Content uploaded by Marie-Theres Weißgerber
Author content
All content in this area was uploaded by Marie-Theres Weißgerber on Jul 05, 2024
Content may be subject to copyright.
Schwa optionality in verbal inflection in German: the effects of stress and
phonetic context
Marie-Theres Weißgerber1
1Humboldt University Berlin, Germany
marie-theres.weissgerber@hu-berlin.de
Abstract
Speech sounds serve the function of distinguishing meaning.
However, in German, schwa sometimes has no such semantic
function and is optional in certain cases, like first-person singu-
lar inflectional suffixes. Nonetheless, this optionality has not yet
led to a total removal, or obligation to articulate, schwa in these
suffixes. The present study investigates the effects of phonetic
context and stress on schwa optionality. The data set consists of
two registers, formal and informal, of German spoken in Ger-
many and Namibian German. The following speech sound and
the stress of the following syllable, which are thought to affect
the likelihood of word-final schwa production, were analysed
for 44 speakers using Praat. Significant effects were found for
stress, with less schwa productions before unstressed syllables.
Significantly less schwa instances were observed before vowels.
Overall, register had a significant effect, with more schwa pro-
ductions in the formal condition. The impact of stress and the
effect of register were more marked in the subset of Namibian
German. These findings highlight the importance of investigat-
ing the interaction of phonetic features and register and em-
phasise the value of exploring different varieties in the study of
speech production phenomena.
Keywords: spontaneous speech, phonetic context, stress pat-
terns, register, Namibian German
1. Introduction
Schwa is optional in German first-person singular verbal inflec-
tional suffixes. Variation in schwa realisations has been docu-
mented as a constant feature of the German language system for
centuries (Fleischer et al. 2018; Nübling et al. 2013; Eisenberg
2020). The use of schwa entered the phonemic system during
the Old High German period1. Old High German word-final un-
stressed vowels were the full vowels <a>, <e>, <i>, <o>, <u>
(Fleischer et al. 2018). Example (1) demonstrates the shift from
full vowels in the final syllable to the inflection with schwa in
New High German.
(1) OHG suochu ‘I search’ (1. ps. sg. ind. pres.), suochi
‘search’ (imp. sg.), suoche ‘may he search’ (3. ps. sg. subj.
pres.) > MHG suoche > NHG such(e)
(Fleischer et al. 2018)
As can be deduced from example (1), schwa in inflection
has become markedly more flexible in present-day German. In
some cases, like in inflectional paradigms used to form the past
1The period of Old High German with attested writing is dated to 750-
1050 AD. The Middle High German (MHG) epoch is dated to 1050-
1350, and the time period of New High German (NHG) began in 1650
and continues to this day (Nübling et al. 2013).
tense without ablaut, a schwa suffix is obligatory. For instance,
the verb “sehen” (‘to see’) forms the third-person singular with
“sieht” in the present tense and with “sah” in the preterite. In
other cases, word-final schwa can be either pronounced or omit-
ted without yielding any semantic change. Figure 1 shows ver-
bal inflection for a weak verb with schwa in the first-person
singular in present tense as an optional suffix. The third-person
singular in present tense and the stem of the first-person singular
in preterite tense are identical on the surface level. Schwa is not
optional in the first and third-person singular in preterite tense.
If schwa was not realised in these cases, the result would be a
semantic change of the verb form towards the present tense.
Figure 1: Inflection of the weak verb “legen” (‘to put’) in
present tense (left side) and preterite (right side), modified with
green boxes by author (Eisenberg 2020).
Schwa variation is driven by a wide range of factors,
from segmental and supra-segmental parameters to articula-
tion rate (Ernestus, Hanique, and Verboom 2015; Kienast and
Sendlmeier 2000) and word frequency (Pluymaekers, Ernes-
tus, and Baayen 2006; Kohler and Rodgers 2001; Jurafsky et
al. 2001). Research on schwa in adverbs provides further in-
sights into why such variation might occur. In a study by Fleis-
cher et al. (2018) optionality in word-final schwas is exam-
ined in adverbs. The authors investigate heut(e), gern(e) and
bald(e) in the letters of Goethe. For heut(e) and gern(e), a
highly significant impact of the following segment was found.
For both adverbs, a following vowel led to significantly fewer
schwa occurrences. In the case of gern(e), a sonority contin-
uum is observed: while vowels in the following segment corre-
late with less schwa occurrences, final schwas occur more fre-
quently when followed by a sonorant, and slightly more often
when followed by an obstruent. These results might be rooted
in a preference for a balanced alternation between vowels and
consonants, whereby consonantal clusters and vowel hiatus are
prevented (Fleischer et al. 2018). On the supra-segmental level,
word stress plays a crucial role in triggering the presence or ab-
sence of word-final schwa. Research that considers stress pat-
terns and their influence on schwa alternations often uses the
ISSP 2024 - 13th International Seminar on Speech Production
13-17 May 2024, Autrans, France
59 10.21437/issp.2024-16
terms ‘stress clash’ and ‘stress lapse’. For example, Kentner
associates ‘rhythmic alternation’ with the avoidance of ‘stress
clash’ and ‘stress lapse’ (Kentner et al. 2018). Another cru-
cial concept is the tendency of German rhythm to adhere to a
pattern of sequential trochees, which is referred to “as an opti-
mal template regulating the shape of words” in German (Kent-
ner et al. 2018, p. 120). However, it is not only within words
that the trochee plays an important role. On the sentence level,
a balanced juxtaposition of trochees prevents a “clustering of
strong syllables (*CLASH)” as well as “sequences of weak syl-
lables (*LAPSE)” (Kentner et al. 2018, p. 120). Within the
framework of Prosodic Parallelism Theory, Wiese and Speyer
(2015) ascribe an important role to schwa. As part of this
framework, the authors argue that “a form that is invariably
(non-)trochaic causes another form in the same dominating cat-
egory2(the phrase) to be (non-)trochaic as well.” (R. Wiese
and Speyer 2015, p. 528). While the framework allows for ex-
ceptions where lexical options are limited (“as in sehr langsam
‘very slow”’), the authors assume that prosodic parallelism is
the favoured choice wherever feasible. They claim that word-
final schwa optionality offers the opportunity for such a selec-
tion.
Yet, contributing factors for schwa-zero alternations are not
only found on a purely linguistic level – whether or not schwa is
articulated also seems to depend on situational and task-based
factors. A small number of studies found effects of different
registers of spoken language on schwa realisations. Kohler and
Rodgers (2001) examine schwa in both read and spontaneous
speech and find that the segment articulated after a potential
word-final schwa influences whether or not it is realised. They
report that verbs and function words often have a non-realised
schwa in word-final position, particularly when preceding a
vowel. Within that group, most unrealised schwas are found
in function words and verb suffixes in the first person singular
(Kohler and Rodgers 2001). Ernestus et al. find that the formal-
ity of a communicative situation affects the frequency and du-
ration of prefixal schwas in Dutch, with less schwa realisations
in “casually articulated speech” (Ernestus, Hanique, and Ver-
boom 2015). Lange et al. discover differences in the frequency
of schwa productions between the registers of free speech and
task-based dialogue, with significantly more schwa productions
in free conversation (Lange et al. 2023).
Data on schwa optionality in different varieties of German
are relatively scarce. To address this gap, the current study in-
vestigates two different varieties of German, German spoken in
Germany (GGER) and German spoken in Namibia (NamGER),
to generate new findings in this area. Wiese and Bracke find that
there is a differentiation in register between standard German
and Namibian German variants (H. Wiese and Bracke 2021).
The majority of Namibian German speakers also speak at least
two other languages, most commonly Afrikaans and English
(Zimmer 2021). Kellermeier-Rehbein identifies the close re-
latedness of Afrikaans and English to German as a major fa-
cilitator for the incorporation of loan words and grammatical
structures into Namibian German (Kellermeier-Rehbein 2016).
Wiese and Bracke assert that the societal context in Namibia,
which is characterised by multilingualism, makes the language
receptive to the integration of diverse linguistic resources (H.
Wiese and Bracke 2021).
This study investigates how schwa in the first-person singu-
lar is distributed in spontaneous speech in two registers, formal
2This assumption is based on a hierarchy of phrase, word, foot and
syllable (in descending order) (R. Wiese and Speyer 2015).
and informal, in two varieties of German. Based on previous
findings (Fleischer et al. 2018; Kohler and Rodgers 2001), it
is hypothesised that schwa should be produced less frequently
when the following syllable is unstressed. This effect is ex-
pected to be particularly marked when the following segment is
an unstressed vowel and to be weaker for following sonorants
or obstruents. It may be assumed that stimuli produced in the
formal register will stay closer to the canonical form found in
written productions and will therefore contain more schwa real-
isations.
2. Methods
2.1. Corpora
Speech recordings were obtained from two corpora containig
spontaneous speech recordings. Data of native speakers of Ger-
man residing in Germany stem from the RUEG corpus (H.
Wiese, Alexiadou, et al. 2021). Recordings of native speakers
of Namibian German have been made available by the research
group Namdeutsch (‘Namibian German’) within the scope of
the corpus DNam (Deutsch in Namibia) (‘German in Namibia’)
(Zimmer et al. 2020).
2.2. Participants and Tasks
Participants were presented with visual material, either in the
form of a video or a photograph story, of an accident. After
viewing the material, speakers provided two summaries of the
events that had taken place. In the formal condition, GGER
participants were asked to provide a witness report to a police
officer in the form of a voice message. In the informal con-
dition, participants summarised events to a friend in a voice
message (H. Wiese, Alexiadou, et al. 2021; H. Wiese 2020).
NamGER speakers spoke to a German teacher, impersonated
by a researcher, in the formal condition (Zimmer et al. 2020).
In the informal condition, speakers provided a summary of the
events to a family member or friend present during the record-
ings (Zimmer et al. 2020). 88 recordings of 44 speakers (20
female) are analysed. The age range of the speakers is between
13 and 40.
2.3. Stimuli and Analysis
A total of 218 instances of verbs in the first-person singular are
analysed in this study. The average speaking time of all par-
ticipants analysed here is 68 seconds. Annotations were done
manually in Praat (Boersma and Weenink 2023) on five differ-
ent tiers (see Figure 2). Tier one contains transcriptions of the
spoken materials. Tier two contains the relevant stimulus with
first-person singular inflection. Tier three shows whether or not
schwa is articulated in the relevant stimulus. Here, 1refers to a
realisation of schwa and 0denotes a non-realisation. In tier four,
the phonological context preceding and following the expected
schwa realisation is described. In tier five the stress pattern of
the following syllable is annotated. Factors influencing schwa
realisation were tested using chi-square tests.
3. Results
3.1. Phonetic context
To assess the effect of phonetic context, the data were subset
into instances of verbs preceding vowels, sonorants, obstru-
ents and pauses (see Figure 3). As can be seen, a following
pause, indicating a prosodic boundary, increases the frequency
60
Figure 2: Textgrid example.
of realised schwas. However, only 14 instances with following
pauses were found. The subset of unrealised schwas preced-
ing vowels are distributed to 94.3% before unstressed vowels,
non-realised schwa before sonorants can be found to 92.3% be-
fore unstressed sonorants, and the subset of unrealised schwas
before obstruents are distributed to 82.8% before unstressed
obstruents. This result indicates a slight sonority continuum
within the distribution of schwa realisations and their interac-
tion with the following context.
5
9
59
23
26
10
70
16
0.00
0.25
0.50
0.75
1.00
p: obstruent sonorant vowel
Following context
Percentage
schwa
0
1
Schwa realisations per following sound
Figure 3: Schwa-realisations coded by following segment, ab-
solute numbers in bars.
3.2. Stress
Out of all instances, merely 32 (14.7%) are followed by a
stressed syllable, 172 (78,9%) precede an unstressed syllable,
and 14 (6.4%) are followed by a pause. Excluding items with
following pauses, a Pearson’s Chi-squared test across the Ger-
man and Namibian German varieties shows that the word stress
of the following syllable has a significant influence on whether
or not a schwa is articulated (χ2= 12.399, df = 1, p< 0.001).
Most instances of first-person singular verbs are pronounced
without schwa when the following syllable is unstressed. In
the GGER data frame, 60% of potential word-final schwas are
articulated before stressed syllables. This is the case in only
45.5% in the NamGER subset (see Figure 4).
57
13
4
6
82
20
12
10
GGER
NamGER
0 1
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Stress of following syllable
Percentage
schwa
0
1
Schwa real. coded by stress and variety
Figure 4: Schwa realisations per variety coded by the stress pat-
tern of the following syllable, absolute numbers in bars. Pauses
are excluded.
3.3. Register
Verbal suffixes are produced without schwa in 63.3% of cases
in the formal condition, and in 87.8% of cases in the informal
condition (χ2= 15.009, df = 1, p< .001). In the formal regis-
ter, NamGER verbs are pronounced with schwa in 63.4%. The
proportions are different between the two varieties in the infor-
mal condition. As can be seen in Figure 5, there is a difference
in the informal condition: in NamGER 10% of the schwas are
realised, whereas in GGER it is 15%.
29
17
34
6
52
30
45
5
GGER
NamGER
f if f if
0.00
0.25
0.50
0.75
1.00
Register
Percentage
schwa
0
1
Schwa realisations in the two varieties
Figure 5: Schwa realisations per variety coded by register, ab-
solute numbers in bars.
4. Discussion and conclusion
In summary, the results demonstrate that the most common re-
alisation of the verbal inflectional ending in the first-person sin-
gular is without schwa in 73.4% of all cases (n= 160), similar
to Lange et al. (2023, accepted). Based on the literature (Fleis-
cher et al. 2018; Kohler and Rodgers 2001), it was expected that
schwa should be realised less often when it precedes a vowel.
This can be confirmed with the data set analysed in the present
61
study, where the effect of following vowels on schwa realisa-
tions is statistically significant. The stress of the following seg-
ment has a significant influence on whether or not a schwa is
realised. In general, the realisation of schwa is significantly in-
fluenced by the stress of the following syllable. However, there
appears to be no interaction between stress and register. The fact
that adjacent rhythmic context affects schwa realisations sup-
ports the findings of Wiese and Speyer and Kentner (R. Wiese
and Speyer 2015; Kentner et al. 2018).
Comparing the two varieties, the results show that schwa
productions are evenly distributed across the formal register. In
the informal register, NamGER exhibits only 10% schwa real-
isations compared to 15% in GGER. This discovery is of par-
ticular interest in the light of the variety’s linguistic openness
identified by Wiese and Bracke (2021), and its inclination to
advance internal structural phenomena of German as noted by
Wiese et al. (H. Wiese, Simon, et al. 2014). Is schwa-zero al-
ternation, which seems to be an inherent structural feature of
German, further progressing in informal Namibian German?
Schwa is indeed optional in first-person singular verbal in-
flectional suffixes in German. This study provides further evi-
dence demonstrating that this optionality is not random. In fact,
schwa optionality is found in both the formal and the informal
register. The presence of optionality in the informal register
highlights that register has an effect on schwa realisations in
the German language. Variation in the formal register, on the
other hand, demonstrates that this variation is not confined to
informal communicative settings – schwa optionality seems to
be ingrained in first-person singular verbal inflectional suffixes
of German. Additionally, this study offers insights into how
this phenomenon operates in Namibian German. The results
indicate that, given that particular elements are even more pro-
nounced in this variety, Namibian German might be advancing
an internal structural phenomenon of the linguistic system of
German. This study confirms previous findings on the interac-
tion of phonetic context and stress with word-final schwa. It will
be a worthwhile endeavour to further examine which individual
elements cause these marked differences in schwa realisations
by investigating other registers and varieties.
5. Acknowledgements
This research was funded by the Deutsche Forschungsgemein-
schaft (DFG, German Research Foundation) – SFB 1412,
416591334. The author would like to thank Prof. Christine
Mooshammer, Prof. Heike Wiese and Dr. Malte Belz for guid-
ance and feedback as well as Robert Lange, Daniela Palleschi
and Georg Lohfink for technical support.
6. References
Boersma, Paul and David Weenink (2023). Praat: doing phonetics by
computer.URL:http://www.praat.org/..
Eisenberg, Peter (2020). Grundriss der deutschen Grammatik: Das
Wort. 5th ed. Stuttgart: J. B. Metzler.
Ernestus, Mirjam, Iris Hanique, and Erik Verboom (2015). “The effect
of speech situation on the occurrence of reduced word pronuncia-
tion variants”. In: Journal of Phonetics 48, pp. 60–75.
Fleischer, Jürg, Michael Cysouw, Augustin Speyer, and Richard Wiese
(2018). “Variation and its determinants: A corpus-based study of
German schwa in the letters of Goethe”. In: Zeitschrift für Sprach-
wissenschaft 37.1, pp. 55–81.
Jurafsky, Daniel, Allan Bell, Michelle Gregory, and William Raymond
(2001). “The effect of language model probability on pronuncia-
tion reduction”. In: IEEE International Conference on Acoustics,
Speech, and Signal Processing. Vol. 2, pp. 801–804.
Kellermeier-Rehbein, Birte (2016). “Sprache in postkolonialen Kontex-
ten. Varietäten der deutschen Sprache in Namibia”. In: Sprache
und Kolonialismus. Eine interdisziplinäre Einführung zu Sprache
und Kommunikation in kolonialen Kontexten. Ed. by Thomas Stolz,
Ingo H. Warnke, and Daniel Schmidt-Brücken. Berlin, Boston: De
Gruyter, pp. 213–234.
Kentner, Gerrit, Christiane Ulbrich, Alexander Werth, and Richard
Wiese (2018). “Schwa-optionality and the prosodic shape of words
and phrases”. In: Empirical approaches to the phonological struc-
ture of words, pp. 121–151.
Kienast, Miriam and Walter F Sendlmeier (2000). “Acoustical analysis
of spectral and temporal changes in emotional speech”. In: ISCA
Tutorial and Research Workshop (ITRW) on Speech and Emotion.
Kohler, Klaus J and Jonathan Rodgers (2001). “Schwa deletion in
German read and spontaneous speech”. In: Spontaneous German
speech: Symbolic structures and gestural dynamics 35, pp. 97–123.
Lange, Robert, Bianca Sell, Megumi Terada, Malte Belz, Christine
Mooshammer, and Anke Lüdeling (2023). “The phonetic realisa-
tion of verbal inflection in two dialogue registers of German spon-
taneous speech”. In: (Submitted for publication).
Nübling, Damaris, Antje Dammel, Janet Duke, and Renata Szczepaniak
(2013). Historische Sprachwissenschaft des Deutschen: Eine Ein-
führung in die Prinzipien des Sprachwandels. 4th ed. Tübingen:
Narr Francke Attempto Verlag.
Pluymaekers, Mark, Mirjam Ernestus, and R. Harald Baayen (2006).
“Effects of word frequency on articulatory durations of affixes”.
In: Proceedings of Interspeech, pp. 953–956.
Wiese, Heike (2020). “Language Situations: A method for capturing
variation within speakers’ repertoires”. In: Methods in dialectology
XVI. Vol. 59. Frankfurt Main: Peter Lang, pp. 105–117.
Wiese, Heike, Artemis Alexiadou, Shanley Allen, Oliver Bunk, Na-
talia Gagarina, Kateryna Iefremenko, Jahns Esther, Martin Klotz,
Thomas Krause, Annika Labrenz, Anke Lüdeling, Maria Mar-
tynova, Katrin Neuhaus, Tatiana Pashkova, Vicky Rizou, Rose-
marie Tracy, Christoph Schroeder, Luka Szucsich, Wintai Tse-
haye, Sabine Zerbian, and Yulia Zuban (2021). “RUEG Corpus”.
In: UR L:https : / / zenodo . org / record / 3236069 #
.ZEKiFS-23fY. (visited on 09/13/2023).
Wiese, Heike and Yannic Bracke (2021). “Registerdifferenzierung
im Namdeutschen: Informeller und formeller Sprachgebrauch in
einer vitalen Sprechergemeinschaft”. In: Kontaktvarietäten des
Deutschen im Ausland. Tübingen: Narr, pp. 273–293.
Wiese, Heike, Horst J. Simon, Marianne Zappen-Thomson, and Kath-
leen Schumann (2014). “Deutsch im mehrsprachigen Kontext:
Beobachtungen zu lexikalisch-grammatischen Entwicklungen im
Namdeutschen und im Kiezdeutschen”. In: Zeitschrift für Dialek-
tologie und Linguistik, pp. 274–307.
Wiese, Richard and Augustin Speyer (2015). “Prosodic parallelism ex-
plaining morphophonological variation in German”. In: Linguistics
53.3, pp. 525–559.
Zimmer, Christian (2021). “Siedlungsgeschichte und Varietätenkon-
takt”. In: Zeitschrift für Dialektologie und Linguistik 88 H. 3,
pp. 324–350.
Zimmer, Christian, Heike Wiese, Horst J Simon, Marianne Zappen-
Thomson, Yannic Bracke, Britta Stuhl, and Thomas Schmidt
(2020). “Das Korpus Deutsch in Namibia (DNam): Eine Ressource
für die Kontakt-, Variations-und Soziolinguistik”. In: Deutsche
Sprache 48.3, pp. 210–232.
62
