Embed Size (px)
Citation preview
DOCUMENT RESUME
ED 119 504 FL 007 498
AUTHOR Hammerly, HectorTITLE The Relative Frequency of Spanish Pronunciation
Errors.PUB DATE [75]NOTE 21p.
EDRS PRICE MF-$0.83 HC-$1.67 Plus PostageDESCRIPTORS Contrastive Linguistics; *Error Patterns;
*Interference (Language Learning); LanguageInstruction; Language Patterns; Phonology;*Pronunciation; *Second Language Learning; *Spanish;Spelling; Suprasegmentals; Teaching Methods;Vocabulary
IDENTIFIERS *Error Analysis
ABSTRACTTypes of hierarchies of pronunciation difficulty are
discussed, and a hierarchy based on contrastive analysis plusinformal observation is proposed. This hierarchy is less one ofinitial difficulty than of error persistence. One feature of thishierarchy is that, because of lesser learner awareness and fierylimited functional load, errors involving allophones are morepersistent than errors involving phonemes. Phonetic difficulty andsuprasegmental and spelling interference are aggravating factors thatincrease the relative difficulty of particular problems in thehierarchy. To test the proposed hierarchy and three other hypotheses,a thorough pronunciation test was administered to 50 English-speakingstudents of Spanish (almost all of whom also knew some French). Thetest was given at the end of 45 hours of mostly oral instruction,which included 4 to 5 hours of specifically pronunciationinstruction. The resulting data formed the basis for error analysis.The results generally confirmed the proposed hierarchy; they alsoshowed the "sacredness" of word boundaries for speakers of English,the negative effects of spelling on pronunciation, greater frequencyof errors in cognate than in non-cognate words, evidence ofsuprasegmental interference, and more frequent errors with knownwords than with unknown words. Conclusions include severalsuggestions for the teaching of Spanish pronunciation.(Author/CL4,
***********************************************************************Documents acquired by ERIC include many informal unpublished
* materials not available from other sources. ERIC makes every effort ** to obtain the beSt copy available. Nevertheless, items of marginal ** reproducibility are often encountered and this affects the quality ** of the microfiche and hardcopy reproductions ERIC makes available ** via the ERIC Document Reproduction Service (EDRS). EDRS is not* responsible for the quality of the original document. Reproductions ** supplied by EDRS are the best that can be made-from the original.***********************************************************************
NE RELATIVE FREQUENCY OF SPANISH PRONUNCIATION ERRORS
Hector HammerLy
Simon Fraser University
Hierarchies of Pronunciation difficuLty can be estabLished
in two ways: they are a priori when they are based on a theo-
retical comparison of two sound systems and a posteriori when
they are,the result of error anaLysis.
An example of a hierarchy estabLished a priori (although
it was also partly based on informaL observations over a period
of severaL years) is that of StockweLL and Bowen. L Substitu-
ting the words aLLophone for nobLigatory" and phoneme for
"optionaLl" the first six probLems in their hierarchy, in
order of decreasing difficulty, are:
Order English Spanish-
L 6 aLLophone
2 6 phoneme
3 phoneme aLLophone
4 aLLophone phoneme
5 aLLophone 6
6 phoneme 0
CHART I
An example of a hierarchy of difficulty made a posteriori
is that established by Briere2after teaching monolinguaL
EngLish speakers a composite Language" made up of Arabic,
French, and Vietnamese utterances. He found no correLation
between the frequency of occurrence of a Phoneme and its
Position in the hierarchy of difficulty. He found the order
2
2
of difficulty for fourteen sound-sto be the folLowing:
L (easiest) Ei
2 [ey, x]
3 [e,
4 [h, r, U, g, t ]
5 (hardest) [1, h (Laryngeal fricative)]
le concLuded, among other things, that the syLLabLe is better
th9n the word as a basis for contrastive anaLysis; that foreign
sounds with close native counterparts are easier to Learn
than those without such equivaLents; and that a prediction of
a hierarchy of difficulty must be based on information at the
phonetic Leve L.
he author, on the basis of contrastive analysis plus the
informal_ observation of Spanish Pronunciation errors over a
period of fourteen years, proposes the foLLowing a priori
tentative List of types of phonoLogicaL interference in
decreasing order, not of initiaL difficulty, but of persistence
as a source of errors (that is, difficulty in the Long run):
Order Native Language- - -
Second Language
L aLLophone 0 !P0t.r.i
7
Zea)
ot6) Z(1, CD
ti PIca
Pa,
m
Q.0Z
o4-I
0o
bZ
n-I
0aw
.
different distribution or function'of native Language' aLLophones
fo
4-I
..-ivo
r-I4-,a)
zo
..ca
.3 b alLophone
4 0 phoneme
5different distribution or functionof native Language phonemes
/o Phoneme 0
CHART II
3
3
ExamoLes (from .-ngLish as the native Language and Spanish as the
second Languarce):
L. English [pa, t4, kc], aspirated in initial position.
2. English flapped [t] and 01 phonetically identical
to Spanish M.
3. Spanish [g] and [b.).
4. Spanish /Z./ and hi/ (also /x/ in some dialects).
5. English15/, phonetically acceptable as the [4] allo-
phone of Spanish /4/.
6. English /-W, /a/9 /v /, etc.
Phonetic difficulty: new Spanish a totally new type
of articulation, much harder than nevi Spanish /fi/,
the conjoining of two known articulations, /n/ and /y/.
Suprasegmdhtal interference: English tendency for /a/
to appear in unstressed syllables; this expLains why
[a] is by far the English sound most intruding in
Spanish.
Spelling interference: Spanish v., 4h, etc.
(The reason why allophonic problems are more difficult than
phonemic ones is that the correct use of allophones is not as
important for communication as that of phonemes; as a resuLt of
their being functionally less important, more errors are made
with them. The reason why stopping the use of native Language
--NL,- allophones is harder than learning new second Language
- -SL-- allophones is that, while NL allophones are mostly
below the level. of awareness and are used automatically, the
Learner can be more easily made phonetically aware of SL
4
4
allophones. On the other hand, learning a SL phoneme is harder
then stopping the use of a NL phoneme, because the former in-
volves perceptual and articulatory difficulties, while the
Latter merely means that the speaker must stop using a sound
of which he is well aware.)
Several scholars have pointed out that pronunciation
errors can only be partially predicted through contrastive
analysis and that error anaLysis should therefore be used as
the main predicting tool. Ideally, such studies would be based
on spontaneous speech; this, however, would require a very
Large sample if it is to include all problems.. Such a sample
would be very difficult to obtain if the subjects are beginning
Language students; and even if it could be obtained, it would
include a very high percentage of repetitive material.
A solution to this problem is to obtain, by means of a
test, a sample containing the, problems that can be predicted
by contrastive analysis plus any other problems that may have
been observed in the course of teaching. Such a test, adminis-
tered at different points in a Language program, would give a
good idea of the nature, prevalence and persistence of errors.
Pine rest of this article is a discussion of such a test
and its results when it was administered, to students of Spanish- -qs mtr's
after nine weeksAof instruction with an oral emphasis.
METHOD
Subjects
The subjects were the 62 students in the first semester
Spanish course at the end of its ninth week. Although the
5
5
pronunciation test was administered to all, the analysis of the
results was based on the pronunciation of 50 students, nine not
being considered because their native Language was not English
and three being eliminated at random. Still, the subjects were
not absolutely monolingual, since almost all of them had taken
several years of secondary school French, a requirement in
British Columbia for students planning to enroll in a univer-
sity.
Instruction
The method of Language instruction was "cognitive audio-
lingual," that is, audio lingual with grammatical explanations
and discussion preceding the study of each grammatical point.
Instruction in pronunciation took place during the first
eight weeks of the course and consisted of selected exercises
from Bowen and Stockwell's Drillbook,3 in periods of ten to
fifteen minutes, two or three times a week, for a total of
four to five hours during the eight weeks. In addition, most
of the students had a copy of the Drillbook and practiced
Spanish pronunciation patterns further in the language labo-
ratory, where several copies of tape recordings based on the
Drillbook were available. There was no prereading period, so
the students read orally-learned materials in standard Spanish
sPeLling from the first day of classes. The instructors were
native speakers of Latin American Spanish.
Test
The pronunciation test had four parts, ail of which were
recorded on tape at 7 1/2 i.p.s. In Pert A the students
6
0
imitated known words and phrases. In Part B they imitated un-
known words and phrases. In Part C they read aloud (after a
silent reading) known words and phrases. In Part D they read
aloud (again after 0 silent reading) unknown words and phrases.
The students had no way of knowing which sound or sounds were
being tested in each word or phrase.
A total of 45 pronunciation problems were tested, all
dealing with segmental phonemes, some of which aopea73ed in
several environments. Most pronunciation problems were pres-
ented in the four °arts, but some appeared in only three or
even two parts of the test.
HYPOTJESES
(1) The first hypothesis was that the errors would group
themselves 1P7 .tpe in the order of difficulty given in Chart II
on parre 2.
(?) Secondly, it was hypothesized that due to the inter-
ference of spelling on pronunciation, already shown clearly
in another study,4 there would be more errors in parts C and D
(rendin7 Al.oud) than in parts A and B (imitating) of the test.
(3) Another hypothesis was that problem sounds in cognate
words would be misoronounced more often than those in non-
co,mate words.
(4) The fourth and final hypothesis was that the sounds
tested in known words and phrases (parts 1 and C) would be
;Mispronounced more frequently than the same sounds in unknown
words and phrases (parts B sand D); the reason for this beLief
wos the observation that once students adopted an utterance vs
7
7
their own, they used it without much attention to its phonic
:end according to a neither-English-nor-Spanish pronun-
ciation parasystem, while they paid more careful attention to
the phonic detail of unknown utterances.
RESTL2S5 AND DISCUSSION
Problems in Decreasing Order of Difficult/
The 45 problems treated in the test appear, in order of
decreasing difficulty, in Chart III on pages 8-14. First
it should be noted that the 45 problems oresented are by no
means all the ootentiaL pronunciation problems of an English
speaker Learning Spanish. Kany problems have been Left un-
exoLored and could be the subject of another study; for ex-
amole: /II does not appear rafter /n/ or after a stressed
vowel, /4/ does not apoear in final position, and only three
of the many possible reductions of Spanish vowels to English
[a] have been studied.
Examples of errors due to spelling abound. Compare, for
instance, parts C and D with A and B in problems 23, 25 and 35.
Spelling-interpretation errors may show in the different types
of errors in the various parts of a problem: compare 8D with
33, and 37 C and D with 37 A and B. Spelling interference,
however, may vary according to the environment in which a letter
appears; thus causes a .67 mean error in word-medial
oosition (orobLem 2 --compare with problem 21) but only a .03
mean error in initial position (problem 45).
An example of greater difficulty with known than with un-
known words is problem 28, where there were more errors in part
A than part B end also more errors in part C than part D.
8
PRO
BL
EM
S IN
DE
CR
EA
SDIG
OR
DE
R O
F D
IFFI
CU
LT
Y
Ord
er
Prob
Lem
Exa
mo
Le
A
Imitr
-itio
n of
know
n w
ords
3Im
itatio
n of
unkn
own
wor
ds
CR
eadi
ng a
Lou
d of
know
n w
ords
DR
eadi
ngun
know
nL
oud
ofw
ord
s
Mea
n E
rror
;
k06
k04
k24
k08
Lk'
7/-
(Zu4
...?
10-
94lc
,96
k476
kc84
.90
kw08
(k`-
....
.88)
__-
____
____
____
_ _
_._-
wor
d-la
-04
la-
L6la
-L
42
med
ia 1
.fa
vor
v86
v54
v62
.89
41e.
) as
bL
Ob
30b
24(v
67)
/n/-
/m/
m30
/ m06
m04
3be
fore
un p
ace
Ln
70n
94n
92
.87
i /p/
If04
920
p26
p1.
0p
124
p/pa
sa9.
78Pt
74pc
90P`
88.8
34.
....
.. .0
2t
L6
tL
6t
1.8
t20
5t/
tiatc
82t
84tc
82tc
78.8
310
02d
02;(t
c....
.82)
wor
d-4 6
initi
al.
4b>
as
estd
boni
tala
- b28 72
la- b
26 74la
- b00
1.00
.82
wor
d-in
itia
L4
2408
307
4d>
asto
disc
od
4890
d8
.80
[4]
i428
f02
i.02
CC
IOrder' ProbLem.
EXamole
AB
CD
au
o other...
00 80 .20
Mean Error
78(Note: there
were no known
words with
Sp.
/4w
/)
8/6
11/ a
sCal],
not
[au]
etc.
au La
au au other...
44 43 .03
9fines L /o/
[o],
not
[on]
periddi cp..
o Jg
22 78o 0,4
22 78o O
R26 74
o 0 T
A
?LP
76 08 90 02
_ __
36 64
77
lo
word-
initial
4g:'
as
Es]
me Frusta
g g 1;.
50
48 02 14 84 02
e eI
1.6
84
,.. e&
1.6
84
36 64
fs g x _ _ ___
e ei
.75
__
.75
1L
fi n
al. /4/
as [e],
not [ej,
]no
s4e ej
1
12/V
sV/ a
s[s
7,not
[z], in
cognates
,r4 "'
s z22 78
s z20 8o
s z24 76 28 38 24
LO
s z il 1-' a ai.'
R Hi"
46 54 32 32 24 02
o6
04
72
1.3
word-
initial
r as
CIV
7.a rises
I'
I' .7 xi
"hi
'R
24 40 1.2
16 06 02
1-1
i'.' a ai"
hi" A Xr.
R Ril
fli"
28 22 02
08.06 16 04 14
I-1 i" a R
.72
(i"
40)
(.z
22)
...
O
Order
ProbLem
Example
AB
_______
14
86
D
38
08 L2
.L8
.24
Mean Error
.72
14
/17y/ as
[VI],
not [VO,
etc.
hay
(ea3T')
al......
az 4
.3854
08
Vai.)
ai
rii
a8
..
..
e
24
60 12
04
(goy
)
of
di;
(601M
ofoiob
Ita
ef
Ile.
other...
126
a14
-14
a30
CoC(C)/
a72
ao0
a62
a48
L5
as [a],
todavfn
02
.2
06
-e
02
06
.69
not [a],
02
i10
etc.
e04
ye
02
(P
6L)
i.
38
I'
46
i.....
'L8
14
28
a40
J34
aLO
a38
L6
ViltI
divertido
bL8
o2
b iL6
b66
b32
.68
R02
Ro6
R02
ail
02
8, b.
.64)
a14
443
426
436
L7
114/
oLvid6
r56
10
d 14
36 L4
d74
d i..
5608
.64
4...,...
.02
i.
oore
~ r36
i4L
oi"
20i"
463'
';84
13
164/
wzrroll
rj
02
02
J s i
hi:
0202 04
a ila
o4 02
.63
R02
El
06RI.
06
Ri...
04
(1.*
55)
RiI
02
(a
04)
4o
4:c
30
g30
g22
L9
V4:07
Luego
53
pL
8g
68
g76
.57
1;.
02
L02
k02
k02
Order
Problem
ticamoLe
Ae
weole
CD
wean Error
20
/...o/
/e.../
-" /we/
encuentro
en
weote
:36
64
90 LO
'le
ole
> 3
72
weole
26
74
3362
.55
.54
r
2L
word-
mediaL
,b;, as
[4-]
obueLo
U. b.......
23.72
la-
b v other...
76
06 L6
.02
b-.
b42
53
b b
22
4o
cansodo
14 d I.
bo
-io
0604
4 il d L v 6
36 20
03
32
02
02
4 r d 4i-.......
28
6604
.02
4 i4
6624 LO
.53
2?
/C(C) as
paLabras .
a :
78
020
f,
6420 L6
q a 0 n v
26
72
02
8020
, n V
34
64
02
6040
(9
.50
.48
44)
24
/ntV
tanto
n /40
60
n /2872
/eY/as
25
[E;Tbnot seis
[el ]
, etc
.
el eA . .....
eO
73 .1.9
,04
e1 el e6
68 L6
L6
el
eg ye ee
other...
32
'42
2404
.06
ej1
eI em
ye
other...
32
1.4,
28L2
.L4
.48
26
VnsI
cansado
n78 22
n V40
60
n V70
30
n V6832
.36
2?
l7tV
fritss
1
tL.00
t te
78
22
t tc
14 other...
40
5202
.06
t tc
14
44
43
08
.35
Ord er
c
Prob Lem
Dcam-ple
AB
CD
Mean
Error
28
... el.
pope'.
. .74
26
i.
-4:
90 Lo
1
-4-..
'32
64
1
-I:...
.. .
70
.3o
.33
29
/6/ a s
[a], not
[Fe]
vera no
1 T a h P.; k
A X
.. .
(8Xk.
6038
02
a94
06
1 it68 32
a .2 b...
.. .
4850
.02
(.33
32)
30
<6,-). as
Ch
3or
[X3
and
not a s
[2
],[f
l,[ g
], etc.
Argentina
.92
02
02
.04
\\
hA
-
z J g t
. . .
.74
04
02 LO
04
06
h8-
X...
z J g k y
.38L6
08
132
04
02
(77,,
.32
S 8:
g.25)
3L
ti.'e...
as
[±1]
trece
'24 .1 R 5:
7' 2
74L6
060202
' a R 11
i2
8o LO
04
04
02
i.' a R b
6822
08
02
±".
.1 R..,
.....
2 0
7o to
.08 LO
02
.1
(h
.27
a- 2
.19)
.27
25)
32
4h> as
bhe Lado
1 00
I..00o
0 h d&
a...
20
76
.04
33
ailik
estudiar
il .1 a 2
82 L4
0202
il a
'R b
72 22 04 02
i. ..1 R
7224
04
i-' .1 2 0 i:a
70 24 02
O2 02
( -I
.26
29
34tz
> a
s[s
]em
pi e
za
I
s z88 12
s z is78 L4
08
s z6o 40
.25
Ord
erPr
oble
mE
xam
° le
AB
98 .02
C
a al a
.....
60
.40
a a I
a
D
56 44
Mea
n er
ror
.23
35/..
.a /+
/a...
/-4
/a/
La
acci
dn1 a
a....
.96
.04
1 a a
.....
36C
PC...
cerc
ae a
R-
E...
.74 26
e R
:a
E...
.98 02
e x
C...
a ye
.74 24 -02
e e,
aC
...72 28
.21
37ii
asC
ii]ch
ampa
kia
ii ny
92 08
Fi ny
72 28
15 n84 i.1
6fi n ny
76 22 02
. L9
LO
)
38
Vif
.IT
inco
gnat
esam
eric
ana
P a74 22 04
-
14 1 R.
11
92 02 02 04
i a R
,76 20 04
i''' a P
92 o6 02
.17
(aL
3)
39
Cas
t...
Eli
[a ]
,no
t [.-
e],
etc.
14S
ti M
a
1 T a
92 06
02
1 ,o, a a
88 04 04 04
a b a a
66 28 02 04
a a
92
08
.16
( no
ncog
-na
tes.
.09)
40T
7sV
inno
ncog
-m
ites
cose
ns z
86 L4
94-
04 02
.14
4L
,5r
PIT
cara
a94 04 02
P 1 R 4
90 02 02 06
P .. R
88 ,08
04
a.0
9(
a05
-)
L
VI
very
; no
P `
L. 0
0ic .1
.
R 5-7 .x in
b
82 06
06 02 02 02
f. .x R r
90 o4 04 02
R
94 04 02.0
9
(.z
04)
Order
fProblem
Example
AB
1
CI
DMean Error
43
u> as
6 4ueda
[k]
q(-...00
kL
k kw
R,
k-...6...
.98 02
k.9, k
kw. 84
L6
.06
44
<--p> as
[h] or
UK]
jueves
h°- X...
.96
014,
h v k
A 55:..
.96
02
02
h R- X..
b k y
.94
02
02 02 92
08
.05
(0 &
g..
i00)
.03
45
utterance-
initia I
as[b]
vamos
b1.
00b v
98 02
b v
CHART III
1.-- .
Other:
8B: ad... .06, ab... .02.
8D: as... .08, ad... .08, 915...
.04.
Zit
L4D: of... .20, of... .04.
2LB: X.... .02 (the word
was abeja, so this is an assimilation).
250: ef... .04,
n.... .02.
25D: et... .14.
27C: 6.... .06 (the word
was sitio).
15
The same NL phoneme may cause different degrees of diffi-
cuLty in different environments. Thus, English /V appears more
often in the environment CaC(C)/ (.61, problem 15) than in
1C(C)as, an ending (.44, problem 23), and even Less freauentLy
in CHC (.2L, orobLem 36). The same SL phoneme win also show
different degrees of difficuLty in different environments; thus,
/i../ is hardest to produce before a consonant (.63, problem L6),
easier after a consonant (.27, orobLem 3L) and in finaL position
(.26, orobLem '33), and easiest intervocaLicaLLy (.09, problems
4L and 42).
Some errors showed uo that cou Ld not be predicted by con-
trastive anaLysis. A good exampLe is 22B, where the most fre-
quent erroneous rendition of the [4] in /...64o/ is [L] (.32).
The fact that [L] is aLso n voiced continuant may help explain
the error, though not fulLy9 since [4] also occurs in the en-
vironment / V in EngLish, as in Lather.
An interesting error is the rendition of intervocaLic /i1/
9S [4] in problem 41.8. A LikeLy explanation is that the subjects
Perceived the Spanish [r] in the native Language, as an aLLophone
of EngLish /d/, and then went on to produce it accorhng to
Soanish distribution ruLes; that is, it seems to be a case of
perceotion in the native Language foLlowed by production in the
second Language.
VaLidation of the Hypotheses
(0 The first hypothesis, concerning the different degrees
of difficuLty of the various types of problems, is generaLLy con-
firmed, as the reader can see by comparing Chart II on page 2
16
L6
with Chart IV on pgge 17. It shouLd be noted that some of the
ranking in Chart IV cannot be definitive without knowledge
--not yet avaiLabLe, as far as the author knows-- of the fre-
quency with which each phoneme (/i1/ is a good example) appears
in each of its environments.
The apparent "sacredness" of the word boundary for English-
speaking subjects shouLd be noted. Distribution ruLes across
word boundaries, whether they appLy to allophones or,,Oonemes,
cause a greater percentage of errors .(.81) than diStribution
ruLes within word boundaries-60.
As to new aLLophones, if the influence of Gv is discounted
(problem 2), [1.g.] and [g] are about equaL in difficuLty (.68 and
.66 respectively). Surprisingly, [4], even though it exists in
EngLish, presents a very similar degree of difficuLty (.69) in
the situations studied -- problems 7 (17, 22). 6
Of the two new phonemes tested, riV, as expected, proved
to be much more difficult (.68, problems 1.3 and L8) than /h7
(.19, probLem 37).
Of the EngLish sounds mistakenly appearing in Spanish, the
most Persistent was [a] (.42) , foLLowed by [v] (.35), [z] (.34),
[a] (.23), [P2] (.21), and [g] or [S] (.05).
Of the Letters misinterpreted, w>> Led the way (.35), with
4z., 4 0> and 4fp not far behind (.25).
(2) The second hyoothesis, which concerns spelling pronun-
ciations, is statisticaLLy confirmed by the finding that the
mean error of C D (104.84) was significantly higher (at the
.05 LeveL of confidence) 4,han the mean error of A B (88.0) in
17
Mean of
Error Means
native Language:
En9
/ish
Second Language:
Span
t'sA
Conditions
Problems Included
.84
a 11
.0phone
6when no other
sound is affected
1.,
4,
5
.31
new distribution ru Les across word boundaries
3, 6
, 7,
LO
.71
b:-
a I_Lophone
(2,
6,2L)
( 10
,1.
9)
.64
phonetic difficu Lti es
8-B,
9,L
i,L
4, 2
5.51
new distribution ru Les within word boundaries
(2,
21)
[1.6
, 3L
, 33
(38,
41,4
2) ]
( L
7, 2
2)1.
9, 2
7.4
3b
phoneme
(L3,
18)
37
.39
bsync Le ph a
20, 3
5.3
8a
1.Lophone
bwhen another
sound is rtf'ected
( 24
, 26)
28
.27
phoneme
b(2
, 45)
[(1
2, 4
0) 3
4][1
3, 2
6, 1
8,31
., 33
( 3
8, 2
.1-.
1.,2
.1-2
) ]
(152
23,
36)
(29
, 39)
(30
, 44)
.224
-spelling misinterpretations
(2, 4
5) 3
0, 3
2, 3
4, 3
7, 4
3, 4
4
CH
AR
T I
V
C1C
-71
18
the 31 problems in which they could be paired (t=2.31.4). Sig-
nificantly greater error means were also found in C vs. A
(p <.05), D vs. A (p <.01) and D vs. B (c) <.05) . In other
words, it is much harder to pronounce Spanish correctly when
reading it aloud than when imitating it orally.
(3) The hypothesis that stated that sounds in cognate words
would be mispronounced more frequently than those in noncognate
words is confirmed by the Limited data available for gl/ (prob-
lems 38 vs. [41.4-42]), /s/ (12D vs. 40D) and /a/ (39C vs. 39D).
Their mean difficulty in cognates was .33, in noncognates .08.
(4) Finally, the hypothesis that known words would be pro-
nounced with more errors than unknown words is only partly con-
firmed by the data. ,4hile the mean error of A C (L00.39) is
higher than that of B D (92.45) in the 31 figures that coud
be paired, the difference comes clOse to, but does not attain,
statistical significance (the t=4.68 obtained falls short of
the t=2.04 required for statistical significance at the .05
Level of confidence). Further experimental study of this
hypothesis seems in order.
CONCLUSIONS
In general, Problems involving the use or non-use of allo-
phones are harder than those involving the use or non-use of
phonemes. Moreover, difficulties depend partly on phonic en-
vironment (e.g., /IV before consonants harder than between
vowels), on Phonetic characteristics (e.g., /iV harder than
/ff/), on the interference of spelling (e.g., 4v
on [b-]) , on suprasegmental interference (e.g., [a] more fre-
quent in unstressed syllables), and on word boundaries
(e.g., [g] harder initially than medially).
19
L9
Spanish pronunciation drillbooks and beginning textbooks
should include more driLLs on the pronunciation probLems that
have been found to be most persistent in this study; in partic-
ular, they shouLd pay more attention to phonetic probLems and
to probLems invoLvinz sound distribution rules across word
boundaries. (IdeaLly, no sound wouLd be produced in connected
speech until its pronunciation has been Learned.)
The Sonnish Language has a reLativeLy good, mostLy unidi-
rectional fit between letters and sounds; however, this fact
seems to be of benefit primariLy to Spanish speakers --EngLish
speakers are very frequently misLed by Spanish speLLing into
various types of oronunciation errors. However, since students
of Sonnish wi LL read a great deaL, Spanish pronunciation drill-
books and beginning textbooks should include, after the pureLy
oraL pronunciation driLls, a thorough section with exercises
(for rending aLoud) designed to tench speLLing-sound correLa-
tions. (Idealy, no Letters wouLd be read untiL their phonic
symbolism has been Learned.)
Pronundiation tests based on reading aLoud do not present
--since reading aLoud is considerabLy harder than imitating-- a
fair picture of a subject's command of pronunciation in speech.
FinaLLy, since cognates increase -- fourfold ? -- the inci-
dence of misoronunciations, it seems that they shouLd be
avoided as Long as oossibLe, certainty untiL after the students
have acquired A good command of pronunciation both oral.Ly and
when rending =a Loud.
20
4
20
No tes
L Robert P. Stockwe L L and J. Dona Ld Bowen, The Sounds of
LiSh and Spanish ( Chicago: University of Chicago Press, L965) ,
PP. 9- L8.
Eugene J. Briere, "An Investigation of Phono Logical Inter-
ference," Language, Vo 1.. XLII, No. ( December, L966) , pp.
768-796.
J. Dona ld Bowen and Robert P. Stockwe I. L, Patterns of
Spanish Pronuncia tto n --A Dri L Lbook ( Chicago: University of
Chicago Press, 1.960) .
Hector Hammer Ly, "And Then They Disbe I.ieved Their
Ears," Hispania, Vol.. 53, No. L (March L970), pp. 72-75.
5 ;There computations of statisti caL significance were
needed, the author is indebted to Professor Cesario Vi I Lega s
of the Dept. of Mathematics, S. F. U. , for his advice, and to
Mrs. Constance T. Dwyer of the Ste tistica I. Laboratory, Dept.
of Mathematics, S. F. U. , for carrying out the computations.
6The oarentheses around 17 and 22 signify that the mean
errors of these two orob Lems were averaged separately before
averaging the resu Lt with the mean error of' prob Lem 7 --that(17 + 221
s, the Tomo La was . the parentheses and
2
brackets under "Prob Lems Inc Lud ed" in Chart IV ( page 1.7) and
on page L8 have the same function.
21
