The Phonetic Properties of Stops

in the following. The phonemes relevant for the present investigation are listed again in table 2.2, which demonstrates that both Standard German and Swiss German stops are produced at three different PoAs: labial, alve-olar and velar. Both languages have an additional two-way contrast in their respective stop systems. In Standard German this contrast is a laryngeal one: /b, d, g/differ from/p, t, k/with respect to their voicing quality – or, to be more precise, in the amount of aspiration that follows the release (see below). In Swiss German, on the other hand, there is no laryngeal contrast.

The stops do not differ in voicing/aspiration but in length. Thus, other than Standard German, Swiss German employs a quantity contrast to distinguish two stop series.

Standard German Swiss German

labial alveolar velar labial alveolar velar

tense /p/ /t/ /k/ /p:/ /t:/ /k:/

lax /b/ /d/ /g/ /p/ /t/ /k/

Table 2.2: Oral stop phonemes in Standard German and Swiss German.

After some general remarks on the phonetic properties of stop consonants, the remainder of this chapter will focus on the phonetic properties and peculiarities of the Standard German laryngeal contrast (section 2.2) and the Swiss German length contrast (section 2.3). The respective phonological representations will be addressed in chapter 3.

2.1 The Phonetic Properties of Stops

A stop (sometimes also termed ‘plosive’ or ‘occlusive’) is a speech sound characterised by a complete closure of the oral tract which prevents the airstream from escaping through the mouth.⁸ The stops occurring in the

8Note that from the point of view of phonology a further distinction can be made between oral stops and nasal stops (also cf. Ladefoged 2006:13f.). Oral stops are marked by acomplete disruption of the airstream (hence the name ‘stops’). In addition to the articulatory oral obstruction, the soft palate is raised, blocking off the nasal tract so that there is no way for the air to escape. Unlike oral stops, nasal stops are produced with a lowered velum. Thus, in spite of the oral obstruction, the airstream can escape through the nose. Since the disruption of the air flow is not complete, nasals are not considered as stops from a phonetic perspective. Phonologically, both kinds of stops are assumed to be characterised by the feature [–continuant]. Nasals are not an issue in the present thesis and in subsequent discussions the term ‘stops’ refers to oral stops only.

phoneme inventories of German and Swiss German, respectively, are listed in table 2.2 above.

Structurally, a stop can be divided into three phases. The first phase consists of the formation of an obstruction in the oral tract, marked by an abrupt decrease in spectral energy. In German and Swiss German, the ob-struction is realised by closing the lips for the bilabial stops /b, p, p:/, by raising the tongue body to the alveolar ridge for the coronal stops /d, t, t:/

and by raising the back of the tongue to the velum for the dorsal stops /g, k, k:/. Next comes the closure phase or ‘occlusion’. The obstruction in the vocal tract is maintained for a short while, leading to an air pressure build-up inside the mouth. In voiceless stops, the closure phase is a period of complete silence. In voiced stops, vocal fold vibration continues during closure: A so-called ‘voice bar’ is visible in the lowest frequency ranges of the spectrogram and regular wave motion can be observed in the oscillo-gram. The final phase is the burst or release phase. It refers to the opening of the closure which is accompanied by a burst sound when the airstream is released, and by an abrupt rise of spectral energy. In the oscillogram, a spike in the waveform usually indicates the release of the stop, that is, the beginning of the burst. Burst intensity is weaker in voiced than in voiceless stops. The release duration – that is the burst plus any potential aspiration preceding the onset of glottal pulsing of the following sound – is compar-atively short in voiced and voiceless unaspirated stops. It is clearly longer in aspirated stops, where the burst is followed by a period of voicelessness.

The waveform and spectrogram of a labial voiceless aspirated stop is given in figure 2.1.⁹

A crucial notion in connection with stops is the concept of voice onset time (VOT), defined as “the interval between the release of the stop and the onset of glottal vibration, that is, voicing” (Lisker & Abramson 1964:389).

The reference point (‘zero-time’) to which the onset of voicing is related is the instant of release. There are three possible configurations that result in either voiced, voiceless or voiceless aspirated stops. In voiced stops, vocal fold vibration starts before the release of the stop’s closure. Such a timing relation is calledprevoicing orvoicing lead and is also referred to asnegative VOT. Voiceless as well as voiceless aspirated stops are characterised by a voicing lag, with vocal fold vibration starting some time after the release burst. Stops of this kind are said to have a positive VOT. In voiceless

9The abbreviation ‘ACT’ used in figure 2.1 refers to ‘after closure time’, a notion that is used instead of (positive) VOT. The reasons that led to the adoption of this terminology are outlined in the following paragraphs. ‘CD’ stands for ‘closure duration.’

2.1 The Phonetic Properties of Stops 19 unaspirated stops the voicing lag is rather short and voicing of the following sound sets in together with the release burst or shortly afterwards. Voiceless aspirated stops exhibit a relatively long voicing lag, which is often perceived as a kind of ‘puff’ following the stop.

Figure 2.1: Voiceless aspirated labial stop (pictured sequence: 350 ms).

In a study on utterance-initial stops in eleven languages, Lisker & Abram-son (1964) show that these three categories, voiced, voiceless and voiceless aspirated, relate to three more or less stable ranges on a VOT continuum, the precise values for which are presented in table 2.3. A schematic repre-sentation of the three categories which differ with respect to their laryngeal configuration is given in figure 2.2, where the time scale is centred on the moment of release.¹⁰ Although there are language-specific differences (also cf. Ladefoged 2006), stops can be categorised according to these VOT ranges.

A more recent study by Cho & Ladefoged (1999) analyses voicing lags in utterance-initial stops in 18 languages and reports on considerable language-specific variation. The authors suggest that there are rather four voicing lag

10Note that the values are averaged over different PoAs. When voicing lags are con-sidered separately for the different PoAs, an increase in VOT is observed from labial to dental/alveolar to velar PoA (see, e.g., Lisker & Abramson 1964; Cho & Ladefoged 1999).

VOT Range Mean VOT

-125 ms to -75 ms -100 ms voicing lead

0 ms to 25 ms 10 ms short voicing lag

60 ms to 100 ms 75 ms long voicing lag

Table 2.3: VOT categories as reported by Lisker & Abramson (1964).

categories than the commonly assumed two (unaspirated vs. aspirated). For velar PoA, they assign stops with a VOT of approximately 50 ms to a cate-gory ‘slightly aspirated’, ‘aspirated’ refers to stops with a VOT of approxi-mately 90 ms and a category ‘highly aspirated’ is mentioned for stops with very long voicing lags above 100 ms. Voiceless unaspirated velars, so they suggest, have a VOT of approximately 30 ms (Cho & Ladefoged 1999:223).

Yet, most of the languages investigated by Cho & Ladefoged (1999) with two laryngeal categories display a contrast of unaspirated stops versus one of the aspiration categories. Thus, instead of speaking of four lag categories, it would be more plausible to speak of two main categories, short and long voicing lag. Language-specific phonetic realisations could then be defined more precisely by reference to three sub-categories for aspirated stops.¹¹

Figure 2.2: VOT-continuum with three distinct stop categories (values adopted from Lisker & Abramson 1964).

Given that velar/uvular PoA involves longer voicing lags than labial or den-tal/alveolar, the data presented by Cho & Ladefoged (1999) are largely in line with the findings by Lisker & Abramson (1964). Cho & Ladefoged

11Note that neither the study by Lisker & Abramson (1964) nor the one by Cho &

Ladefoged (1999) reports of a language which has more than two stop categories on the lag side of the laryngeal continuum, thus making the four-fold division of the lag region a phonetic accuracy, which, however, does not conflict with the division of the laryngeal continuum into one lead and two lag categories suggested by Lisker & Abramson (1964).

2.1 The Phonetic Properties of Stops 21 (1999) report a significant difference in VOT between coronal and dorsal PoA (∆18.9 ms for unaspirated stops, ∆16.7 ms for aspirated stops). The difference in VOT between labials and coronals, in turn, is not significant.¹² Cho & Ladefoged’s (1999) data suggest that for velars the boundary that separates unaspirated and aspirated voiceless stops is located around a VOT value of 40 ms. Consequently, it can be inferred that for labial and coronal stops the boundary is found somewhere around 20 ms to 25 ms, a value well compatible with Lisker & Abramson’s (1964) generalised categories (table 2.3). Furthermore, both studies suggest that the long lag category shows more variation than the short lag category, which seems to be refined to a smaller VOT range. Also, since Cho & Ladefoged (1999) do not find a phonological explanation for why a given aspiration-subcategory is se-lected in a particular language, they stick to the conventional tripartition of VOT contrasts and speak of “three modal values of VOT, [voiced], [voiceless unaspirated], and [aspirated]” (Cho & Ladefoged 1999:226), of which only the latter two are examined in their study.

Lisker & Abramson’s (1964) cross-linguistic study revealed that lan-guages employing only two stop categories do not use the most distinct categories to make a contrast, that is, stops with a voicing lead on the one hand are not contrasted to stops with a long voicing lag on the other hand.

Instead, all examined languages with a two-way stop opposition contrast either stops with a voicing lead to stops with a short voicing lag or stops with a short voicing lag to stops with a long voicing lag. In other words, the middle category is always present.¹³ Although not explicitly mentioned, the same is true for the languages examined by Cho & Ladefoged (1999).¹⁴ 2.1.1 Some Terminological Remarks

It has been shown that VOT is not always an appropriate notion to describe the phonetic attributes of stops. Consider the fact that voicing does not al-ways continue throughout the whole closure phase. Especially in lax stops

12Labial initial stops are reported to have a mean VOT of 15.3 ms, a value similar to the values found for voiceless (unaspirated) stops in German and in Swiss German, as will be seen in section 2.3 as well as in chapter 5.

13One of the four speakers recorded by Lisker & Abramson (1964) marks an exception in that this person consistently produced lax stops with prevoicing, while tense stops were aspirated.

14Accordingly, the voiceless unaspirated stops are regarded the least marked stops in the languages of the world. Languages which have one stop series only usually include stops of this kind in their inventory and not voiced or voiceless aspirated ones (see Ladefoged

& Maddieson 1996:53; Hall 2000:81, 88.)

following voiceless segments, the beginning of the closure might be voiceless and glottal pulsing might start only shortly before the stop is released. In this case as well as in utterance-initial position, negative VOT can be mea-sured easily. The situation is different if the stop follows a sonorant. There might be some voicing at the beginning of the stop’s closure which might decrease to zero before the stop is released. Thus, there is a ‘voicing inter-ruption’, a short voiceless phase between voicing during closure and voicing of the following segment (see figure 2.3). Furthermore, voicing into closure is not restricted to lax stops but can be observed sometimes also in tense stops – although it is usually more pronounced in lax ones (cf. Mitleb 1981).

Similarly, Mikuteit (2006) argues that in a voiced context (e.g., in in-tervocalic position), it is impossible to determine where sonorant voicing ends and obstruent voicing starts. Hence, according to her, it is incorrect to speak of voice onset time since glottal pulsing already starts in a preceding segment. She relinquishes the notion of VOT and introduces two notions to replace it. The first one, closure voicing, is used to refer to the presence of vocal fold vibration before the stop’s release, that is, a voicing lead, formerly referred to as negative VOT. Secondly, a voicing lag is called after closure time (ACT) instead of positive VOT.¹⁵Furthermore, Mikuteit (2006) points out that even in stops displaying prevoicing, glottal pulsing of the following vowel does not always start immediately after the burst and there might be a slight lag of time between the release burst of the stop and the onset of periodicity of a following sound, such that there might be an interruption in vocal fold vibration even in voiced stops. In other words, prevoiced stops may exhibit both a voicing lead and a short voicing lag, as shown in the example of a prevoiced labial stop in figure 2.3. Similar observations are found in Jessen & Ringen (2002). Accordingly, the authors present positive VOTs for both tense and lax stops in their experiment and they indicate separately whether or not (full or partial) voicing is present during a stop’s closure phase.¹⁶ Similarly, when lax stops are produced with prevoicing, Ke-hoeet al.(2004:76) distinguish between ‘lead VOT’ and ‘lag VOT’, pointing out that prevoiced stops can be characterised by both measures.

15Mikuteit’s (2006) main argument for the introduction of the new terminology is based on East Bengali data. With ACT and a second new measure, ‘superimposed aspiration’, she analyses the East Bengali four-way contrast without having to refer to ‘breathy voice’

to distinguish voiced aspirated stops. Voiced aspirated stops do not play a role in the stop systems of German and Swiss German.

16Note that van Alphen (2004) reports that in cross-modal priming experiments with adults, variation in the amount of prevoicing in primes with initial voiced stops has no influence on lexical access. Thus, it seems to be rather the general presence of glottal vibration than the precise amount of prevoicing that affects lexical processing.

2.1 The Phonetic Properties of Stops 23

Figure 2.3: Intervocalic voiced labial stop with both a voicing lead (closure voicing) and a short voicing lag (ACT).

In his study on German stop contrasts, Braunschweiler (1997) avoids the term ‘VOT’. When referring to voicing lags, he speaks of release duration.

Like Mikuteit (2006) and Jessen & Ringen (2002), he does not specify voic-ing leads in absolute values. He simply refers to the presence or absence of closure voicing. Thus, even though the dimension is often regarded to be the same (i.e., the onset of vocal fold vibration relative to the stop’s release to be measured on a voicing continuum), it appears to be more appropri-ate to think of voicing lead and voicing lag as two separappropri-ate cues instead of seeing lead and lag as complementary parts of the same cue. Braun’s (1996:20) claim that VOT pools the originally distinct categories of voicing and aspiration in one single dimension is of the same tenor. For these rea-sons, Mikuteit’s (2006) terminology is adopted in the present work. Instead of negative VOT and positive VOT, the terms closure voicing and after closure time (henceforth ACT) will be used in the remainder of the thesis.