4
Underlying and surface representations
44 INTRODUCTION.
In this chapter we discuss the need fora
morpheme
pronunciation of segments
‘allophonic’ information is assumed
tions of morpheme:
automatically sup
variation from situation-depende
the need for an underlying representation of a morphere can
he basis of morphemes that havea numberof contextdependent
ural suffix, with is alternants
the underlying form expresses the
‘Third, the existence ofan underlying represent
on the bass of the impossibility of expres
reference tothe surface representation. W
4.2_ALLOPHONIC VARIATION
‘The'Segments referred to so far inthis book typically splay a good deal of variation,
speech ofa single speaker. Some of this variation is apparently random, but
traspeaker variation. Oneis
of the speech situation. For
frequently have an unreleased final plosive (symbol
ol forthe plosiveand a superscript a in [rat
| may well be released (so that you can hear a we
istic studies the stylistic v
values of a phonological variable (t), whose frequencies of
x across diferent speech styles for diferent groups of speakers
mal speech
und after
ingly seen as the
second type of variation is due to variation in the phonological context
Te tdependent variants arecaledallophones. The term phonemes sed to referto
ent category that the various allophones are variants of. To distinguish the
nguebody towards the uvulasoun
‘many languages, coronal consonant altermate
Iatoatelar one appears before
1s, in Korean, underlying [sipsam] is pro-
isewhere,as shown by
ra be washed! (Jones
suggested by thee
guages wl ile in
toxphems, That so ien segment
a i the difference between [s] and [{] is allophonic in.
(022. Far each of the tree folowing languages, say whether te afforence between [7]
and (1s) aloponi, (phoneme o (syste Matte your answers.
4. nsome ipes of Souther Swedish] alvays appears atthe begining of a yabie and
Matways at the end
2. Insome types of Dut, is the usual realzstion in words ike (proyrama]
‘programme’ Some speakers of hose varieties use this pronunciation in everyday
He, but use [x sin pao ynomas, when announcing programmes on natonl
3, In Provengal the word for ‘evening’ is (sro), ad the word for ‘sa
anes 1967.50 Undenying and su
representation
ns
jostulated that adds the velarization tothe speci
Lveunzxion
ou
Allophonie differences typically arise be
‘mentary distribution. The phonological
complement of the phot
tino
rmotogo ‘wash hands’
ag ‘snake’
| membembean “otrenbie’
bia ‘ago’
distribution?
adaptations in the production of the segment. Asa result
veo lateral approximant [sin complementary istbuton with a
rete ateral ap glvon as]. What determines ther dsbuton (Hmmelman 1801
— salhwa
fd Eng x] arin compe
“_wich each of hese allophones dca
‘toy
“manure!
“roste’
| “ale
aybreake
cause particular contexts imite pati
phones are in comple-
ars in Korea
context in which
nological
ge: together they make up the phon
tout! "al
‘clo
‘mongiaan ‘wehote
Ielembatan “t.camy onthe shoulder
witha stk
(25 tn Korean and] occur in complomentaryestibuton. What determines tet
"ban
‘aitance’
‘ang?
‘reaches
eet |
iment distibuton. Stato the contexts in
tex@
boat
nella
boxta
bis,
of segments can also arise ina dilferent way,such as
wand another to the
Tino loves of representation 5}
depend on the si
‘neo segments vi 1 speaker
contextual variants ofthe same underlying segment. A W
ity of the
ion ~ teats them as
onologiss assume that there ae two levels of represen
face one? Thee argument have been advanced
entries ifitcan be stated in a set ofa
phemes ofthe lexicon? The strength ofthis argument has been
tion by Kenstowvicr (1994a: 69), who points out that
‘occurs. However,
ysbut also, or even mainly search time
o be more difficult to
hence com.
for of two levels of repre
witha single level it would not be possible to express the pho
ness of morpheme alternants. We ake the English plural suffix as an exaimpl
igthat there is a morpheme [7] which indifferent
phonological
straints of Engl
y is
can have
compar
expressed by the ives, like nice, but by
rmeans of the'periphrastic more Inthe case
ofthe plural sufi, the availab makes it52 Underlying and suave representations
possible to state that under
surface representations
formis (2
ng form ina way that differentiates
mn from cases in which the different pho
generalization which could account forthe
While the case for a single unde English plural
very clear there has be the general question wl
ferent forms should be related toa single underlying form and wh
must be listed as separate word forms. Chomsky and Halle (1968) derived
27 nic fthe following pas of English words would you sey contains a common
tunderying form?
SoH — sowage ‘cork corkage
ow— Hew ‘conceive conception
ta taker fraternal brother
3. The third reason for the post
mn that sequences of.
le forthe fact th
(or *[bass)). The general
plural of kiss is kiss{az] and that of nose is nos{27
verbal forms before vovie:
ing. (The assumption hee is that
Neutalvation 53
fon, not forthe surface reps
8
1 Hitwareassuned |
wou be made a |
2. The past pain of sh verbs is formed by sutxng [tothe
| stom, as itis devoiced afer viceless
‘would be made abouts
make te caret pred
Surface forms that contradict a phonolog
of 2), are opaque, that is, no