Vowel and consonant quantity in two Swiss German dialects and their corresponding varieties of Standard German: effects of region, age, and tempo

1.1 Diglossia in German-speaking Switzerland

The dialects spoken in German-speaking Switzerland comprise Low Alemannic, which is limited to the city of Basel, High Alemannic, spoken in the north, and Highest Alemannic spoken in the south (Arquint et al. 1982; Christen 2019). The two dialects investigated in this study belong to the High Alemannic group and will simply be referred to as Alemannic in this paper.

It is typical in German-speaking Switzerland to speak Alemannic in most situations in everyday life. SSG is learned in school and spoken only in specific formal situations, e.g., parliament sessions or formal news reports. For written communication, mainly SSG is used with some exceptions, e.g., written commercials, which may also be written in dialect. However, the diglossic configuration has been changing since 1980, when it has become increasingly common to also communicate in written Alemannic, e.g., in personal letters, emails, and, more recently, in online chats or on social media (Siebenhaar 2006). While it has been shown that one cannot infer a specific Alemannic dialect when analyzing vowel and consonant durations in SSG, recent findings revealed that speakers use the same durational categories in SSG as in Alemannic, suggesting that, in general, SSG is influenced by Alemannic rather than the other way around (Zihlmann 2020b). The predominant use of Alemannic in everyday life and the fact that it is used increasingly in the written language implies that the diglossic situation in German-speaking Switzerland is relatively stable. Even so, the impact of GSG has not been investigated so far. This is why this study focuses on an urban area with high contact to GSG and a rural area, as well as two age groups.

1.2 Articulation rate, vowel durations, and consonant durations

2.1 Speakers

Of the 40 speakers recorded, 20 were from LU, 10 speakers (5 female) were younger with ages between 25 and 32 years (mean = 28.7, SD = 1.94) and 10 speakers (5 female) were older with ages between 47 and 64 (mean = 58.0, SD = 6.30) at the time of the first recording. The other 20 speakers were from ZH, 10 speakers (5 female) in the younger group with ages between 18 and 28 (mean = 23.0, SD = 2.98) and 10 speakers (5 female) in the older group with ages between 58 and 69 (mean = 64.1, SD = 3.75) at the time of the first recording. All of the speakers grew up in either LU or ZH and at least one but in most cases both of their parents also come from the same canton. With the exception of four speakers in LU and five speakers in ZH (one of them still going to secondary school) the speakers either finished university or were still studying at university at the time of the recordings.

2.2 Stimuli

2.3 Procedure

Whenever possible, the participants were recorded in a soundproof booth at the Phonetics Laboratory of the University of Zurich using a personal computer with the interface USBPre^® 2 (Sound Devices) and the microphone NT2-A (RØDE). Otherwise, they were recorded in a quiet room using a laptop computer with the same interface and the microphone Opus 54.16/3 (BeyerDynamic). For the recording software, SpeechRecorder 3.8.0 (Draxler and Jänsch 2004) was used, except for the interviews (cf. below), which were recorded using Audacity (Audacity Team 2017). All recordings had a sample rate of 16 bit/44.1 kHz and were saved as .wav-files. Prior to the first recording session, participants signed a declaration of consent. They had two appointments, the first one lasting about 75 min, the second one about 105 min. The participants received a reimbursement of 15CHF per 30 min, resulting in a payment of 60CHF for most participants.

At the first appointment, an interview of about 10 min was conducted to gather information about the sociolinguistic background of the participants and for them to get used to the recording situation. After the interview, the training phase began. Participants were instructed to read three sentences with different target words both in normal and in fast speech tempo. These recordings served as the basis for establishing each speaker’s time limit for each sentence during the rest of the experiment, which was implemented using a time bar. The average duration for normal and fast speech tempo was calculated separately for each speaker. This value plus 400 ms served as the basis for the time bar. The speakers read four further test sentences in each speech tempo first silently, and when the time bar appeared, aloud. Participants were instructed to repeat the sentences before the time indicated by the bar was over. During the appearance of the time bar, the written sentences were not presented anymore. An additional purpose of the training phase was for the participants to get used to the Swiss German spelling system by Dieth (1986), which was used for the Alemannic production task. After the training, the actual experiment took place. Each sentence was repeated five times in both tempo conditions, resulting in 340 recordings for each participant from LU and 310 recordings for each participant from ZH. All stimuli were recorded in one session and presented in semi-random order, with two of the same stimuli never appearing twice in a row. Participants were instructed to start the sentences from the beginning in case they made a mistake.

At the second appointment, the same participants’ productions of SSG were recorded. With the first six training sentences (three for each tempo), again, the durations for the time bars were calculated for each speaker. After this, the participants read four further training sentences in normal and fast speech tempo, first silently, then aloud with the time limit given by the time bar. For the SSG stimuli, the sentences were written in the official standard German orthography. After the training, the experiment began. Each sentence was repeated five times for each tempo, resulting in 230 recordings for each participant from both LU and ZH. All stimuli were recorded in one session and presented in semi-random order, with two of the same stimuli never appearing twice in a row. Participants were instructed to start the sentences from the beginning in case they made a mistake. After the production experiment, the speakers participated in a perception experiment also focusing on segment durations, which is not part of this study. Lastly, the conductor of the experiment filled out the main information about the speakers, i.e., name, speaker ID, place of residence, date of birth, and date of recording, in a form, before further preparing the data for the analysis.

2.4 Data preparation

After the recordings were saved, they were automatically segmented using WebMAUS (Schiel 1999), selecting in the language annotation settings the option German Dieth (CH) for the Alemannic recordings and German (DE) for the SSG recordings. The phonetic segments were manually adjusted with the EMU-webApp (Winkelmann and Raess 2014), using the following procedure: The beginning and end segments of each sentence were corrected if necessary. For the sentences beginning with a plosive, the release was used as the beginning of the first segment. If speakers paused within a sentence, the pause segment was removed and its duration was subtracted from the duration of the sentence. If there was more than one longer pause, the recording was not included in the analysis. Each of the phonetic segments of the target words was precisely adjusted. The plosives were segmented into two phases: closure, indicated by /b̥ d̥ ɡ̊/ for lenis and /p t k/ for fortis, and release (VOT), indicated by _h (which was not included in the analyses). As VOT is known to not significantly differ between lenis and fortis plosives, this study focuses on closure duration only. An example for the segmentation can be seen in Figure 1.

Figure 1:

Segmentation of the target word Kater (tomcat), pronounced as [ˈkxɒːtʰər], spoken by LU_0001, a young female speaker from LU, in SSG.

If participants produced a target word in a different way than expected, i.e., accidentally produced another word or pronounced the word incorrectly, the recording was excluded from the analysis. Further recordings were excluded if participants made mistakes that resulted in an alteration of the number of syllables, the recording was cut at the beginning or end, or the quality of the recordings that were not conducted at the University of Zurich was not sufficient. Unfortunately, for some speakers from LU, there were quality issues that resulted in a relatively high number of exclusions. The rest of the aforementioned exclusion criteria only included a small number of recordings. Ultimately, from the total of 13,000 of the Alemannic recordings (6,800 for LU, 6,200 for ZH), 11,182 were used for the analysis (5,514 for LU, 5,668 for ZH). From the total of 9,200 of the SSG recordings (4,600 each from LU and ZH), 8,431 were used for the analysis (4,310 for LU and 4,121 for ZH).

2.5 Measurements

The duration of each sentence, the duration of each target word, and the word-medial vowel and closure durations were measured using the emuR package (Winkelmann et al. 2021) in RStudio (version 2022.07.2; R version 4.0.2; RStudio Team 2020), which provides a direct link between the EMU-webApp and R. In this study, only the measurements of the closure duration of the plosives were included, as VOT is assumed not to differ between lenis and fortis plosives. The sentences were measured in milliseconds, and afterwards, the AR was calculated in terms of MSD by dividing the duration of the sentences by the number of syllables of each sentence (pauses had already been excluded). To obtain the relative vowel and closure durations, their absolute duration was divided by the word duration. This normalization procedure is appropriate considering the similar phonotactic and prosodic makeup of the words used for the stimuli, i.e., disyllabic trochees. Lastly, V/(V + C) ratios were calculated by dividing the duration of the vowel by the duration of the whole V + C sequence.

2.6 Statistical analyses

Statistical analyses were performed in RStudio (RStudio Team 2020). Linear mixed-effects models were fitted for all the analyses using the lme4 (version 1.1-29) and the lmerTest (version 3.1-3) packages (Bates et al. 2015; Kuznetsova et al. 2017). For the model regarding AR, the dependent variable was MSD. Fixed effects were variety (Alemannic vs. SSG), tempo (normal vs. fast speech tempo), region (LU vs. ZH), and age (younger vs. older), including two-way interactions. Random intercepts were added for speaker and word.

For the models regarding the durations, either relative vowel duration, relative closure duration or V/(V + C) ratio were defined as the dependent variable. Fixed effects were variety, VC combination, region, age, and tempo, including two-way interactions. Random intercepts were added for speaker as well as word nested within VC combination.

A Type II ANOVA was calculated for each model using the R package car (version 3.1-0; Fox and Weisberg 2018), which yielded the Chi-square and p values reported in the Results section of this paper. For the interactions that turned out to be significant, pairwise comparisons using Tukey’s tests were calculated using the R package emmeans (version 1.7.2; Lenth 2021). Additionally, Tukey’s tests were calculated in case of a significant effect of VC combination in order to compare the durations of each combination to each other.

The .csv files as well as the R script for the analyses will be available under osf.io/y5c7d.

3.1 Articulation rate

The means and standard deviations of the MSD per group are shown in Tables 1a (Alemannic) and 1b (SSG).

The linear mixed-effects model revealed significant main effects of age and tempo on MSD, as can be seen in Table 2. As expected, older speakers had an overall longer MSD compared to younger speakers, which can also be seen in Figure 2.

Figure 2:

MSD in ms (y axis) in normal (left) and fast (right) speech tempo in both dialect and standard speech of LU and ZH speakers (x axis); younger speakers in dark gray, older speakers in light gray.

Pairwise comparisons of the significant interactions revealed that LU and ZH speakers did not differ significantly from each other when comparing them in dialect and standard speech separately as well as in fast and normal tempo separately, although a strong trend occurred: Similarly to the findings by Zihlmann (2020a), speakers from the western area of LU did not have a slower AR than those from ZH. Instead, the opposite pattern occurred not only in Alemannic, in accordance with the findings in Zihlmann (2020a), but also in SSG, which is even more surprising. This difference might be due to the fact that read speech, which was also investigated by Zihlmann (2020a), was analyzed, while previous research mainly focused on spontaneous speech (Leemann and Siebenhaar 2010). Another surprising result was that the age difference turned out to be significant only in SSG (z = 3.266, p = 0.0011), probably due the large amount of variability in Alemannic for the younger speakers, which can be seen in Figure 2. Comparing Alemannic to SSG, ZH speakers had a significantly higher MSD in Alemannic than in SSG (z = 7.191, p < 0.0001), while it was the other way around for LU speakers (z = −7.051, p < 0.0001).

As proposed by Quené (2007), the JND, i.e., the perceptual threshold, for AR lies at 5 %. The percentage values of increase (positive numbers) and decrease (negative numbers) are therefore also shown in Tables 1a and 1b. Despite not all being significant, most of the regional and age-related differences are above the JND, particularly in the normal speech condition. The difference between Alemannic and SSG did not reach the 5 % in any case. To conclude, the age-related differences in AR were most stable and confirm H1a, while the regional differences yielded surprising results, leading to a rejection of H1b. The differences between Alemannic and SSG are significant but likely not perceivable. Therefore, when taking the JND into account, H1c ultimately must be rejected.

3.2 Relative vowel durations

The means and standard deviations of the relative vowel durations per group are shown in Tables 3a (Alemannic) and 3b (SSG).

The statistical model revealed significant main effects of variety, VC combination, age, and tempo, also shown in Table 4. Relative vowel durations were overall longer in dialect compared to standard speech. Regarding age, older speakers produced longer vowel durations compared to younger speakers. Comparing the two tempo conditions to each other, the relative vowel duration turned out to have higher values in fast compared to normal speech.

The pairwise comparisons for the significant interaction revealed the following: There was no significant regional difference in either dialect or standard and in either normal or fast tempo for none of the VC combinations. An age-related difference was found for both Alemannic (z = 3.893, p = 0.0001) and SSG (z = 2.488, p = 0.0129), with older speakers producing longer vowels, but only significantly so in the VC combinations VːC (z = 6.197, p < 0.0001) and VːCː (z = 4.075, p < 0.0001). The combination VːC was significantly longer in normal compared to fast speech (z = −4.827, p < 0.0001), while VC (z = 3.727, p = 0.0002) and VCː (z = 1.169, p < 0.0001) were longer in fast compared to normal speech. This explains the direction of the main effect of tempo and is also not surprising, as longer segments are known to be shortened with acceleration in AR a lot more in comparison to short segments (e.g. Arvaniti 1999; Klatt 1973).

To have a better overview on how many vowel categories there are and on possible differences between the factors of interest, pairwise comparisons were made between the consecutive VC combinations from shortest to longest. This is shown in Tables 5a (Alemannic) and 5b (SSG) and is also visible in Figures 3a (normal tempo) and 3b (fast tempo).

Figure 3a:

Relative vowel durations (y axis) in normal speech tempo in Alemannic (left) and SSG (right) for all four VC combinations (x axis); younger speakers in dark gray, older speakers in light gray.

Figure 3b:

Relative vowel durations (y axis) in fast speech tempo in Alemannic (left) and SSG (right) for all four VC combinations (x axis); younger speakers in dark gray, older speakers in light gray.

The results show that there are three phonetic categories of relative vowel durations in Alemannic, which are rather stable across tempo, at least for the ZH speakers. Still, they differ from each other between regions: While LU speakers differentiate so much between VCː and VC that VC and VːCː do not differ significantly from each other, ZH speakers produce the short vowels in VCː and VC almost identically. The categories for ZH speakers are consistent with the evidence from Zihlmann (2020b), who found the same for Bernese and ZH speakers and invoked the possibility for more than two (i.e., long and short) vowel categories. The findings of the present study seem to confirm the existence of three categories, at least for Alemannic. Still, the categories for LU speakers seem to be unique and are one of the major differences between the two regions in Alemannic. The categories are much more similar in SSG. In addition, it is apparent that the vowel durations, specifically of the combination VːC, are shorter in standard speech, leading them to be closer together and to a reduction to only two categories, i.e., short and long vowels. Although this is what the analysis revealed, it is clear that there are still quite large differences between VC combinations. The main question here remains if these differences are perceivable.

3.3 Relative closure durations

The means and standard deviations of the relative closure durations per group are shown in Tables 6a (Alemannic) and 6b (SSG).

Table 7 shows the model output, which revealed main effects of variety, VC combination, region, age, and tempo. Similarly to the results on the relative vowel durations, closure durations were significantly longer in dialect compared to standard productions. Generally, LU speakers produced significantly longer closure durations compared to ZH speakers, which confirms H2, while older speakers produced longer durations than younger speakers. As for tempo, the normal tempo condition resulted in longer relative closure durations compared to the fast tempo condition.

Moving on to the pairwise comparisons, the difference between fast and normal speech is significant for all combinations: While the relative closure durations are longer in fast speech for VːC (z = 3.060, p = 0.0022) and VC (z = 2.373, p = 0.0176), they are longer in normal speech for VːCː (z = −3.009, 0.0026) and VCː (z = −12.472, p < 0.0001). This could be expected as longer segments are shortened more in fast speech than short segments (Arvaniti 1999; Klatt 1973). When comparing younger to older speakers, older speakers produce longer closures only in SSG (z = 2.672, p = 0.0075). Looking at Tables 6a and 6b and at Figures 4a and 4b, it is apparent that this is probably due to the productions from the younger ZH speakers. There is a much larger age-related difference for ZH speakers than for LU speaker in both Alemannic and SSG. Furthermore, the productions from the older ZH speakers look much more similar to those from the LU speakers. An explanation for this could be that younger speakers from the urban area of ZH, have increasing contact to GSG. While GSG speakers’ productions of word-medial fortis consonants are also longer than those of lenis consonants (Jessen 1998), this difference is much smaller than for Swiss-German speakers. Instead, in GSG, aspiration is the main cue to differentiate between lenis and fortis (Jessen 1998).

Figure 4a:

Relative closure durations (y axis) in normal speech tempo in Alemannic (left) and SSG (right) for all four VC combinations (x axis); younger speakers in dark gray, older speakers in light gray.

Figure 4b:

Relative closure durations (y axis) in fast speech tempo in Alemannic (left) and SSG (right) for all four VC combinations (x axis); younger speakers in dark gray, older speakers in light gray.

Similarly to the vowel durations, pairwise comparisons for the VC combinations were calculated, which are shown in Tables 8a (normal tempo) and 8b (fast tempo). Firstly, there are clearly three phonetic categories in Alemannic, i.e., lenis, fortis, and extrafortis, which are stable across tempo. With the exception of the LU speakers in fast speech tempo (where VːCː and VCː still are close to being significantly different from each other), those three categories are also found in SSG. This also means that, although younger ZH speakers behave differently from the other groups, they still maintain these three categories at this point.

As stated above, some compensation between vowel and closure duration is expected. Thus, although some significant differences in vowel and consonant categories have been revealed at this point, the entire V + C sequence should also be taken into account. That’s why the additional measure of V/(V + C) ratios was selected for this study and its results presented in the following section.

3.4 V/(V + C) ratios

The means and standard deviations of the V/(V + C) ratios per group are shown in Tables 9a (Alemannic) and 9b (SSG).

As can be seen in Table 10, the model for the V/(V + C) ratios revealed significant main effects for variety, VC combination, and tempo. In general, the ratios in dialect speech have higher values compared to standard speech. As for tempo, the values of the ratios are higher in fast speech tempo.

Looking at the significant interactions, the tempo difference turned out to be significant for VːC (higher ratios in normal speech) (z = −3.892, p = 0.0001) and VCː (higher ratios in fast speech) (z = 13.656, p < 0.0001), confirming, again, that longer segments are shortened more than shorter ones with an acceleration of speech tempo. Furthermore, it was revealed that the higher values of the ratios in Alemannic compared to SSG are only significant for the combinations VC (z = 3.097, p = 0.0020) and VCː (z = 7.586, p < 0.0001). Furthermore, there are significant regional differences in SSG, with higher ratios in ZH compared to LU speakers (z = −2.906, p = 0.0037). As for age, only VːC (z = 2.118, p = 0.0341) turned out to be significantly different, with older speakers producing higher ratios.

Overall, the V/(V + C) ratios are relatively similar to each other when comparing varieties, age groups and tempo conditions, which is also shown in Figures 5a (normal tempo) and 5b (fast tempo).

Figure 5a:

V/(V + C) ratios (y axis) in normal speech tempo in Alemannic (left) and SSG (right) for all four VC combinations (x axis); younger speakers in dark gray, older speakers in light gray.

Figure 5b:

V/(V + C) ratios (y axis) in fast speech tempo in Alemannic (left) and SSG (right) for all four VC combinations (x axis); younger speakers in dark gray, older speakers in light gray.

The most obvious difference seems to be that LU speakers produce four categories of VC combinations, which can also be seen in Tables 11a and 11b. The difference between VːCː and VC is not entirely stable, as it is not maintained in fast tempo (for older speakers) and in SSG. It is not clear if VːCː and VC can be considered two separate categories because they are also much closer together than VCː to VːCː or VC to VːC. At the same time, ZH speakers clearly produce three categories of V + C sequences in both Alemannic and SSG. For now, it seems more reasonable to assume three categories in both LU and ZH, in accordance with the analysis proposed by Zihlmann (2020b). All in all, these findings confirm H3a, stating no regional differences, and H3b, expecting three categories of V/(V + C) ratios.

The reported results will be further discussed in the next section.

4.1 Discussion of the results

Based on the results of this study, no strong conclusions can be formed on the differences in AR between Alemannic and SSG or between LU and ZH. Even the age-related differences, which turned out to be the most stable, are not as strong as expected, at least in Alemannic, where there is a lot of variability in the group of younger speakers. Ultimately, AR alone does not seem to be a reliable measure to distinguish between Alemannic and SSG or between LU and ZH. At least in the case of this study, it is rather a measure to control for when analyzing segment durations that are dependent on speech tempo.

The most prominent difference between Alemannic and SSG turned out to be that both vowel and closure durations are significantly longer in dialect compared to standard speech. Regarding the vowel durations, LU and ZH speakers clearly differ in terms of their vowel categories in Alemannic: While the three categories for LU speakers consisted of (1) VCː, (2) VC and VːCː, and (3) VːC, those for ZH speakers consisted of (1) VCː and VC, (2) VːCː, and (3) VːC. There seems to be a specific vowel pattern used by the speakers from ZH, which they only produce in Alemannic.

Note that the relative closure durations are the only measure with a main effect of region, which was due to LU speakers producing longer closures than ZH speakers. As stated above, this was most probably due to the shorter closure durations produced by younger speakers from ZH. This is possibly the most important finding in this study, as it might be due to a contact to GSG, which, in turn, could lead to a sound change in progress, reducing the phonetic consonant categories to lenis and fortis. It is particularly striking that younger speakers from ZH do not only produce shorter closure durations in SSG, but also in Alemannic. As for now, the distinction between lenis, fortis, and extrafortis is still found in all groups, but the trend found here might develop into consonant productions much closer to those found in GSG in those regions that have high contact to GSG. This result offers a new perspective on Alemannic and SSG, as it has been assumed, until now, that SSG is only influenced by Alemannic (Hove 2002; Zihlmann 2020b). The next step here would be to investigate not only closure duration, but also VOT, which is known to be the primary cue to distinguish between lenis and fortis plosives in GSG. If younger speakers from ZH produce shorter closure durations and longer VOTs, there might be a change due to a trading relation between the two, meaning the more they aspirate, the shorter the closure becomes. A similar trend has been found for German speakers of the Bavarian and Saxon dialect, where VOT becomes increasingly important for younger speakers (Kleber 2018). At this point, it is speculative if the situation is similar for ZH speakers, but it is worth further investigation in both production and perception.

All in all, although some remarkable differences – mainly the difference in segment durations between Alemannic and SSG, the different vowel categories in LU compared to ZH speakers in Alemannic, and the shorter closure durations produced by younger ZH speakers – were found when looking at the segment durations on their own, the V/(V + C) ratios are remarkably stable.

4.2 Remarks/limitations

There are some limitations due to practicability and time reasons that need to be mentioned, since they might have influenced the results to a certain amount.

Firstly, the number of speakers might not have been enough to detect either regional or generational differences in speaking behaviors. Yet, this limitation should be partly compensated by the high number of tokens per speaker. For future research, it is worth considering more participants and a lower number of tokens per speaker, considering that the inevitable manual adjustment of the segmentation process is a time-consuming task.

The speaker groups could also have been more homogenous in terms of age. The mean ages of the older speaker groups differed six years from each other between LU and ZH. Not only were the older speakers from LU younger than those from ZH, but they also differed more in age within the group. Furthermore, the younger LU group was, on average, almost six years older than the younger ZH group, meaning that the age difference between the younger and older speakers was somewhat smaller in LU than in ZH. Despite that, there was still a mean difference of nearly 30 years between the age groups in LU, which certainly can be sufficient for generational comparisons. The mean age difference between the groups from ZH was about 40 years, which is even more suitable for age-dependent comparisons.

Furthermore, some of the speakers tended to produce the fast sentences as fast as they possibly could, instead of speaking in a fast but natural speech tempo. It remains unclear how this may have influenced the results. Since it was the case for speakers from both regions, the groups are still comparable to each other.

Concerning the stimuli, there are three main remarks. First, it would have been optimal for the dialect data to have the same target words for both regions to make a better comparison. This, however, would result in a smaller number of stimuli, as both lexical and phonetic properties differ between the two regions. Therefore, the comparisons between regions concern groups of VC combinations rather than words. Second, the SSG stimuli did not have a category for bilabial plosives + [i], or one for velar plosives + [u]. This could have influenced the results. Since no minimal pairs could be found in these categories of SSG, the only solution would have been to remove these categories from the Alemannic stimuli as well and this, in turn, would have led to a further reduction in stimulus number. It ultimately was decided that the higher stimulus number was more important in this case. Third, the usage of the spelling system by Dieth (1986) for the Alemannic recordings but the standard German orthography for the SSG recordings might have influenced some of the participants’ productions. Yet, this effect should be minor because the speakers read the sentences silently before speaking out loud and did not have the text in front of their eyes while speaking. Additionally, all unexpectedly pronounced tokens were excluded from the dataset.

To summarize, for future studies, a higher number of speakers per group is considered most important. If possible, a more homogeneous group in terms of age is also preferable; yet, an age difference of 30 years for one regional group and of 40 years for the other one is also acceptable considering the difficulties of participant recruiting.

4.3 Outlook

To conclude, the main findings of this study show that the most prominent difference between Alemannic and SSG lie in the longer segment durations in dialect compared to standard speech. Furthermore, there are three phonetic vowel categories in Alemannic that differ between LU and ZH. Regarding SSG, perceptual data would be needed to obtain the number of vowel categories, but there are at least two. The three V/(V + C) ratio categories and the phonetic consonant categories lenis, fortis, and extrafortis are stable across tempo, age, and region in both Alemannic and SSG. Most importantly, younger ZH speakers produced overall shorter closure durations, even if they still conserve the three consonant categories. This should be investigated further by focusing on both the closure duration and the VOT of plosive consonants in ZH German in both production and perception.