprechen Ø starke Dialekt. (17_k_24_c)
		‘A few people speak a Ø strong dialect.’
	b.	Ich denke, dass Salzburg eine Ø schöne Stadt ist. (09_m_21_c)
		‘I think that Salzburg is a Ø pretty city.’
	c.	Im Süden sprechen sie ein Ø schwieriger Dialekt. (17_a_21_c)
		‘In the south they speak a Ø difficult dialect.’
	d.	Ich spreche mit ein ganz normale nördliche Amerikaner Dialekt. (20_c_31_c)
		‘I speak with a completely normal northern American dialect.’
	e.	Wir haben so viele Menschen in London. (07_a_40_c)
		‘We have so many people in London.’

One might note the clear differences in the syntactic complexity of the utterances in addition to the less-proficient L2 learners’ noticeable lack of use of intensifiers of adjective phrases in attributive function. When less-proficient learners did intensify attributive adjectives, it was typically found for instances of frequent phrases (i.e., reproduction of chunks; see Ellis, 1996), as shown in (4d), and for syntactic intensification patterns similar to the L1 (i.e., English), as shown in (4e). Note that coding for frequency of intensifier–adjective combinations would have exceeded the scope of the current analyses, so our conclusions regarding frequency effects remain qualitative in nature and speculative.

Figure 4 also illustrates that intensification of adjectives in predicative or adverbial position increases—again, non-significantly—in relation to global L2 proficiency. To exemplify this general trend, Example (5) provides excerpts of predicative adjectives among high-proficiency learners, and Example (6) several instances as used by low-proficiency learners.

(5)	a.	Ich find MLE [Multicultural London English] total schön. (05_k_36_c)
		‘I find MLE [Multicultural London English] totally beautiful.’
	b.	Der Dialekt is halt voll witzig. (13_e_21_c)
		‘The dialect is just fully funny.’
	c.	Vor allem die Stadt Salzburg find i eigentlich a relativ schön. (15_p_30_c)
		‘Especially the city of Salzburg I actually find to be relatively pretty.’

(6)	a.	In Schottland es is Ø schwierig. (07_a_40_c)
		‘In Scotland it is Ø difficult.’
	b.	Aber die Unterschied is Ø klein. (11_t_31_c)
		‘But the difference is Ø small.’
	c.	Es [der Dialekt] ist … aber ein bisschen schwierig für gut Deutsch lernen. (07_a_40_c)
		‘It [the dialect] is … a bit difficult for good German learning.’
	d.	Ja das is ganz normal. (11_t_31_c)
		‘Yes, that is completely normal.’
	e.	Österreich is sehr schön. (17_t_42_c)
		‘Austria is very beautiful.’
	f.	Weil den USA is so groß. (15_m_26_c)
		‘Because the USA is so big.’

Whereas Example (5) demonstrates the diversity of intensifiers used among high-proficiency learners, excerpts (6a) and (6b) show that use of intensifiers is often avoided among low-proficiency L2 learners. When low-proficiency learners did intensify adjectives in predicative function, it was, again, found primarily in cases of frequent phrases (excerpt [6d]) and when the intensification pattern resembled structures especially common for their L1 English (excerpts [6e] and [6f]).

4.3. RQ3: INTRA-Individual Variation

Finally, we questioned whether intra-individual variation in adjective intensification—that is, the weighted diversity in lexical forms L2 learners employed to intensify an adjective—may relate to any sociodemographic factors or individual learner differences. We adapted Wirtz et al.’s (in press) ‘variation intensity’ measure in order to elucidate how variable each learner is in terms of intensifier use. Based on their formula, L2 learners employing intensifiers only infrequently and without much lexical diversity receive lower scores, whereas L2 speakers who frequently make use of a wide range of intensifiers will receive higher scores. Recall that, in Section 3.2, we established that the diversity in adjective intensification differed between VR interlocutors, with a higher diversity in interaction with the standard German-speaking VR interlocutor. In what follows, we assess whether the individual difference variables correlate with the lexical diversity in adjective intensification at the global level (i.e., combined based on data from both VR interlocutors, for which we focus on the 37 individuals who produced intensifiers; see Section 4.1), as well as with each individual VR interlocutor. Note, however, that not each L2 learner intensified an adjective with each VR interlocutor; thus, the analyses of the diversity in adjective intensifiers with the individual VR interlocutors rely on subsets of data, each with n = 33.

At the global level, Pearson correlational analyses (Figure 5) revealed weak positive correlations (albeit non-significant) between higher diversity in adjective intensification and length of residence in Austria (r = 0.23, CI = [−0.10, 0.51], p = 0.17), intensity of L2 exposure (r = 0.19, CI = [−0.14, 0.49], p = 0.25), global L2 proficiency (r = 0.22, CI = [−0.11, 0.51], p = 0.18), and participant age (r = 0.25, CI = [−0.08, 0.53], p = 0.14). Smaller effects were found for the relationship between diversity in adjective intensification and L2 socioaffect (r = −0.17, CI = [−0.47, 0.16], p = 0.32) and gender (t(33) = 0.85, CI = [−0.09, 0.24], p = 0.40), where women demonstrated a slightly higher, albeit non-significantly so, mean (M = 0.53) compared to men (M = 0.47).

When considering L2 learners’ diversity in adjective intensification disaggregated by the VR interlocutor condition, we found no positive correlations in the context of the dialect-speaking VR interlocutor. With the standard German-speaking VR interlocutor (see Figure 6), however, we found significant positive correlations between length of residence in Austria (r = 0.48, CI = [0.17, 0.71], p = 0.004), intensity of L2 exposure (r = 0.35, CI = [0.01, 0.62], p = 0.04), participant age (r = 0.41, CI = [0.08, 0.66], p = 0.02), as well as a marginal correlation with L2 proficiency (r = 0.33, CI = [−0.01, 0.60], p = 0.06).

5. Discussion and Conclusions

In this study, we set out to document, for the first time, the rates of, and the lexical diversity in, adjective intensification among L2 learners of German (L1 English). We additionally attended to the issue concerning whether sociodemographic variables (i.e., length of residence, age, and gender) and individual learner differences (i.e., L2 language proficiency, intensity of exposure to the L2, and L2 socioaffect) can predict (a) the inter-individual variation in syntactic adjective intensification, and (b) the observed intra-individual variation based on a weighted measure of intensifier lexical diversity. We now discuss these issues in accordance with our three RQs and corresponding hypotheses.

5.1. RQ1: General Patterns of Syntactic Adjective Intensification

We first sought to determine at what rates learners of L2 German intensify adjectives, and how the lexical diversity in their intensification practices measures up to what has been reported for L1 German communities. Of the 442 intensifiable adjectival heads in our corpus, 193 were intensified, which corresponds to an intensification rate of 44%. In spoken (standard) German among speakers from Germany, Stratton’s (2020) analysis revealed an intensification rate of 37%; Stratton and Beaman’s (2025) analysis of Swabian German revealed a rate of 27%; Pheiff’s (in press) analysis of Swiss German revealed a rate of approximately 22%. The L2 learners in the present study, it would seem, demonstrate comparatively high rates of intensification. This confirms our first hypothesis, namely that, as has been frequently attested (e.g., Lorenz, 1998; Philip, 2007; Recski, 2004), L2 learners often ‘overuse’ intensifiers when compared to patterns observed in L1 speech communities (Czerwionka & Olson, 2020). Several rationales for this overuse have been put forth. According to Philip (2007), it may be a way for learners to enhance their (perceived) fluency in a communicative instance, essentially functioning as a compensatory strategy to reinforce their meaning due to limited vocabulary or pragmatic awareness. Czerwionka and Olson (2020, p. 24) also maintain that an overreliance on intensifiers in the L2 may result from “the desire to communicate emotion and involvement with others, key pragmatic meanings associated with intensifiers that positively impact interactions” (see also, e.g., Baños, 2013). Especially within a sample of L2 learners of German with L1 English, the higher rates may also reflect the fact that the syntactic patterns involving intensifiers are cross-linguistically similar, thus encouraging transfer from the L1 to the L2 (we revisit this more extensively in the next section). Relatedly, the pragmatic meaning of intensifiers is comparable between languages, in that they function as pragmatic devices which can express heightened emotion and involvement. Given the informal environment of data collection (i.e., via a VR experiment), the observed overuse may also reflect L2 learners’ attempt to communicate emotion, subjectivity, and involvement; indeed, as Taguchi (2022, p. 326) notes, VR tasks require “greater social-interpersonal demands” and invoke strong affective responses in participants (see also Taguchi, 2021). That said, we must bear in mind that the reference studies outlined here stem from different corpora (see Section 2.1), which emphasizes that any comparisons in terms of rates of intensification must be made with caution.

Despite the general overuse of intensifiers, we identified two trends consistent with L1 speech communities. First, rates of amplification (i.e., scaling a quality upward) and downtoning (i.e., scaling a quality downward) were generally consistent with reports from the literature attesting a 70% amplifier and 30% downtowner divide in L1 German (see, Stratton, 2020, p. 200; Stratton & Beaman, 2025, p. 122; though cf. Pheiff, in press, who found amplification rates of approximately 80%). This finding moreover stands in agreement with the cross-linguistically stable insight that amplifier expressions are more frequent than downtoners in terms of both tokens and types (D’Arcy, 2015, p. 460; Pintarić & Frleta, 2014, p. 44; Stratton, 2020, p. 200; 2021, p. 41; Stratton & Sundquist, 2022, p. 400; Van Olmen, 2023, pp. 26–27). Second, the lexical manifestations of intensifiers and their relative frequencies bear clear parallels to reports from L1 German communities. In line with Stratton (2020), Stratton and Beaman (2025), and Pheiff (in press), intensifiers such as sehr ‘very’ and ganz ‘completely’ rank among the most frequent, alongside ziemlich ‘rather, quite’, ein bisschen ‘a bit’ and so ‘so’.

5.2. RQ2: Predictors of Adjective Intensification

Our second goal was to determine whether any of the sociodemographic variables (i.e., length of residence, age, and gender) or individual learner differences (i.e., L2 language proficiency, intensity of exposure to the L2, and L2 socioaffect) can predict the inter-individual (i.e., between-person) variation in adjective intensification. The Boruta algorithm (i.e., a classification method which aims to provide the best-possible and non-redundant set of constraints that predict the outcome variable) suggested that the effects of the sociodemographic variables were negligible, while the individual learner differences may more successfully predict intensification among L2 speakers of German. We thus submitted the three individual learner difference variables to mixed-effects regression analysis in order to determine their respective predictive power while controlling for the effects of the other covariates and accounting for idiosyncratic variance in the data via random effects (i.e., issues which are not taken into account in the Boruta algorithm). As it turns out, the effects of these three variables—in interaction with the syntactic context of the adjectival head—were non-significant, which, strictly speaking, refutes our second hypothesis. That said, several of the effect sizes were notable; given the data sparsity by virtue of the small corpus, we find it prudent to discuss some of the more pronounced trends we observed and to provide potential explanations for them, also because the Boruta analysis did indicate that these factors are not irrelevant where L2 adjective intensification is concerned.

The most pronounced—but again, non-significant—trends were identified in relation to L2 learners’ intensification of adjectives in relation to global L2 proficiency. That is, more proficient learners tended to engage in higher rates of intensification. This finding is generally in line with previous variationist SLA work attesting that more proficient learners will accommodate, at least to some extent, to the sociopragmatic norms of the L2 community (e.g., Davydova, 2024; Edmonds & Gudmestad, 2014; Kanwit et al., 2018; Czerwionka & Olson, 2020), even if they do overshoot in terms of intensification rates. Furthermore, the trend that higher L2 proficiency favors intensification of adjectives in attributive context is conform with insights from diverse SLA scholars focusing on L2 grammatical variation. For example, adjectives in attributive function typically require morphological agreement and specific word order constraints—features of the target language which are acquired later in the developmental process. Vyatkina et al. (2015) exemplified in her corpus of written texts by L2 learners of German that beginning learners primarily rely on the predicative use of adjectives, whereas increasingly proficient learners increase their rates of usage of adjectives in attributive function. This trend is not necessarily restricted to L2 German, either, as her findings confirm results from other L2 contexts (e.g., L2 French, L2 English) that associate high frequencies of predicative adjectives with lower L2 proficiency (e.g., Hinkel, 2002) and those of attributive adjectives with higher L2 proficiency (Granfeldt & Nugues, 2007; Grant & Ginther, 2000). Our results add to this scope of research, corroborating that, not only does the use of attributive adjectives generally increase in relation to L2 proficiency, but so too does the ability to augment them via sociopragmatic devices (i.e., intensifiers).

While non-significant, the finding that L2 socioaffect was negatively correlated with adjective intensification was indeed curious, although not necessarily unprecedented in the grander scheme of SLA. Specifically, low L2 socioaffect was associated with extreme overuse of intensifiers, with the model predicting rates as high as 70% in the context of predicative adjectives. L2 learners with a more pronounced affective orientation towards the L2 were, in contrast, predicted to intensify predicatively used adjectives at rates of approximately 40% (see Figure 4). As we see it, L2 learners reporting higher affective orientation towards the target language may exhibit greater sensitivity to social norms and communicative appropriateness in the target language. Since intensification can carry emotional weight—expressing emphasis, exaggeration, or heightened affect (e.g., Baños, 2013)—those who are more socially attuned may avoid overuse to maintain perceived politeness and appropriateness during interactions in the L2. In other words, learners reporting higher affective engagement with the L2 are likely to prioritize pragmatic adaptability, leading them to opt for more measured and contextually appropriate expressions rather than overly frequent intensification. This trend aligns well with insights on the effects of related psychological and affective variables on L2 performance; specifically, factors relating to cultural empathy, open-mindedness, and flexibility are associated with high performance in terms of verbal fluency, vocabulary, phrasal knowledge, and more generally with successful intercultural adaptation and L2 learning (e.g., Forsberg Lundell et al., 2018; Moyer, 2021; Ożańska-Ponikwia & Dewaele, 2012). For example, in their study of late French L2 learners, Forsberg Lundell and Sandgren (2013) emphasized that individuals “who had an open mind and were able to empathize with others led to their engagement with successful learning strategies (including exposing themselves to the L2 community), in order to integrate well professionally and socially” (p. 249).

Collectively, our inter-individual analyses demonstrate that lower-level learners (and, similarly, those with low affective engagement with the L2, i.e., low scores on L2 socioaffect), are less sensitive to target-like frequency (i.e., they tend to either under- or overuse intensifiers compared to L1 speakers) compared to more advanced language learners. One pattern that our qualitative inspection of participant utterances revealed was that, in cases where low-proficiency learners intensified adjectives—both in predicative and attributive function—they did so by way of frequent phrases (e.g., ganz normal ‘completely normal’) and/or when the intensification pattern resembled structures especially common for their L1 English (e.g., sehr schön ‘very beautiful’, so groß ‘so big’). The latter effects, we hypothesize, may be a result of cross-linguistic influence, a finding which is by no means unheard of in variationist SLA. Spina et al. (2025), for example, who investigated L2 learners of Italian in South Tyrol (the majority of whom spoke L1 German) found that a dominant German-speaking linguistic environment was a significant predictor of learners’ preferences for a syntactic over a morphological intensification type. Among Croatian learners of L2 English, Lovrović and Pintarić (2019) similarly maintain that their students employed common adverb–adjective collocations in their L2 English as found in their L1 Croatian; this again points to potential mechanisms of cross-linguistic influence as speakers learn to harness sociopragmatic variation in their L2.

5.3. RQ3: Explaining Intra-Individual Variation in Adjective Intensification

Perhaps the most unexpected finding emerged in relation to the significant differences in the lexical diversity in adjective intensification as a function of the VR interlocutor. Indeed, this analysis did not feature in our original RQs for several reasons: (1) We were interested in L2 learners’ global diversity in adjective intensification, i.e., regardless of potential external influences such as VR interlocutor; (2) rates of adjective intensification did not significantly differ between interlocutors; and (3) in no other analyses has this external factor constrained these L2 learners’ linguistic behavior (see, e.g., Wirtz et al., 2024; Wirtz, 2025). Here, however, we found that the diversity in adjective intensification was higher in interaction with the standard German-speaking VR interlocutor. It seems feasible that we may be observing certain psycholinguistic processes at play. Specifically, standard German is the variety typically employed in, for instance, testing situations, classroom settings, formal interactions, etc., in which enhanced attention to form is encouraged, often resulting from the increased cognitive complexity of the respective activity. From psycholinguistic and SLA research, we know that lexical variety benefits from this increased attention to form induced by task complexity (e.g., J. C. Mora et al., 2024; Robinson, 2001), which may explain the effects of VR interlocutor in terms of L2 learners’ lexical diversity in adjective intensification. Importantly, while the overall rate of intensification did not differ by VR interlocutor, lexical diversity requires not only frequent use but also variation across types. This distinction may help explain why diversity—but not frequency—was affected by VR interlocutor condition.

The third and final question we asked was whether intra-individual variation, operationalized as the diversity in adjective intensification based on Wirtz et al. (in press) measure of variation intensity, can be predicted by the same set of sociodemographic variables and individual learner differences as used in the inter-individual analyses. At the global level (i.e., variation intensity computed on the basis of L2 learners’ data in interaction with the standard German-speaking and dialect-speaking VR interlocutor), we found only weak and non-significant correlations between the sociodemographic variables and the individual learner differences of interest, which, under a strict interpretation, refutes our third hypothesis. Again, in light of the data sparsity, and also because the weak effects we observed at the global level turned out to be significant in interaction with the standard German-speaking VR interlocutor, a brief discussion of the most prominent trends and their potential explanations seems warranted.

Length of residence in Austria and speaker age exhibited the comparatively strongest positive effects. The effect of length of residence is intuitive when considering previous variationist SLA findings that prolonged exposure to L1 input and increased opportunities for interaction in varied social settings are often associated with more successful acquisition of stylistic and pragmatic variation (e.g., Bardovi-Harlig & Bastos, 2011; Davydova et al., 2017; Howard et al., 2013; George, 2014). Our finding also appears reminiscent of the trend that collocational knowledge and lexical diversity are especially susceptible to development in relation to extended stays abroad (e.g., Edmonds & Gudmestad, 2014, 2021; Czerwionka & Olson, 2020). Similarly, the positive—albeit non-significant at the global level—effect of age implies that older speakers tend to demonstrate greater intra-individual variation in adjective intensification, especially in interaction with the standard German-speaking VR interlocutor. Arguably, this may relate to differences in cumulative linguistic experience, i.e., the relationship between cognitive aging and the linguistic repertoire. In this vein, we know that certain cognitive abilities, specifically those relating to fluid intelligence such as working memory, processing speed, and spatial ability decline across the lifespan; meanwhile, crystallized functions such as verbal ability, vocabulary depth, the ability to draw on previously acquired knowledge and experiences, and comprehension tend to increase well into late adulthood (e.g., Park & Reuter-Lorenz, 2009; Murman, 2015; Rohwedder & Willis, 2010; Salthouse, 2010). Indeed, developmental psychologists have found that older adults generally demonstrate richer vocabulary and employ more complex grammatical structures compared to younger individuals (e.g., Luo et al., 2019). This cumulative acquisition of vocabulary, arguably also including pragmatic devices such as intensifiers, likely transfers to the L2 as well, which may aid in explaining the, albeit weak, correlations between intra-individual variation in adjective intensification and speaker age.

5.4. Limitations and Future Directions

The present paper has made first strides in documenting the extent to which L2 learners of German acquire patterns of adjective intensification reported for L1 communities. There are, however, several (methodological) limitations which warrant consideration.

First, our corpus was built based on an experimental task originally intending to assess the L2 acquisition of standard German and Austro-Bavarian vernacular varieties. While the VR experiment was useful in facilitating an informal, conversational setting (see also Wirtz, 2022), arguably representing an optimal site for the analysis of adjective intensification (e.g., Ito & Tagliamonte, 2003; Stratton, 2020), the VR task was relatively short (approximately 15 minutes per participant) and thus resulted in a comparatively small corpus of intensifiable adjectival heads. The present dataset was nevertheless advantageous—and was thus chosen—because it facilitated first hints as to how migrants living in Austria acquire patterns of intensification, and came with an extensive battery of sociodemographic variables and individual learner differences that could be subjected to systematic correlational analyses.

Second, comparisons of L2 learners’ acquisition of intensifier variation with rates of intensification among Austrian speakers could not be made given the dearth of research on the matter in Austria-centered sociolinguistics. Considering the accumulating interest in intensifier variation in L1 German communities (e.g., perceived standard German in Germany (Stratton, 2020), Swabian German (Stratton & Beaman, 2025), and Swiss German dialects (Pheiff, in press)), findings from the Austro-Bavarian and Austro-Alemannic context will undoubtedly aid in identifying salient cross-varietal patterns.

Finally, it has been asserted that L2 overuse of certain intensifiers can negatively impact communication. Philip (2007) and Czerwionka and Olson (2020) among others maintain that this overuse may be identified as inappropriate or unexpected, too colloquial, or communicating a sense of overstatement, but that these negative effects can often be mitigated by an increase in intensifier lexical diversity—a trend we also observed among high-proficiency learners. These assumptions provide us with a number of testable hypotheses. Much like DuBois (2023) questioned whether L2 learners’ use of colloquial lexical items affects L1 speakers’ judgements of their proficiency, future research might explore the extent to which L2 overuse of intensifiers is perceived to hinder effective communication in the target language. It is of additional interest whether different rates of intensification, varying in lexical diversity, modulate these perceptions. Such studies will necessarily address topics of inter- and intra-individual variation, L2 use of lexical and pragmatic devices, and, more broadly, the potential existence of ‘double standards’ in terms of whether L2 speakers suffer the same negative consequences of colloquial language as do L1 speakers without reaping the same benefits.