. 2
( 12)


rule (IP stands for “intonational phrase”):20

(44) Aspiration ± 
’continuant ’ [+aspirated] /_____[’consonant] 
+spread glottis _____[+sonorant, ’syllabic]
 

The Aspiration rule states that stops that are [+spread glottis] are aspirated
when they occur before a vowel, a non-syllabic liquid or nasal, or a pause
(at the end of an intonational phrase).21 The examples in (45) have aspirated
stops in prevocalic position (before stressed and unstressed vowels); those
in (46) show aspirated stops before non-syllabic liquids and nasals;22 those in
(47) have aspirated stops before a pause.

Pass ˜passport™ /"p as/ ’ ["p as]
Miete ˜rent™ /"mi¦t ™/ ’ ["mi¦t ™]
Br¨ cke ˜bridge™ /prk ™/ ’ [pʁk ™]
Pr¨ fung ˜test™ /p ry¦f…N/ ’ [p ʁy¦f…N]
(46) u
weltlich ˜wordly™ /vµlt l©x/ ’ [vµlt l©c]
Knie ˜knee™ /k ni¦/ ’ [k ni¦]
24 German

schlapp ˜worn out™ /ʃlap / ’ [ʃlap ]
nett ˜nice™ /nµt / ’ [nµt ]
Rock ˜skirt™ /r”k / ’ [ʁ”k ]
Because the rule of Aspiration does not apply to [+spread glottis] stops that
occur before syllabic liquids and nasals, it prevents aspiration in words like
leiten ˜to lead™ when pronounced without schwa, since the absence of schwa
requires a syllabic pronunciation of /n/: /la©t ™n/ ’ [la©tn] (see the discussion
in sections and
Notice that stops that are phonemically lenis ([’spread glottis]) are realized
as [+aspirated] when they occur at the end of a phonological word and in an
environment for aspiration.24
handlich ˜handy™ /hantl©x/ ’ [hant l©c]
(48) ¸
Neid ˜envy™ /na©t/ ’ [na©t ]
Because they occur at the end of a phonological word, they undergo Fortition,
which causes them to become [+spread glottis]. They are then eligible for the
rule of Aspiration “ as long as they occur in one of the proper environments.
In handlich, /t/ undergoes Fortition and is then eligible for Aspiration because
it occurs before a non-syllabic liquid. The /t/ in Neid is eligible for Aspiration
(after Fortition) because it occurs before a pause.
Stops that occur in an onset with a preceding fricative are not aspirated,
as demonstrated by the word Stahl ˜steel™ [ʃta¦l]. Because the fortis/lenis dis-
tinction is neutralized in this position, it is not clear whether these stops are
allophones of fortis or lenis stops. We will simply assume that they are lenis
([’spread glottis]), in which case the rule of Aspiration will not apply to them.
A word like Stahl, for example, is represented phonemically as /ʃta¦l/. Fricative Devoicing Traditionally, fricatives in German are
treated together with stops (they both belong to the natural class of obstruents),25
and rules that affect stops (Fortition, for example) apply to fricatives as well.
However, it turns out that German fricatives differ from stops. In particular,
unlike their stop counterparts, they contrast in voicing. The fortis fricatives, /f
s ʃ c x h/, are [’voice]; the lenis fricatives, /v z j/, are [+voice]. Minimal
pairs (or near-minimal pairs) demonstrate this contrast in onsets (word-
initially as well as intervocalically):26
(49) vier ˜four™ [fi¦ɐ8] wir ˜we™ [vi¦ɐ8]
Grafen ˜counts™ [kʁa¦.fn] braven ˜well-behaved™ [pʁa¦.vn]
(50) Seal ˜seal™ [si¦l] Siel ˜sluice™ [zi¦l]
reißen ˜to tear™ [ʁa©.sn] reisen ˜to travel™ [ʁa©.zn]
In word-¬nal position, however, this contrast is neutralized by Fortition:
Phonetics and phonology 25

reisen ˜to travel™ /ra©.z™n/ ’ [ʁa©.zn]
reis! ˜travel!™ /ra©z/ ’ [ʁa©s]
reißen ˜to tear™ /ra©.s™n/ ’ [ʁa©.sn]
(ich) reiß™ ˜(I) tear™ /ra©s/ ’ [ʁa©s]

The contrast in voicing is also neutralized when a fricative occurs before an
reist ˜travels™ /ra©zt / ’ [ʁa©st ]
reißt ˜tears™ /ra©st / ’ [ʁa©st ]

The rule of Fricative Devoicing accounts for the [’voice] realization of lenis
fricatives when they occur before an obstruent.

(54) Fricative Devoicing
[’sonorant] ’ [’voice] /_____ [’sonorant]

This rule states that all obstruents (stops as well as fricatives) are realized
as [’voice] when they occur before an obstruent. The fricative /z/ in reist
˜travels™(/ra©zt /), for example, is realized as [s] ([ʁa©st ]) because it occurs
before the obstruent /t /. It is not necessary to add the feature [+continuant]
and restrict the rule to fricatives, since all stops in German are phonemically
[’voice] and therefore cannot be affected by the rule.27 Velar Fricative Assimilation The phonemic status of Ger-
man [c] and [x] has long intrigued scholars and has resulted in extensive
discussion in the literature (see Robinson 2001 for an overview of the research
on this topic). The following examples, which show the two sounds to be in
complementary distribution, with [x] after non-front vowels and [c] elsewhere
(after front vowels, after consonants, word-initially), would argue for a single

(55) Buch ˜book™ [pu¦x]
doch ˜but still™ [t”x]
Bach ˜brook™ [pax]
Brauch ˜custom™ [pʁa…8x]
(56) ich ˜I™ [©c]
B¨ cher ˜books™ [py¦cɐ]
u ¸
brechen ˜to break™ [pʁµcn]¸
h¨ chst ˜highest™ [hø¦cst ]
o ¸
Leiche ˜corpse™ [la©c™]
Br¨ uche ˜customs™ [pʁ”©c™]
a ¸
26 German

(57) Milch ˜milk™ [m©lc]
durch ˜through™ [t…ɐ8c]
manch ˜many a™ [manc] ¸
(58) Chemie ˜chemistry™ [cemi¦]
China ˜China™ [ci¦na]
However, minimal pairs like those in (59) and word-initial [x], illustrated in
(60), raise doubts about the validity of a single-phoneme analysis.
(59) Kuchen ˜cake™ [k u¦xn] Kuhchen ˜little cow™ [k u¦cn]
tauchen ˜to dive™ [t a…8xn] Tauchen ˜little rope™ [t a…8cn]
(60) Chanukka ˜Hanukkah™ [xan…k a¦]
Junta ˜junta™ [x…nt a]28
Notice, though, that the differences in pronunciation of <ch> in the word
pairs in (59) correlate with differences in word structure. Kuhchen ˜little cow™
and Tauchen ˜little rope™ contain the suf¬x -chen, where <ch> occurs at the
beginning of a phonological word: Kuchen, Kuh+chen, tauch+en, Tau+chen.
In addition, word-initial [c] in the examples in (58) appears before front vow-
els, whereas word-initial [x] in (60) is found before non-front vowels.29 The
predictable differences in distribution between [c] and [x] allow us to account
for the data by assuming a single phoneme for German, /x/.30 The palatal real-
ization of this phoneme, [c], is accounted for by the rule of Velar Fricative
(61) Velar Fricative 
Assimilation ± 
+consonantal ’ [’back]/  +sonorant _____ 
 
 
°+continuant » ’back
+back  {_____ +sonorant 
 
 
This rule realizes the velar fricative, /x/, as [c] when it follows [’back] vowels
and /r/, /l/, and /n/ (the consonantal sonorants) and when it occurs at the
beginning of a phonological word before these same segments.32 It accounts
for the [c] pronunciation of /x/ in words like ich, Milch, China, and Kuhchen.
This rule represents a process of assimilation whereby a [+back] fricative is
realized as [’back] when it follows a [’back] sonorant (vowel or consonant) or
when it precedes a [’back] sonorant at the beginning of a phonological word.
When /x/ does not occur in these environments it will be realized as a [+back]
fricative, that is, as [x] (/x/ in words such as Buch, Chanukka, and Kuchen). I
assume that schwa is phonologically [’front, ’back], which accounts for the
[c] pronunciation in the suf¬x -chen when schwa is pronounced: [c™n]. I also
¸ ¸
assume that /a/ and /a¦/, which are central vowels phonetically, are speci¬ed
Phonetics and phonology 27

phonologically only for the feature [front]. Because they are [’front] (and not
speci¬ed for the feature [back]), they will not cause /x/ to be pronounced as [c]
in words such as Bach and Chanukka. Velar Spirantization As the following examples illustrate,
[k] can alternate with [c] in related word forms.
ruhiger ˜calmer™ [ʁu¦©kɐ]
(62) ruhig ˜calm™ [ʁu¦©c]
beruhigen ˜to calm™ [p™ʁu¦©k™n] beruhigt ˜relieved™ [p™ʁu¦©ct ]
K¨ nige ˜kings™ [k ø¦n©k™]
o K¨ nig ˜king™ [k ø¦n©c]
o ¸
If we assume that /k/ is the phoneme in all these forms, we can derive the
forms with [c] using the rule of Velar Spirantization (where Cø stands for zero
or more consonants and $ stands for a syllable boundary) and Velar Fricative
(63) Velar Spirantization ® 
® 
’continuant ’ [+continuant] / ’consonantal _____ Cø $
+high 
°’spread glottis »  
’back 
+back  
’tense 
 
°’round »
Condition: Does not apply to words in which /k/ is followed by a
syllable containing /x/.
The rule of Velar Spirantization turns the velar stop /k/ into a voiceless velar
fricative (spirant), [x], when it is in coda position following unstressed /©/ (/k/
is phonemically voiceless), and Velar Fricative Assimilation realizes it as [c], ¸
since it follows /©/, a [’back] vowel.
A condition on Velar Spirantization speci¬es that the rule does not apply to
words that contain /x/ in the syllable following /k/. This condition prevents the
rule from applying to words like K¨ nigreich ˜kingdom™, k¨ niglich ˜royal™, and
o o
ewiglich ˜eternally™, which have a /k/ that is realized as [k ] rather than [c]:
K¨ nigreich ˜kingdom™ [k ø¦n©k ʁa©c]
(64) o ¸
k¨ niglich ˜royal™ [k ø¦n©k l©c]
o ¸
ewiglich ˜eternally™ [e¦v©k l©c] (cf. ewig ˜eternal™ [e¦v©c])
¸ ¸
Words like those in (64) contain [k ] rather than [c] because the standard
language apparently does not favor two successive syllables ending in [c] ¸
(Siebs 1969:100).
The rule of Velar Spirantization does not occur in southern German, where
/k/ in words like ruhig ˜calm™ and beruhigt ˜relieved™ is realized as a stop:
28 German

(65) Southern German
ruhig ˜calm™ /ru¦©k/ ’ [ʁu¦©k ]
beruhigt ˜relieved™ /p™ru¦©kt / ’ [p™ʁu¦©kt ]

In northern and central German, on the other hand, the rule of Velar Spiranti-
zation is more general, applying to all instances of /k/ in coda position, not just
those following unstressed /©/:34

(66) Northern and central German
Sieg ˜victory™/zi¦k/ ’ [zi¦c]
gesagt ˜said™ /k™za¦kt / ’ [k™za¦xt ]

Standard German is thus an interesting mix of southern and non-southern
pronunciation: In southern German (S), /k/ in coda position is realized as a
stop; in northern and central German (N/C), /k/ in coda position is realized as
a fricative. In Standard German, /k/ in coda position is realized as a stop in all
positions except in the sequence /©k/, where it is realized as a fricative:35

(67) /k/ in coda position S N/C Standard
After central/back vowel Zug ˜train™ [k ] [x] [k ]
After front vowel Sieg ˜victory™ [k ] [c]
¸ [k ]
In the sequence /©k/ wichtig ˜important™ [k ] [c]
¸ [c]
¸ Glottal Stop Insertion Although the glottal stop is not con-
sidered a phoneme of German, its distribution is predictable. The glottal stop
occurs before vowels and is most common at the beginning of a phonological
word with initial stress and at the beginning of utterances (Kohler 1994). We
can thus express the rule of Glottal Stop Insertion in terms of the phonological
word ({), the foot (F), and the intonational phrase (IP). (Recall that a foot is a
stressed syllable and any following unstressed syllables that intervene before
the next stressed syllable.) According to Kohler (1994:42), the glottal stop is
more commonly pronounced as glottalization (creaky voice) of the following
vowel (with or without long glottal closure preceding the vowel) than as a sim-
ple glottal stop (closure and release).36 We will express Glottal Stop Insertion
as the insertion of the feature [+constricted glottis] and leave the realization of
this feature (creaky voice or glottal stop) to the phonetics.

(68) Glottal Stop Insertion
˜ ’ [+constricted glottis] / {F [_____
IP [_____
Condition: Optional.
Phonetics and phonology 29

The rule of Glottal Stop Insertion inserts the feature [+constricted glottis]
before a vowel at the beginning of a phonological word when that word bears
initial stress or is utterance-initial. Because the glottal stop (glottalization) is
not used universally, it is expressed as an optional rule. This rule accounts for
the distribution of glottal stops in a phrase like the following (from C. Hall
(69) Er hat die anderen beeindruckt. ˜He has impressed the others.™
[”eɐ8 hat ti "”ant™ʁ™n p#™"”a©ntʁ…kt ]
The ¬rst glottal stop in this phrase is utterance-initial; the second two glottal
stops occur at the beginning of phonological words and are also foot-initial.
Words with stressed vowels that are morpheme-internal do not undergo Glottal
Stop Insertion.38
Theater ˜theater™ [t e"a¦t ɐ]
Kloake ˜sewer™ [k lo"a¦k ™]
Although these vowels are foot-initial, they do not occur at the beginning of a
phonological word. Schwa Deletion The status of schwa in German has received
a good deal of attention in the literature, with some authors arguing that schwa
is a phoneme of German that is deleted in some contexts (Kloeke 1982, Strauss
1982, Benware 1986), and others arguing that schwa is not a phoneme, but that
it must be inserted in certain contexts (Wiese 1986, 1988; Giegerich 1987).39
A third approach, expressed by Becker (1998), treats schwa as an allophone
of /e/. I assume that schwa is a phoneme of German and that a deletion rule is
necessary to account for schwa“zero alternations like the following:
Cochemer ˜person from Cochem™ [k ”x™mɐ], Cochem ˜Cochem™
[k ”xm]
ebene ˜smooth™ [e¦p#™n™], eben ˜smooth™ [e¦p#m]
Eselei ˜stupidity™ [e¦z™la©], Esel ˜donkey™ [e¦zl]
bittere ˜bitter™ [p©t ™ʁ™], bitter ˜bitter™ [p©t ɐ]
Assuming that both forms in each pair in (71) have a schwa phonemically, the
rule of Schwa Deletion accounts for the forms in which schwa is not realized
(72) Schwa Deletion ± 
 
’sonorant  
 

/™/ ˜/ _____ /m/  
 
[’sonorant ] _____ /n/ Cø $
 
 
 
[+consonantal] _____ /l/  
 
_____ /r/
30 German

This rule deletes schwa before /m/, /n/, /l/, and /r/ when these occur in a coda, as
long as certain conditions are met on the segments that precede schwa. Schwa
is deleted before /m/ only if a fricative precedes schwa:

tiefem ˜deep™ /t i¦f™m/ ’ [t i¦fm]
Cochems ˜Cochem™s™ /k ”x™ms/ ’ [k ”xms]
bl¨ dem ˜stupid™ /plø¦t™m/ ’ [plø¦t™m] (stop precedes schwa)

If schwa occurs before /n/, an obstruent (a fricative or a stop) must precede
schwa. Thus the conditions on Schwa Deletion before /n/ are less restrictive
than those before /m/.

reisend ˜travelling™ /ra©z™nt/ ’ [ʁa©znt ]
Faden ˜thread™ /fa¦t™n/ ’ [fa¦tn]
nehmen ˜to take™ /ne¦m™n/ ’ [ne¦m™n] (nasal precedes schwa)

The conditions on Schwa Deletion before /l/ are even less restrictive. The
segment preceding schwa must be a consonant.41

L¨ ffel ˜spoon™ /l“f™l/ ’ [l“fl]
(75) o
Mittel ˜means™ /m©t ™l/ ’ [m©tl]
angeln ˜to ¬sh™ /aN™ln/ ’ [aNln]
Greuel ˜horror™ /kr”©™l/ ’ [kʁ”©™l] (vowel precedes schwa)

There are no conditions on the segment that precedes schwa when it occurs
before /r/; schwa is simply deleted before /r/ when it occurs in a coda (and /r/
is realized as [ɐ]).

bitter ˜bitter™ /p©.t ™r/ ’ [p©.t ɐ]
verbittert ˜embittered™ /fµr.p©.t ™rt / ’ [fµɐ8.p#©.t ɐt ]

Notice that /m/, /n/, /l/, and /r/ are all realized as [+syllabic] after a preceding
schwa has been deleted. In addition, /r/ is realized as [’consonantal]. These
details are addressed in the following section, which treats the realization
of /r/. r-Vocalization The feature [’back] is one of the important
features of /r/, since this is the feature that plays a role in Velar Fricative
Assimilation, causing a [+back] fricative to become [’back]: durch ˜through™
/t…rx/ ’ dur[c]. Although the vocalic realizations of /r/ are [’back] ([ɐ]
and [ɐ8]), the consonantal realization ([ʁ]) is [+back]. The rules that state
the distribution of the allophones of /r/ take this into account. The consonantal
allophone of /r/ is produced by a rule that realizes /r/ as a voiced uvular fricative
when it occurs in an onset:
Phonetics and phonology 31

(77) Consonantal Realization of /r/
/r/ ’ +back / $Cø _____

This rule yields the consonantal pronunciation of /r/ in words like those
in (78).

Rede ˜speech™ /re¦t™/ ’ [ʁe¦t™]
leeren ˜to empty™ /le¦.r™n/ ’ [le¦.ʁ™n]
Schrift ˜writing™ /ʃr©ft / ’ [ʃʁ©ft ]

The vocalic allophones of /r/ are accounted for (in part) by the rule of r-
Vocalization, which changes the [+consonantal] feature of /r/ to [’consonantal]
when /r/ occurs in a coda:42

(79) r-Vocalization
/r/ ’ [’consonantal] / _____ Cø $

The rule does not change the [’syllabic] feature of /r/; thus it realizes /r/ in
coda position as [ɐ8].

leer ˜empty™ /le¦r/ ’ [le¦ɐ8]
leert ˜empties™ /le¦rt / ’ [le¦ɐ8t ]
verkauft ˜sold™ /fµr.k a…8ft / ’ [fµɐ8.k a…8ft ]

The rule that accounts for the [’consonantal, +syllabic] realization of /r/
also accounts for the [+syllabic] realizations of the other consonantal sonorants,
namely, /m/, /n/, and /l/.

(81) Sonorant Syllabi¬cation
[+sonorant] ’ [+syllabic] / $Cø _____ Cø $

This rule causes /m/, /n/, /l/, and /r/ to become syllabic and thus form the nucleus
of a syllable when they ¬nd themselves in a syllable without a nucleus because
of Schwa Deletion. This is exempli¬ed for /l/ in (82).

(82) /l“.f™l/ L¨ ffel ˜spoon™
’ [l“.fl] Schwa Deletion
’ [l“.fl] Sonorant Syllabi¬cation

When Schwa Deletion occurs before /r/, the rule of r-Vocalization will yield a
form with [ɐ8]; Sonorant Syllabi¬cation will turn [ɐ8], which is [’syllabic], into
[ɐ], which is [+syllabic].43
32 German

(83) /li¦.f™rn/ liefern ˜to supply™
’ [li¦.frn] Schwa Deletion
’ [li¦.fɐ8n] r-Vocalization
’ [li¦.fɐn] Sonorant Syllabi¬cation Nasal Assimilation Schwa deletion has repercussions in
addition to those that affect the syllabicity of a following consonant. In partic-
ular, when a stop precedes a schwa that is deleted before /n/, /n/ takes on the
place of articulation of the preceding stop. It is realized as a bilabial nasal, [m],
following a bilabial stop and as a velar nasal, [N], following a velar stop:
haben ˜to have™ /ha¦p™n/ ’ [ha¦p#m]
backen ˜to bake™ /pak ™n/ ’ [pakŋ]

The rule of Nasal Assimilation accounts for this change in place of articulation
of /n/. This rule uses “alpha notation,” where the Greek letter ± serves as a
variable and must have the same value each time it occurs.44
(85) Nasal Assimilation ® 
+nasal ’ [± artic]/ ’continuant _____
°’sonorant »
± artic
The rule of Nasal Assimilation states that a syllabic nasal preceded by a stop
must have the same place of articulation as the preceding stop.45 For example,
if the stop is bilabial, /p/ or /p /, the nasal has to be bilabial: the /n/ in haben
˜to have™ /ha¦p™n/ is realized as the bilabial nasal [m] because it is preceded by
the bilabial stop /p/ following Schwa Deletion.

1.2.5 Phonotactic constraints
Phonotactics deals with the permissible sequences of segments in a given lan-
guage. What is permissible in one language may not be permissible in another.
The sequence /k n/, for example, is perfectly acceptable in an onset in German,
but not in English; compare German Knie ˜knee™ /k ni¦/ and English knee /ni/.
The syllable plays an important role in phonotactics, since the constraints on
permissible sequences of segments in a language can often be stated in terms
of permissible onsets and codas.46 Syllable structure Although some languages allow only very
simple syllables (Hua, a Papuan language of New Guinea, allows only CV
syllables; Blevins 1995:217), German allows a variety of syllable types ranging
from syllables that consist solely of a nucleus to those that contain onsets and
Phonetics and phonology 33

codas made up of multiple segments. In German, all combinations of onset,
nucleus, and coda are possible (each syllable, by de¬nition, must contain a
(86) nucleus Kakao ˜cocoa™ [k a.k a¦.o]
onset + nucleus liefern ˜to supply™ [li¦.fɐn]
nucleus + coda Chaos ˜chaos™ [k a¦.”s]
onset + nucleus + coda kann ˜can™ [k an]
Onsets can consist of up to three segments, and codas can have up to four
(87) Onsets
one segment lebt ˜lives™ [le¦pt ]
two segments klebt ˜sticks™ [k le¦pt ]
three segments Sprache ˜language™ [ʃpʁa¦.x™]
(88) Codas
one segment mit ˜with™ [m©t ]
two segments kalt ˜cold™ [k alt ]
three segments ¬lmt ˜¬lms (3rd person sg.)™ [f©lmt ]
four segments hilfst ˜help (2nd sg.)™ [h©lfst ]
Words in German can be syllabi¬ed using universal principles and language-
speci¬c constraints. Each vowel is assigned to a nucleus. The longest sequence
of segments to the left of each nucleus that does not violate the phonotactic
constraints of the language (see the discussion below) is assigned to the onset
of the syllable with that nucleus. The remaining segments to the right of each
nucleus are assigned to the coda. These principles yield syllabi¬cations like
those indicated in the following words.
(89) beschleunigen ˜to accelerate™ /p™.ʃl”©.n©.k™n/
gesprochen ˜spoken™ /k™.ʃpr”.x™n/
Wagner ˜Wagner™ /va¦.kn™r/
The onsets of the second syllable in each of the words in (89) do not violate the
phonotactic constraints of the language, since these are possible word-initial
consonant clusters, as the examples in (90) demonstrate:
(90) schlimm ˜bad™ /ʃl©m/
Sprung ˜leap™ /ʃpr…N/
Gnade ˜mercy™ /kna¦.t™/
An additional principle must be taken into account when syllabifying words
like T¨ r¨ ffner ˜doorman™. The above principles would yield the syllabi¬cation
in (91), which would result in an unacceptable pronunciation.48
34 German

T¨ r¨ ffner ˜doorman™ /t y¦.r“f.n™r/ ’ — [t y¦.ʁ“f.nɐ]
(91) uo

In this syllabi¬cation, the /r/ in T¨ r¨ ffner is not in a coda “ but it must be, because

it must be vocalized (T¨ [ɐ8]offner). The principle necessary for producing the
correct syllabi¬cation, which has precedence over all other principles of syllab-
i¬cation, requires syllable boundaries to coincide with word boundaries. This
principle locates a syllable boundary between the two words in the compound
T¨ r¨ ffner. This yields the syllabi¬cation in (92), which results in an acceptable
T¨ r¨ ffner ˜doorman™ /t y¦r.“f.n™r/ ’ [t y¦ɐ8.“f.nɐ]
(92) uo
Many scholars assume that intervocalic consonants in German can be ambi-
syllabic, that is, that they can belong to the coda of one syllable and the onset
of a following syllable (e.g., Sievers 1893, Jespersen 1904, Trubetzkoy 1939,
Giegerich 1985, Benware 1986, Kohler 1995, Wiese 1996). In particular, an
intervocalic consonant following a short vowel (the /t / in bitte ˜please™, for
example) is assumed to be ambisyllabic. One argument in favor of such an
analysis is based on the fact that word-¬nal syllables in German do not end
in a short (lax) vowel.49 This leads to the conclusion that there must also be a
constraint against word-internal syllables ending in a short (lax) vowel. Ambi-
syllabic consonants are thus required in order to satisfy this constraint. As Hall
(1992:52) points out, however, a ¬‚aw in this line of reasoning is the assumption
that generalizations that hold for word edges also hold for syllable edges. He
notes that a number of studies show that this is not necessarily the case (Halle
and Vergnaud 1980, Clements and Keyser 1983, Itˆ 1986). In addition, there
does not appear to be clear-cut phonetic evidence in favor of ambisyllabicity.50
The analysis of German syllable structure presented here thus does not
recognize ambisyllabic consonants as being relevant for the phonology of
German. Onsets The acceptability of word-internal onsets is typically
determined by appealing to the acceptability of onsets in word-initial posi-
tion. For example, because /t l/ does not occur word-initially in German, this
sequence cannot function as an onset in a word like Atlantik ˜Atlantic™. The
onset of the second syllable in this word can consist maximally of one seg-
ment, /l/: /at .lan.t ©k /, not — /a.t lan.t ©k /. However, word-initial and word-
internal onsets do not always behave identically when the onset consists of
a single segment. For example, we do not ¬nd word-initial /s/ before vowels
in native German words. Even initial /s/ in words of foreign origin is typi-
cally integrated into the phonology of German and pronounced as [z]: Safe
˜safe™ [ze¦f], Sex ˜sex™ [zµks] (Mangold 2005:691, 722). However, /s/ is a pos-
sible onset word-internally: reißen ˜to tear™ /ra©.s™n/. Similarly, /N/ does not
Phonetics and phonology 35

occur in an onset word-initially, but does word-internally: bringen ˜to bring™
Except for /s/ and /N/, which may occur in a one-member onset only word-
internally, all the consonant phonemes of German may occur in one-member
onsets word-initially and word-internally. This holds, for example, for /x/:
(93) Chemie ˜chemistry™ /xe¦.mi¦/ Kuchen ˜cake™ /k u¦.x™n/
The phonemes /h/ and /j/ also occur in both positions, but they never occur in
onsets that consist of more than one segment.
(94) Hand ˜hand™ /hant/ Ahorn ˜acorn™ /a¦.h”rn/
Jahr ˜year™ /ja¦r/ Boje ˜buoy™ /po¦.j™/
Only certain combinations of phonemes are possible in two-member onsets.
Obstruents (stops, fricatives, and affricates) are permissible as the initial seg-
ment in such onsets; sonorant consonants (liquids and nasals) are typically the
second segment.
stop + liquid
(95) Brauch ˜custom™ /pra…8x/
stop + nasal Knecht ˜servant™ /k nµxt /
fricative + liquid ¬‚ach ˜¬‚at™ /flax/
fricative + nasal Schnee ˜snow™ /ʃne¦/
affricate + liquid P¬‚ug ˜plow™ /pflu¦k/
In addition, a few obstruents (the fricative /v/ and the lenis stops) may occur as
the second segment in a limited number of two-member onsets.51
(96) /ʃv/ Schwein ˜pig™ /ʃva©n/
/tsv/ zwei ˜two™ /tsva©/
(97) /ʃp/ Spiel ˜game™ /ʃpi¦l/
/ʃt/ Stein ˜stone™ /ʃta©n/
/sk/ Skat ˜skat™ /ska¦t /
Not all two-member onsets that are possible word-initially are also possible
word-internally in words made up of a single morpheme. For example, /fr/ can
form an onset word-initially and word-internally in a morphologically simple
word, but this is not the case with /ʃl/ (Hall 1992:68):
(98) word-initial frei ˜free™ /fra©/
morpheme-internal Afrika ˜Africa™ /a¦.fri¦.k a¦/
(99) word-initial Schlaf ˜sleep™ /ʃla¦f/
word-internal be-schlaf-en ˜to sleep on™ /p™.ʃla¦.f™n/
morpheme-internal NA
36 German

These distribution facts would argue against positing morpheme-internal onset
clusters that do not also occur word-initially. In particular, /tl/ would be ques-
tionable as an onset in words like adlig ˜noble™ and Handlung ˜action™ because
it does not occur word-initially. These words would have to be syllabi¬ed with
only /l/ in an onset and /t/ in the preceding coda:

adl+ig /a¦t.l©k/ ’ [a¦t.l©c]
(100) ¸
Handl+ung /hant.l…N/ ’ [hant.l…N]

This syllabi¬cation poses no problem for our analysis of German stops, how-
ever. The /t/ in both words is phonemically [’voice], but undergoes voicing
assimilation (and is realized as a [+voice] sound, [t]) because it occurs between
two [+voice] sounds (see the discussion of voicing assimilation in section
There are only ¬ve different three-member onsets that are possible. The ¬rst
consonant in such a cluster is either /s/ or /ʃ/, the second is a stop, and the third
is a liquid.

(101) /skl/ Sklave ˜slave™ /skla¦.v™/
Skrupel ˜scruple™ /skru¦.p ™l/
Splitter ˜splinter™ /ʃpl©.t ™r/
/ʃpr/ Spruch ˜saying™ /ʃpr…x/
/ʃtr/ Strom ˜current™ /ʃtro¦m/

These clusters occur only word- or morpheme-initially. They do not occur as
onsets in morpheme-internal position (Hall 1992:69). Codas With the exception of /h/, /j/, and /t /, all consonant
phonemes may occur in one-member codas in German. All single-member
codas that are possible word-¬nally are also possible word-internally. How-
ever, some single-segment codas occur word-internally only in morpheme-¬nal
position, that is, not in morpheme-internal position (Hall 1992:111).

(102) word-¬nal mit ˜with™ /m©t /
morpheme-internal Atlas ˜atlas™ /at .las/
(103) word-¬nal Fisch ˜¬sh™ /f©ʃ/
misch-te ˜mixed™ /m©ʃ.t ™/
morpheme-internal NA

There are many combinations of consonants that can occur in two-member
codas. Examples of those beginning with a sonorant consonant are listed in
(104); those beginning with an obstruent are in (105):
Phonetics and phonology 37

liquid + liquid Kerl ˜guy™ /k µrl/
liquid + nasal Helm ˜helmet™ /hµlm/
liquid + obstruent halb ˜half™ /halp/
nasal + obstruent Hemd ˜shirt™ /hµmt/
stop + fricative
(105) Mops ˜pug (dog)™ /m”ps/
stop + stop nackt ˜naked™ /nak t /
fricative + fricative Kochs ˜cook™s™ /k ”xs/
fricative + stop Lust ˜pleasure™ /l…st /
affricate + fricative Kopfs ˜head™s™ /k ”pfs/
affricate + stop rutscht ˜slides™ /r…tʃt /

Codas made up of three segments that begin with a sonorant consonant
are followed by another sonorant consonant and an obstruent or by two

sonorant consonant + sonorant consonant + obstruent
/rls/ Kerls ˜guy™s™ /k µrls/
/lmt / ge¬lmt ˜¬lmed™ /k™f©lmt /
sonorant consonant + obstruent + obstruent
/lps/ R¨ lps ˜burp™ /rlps/
/nft / sanft ˜gentle™ /zanft /

Codas made up of three segments that begin with an obstruent are followed by
the obstruents /st / or /t s/:

obstruent + obstruent + obstruent
/fst / schaffst ˜create (2nd sg.)™ /ʃafst /
/ft s/ Stifts ˜pencil™s™ /ʃt©ft s/

Although there are three-member codas that can occur within a single mor-
pheme, many occur only across morpheme boundaries:

Kunst ˜art™ /k …nst /
(109) /nst /
/rxt / Furcht ˜fear™ /f…rxt /
(110) /nx+s/ M¨ nchs ˜monk™s™ /m“nxs/
/x+st / suchst ˜look for (2nd sg.)™ /zu¦xst /

Codas that contain four segments always begin with a sonorant consonant
and end with either /st / or /t s/:
38 German

sonorant consonant + sonorant consonant + /st /
/lmst / ¬lmst ˜¬lm (2nd sg.)™ /f©lmst /
sonorant consonant + obstruent + /st / or /t s/
/lfst / hilfst ˜help (2nd sg.)™ /h©lfst /
/rkt s/ Markts ˜market™s™ /markt s/
Four-segment codas occur only rarely within a single morpheme. Two examples
that Hall (1992:121) notes are /rpst / in Herbst ˜autumn™ /hµrpst / and /rnst /
in ernst ˜serious™ /µrnst /.53 The complexity (length) of German codas is thus
due in large part to the in¬‚ectional morphology of the language. The second
person singular ending -st on present tense verb forms in particular accounts
for the majority of four-member codas.

1.2.6 Stress Word stress A stressed syllable is one that is perceived as
more prominent than other syllables. Various factors that play a role in deter-
mining prominence include length (duration), loudness, and pitch (the audi-
tory property of a sound that allows listeners to place it on a scale ranging
from low to high). In German, the strongest correlate of word stress is length;
stressed syllables are longer than unstressed syllables.54 Although there may
be a number of different degrees of stress phonetically, at the phonological
level we will recognize two: primary stress (represented by the raised verti-
cal stroke, ", before the syllable receiving primary stress) and secondary stress
(represented by the low vertical stroke, Æ); all other syllables are unstressed (no
It is necessary in German to recognize a level of secondary stress because of
words like Demut ˜humility™, biegsam ˜¬‚exible™, and Bahnhof ˜train station™.
All of these words have primary stress on the ¬rst syllable, but the vowels in
the second syllable are all long, an indication that the second syllable in each
word is stressed to some degree, since Vowel Shortening does not apply (Hall
(113) Demut ˜humility™ ["te¦Æmu¦t ]
biegsam ˜¬‚exible™ ["pi¦kÆza¦m]
Bahnhof ˜train station™ ["pa¦nÆho¦f]
In what follows, secondary stress will not be indicated unless it is relevant for
the discussion.
A number of generalizations can be made regarding the placement of stress
in German words. These generalizations are sensitive to the morphological
structure of a word. The relevant categories are simplex words (those that
have no af¬xes and are not part of a compound), complex words (formed with
Phonetics and phonology 39

pre¬xes and/or suf¬xes), and compound words (formed from two or more
lexical items). Simplex words Of the various generalizations that can be
made about the assignment of stress in simplex words in German, three exhibit
few exceptions and are thus highly reliable: the “three-syllable window restric-
tion,” the “closed-penult restriction,” and the “¬nal-schwa restriction” (Jessen
The three-syllable window restriction requires that stress fall on one of the
last three syllables of a word (Jessen 1999:519). This accounts for the placement
of stress in the following words:55
(114) Kle"o.pa.tra ˜Cleopatra™
Har"mo.ni.ka ˜harmonica™
The closed-penult restriction says that the main stress cannot occur farther
to the left than on the penultimate (second-to-last) syllable if the penultimate
syllable is closed (ends in a consonant) (Jessen 1999:520):
(115) Hi"bis.kus ˜hibiscus™
Pla"cen.ta ˜placenta™
Rho.do"den.dron ˜rhododendron™
As Jessen (1999:520) notes, syllables that end in diphthongs behave like closed
syllables with respect to stress assignment:
(116) Ba.la"lai.ka ˜balalaika™
Her.me"neu.tik ˜hermeneutics™
The"sau.rus ˜thesaurus™
The third generalization (the ¬nal-schwa restriction) states that if the ¬nal
syllable contains schwa and has an onset, stress falls on the penultimate syllable
(Jessen 1999:521):
(117) Me"tho.de ˜method™
Ta"pe.te ˜wallpaper™
Ka"nis.ter ˜canister™
If the ¬nal syllable does not have an onset, the antepenultimate syllable is
stressed (Jessen 1999:521):
(118) A"ka.zi.e ˜acacia™
"Li.li.e ˜lily™
"Sta.tu.e ˜statue™
Notice that both the closed-penult restriction and the ¬nal-schwa restric-
tion result in penultimate stress assignment. Kohler (1995) in fact notes the
40 German

tendency for stress to fall on the penultimate syllable in simplex words.
Evidence that penultimate stress could be the default stress pattern in
German comes from stress placement in foreign words and foreign names
that have entered the language in written form (Kohler 1995:187). The word
Niagara, for example, is pronounced by German speakers with penultimate
stress (Nia"gara).56 Wiese (1996:280“281) reports on an experiment involving
words in Japanese that provides further evidence that penultimate stress is the
dominant/unmarked pattern in German. For discussion of additional generaliza-
tions that can be made regarding stress assignment in simplex words, see Jessen
1999. Complex words Suf¬xed words. In¬‚ectional suf¬xes (e.g.,
plural endings on nouns, adjective endings, etc.) are unstressed and typically
do not change the stress pattern of the words to which they are attached.

(119) "K¨ nig ˜king™
o "K¨ nige ˜kings™
"mager ˜meager™ "magere "Ernte ˜poor harvest™

Derivational suf¬xes fall into two basic classes with respect to stress: those that
do not bear stress and those that do.57
The large majority of derivational suf¬xes that do not bear stress (or bear
secondary stress) are Germanic in origin. These suf¬xes typically do not affect
the stress pattern of the words to which they are attached.

(120) "d¨ ster ˜gloomy™
u "D¨ ster-heit ˜gloominess™
"Bruder ˜brother™ "Bruder-schaft ˜brotherhood™
"Abend ˜evening™ "abend-lich ˜evening™
"Spiegel ˜mirror™ "Spiegel-chen ˜little mirror™
(121) be-"greif-en ˜to understand™ be-"greif-Æbar ˜conceivable™
"arbeit- ˜work™ "arbeit-Æsam ˜industrious™

A small number of suf¬xes that do not bear stress can bring about a change in
stress. Some of these suf¬xes are native (-isch, -er); others are not (-ik, -or).58
If -isch, for example, is added to a derived word (a word created using a
derivational af¬x) or compound, it typically does not have an effect on stress:

(122) "M¨ rd-er ˜murderer™
o "m¨ rd-er-isch ˜murderous™
"Haus-h¨ lt-er ˜houskeeper™
a "haus-h¨ lt-er-isch ˜economical™

However, when added to polysyllabic non-derived forms, it can bring about a
change in the stress pattern:
Phonetics and phonology 41

(123) "Chaos ˜chaos™ cha"ot-isch ˜chaotic™
"D¨ mon ˜demon™
a d¨ "mon-isch ˜demonic™
Eu"ropa ˜Europe™ euro"p¨ -isch ˜European™
"Japan ˜Japan™ ja"pan-isch ˜Japanese™
Notice that this change in stress results in and thus preserves a penultimate
stress pattern (Kohler 1995:186).
The derivational suf¬xes that do bear stress are all of Romance origin (Latin,
French). These suf¬xes are stressed on the last syllable that has a full (non-
schwa) vowel as its nucleus (Kohler 1995:187):
Spio"n-age ([a¦ ™]) ˜espionage™
(124) "Spion ˜spy™
"liefer-n ˜to supply™ Liefe"r-ant ˜supplier™
"Haus ˜house™ hau"s-ier-en ˜to peddle™
Hu"mor ˜humor™ Humo"r-ist ˜humorist™
so"lid ˜solid™ Solid-i"t¨ t ˜solidness™
ko"rrekt ˜correct™ Korrek"t-ur ˜correction™
If a derived word contains several of these suf¬xes, only the last suf¬x is
(125) Nat-i"on ˜nation™ nat-io"n-al ˜national™ Nat-ion-al-i"t¨ t ˜nationality™
Pre¬xed words. Stress assignment in pre¬xed words is sensitive to the lexical
category of a word. In particular, verbal pre¬xes behave differently from non-
verbal pre¬xes. Verbal pre¬xes are typically unstressed. The verbal pre¬xes
be-, ent-, er-, ge-, ver-, and zer- (which are never separated from the verbal
stem) are generally unstressed and do not have any effect on the stress of the
words to which they are pre¬xed.
(126) "steig-en ˜to climb™ be-"steig-en ˜to ascend™
"leb-en ˜to live™ er-"leb-en ˜to experience™
"end-en ˜to end™ ver-"end-en ˜to perish™
The forms durch-, hinter-, uber-, um-, unter-, and wider- can behave in the
same way, that is, as unstressed, inseparable pre¬xes, and are analyzed here as
(127) "blut-en ˜to bleed™ durch-"blut-en ˜to supply with blood™
"fahr-en ˜to drive™ uber-"fahr-en ˜to run over™
"nehm-en ˜to take™ unter-"nehm-en ˜to undertake™

These forms can also be stressed (e.g., "durch-blut-en ˜to bleed through™), in
which case they are treated not as pre¬xes but as part of a phrasal verb (like Rad
fahren ˜to pedal™) and are stressed in their citation (dictionary) form according
to the rules that apply to compounds (see section Notice that the
42 German

pre¬xed words in (126) and (127) preserve the penultimate stress pattern of
their non-pre¬xed counterparts (Kohler 1995:187).
The verbal pre¬x miss-, as well as the set of pre¬xes that includes uber- (when
meaning ˜too much™) and unter- (when meaning ˜too little™), are unstressed
unless they are followed by an unstressed syllable, in which case they are
stressed (Kiparsky 1966):60
(128) miss-"trau-en ˜to mistrust™ "miss-ver-steh-en ˜misunderstand™
uber-"sch¨ tz-en ˜to overestimate™ "¨ ber-be-licht-en ˜to overexpose™
a u
unter-"sch¨ tz-en ˜to underestimate™ "unter-be-licht-en ˜to underexpose™
Nominal and adjectival pre¬xes are typically stressed.61 In particular,
the nominal and adjectival pre¬xes erz-, miss-, un-, and ur- are generally
(129) "Erz-bischof ˜archbishop™ "erz-konservativ ˜ultraconservative™
"Miss-brauch ˜abuse™ "miss-mut-ig ˜sullen™
"Un-mensch ˜monster™ "un-reif ˜immature™
"Ur-text ˜original (text)™ "ur-alt ˜ancient™
Interestingly, the stress pattern of these words is identical to the stress pattern
of compounds. As Wiese (1996:294) points out, the stress pattern of a pre¬xed
word like "Miss-gunst ˜resentment™ is identical to that of the compound "Miss-
wahl ˜beauty (Miss) contest™.63 Compound words A compound is a word made up by
adjoining two or more words. The compound Haustier ˜pet™, for example,
is formed from the words Haus ˜house™ and Tier ˜animal™. Although com-
pounds can be made up of four words or more (Kern+kraft+werk+unfall
˜atomic power plant accident™, Nudel+saucen+rezept+sammler+treffen ˜noo-
dle sauce recipe collector meeting™), only two- and three-member compounds
will be discussed here.
In a copulative compound, where all members of the compound are of equal
status (and could be conjoined with und ˜and™), each member is stressed.64
(130) "Schleswig-"Holstein ˜Schleswig-Holstein™
"geistig-"seelisch ˜spiritual-psychological™
"schwarz-"rot-"gold ˜black-red-and-gold™
In a subordinate compound, where one member of the compound further spec-
i¬es or de¬nes another (in the subordinate compound Blutdruck ˜blood pres-
sure™, for example, the ¬rst member, Blut ˜blood™, speci¬es the kind of Druck
˜pressure™), the unmarked pattern is stress on the initial member.65 This holds
for three-member as well as two-member compounds. In three-member com-
pounds, the A constituent is stressed, regardless of the internal structure of the
Phonetics and phonology 43

compound, A(BC), as in Volkshochschule, or (AB)C, as in Autobahnkreuz and
(131) "Haust¨ r ˜front door™
"blaurot ˜purple™
"Uberstunde ˜hour of overtime™
"Volkshochschule ˜adult education center™
"Autobahnkreuz ˜expressway interchange™
"Ubersichtskarte ˜general map™
In some compounds, however, the second (B) member is stressed, not the ¬rst.
One set of compounds with this stress pattern, (AB)C compounds, is illustrated
in (132).
(132) Rot"kreuzschwester ˜Red Cross nurse™
Drei"zimmerwohnung ˜three-room apartment™
Zehn"¬ngersystem ˜touch-typing method™

This type of compound differs in a number of ways from compounds with
initial stress (Benware 1986:108): the A constituent is an adjective or a quanti-
¬er (number); the (AB) constituent shows the stress pattern of a corresponding
noun phrase (rotes "Kreuz ˜red cross™; drei "Zimmer ˜three rooms™); the (AB) con-
stituent does not occur independently as a compound (— Rotkreuz, — Dreizimmer).
Another set of compounds with stress on the B constituent, a set of A(BC)
compounds, is illustrated in (133).
(133) Beethoven-Kla"vierkonzert ˜Beethoven piano concerto™
Donau-"Dampfschiff ˜Danube steamship™
Weltge"sundheitsorganisation ˜World Health Organization™
Benware (1987) argues that the A constituents in these compounds are seman-
tically marked (less common), and thus exhibit a marked stress pattern by
being unstressed. The A constituent in a compound with the structure A(BC)
is usually inde¬nite. In "Autowerkstatt ˜car repair shop™, for example, the A
constituent refers to cars in general, not to a speci¬c car. The same is true for
compounds like "Opernfestspiele ˜opera festival™ and "Kunsthandwerk ˜artistic
handicrafts™. In compounds like those in (133), however, the ¬rst constituent is
de¬nite. There is only one Beethoven, one Donau, etc. The semantic marked-
ness of these compounds is signaled by a marked stress pattern. See Benware
1987 for a discussion of other sets of A(BC) compounds that are stressed on
the B constituent.66 Sentence stress When words occur in stretches of speech
rather than in isolation, not every word will contain a syllable with primary
44 German

stress. Although the ¬rst syllable in meine ˜my™, for example, can in prin-
ciple receive primary stress, it is unstressed in the following utterance (Fox
(134) Meine "Schwester "schreibt mir "lange "Briefe.
˜My sister writes me long letters.™
In German, as in English, stressed syllables play an important role in the rhythm
of speech. A stressed syllable, together with the unstressed syllables that follow
it, forms a foot, and each foot takes up roughly the same amount of time.67
German, like English, is thus a stress-timed language (stressed syllables recur
at regular intervals of time), in contrast to languages like French, which are
syllable-timed (each syllable takes up approximately the same amount of time).
In every utterance, which may contain a number of words with primary stress,
at least one syllable stands out as being particularly prominent, the nucleus.
In the sentence in (134), the syllables Schwes-, schreibt, lang-, and Brief- all
bear primary stress, yet Brief- is particularly prominent and therefore forms the
nucleus of the sentence. The nucleus plays an important role in intonation.

1.2.7 Intonation
The pitch of a speech sound depends on the rate of vibration of the vocal cords.
Sounds with a high pitch are produced with a higher rate of vibration than
sounds with a low pitch. The changes in pitch over the course of an utterance
are known as intonation.
The nucleus in an utterance is associated with a noticeable change in pitch.
In the sentence in (135), for example, there is a fall in pitch (indicated by the
symbol `) between the ¬rst and second syllables of Nachmittag.
Ich "komme "heute `Nachmittag. (C. Hall 2003:120)
˜I™m coming this afternoon.™
In the example in (136), there is a rise in pitch on the word m¨ de ˜tired™
(indicated with the symbol ´).
"Sind Sie ´m¨ de?
˜Are you tired?™
Although the intonation patterns in (135) and (136) are associated with
sentences, intonation patterns can be associated with other grammatical units.
In (137), for example, there are two intonation patterns: falling, associated with
a sentence, and rising, associated with a tag question.68
Er `kommt ´nicht? (Fox 1984:44)
˜He™s coming right?™
Phonetics and phonology 45

Because intonation patterns can be associated with various kinds of grammatical
structures, the term “intonational phrase” is used to refer to the stretch of
speech associated with an intonation pattern.69 Each intonational phrase has
one nucleus. There is one intonational phrase and therefore one nucleus in the
utterances in (135) and (136): Nach- in (135) and m¨ - in (136). The utterance
in (137), on the other hand, has two intonational phrases and therefore two
nuclei: kommt and nicht.70
The nucleus occurs normally on the last lexical word in an intonational
phrase, unless that word is a verb (C. Hall 2003:133).71

"Gerd "schreibt ein "Buch uber `Mozart.
˜Gerd is writing a book about Mozart.™
"Gerd hat ein "Buch uber `Mozart geÆschrieben.
˜Gerd has written a book about Mozart.™

If the nucleus occurs on any other word in an intonational phrase, stress is
being used contrastively. That word is contrasted (implicitly or explicitly) with

(139) Gerd hat ein Buch uber Mozart ge`schrieben (nicht gelesen).
˜Gerd has written (not read) a book about Mozart.™

There are a number of different intonation patterns in German, three of
which will be discussed here: the falling pattern, the rising pattern, and the
level pattern.72 The falling pattern is commonly used for statements (140a),
commands (140b), and questions that begin with a question word (140c) (C.
Hall 2003:124“125).

(140) The falling pattern
a. Du "hast jetzt ge`nug geÆarbeitet.
˜You have worked enough now.™
b. "Gehen Sie "bitte "nicht `weg!
˜Please don™t go away.™
c. Wann "kommst du nach `Hause?
˜When are you coming home?™

The use of the falling pattern is very similar in German and English. In German,
however, the fall is much steeper than in English.
The rising pattern is used in yes/no questions (141a), among other sentence
types. In the example in (141b), the rising intonation turns the sentence into a
question (C. Hall 2003:126).
46 German

(141) a. "Kommst du ´mit?
˜Are you coming along?™
b. "Dieter war in I´talien?
˜Dieter was in Italy?™

The level pattern (indicated by the symbol ¯ ) is characterized by the absence
of movement in pitch directly on or following the nucleus. The level pattern
is used, among other things, for non-¬nal intonational phrases (142a), in lists
for all but the ¬nal element (142b), and for expressing a noncommittal attitude
(142c) (C. Hall 2003:128):
(142) a. Wenn wir nach ¯Hause "kommen, `essen wir was.
˜When we come home, we™ll eat something.™
b. Wir "kaufen ¯Butter, ¯Milch, ¯Eier und `Brot.
˜We™re buying butter, milk, eggs, and bread.™
c. ¯Danke.

Fox 1984 discusses the general meaning conveyed by the two most
important intonation patterns in German, the rising pattern and the falling
pattern. According to Fox (1984:59), the rising pattern makes an “appeal to the
listener, who is being invited, or challenged, to respond.” The rising pattern is
clearly appropriate for questions, as in (141a), but also for sentences that are
statements on the surface but intended as questions and thus invite a response
from the listener, as in (141b). The falling pattern, on the other hand, functions
as an assertion. It makes the utterance self-suf¬cient, since it does not depend
on the listener™s reply. The falling pattern is thus appropriate for statements, as
in (140a).73

Solutions to the exercises can be found in the online answer key at
1. Provide the phonetic symbols for the vowels in German that are phonetically
(a) [+high, +round]; (b) [+front, ’round]; (c) [’high, +back]; (d) [+low];
(e) [+high, ’tense]; (f) [+long, ’tense]; (g) [’high, +front, +round];
(h) [+high, +front, ’round]
2. Describe the following consonants phonetically (for example, [p ] is a voice-
less aspirated bilabial stop; [] is a uvular trill): (a) [p#]; (b) [N]; (c) [t ];
(d) [c]; (e) [r]; (f) [ʃ]; (g) [l]; (h) [k]; (i) [s]; (j) [”]; (k) [m]; (l) [ʁ]; (m) [v];
(n) [t]; (o) [ ]
Phonetics and phonology 47

3. Transcribe the following words phonetically, applying the rules discussed in
section 1.2.4 (transcribe a glottal stop in the positions where glottalization
can occur).
(a) kommen ˜to come™ / "k ”m™n/; (b) Gabe ˜gift™ / "ka¦p™/; (c) lebt ˜lives™
/ "le¦pt /; (d) Teig ˜dough™ / "t a©k/; (e) phonetisch ˜phonetic™ /fo¦"ne¦t ©ʃ/;
(f) Tiere ˜animals™ / "t i¦r™/; (g) Bier ˜beer™ / "pi¦r/; (h) niest ˜sneezes™ / "ni¦zt /;
(i) welcher ˜which™ / "vµlx™r/; (j) Buchhandlung ˜bookstore™ / "pu¦xÆhantl…N/;
(k) lockige ˜curly™ / "l”k ©k™/; (l) sonnig ˜sunny™ / "z”n©k/; (m) Theater
˜theater™ /t e¦"a¦t ™r/; (n) veralten ˜to become obsolete™ /fµr"alt ™n/; (o) ¬‚a-
chem ˜¬‚at™ / "flax™m/; (p) grobem ˜coarse™ / "kro¦p™m/; (q) knappen ˜scarce™
/ "k nap ™n/; (r) Himmel ˜heaven™ / "h©m™l/; (s) Wasser ˜water™ / "vas™r/
4. Syllabify the following words (indicate syllable boundaries) using the prin-
ciples and constraints discussed in section 1.2.5.
(a) halben ˜half™ /halp™n/; (b) umschmeicheln ˜to ¬‚atter™ /…mʃma©x™ln/;
(c) bierernst ˜deadly serious™ /pi¦rµrnst /; (d) beknien ˜to beg™ /p™k ni¦n/;
(e) besuchte ˜visited™ /p™zu¦xt ™/; (f) kopf¨ ber ˜head¬rst™ /k ”pfy¦p™r/;
(g) Heizung ˜heating™ /ha©ts…N/; (h) K¨ nigreich ˜kingdom™ /k ø¦n©kra©x/;
(i) milchig ˜milky™ /m©lx©k/; (j) Milcheiweiß ˜lactoprotein™ /m©lxa©va©s/
5. Indicate which syllable in each of the following words bears primary stress.
(a) Arbeit ˜work™; (b) Zitrone ˜lemon™; (c) Phantasma ˜phantasm™; (d) Wette
˜bet™; (e) Freundschaft ˜friendship™; (f) Linguist ˜linguist™; (g) transportieren
˜to transport™; (h) verbringen ˜to spend (time)™; (i) beschleunigen ˜to accel-
erate™; (j) Spiegelbild ˜re¬‚ection™; (k) Bet¨ ubungsmittel ˜anesthetic™;
(l) Lebensmittelgesch¨ ft ˜grocery [store]™; (m) Dreitage¬eber ˜sand¬‚y
fever™; (n) Taschenw¨ rterbuch ˜pocket dictionary™; (o) Beruhigungsmittel
6. Locate the nucleus (nuclei) in each of the following sentences (assume that
the sentences are uttered without contrastive stress).
(a) Der Wirt kannte den Mann. ˜The proprietor knew the man.™
(b) Er war gegen 3 Uhr in sein Lokal gekommen. ˜He had come into his bar
around 3 o™clock.™
(c) Inzwischen hat ein Projektentwickler das Gel¨ nde gekauft und will dort
ein Hotel bauen. ˜In the meantime a project developer has bought the
site and intends to build a hotel there.™
(d) Die Skip¨ sse werden jetzt elektronisch gelesen. ˜Ski passes are now read
(e) Ich habe Wasserf¨ lle, Elche und eine Stabkirche gesehen. ˜I saw water-
falls, elk, and a stave church.™

1 See Wiese 1996:21“22 for a discussion of the literature on the two a sounds in
German and additional arguments for treating the two sounds as differing only in
48 German

2 German diphthongs are transcribed in a variety of ways in the literature. For example,
the diphthong in H¨ user ˜houses™ is transcribed as [”ø] (Siebs 1969, Krech et al.
1982), [”©] (Kohler 1995), [”] (Wiese 1996), etc. The transcription of German
diphthongs employed here follows Kohler 1995, since the use of [©] and […] as the
second vowel in the diphthong is phonetically most accurate (Mangold 2005:36;
Jessen 2007).
3 See section for further discussion of the vocalic realization of r.
4 Nasal consonants are sounds like [m] in Mutter/mother, [n] in Nacht/night, and [N]
in singen/sing. See section 1.1.3 for further discussion.
5 I do not use the IPA symbols that are typically used to represent voiced stops ([b],
[d], [ ]). Instead, I use the diacritic for voicing (a wedge, ˇ) with the symbols for
voiceless stops. I explain my reasons for this in section
6 Because the turbulence during the articulation of [h] is produced primarily by the air
passing through the entire oral cavity rather than through the glottis, [h] is probably
more accurately characterized as a glottal approximant (see Ladefoged 1971:122,
for example). However, following the general practice in the literature (e.g., Hall
1992, Wiese 1996, C. Hall 2003, Mangold 2005), [h] will be classi¬ed here as a
7 Dark l can be found in some German dialects, for example, in the Rhineland
(Cologne) and in parts of Switzerland (C. Hall 2003:59).
8 There is a good deal of variation in the distribution of [l] and [ ] in the various
dialects of English. Some speakers of American English, for example, have dark l
in all positions. At the other end of the continuum are dialects like southern Irish
English, where clear l may occur in all positions (see Cruttenden 2001:204).
9 The morphologically conditioned rules of e-Epenthesis and umlaut are treated
only very brie¬‚y in chapter 2. See Wiese 1996 (and references therein) for fur-
ther discussion of morphologically conditioned rules of German. For literature on
sentence-level phonetics and phonology, see, for example, Kohler 1979, 1990, and
10 This is a simpli¬ed description of the distribution of [ʁ] and [ɐ8]. See below (this
section and section for a more detailed description.
11 We follow Vater 1992 and assign the less prominent vowel in German diphthongs
(e.g., […8] in [a…8]) to the nucleus. Hall (1992) takes a different approach, and assigns
this vowel to the coda. See Vater 1992:109 and Hall 1992:142“153 for further
12 Only /a©/, /a…8/, and /”©/ have the status of phonemes. Diphthongs like [i¦ɐ8] arise
from the sequence of a vowel followed by /r/ (see section for details). The
diphthong [io¦] in a word like Emotion ˜emotion™ arises from the sequence /io¦/ and
a low-level phonetic process that results in a non-syllabic pronunciation of /i/ before
vowels (Mangold 2005:42).
13 An affricate is a sequence of a stop followed by a homorganic fricative (a fricative
with roughly the same place of articulation).
14 “Liquid” is a cover term for laterals and various types of r sounds.
15 Various pronunciations are used by speakers of German for consonantal r, for
example, [], [r], and [ʁ]. The choice of these different r-sounds is dependent on
factors such as dialect, style, content, etc. (see, e.g., Kohler 1995:165, Mangold
2005:53“54, and C. Hall 2003:65“66 for further details). I use [ʁ] as representative
of consonantal r, since this pronunciation predominates (Kohler 1995:165).
Phonetics and phonology 49

16 The lenis stop phonemes in German are traditionally represented as /b d / and the
fortis as /p t k/. However, since b, d, and g represent voiced stops in the IPA, and
p, t, and k represent voiceless stops, this suggests that the distinction between lenis
and fortis stops in German is one of voicing. Because both sets are voiceless, I use
the symbols p, t, and k for both sets (lenis as well as fortis). I add the symbol for
aspiration, [ ], to the symbols for the fortis stops to represent the [spread glottis]
feature of these stops.
17 The space between words in phonemic and phonetic transcriptions is employed
simply for ease of reading.
18 See Jessen and Ringen 2002:205“206 for a discussion of the phonetic/physical
factors that account for the gradient nature of voicing assimilation.
19 Wiese (1996:67) argues that all consonant-initial suf¬xes are phonological words.
20 [continuant] is a feature that characterizes sounds made with free or nearly free
air¬‚ow through the center of the oral cavity. Sounds that are [+continuant]
are vowels and fricatives; stops, nasals, and laterals are [’continuant]. The fea-
ture [sonorant] characterizes sounds produced when air ¬‚ows smoothly through the
vocal tract. Vowels, nasals, and liquids are [+sonorant].
21 I follow Hall 1992:55 and use the notion of intonational phrase to indicate that fortis
stops are aspirated syllable-¬nally before a pause.
22 Some authors argue that liquids following fortis stops are voiceless “ that the
stops in this environment are not aspirated (the diacritic indicates that a segment
is voiceless): Kreis ˜circle™ ["kʁ9a©s], Klasse ˜class™ ["klas™] (Kohler 1995:158, C.
Hall 2003:31). However, it is dif¬cult to distinguish experimentally between stop
aspiration and (partial) sonorant devoicing when stops occur before sonorant con-
sonants (Michael Jessen, personal communication, September 19, 2005). I follow
Hall 1992:54, for example, and treat fortis stops before sonorant consonants as
23 The application of Aspiration must follow the deletion of schwa, since the opposite
order of application would yield the incorrect form — [la©t n] (the asterisk indicates
that this pronunciation is unacceptable).
24 We assume that the <d> in words like handlich and Neid is underlyingly (phonem-
ically) lenis, /t/, since it behaves like a lenis stop in other related forms (it is voiced
when it occurs between voiced segments): Handes ˜hand (genitive)™ [hant™s]; Neides
˜envy (genitive)™ [na©t™s].
25 The class of obstruents includes stops, fricatives, and affricates.
26 Fully native minimal pairs for [s] versus [z] occur only intervocalically. The word
Seal, for example, is non-native.
27 For more details on German fricatives, see Piroth and Janker 2004 and Beckman
et al. to appear.
28 Junta can also be pronounced with an initial [j] (Mangold 2005:446).
29 There are some exceptions to this generalization. However, there is typically an
alternative pronunciation for those words that exhibit [x] before front vowels and
[c] before non-front vowels: Cholesterin ˜cholesterol™ [k] or [c] (listed in this order
¸ ¸
in Mangold 2005:241); Xeres ˜sherry™ [x] or [c] (Wahrig 2000:1412; Muthmann
1996:457). This suggests that pronunciations with word-initial [x] before front
vowels and [c] before non-front vowels are non-native. As Robinson (2001:19) points
out, “there is no native historical source for word-initial [x] or [c]. All examples in
Modern German are borrowings.”
50 German

30 My analysis has bene¬ted from the observation that [c] and [x] are largely in
complementary distribution in word-initial position (Michael Jessen, personal com-
munication, September 4, 2007; Jessen 1988).
31 [back] is a feature that involves the placement of the body of the tongue and
characterizes consonants as well as vowels; [+back] sounds are articulated behind
the palatal region in the oral cavity. I assume that [c] has the features [’back, +high,
’coronal], which distinguishes it from [ʃ], which is [’back, +high, +coronal]
([+coronal] sounds are produced with the blade of the tongue raised from its neutral
position). Sounds that are [+consonantal] are produced with major obstruction in
the vocal tract. This feature is necessary in the formulation of this rule to prevent it
from applying to /h/, which is [’consonantal].
32 This rule would also realize /x/ as [c] when it follows /m/, which is [’back], but to
my knowledge there are no sequences of /mx/ in German.
33 Notice that the rule of Velar Spirantization yields a fricative that is [’voice] and
[’spread glottis] (when the rule changes the manner of articulation from stop to
fricative, it does not change the [’voice] or [’spread glottis] feature of the stop).
I assume that a low-level phonetic constraint realizes this and all other [’voice]
fricatives as [+spread glottis]. (See Vaux 1998 for evidence that voiceless fricatives
are [+spread glottis] in their unmarked state.)
34 According to Durrell (1992:13), this regional variation in pronunciation of
-(i)g is followed “almost universally,” whether used in colloquial or more formal
35 The information in (67) is a modi¬ed version of information presented in C. Hall
2003:47. According to Hall (who makes a simple north/south distinction), “millions
of German-speakers whose pronunciation is otherwise perfectly standard consis-
tently use either the northern or the southern pronunciation” (2003:47“48). The
standard language (re¬‚ected in the rule of Velar Spirantization) is actually the out-
come of (arguably arbitrary) decisions made at a conference on theater pronunciation
held in Berlin in 1898 (Robinson 2001:3). There is clearly something less than nat-
ural about Velar Spirantization, evident in the formulation of the rule itself “ the
complicated statement of the environment in which it applies and the condition on
its application. Note that Velar Spirantization must apply before Fortition, since the
opposite order would yield unacceptable forms: saftig ˜juicy™ /zaft ©k/ would be
realized incorrectly as — [zaft ©k ] instead of [zaft ©c].
36 Creaky voice is produced with the arytenoids tightly together, but with a small length
of the vocal cords vibrating (Ladefoged 1971:8).
37 We have modi¬ed Hall™s transcription of stops to correspond to the convention we
follow here.
38 Some authors claim that words such as Theater and Kloake contain glottal stops
(e.g., Hall 1992:58, Wiese 1996:59). However, Kohler (1994) ¬nds no evidence of
glottal stops or glottalization in stem-internal vowel sequences (he mentions the
word Kloake explicitly). Both Theater and Kloake are transcribed in Mangold 2005
without a glottal stop (words like beanspruchen ˜to claim™ are transcribed with a
glottal stop: be[”]anspruchen).
39 Hall (1992:22) argues that almost all instances of schwa are predictable and therefore
epenthetic. He does consider schwa to be a phoneme of German, however, because
its presence word-internally in many morphemes and certain pre¬xes cannot be
accounted for by epenthesis.
Phonetics and phonology 51

40 The distribution of schwa is a complex matter. In the following pairs, which also
exhibit a schwa“zero alternation, the “deleted” schwa is not represented in the
orthography, unlike the deleted schwa in (71): Atem ˜breath™, at˜men ˜to breathe™;
trocken ˜dry™, trock˜nen ˜to dry™. This schwa“zero alternation will not be treated
41 The condition that a consonant must precede schwa when schwa precedes /l/ is a
bit of an oversimpli¬cation. The consonant may not be /r/, since schwa may not be
deleted in a word such as Minstrel ˜minstrel™ [m©nstʁ™l]. However, since there are
only two additional words that I am aware of that end in the sequence -rel (Barrel
˜barrel™, Varel ˜Varel [city name]™), and all three are either borrowings or place
names, I have chosen not to complicate the conditions on Schwa Deletion in order
to accommodate these forms.
42 Recall that [ʁ] and [ɐ8] are in free variation following a short vowel: dort ˜there™
/t”rt / ’ [t”ʁt ] or [t”ɐ8t ]. Thus the rule of r-Vocalization is optional when /r/
follows a short vowel. For the sake of simplicity, however, I do not include this
condition in the statement of the rule. I simply present the consonantal and vocalic
allophones of /r/ as being in complementary distribution in all environments.
43 Note that the [’consonantal, +syllabic] realization of /r/ ([ɐ]) shares the features
[’back, ’high, ’low] with /r/; it is not speci¬ed for the feature [front]. It will thus
differ from the realization of /™/ and /a/, both of which are speci¬ed for the feature
[front] (/™/ is [’front, ’back]; /a/ is [’front]).
44 Alpha notation allows us to collapse several rules into one.
45 I adopt Giegerich™s (1989:19) use of the feature [± artic] to express homorganicity
(having the same point of articulation). This convention for expressing homorganic-
ity goes back to Lass and Anderson 1975:262“263.
46 A number of studies have dealt with the phonotactics of German, for example,
Moulton 1956, Seiler 1962, Tanaka 1964, Benware 1986, Hall 1992, Kohler 1995,
and Wiese 1996. Benware 1986, Hall 1992, and Wiese 1996, in particular, emphasize
the importance of the syllable in the treatment of phonotactics. See these studies for
detailed tables and examples of possible onsets and codas in German.
47 Of the three words with ¬ve-member codas cited by Hall (1992:121), only Herbsts
˜autumn (genitive)™ is listed as a possible form in the Duden spelling dictionary
(Dudenredaktion 2006:495). Wiese (1996:48), however, argues that such words
“are probably not well formed and are clearly avoided.” Hall does say that such
¬ve-member clusters are unpronounceable for many Germans and are made pro-
nounceable by epenthesis of schwa, as in Herbstes (an alternative form also listed
in the Duden dictionary).
48 An asterisk before a word (or a sentence) indicates that the form is unacceptable.
49 See Ramers 1992 for an overview of the various phonological arguments for ambi-
syllabic consonants in German.
50 The evidence for ambisyllabic consonants in German is divided. A study by Fischer-
Jorgensen (1969) shows that an intervocalic consonant after a short vowel has the
tendency to be longer than one after a long vowel. A recent study by Spieker-
man (2000:66), however, shows the opposite pattern, namely, a slight tendency for
consonants after long vowels to be longer than those after short ones.
51 There are some additional two-member onsets with various combinations of stops,
fricatives, and affricates, but these occur only rarely: for example, Xylophon ˜xylo-
phone™ /ksy¦lo¦fo¦n/, Szene ˜scene™ /stse¦n™/.
52 German

52 This syllabi¬cation does pose a problem for analyses of German that assume that
stops and fricatives are devoiced in coda position. Hall (1992:86) makes this assump-
tion and thus argues for a rule of Resyllabi¬cation that yields the syllabi¬cations
A.dlig and Han.dlung, which are impervious to devoicing and thus lead to the
desired pronunciation. Hall 2005, recognizing that such syllabi¬cations are prob-
lematic because of the lack of word-initial dl- clusters, assumes the syllabi¬cations
in (100) and appeals to the notion of Paradigm Uniformity to account for the absence
of devoicing in these forms.
53 Hall (1992:121) also includes the cluster /rtst / in Arzt ˜doctor™. However, I analyze
this as a three-consonant cluster with the affricate /ts/ as its second member.
54 Loudness (measured as “spectral balance” rather than simply as overall amplitude)
has also been shown to be a correlate of stress in German (Claßen et al. 1998).
55 Note that neither the closed“penult restriction nor the ¬nal-schwa restriction account
for the placement of stress in these words.
56 Notice that stress placement in Nia"gara is not accounted for by the closed-penult


. 2
( 12)