Skip to main content
. Author manuscript; available in PMC: 2012 Nov 27.
Published in final edited form as: J Acoust Soc Am. 1989 Aug;86(2):566–581. doi: 10.1121/1.398236

TABLE X.

MRT open-response phonemes accounting for the highest percentage of error in final position.

System Phonemes
(Percentage of error accounted for in each evaluation)
Natural ŋ
(26.67)
m
(13.33)
p
(13.13)
n
(13.13)
ð
(06.67)
Prose 2000 V3.0 k
(28.02)
θ
(11.64)
s
(07.76)
p
(06.47)
t
(05.60)
DECtalk Paul n
(17.20)
k
(13.70)
ŋ
(11.80)
s
(10.30)
m
(05.90)
DECtalk Betty p
(15.50)
ŋ
(12.90)
s
(12.40)
n
(10.40)
t
(08.70)
MITalk-79 n
(23.60)
ŋ
(20.00)
vowela
(10.20)
m
(08.90)
f
(06.20)
Amiga 500 k
(18.64)
t
(14.57)
ŋ
(08.53)
ld
(08.01)
v
(06.43)
Infovox vowel
(09.30)
ŋ
(06.80)
m
(05.70)
n
(05.60)
v
(05.20)
Smoothtalker d
(13.43)
g
(10.45)
t
(09.57)
p
(09.13)
n
(08.25)
Votrax p
(14.05)
k
(13.04)
n
(10.03)
ŋ
(09.03)
m
(08.11)
Echo k
(15.81)
n
(12.99)
p
(12.09)
d
(10.44)
g
(08.89)
a

The MRT contains six words (way, may, say, gay, day, and pay) with the vowel /eˈ/ in final position. Errors for the vowels in these words are also included in the calculation of the error rates for the vowel /eˈ/ in Table IX.