Unstressed versus stressed German additive auch – what determines a speaker’s choice?

1 Introduction

Particles like English also and too, Italian anche, and German auch (e.g., Sæbø 2004 for English; Andorno and De Cesare 2017; De Cesare and Andorno 2015, for Italian; Büring and Hartmann 2001; Dimroth 2004; König 1991; Reis and Rosengren 1997; Krifka 1999; Sudhoff 2010 for German) apply their additive meaning to a smaller or larger constituent of the sentence in which they appear. This constituent is called the associated constituent (AC), and the particle marks its addition to a set of contextually relevant discourse alternative(s) for which some common claim is made. In (1), the AC is marked by square brackets and the set of alternatives is given in braces.

The identification of the AC thus depends on the availability of discourse alternatives and speakers’ needs to explicitly express the additive relation between alternatives and the AC. The particles themselves typically also contribute to the identification of the AC. Italian anche, for example, is nearly always placed left adjacent to its AC (e.g., Andorno and De Cesare 2017; De Cesare and Andorno 2015). In German, the situation is slightly more complicated. In particular, there is an unstressed (auch) and a stressed (AUCH) variant of the particle that seem at first glance to be exchangeable. Compare examples (2a) and (2b)

Note that both versions of auch are perfectly fine in the context above. However, speakers differ with respect to their preference for using unstressed or stressed auch. It is the aim of the present paper to investigate what guides a speaker’s decision to use one version of the particle over the other. From the literature we know that a couple of factors play a role, such as the involvement of alternatives (Reimer and Dimroth 2021), the information structural status of the AC (Dimroth 2004; Krifka 1999; Sæbø 2004), and the syntactic function of the AC (Höhle et al. 2009). Since some of these factors are intermingled, we ask the broader question of whether speakers rely on contextual information or on default mechanisms when they plan their utterance containing the additive particle auch. After giving an overview on German unstressed and stressed auch, we will present a sentence fragment arrangement task which shows that although speakers strongly rely on default mechanisms, they are also influenced by syntactic and information structural properties of the context.

2 Unstressed versus stressed auch

We will start with the unstressed particle auch. Since German has flexible word order, auch and its AC can be positioned in the prefield or in the middle field. Unstressed auch directly precedes the AC,^[1] independently of its syntactic function (subject in [3], direct object in [4]) and independently of word order (middle field in the a-variants, prefield in the b-variants).

In (3), situations with the descriptive properties ‘having eaten peaches’ are the common denominator applying to both, the AC and its discourse alternative; in (4), it is ‘being eaten by Peter’. In all cases, the AC carrying a focus accent is analyzed as the information focus of the sentence containing the particle, and the discourse alternatives in the context sentence function as focus alternatives. That the AC is the focus of the utterance is a feature that unstressed auch shares with the focus particle nur (‘only’; Nur [Maria] hat Pfirsiche gegessen).

Let us now turn to the second version of the focus particle, stressed AUCH. A superficial comparison reveals that AUCH follows its AC instead of preceding it (see Sudhoff 2010 for other patterns), independently of its syntactic function (subject in [5], direct object in [6]) and word order (middle field in the a-variants, prefield in the b-variants).

A comparison of the examples under (3)/(4) with those under (5)/(6) reveals that the position of the particle relative to the AC (preceding vs. following the AC) correlates with its intonational realization (unstressed vs. stressed). When AUCH is stressed, the AC of AUCH is not the focus of the utterance, but the topic (Maria in [2b]). What kind of topic, however, is a matter of debate; the AC of AUCH is sometimes analyzed as a contrastive topic^[2] (Dimroth 2004; Krifka 1999; Sæbø 2004; Sudhoff 2010), and sometimes as an aboutness topic (Dimroth 2004; Sudhoff 2010). For reasons of space, we will not go into further detail here. Compared to unstressed auch, stressed AUCH cannot be replaced by nur, since there is no accented version of the exclusive particle nur (*[Maria] hat NUR Pfirsiche gegessen).^[3] Thus, auch is more flexible than the particle nur.

Furthermore, both variants of the particle seem equally possible when the AC is the subject or the object of the sentence – independent of word order (prefield or middle field).^[4] Nevertheless, speakers have to decide whether to use unstressed or stressed auch when they construct an utterance, and they can either rely on structures that they were faced with prior to making the decision of using unstressed or stressed auch, or they can rely on default mechanisms. The former case is known as structural or information structural priming (see Branigan 2007; Ziegler et al. 2019; Ziegler and Snedeker 2019 for structural priming; see Fleischer et al. 2012 for information structural priming). Specifically, after hearing a context sentence that contains a focus particle and an alternative (e.g., Nur [Peter] hat Pfirsiche gegessen with the alternative Peter) a speaker’s decision to use unstressed or stressed auch can be guided by the position of the particle in the context sentence ( nur [Peter] → auch [Maria]), or by the information structural status of the alternative ([Peter] is integrated as a focus, due to the focus particle nur). Furthermore, they can be guided by the syntactic function of the alternative, in that unstressed auch is preferably used if the AC (following the syntactic function of the alternative) is construed as the object of the sentence, and stressed AUCH if the AC (again like the alternative) is the subject of the sentence (see Höhle et al. 2009). Note, however, that subjects are prototypical topics (Reinhardt 1981), and objects are likely to be the focus of the utterance. Consequently, we cannot disentangle the impact of information structure from the impact of the syntactic function. It is rather a concatenation of circumstances that arises by default: Maria is the subject, which preferably has the information structural status of a topic, which preferably is the AC of stressed AUCH.

In this study, we asked whether speakers are guided by (information) structural properties of the preceding context, or whether they rely on default patterns when they construct an utterance containing auch. In an experiment with a sentence fragment arrangement task (see Gauza 2018), we presented participants with a context sentence containing the particle nur (e.g., Nur Peter hat Pfirsiche gegessen, ‘Only Peter ate peaches’) and with the fragments necessary to build a target sentence containing the particle auch (e.g., Auch Maria hat Pfirsiche gegessen, ‘Maria ate peaches, too’) in written form. While the context, containing the alternative Peter, was thus presented as a whole sentence, participants had to arrange the fragments to gain a meaningful target sentence. Crucially, since the particle auch is more flexible than the particle nur, participants had the choice to integrate auch in a way that it resulted in the unstressed ( auch [Maria]), following the example of the context sentence, or in the stressed version ([Maria] AUCH ).

We had two conditions: a subject and an object condition. In the subject condition with the context sentence Nur [Peter] hat Pfirsiche gegessen, the alternative Peter is the subject, but has the information structural status of a focus and is preceded by the particle. If speakers are guided by the information structural status of the alternative and by the structure, that is, the position of the particle relative to the alternative, they should use unstressed auch, which precedes its AC, which in turn is the focus of the utterance ( Auch [Maria] hat Pfirsiche gegessen). If, however, they rely on default associations, independent of the context, they should integrate the subject as a topic and use stressed AUCH ([Maria] hat AUCH Pfirsiche gegessen). In the object condition with the context sentence Peter hat nur [Pfirsiche] gegessen, the alternative Pfirsiche has the information structural status of a focus, and is preceded by the particle. There is no reason to use stressed AUCH here, so we expect speakers to show a clear preference for unstressed auch (although it is technically possible to use stressed AUCH here, too). Unfortunately, since there is no stressed version of the focus particle nur, it was not possible to manipulate the factor information structure in this context. Table 1 summarizes the conditions of the experiment and the possible outcomes of the target sentence containing auch.

3 Methods

4 Results

We observed six different orderings, three orderings for the subject condition and three for the object condition. In the subject condition, one ordering corresponded to the use of stressed AUCH ([Maria] hat AUCH Äpfel gegessen), and two to unstressed auch ( auch [Maria] hat Äpfel gegessen, Äpfel hat auch [Maria] gegessen). In the object condition, two orderings corresponded to the use of unstressed auch (Maria hat auch [Äpfel] gegessen, auch [Äpfel] hat Maria gegessen), and one to stressed AUCH ([Äpfel] hat Maria AUCH gegessen). Crucially, in the majority of the sentences of the subject condition, the particle followed its AC (69%), which corresponds to stressed AUCH. In the majority of the sentences of the object condition, the particle preceded its AC (92.8%), which corresponds to unstressed auch. We created a new variable and, following Jaeger (2008), performed a general linear model on the unstressed realizations in R (R core Team 2017, package lme4, Bates et al. 2015). The fixed-effects factor was syntactic function (subject vs. object). The results show a significant main effect of syntactic function (β = −3.36, SE = 0.28, t = −12.01, p < 0.001), indicating that unstressed auch is chosen more often when the AC has the syntactic function of the object, than of the subject (see Figure 1).

Figure 1:

The realization of unstressed auch, depending on the syntactic function of the AC.

The audio recordings corresponded to the intended readings, indicating that participants allocated the utterances in a way that they matched the contexts, and that the position of the AC relative to the particle corresponded to the respective stress patterns.

5 Discussion

We investigated whether speakers are guided by syntactic or information structural information from the context, or whether they apply default patterns when constructing an utterance containing the particle auch. We presented participants with context sentences that contained the particle nur. The AC of nur functioned as the alternative to the AC of auch, which had to be integrated in the target sentence by either using unstressed auch or stressed AUCH. Furthermore, the AC of the particle in the context and target sentences were either the subject or the object. For the subject condition, we hypothesized that if speakers are guided by the information structural status of the alternative and by the position of the particle relative to the alternative, they should use unstressed auch, which precedes its focused AC exactly like nur in the context. If, however, they rely on default associations, they should integrate the subject as a topic and use stressed AUCH following its AC. To summarize, if speakers solely rely on (information) structural information from the context, they should show a general preference for unstressed auch. If, however, speakers rely on default mappings (subjects are prototypical topics), they should show a preference for stressed AUCH in the subject condition and for unstressed auch in the object condition. The results speak in favor of the latter case: We found a significant difference between the subject and object conditions, in that speakers preferably used stressed AUCH when the AC of the particle was the subject, and unstressed auch when the AC of the particle was the object. Thus, they relied on the knowledge that subjects are prototypical topics and objects are likely to be integrated as the focus. Furthermore, these results match the results of Höhle et al. (2009). They investigated the comprehension of auch in young children and demonstrated that children relate stressed AUCH to subjects and unstressed auch to objects. Based on our data, we can transfer these findings to language production in German adult speakers.

However, we see that the preference for unstressed auch in the object condition was stronger than for stressed AUCH in the subject condition (92.8 vs. 69%). The former can be explained by the fact that speakers had no reason to use stressed AUCH in this context, since all cues (particle precedes AC, AC = focus, AC = object) pointed towards using unstressed auch and integrating its AC as the focus. Furthermore, using stressed AUCH in the object condition requires a topicalization (keeping the object in the middle field is slightly marked) which was not motivated by our contexts. The latter can be explained by the fact that speakers were primed by the syntax and the information structure of the context. Due to the focus particle nur, which always preceded its AC, which in turn is the focus of the utterance, speakers used unstressed auch (with the exact same properties) in nearly one third of the cases in the subject condition. However, our experimental sentences with “No!” introducing the target sentence were related to a specific interpretation involving a correction. It is an open question whether this influenced the choice of the speakers.

To summarize, the results indicate that speakers strongly rely on default associations (subjects are prototypical topics), but that they are also primed by the (information) structure of the context. Future studies should go into further detail, specifically with respect to the priming effect and the question as to whether speakers rely more on the syntactic structure or on the information structure of the context sentence when constructing their utterances.