Key to questions 289

4. This solution accounts for the fact that all vowels nasalize before [!], because
it is ambisyllabic regardless of whether it is foot-initial or foot-internal.

Chapter 12
Q114
1. It is the [coronal] node that spreads to the place node of a preceding
[t,d,n,l] in English coronal assimilation.
2. It is the place node that spreads from the consonant to the supralar node of a
preceding nasal in Hindi.
Q115 The picture you end up with should look more or less as shown.

root root

LAR
SUPRALAR

PLACE PLACE

Q116 You cannot, because the place tier and the root tier are not adjacent, and
hence do not define a plane.
Q117 Natural feature classes are not phonetically abstract to the extent that the nodes
in a feature tree can be defined in phonetic terms. For instance, the laryngeal node
could phonetically be defined as ‘activity in the larynx’, the surpralaryngeal node as
‘activity in the vocal tract’, etc.
Q118 English place assimilation can be stated in the display format as follows:

SUPRALAR

PLACE

SD: CORONAL [⫺cont]

⫺cont
290 Key to questions

Q119
1. In order to account for the Yorkshire English and Durham English
assimilation facts, we could assume the SPE rules in (1) and (2).
(1) Yorkshire voicelessness assimilation [son] → [voice]/__ # # [voice]
(2) Durham voicing assimilation [son] → [voice]/__ # # [voice]
2. Given that the feature [voice] needs to be referred to in the structural
change and the structural description of Yorkshire voicelessness assimilation,
it cannot be left unspecified and be filled in only at the end of the
derivation.
Q120 Two ordered rules are necessary to derive the correct Klamath output
forms. The first rule should spread the supralaryngeal node of a lateral consonant to
a preceding [n] or [l]. The second rule should delink the supralaryngeal node of the
lateral consonant, which, if it is specified as [voice] or as [constricted], will leave
behind only its laryngeal specification, causing it to end up as [h] or [ʔ]. The rules
can be formalized as (1) and (2) below.

(1) LAR

SUPRALAR

SD: CORONAL [⫹lat]

⫺cont

(2) LAR

SUPRALAR
SD: [⫹lat]

We assume that in the case of [nl] rule (1) creates a geminate structure, as shown.
C C

ROOT

LAR

SUPRALAR

Rule (2) cannot apply to this structure, which will be supplied with the feature
[voice] by default.
 Key to questions 291

Q121
1. The full representation for the four types of initial plosives is the following:
voiced voiceless voiceless preglottalized
aspirated voiced
⎡⫹voice ⎤ ⎡⫺voice ⎤ ⎡⫺voice ⎤ ⎡⫹vvoice ⎤
⎢⫺spread ⎥ ⎢⫺spread ⎥ ⎢⫹spread ⎥ ⎢⫺spread ⎥
⎢⫺constr ⎥ ⎢⫺constr ⎥ ⎢⫺constr ⎥ ⎢⫹constr ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
2. Assuming radical underspecification, the voiced plosives could be
characterized by the presence of the feature [voice] in the underlying
representation, the voiceless aspirated plosives by the feature [spread]
and the preglottalized voiced plosives by the feature combination [voice,
constr]. The features [voice], [spread] and [constr] could then be
left unspecified, to be filled in by default rules. This means that the voiceless
unaspirated plosives are left unspecified underlyingly.
3. Coda plosives can be left unspecified, as their feature values can be provided
by the default rules.

Chapter 13
Q122
1. nasalization in Sundanese should transfer the nasality of a nasal consonant to
any following vowel, but is blocked by any consonant, except when this is a
laryngeal segment.
2. If the rule is formulated such that it spreads [nas] from a nasal consonant
to the supralaryngeal node of a following vowel (V-position), it will be blocked
by any segment that has a supralaryngeal specification. Since [h] and [ʔ] do
not have a supralaryngeal node, these segments are invisible for the spreading
[nas], and their transparancy is straighforwardly accounted for.
Q123
1. One default rule needs to assign [voice] to a stem-initial obstruent and one
needs to assign [voice] to an intervocalic obstruent.
2. The two [voice]-melodies are [voice] and [voice] [voice]. Similar to
the avoidance of the first TBU at the start of the tonal association in Kikuyu
in Chapter 10, the melodies cannot associate to a stem-initial obstruent. The
features must not spread, because a [voice] melody would produce
ungrammatical forms like *[itupi] and *[titupi] if it did.
3. Whereas [υdɔdɔ] does not have an underlying [voice] melody, [ədəpiɔ] is
specified for a [voice][voice] melody.
4. Underlying form /iTuPi/ /TəTəKaKe/

Melody association [–voice] [+voice][–voice]

/ituPi/ /TədəkaKe/
Default rules [itubi] [tədəkae]