In order to calculate the evolution of the learner’s pronunciation skills within each group
from the first recordings to the second recordings, the following mathematical formula was
used:
The source value corresponds to the the pre-training score, and the target value is the posttraining
score. The result gives the pre-training to post-training evolution, expressed in
percentage.
Evolution of Group A (Hypothesis 3 #1)
As expected, the subjects belonging to Group A have obtained better scores after the training
than before the training. This result proves that the segmental training has helped them
increase their pronunciation capacities concerning read speech.
Table 4 below shows the mean pre-training scores and post-training scores that were given
to Group A by the three listener-judges. The last column shows the global evolution in
percentage:
Table 4: Pre-training to post-training evolution of Group A
When looking at the mean scores of Judge 1 and Judge 2, one can observe that Group A
evolved in a similar way for the two listeners. From the pre-recordings to the postrecordings,
the mean scores of Group A as given by Judge 1 changed from 2.61 to 3.37 out of
7, which corresponds to an evolution of +29%. Our claim that the groups will have evolved
positively after their trainings is confirmed with this first result. Judge 2’s scores show a very
similar evolution within Group A of +27%. From 3.37 before the training, the participants of
Group A went up to 4.29 after the training. As to Judge 3’s mean scores, the evolution is
slightly weaker than it is with the two other judges. With a change from 2.83 to 3.24, the
evolution corresponds to +14.5%.
The calculation of the overall mean scores within Group A confirms Hypothesis 3 #1. The
global score of the group has increased from 2.94 in the pre-training productions to 3.63 in
the post-training productions. For that matter, the evolution of +23.5% is quite strong. The
hypothesis is confirmed, and the segmental training has helped the subjects improve their
overall pronunciation. By the same token, it may be concluded that French learners do not
have much difficulty in understanding and applying theoretical knowledge on English
segments.
Evolution of Group B (Hypothesis 3 #2)
Concerning the evolution within Group B, the hypothesis is also confirmed, as can be seen in
Table 5, which presents the detailed mean scores before and after the suprasegmental
training and the evolutions in percentage, whether it is an increase (+) or a decrease (-):
Table 5: Pre-training to post-training evolution of Group B
Surprisingly enough, the mean scores of Judge 1 alone invalidate our hypothesis. They
show that Group B has a lower mean score after the training, however little the difference
may be, i.e. -1.5%. This may be due to the fact that the scores of one subject of Group B
(Subject 7) significantly decreased from the pre-training session to the post-training session,
and according to all three judges. While this surprising decrease has lowered the overall
mean score of the group, it may be explained by many extra-linguistic factors, such as the
excitement, tiredness, or real comprehension difficulty of the participant. All this is discussed
later, in the subsection devoted to the discussion of the results. The evolutions as presented
by the mean scores of Judge 2 and Judge 3 are in agreement. Following Judge 2’s scores,
Group B’s score has changed from 3.79 to 4.24, with an evolution rate of +11.9%. Judge 3’s
scores present a slightly weaker evolution of +13.8%, i.e. from 2.96 to 3.37.
Despite the decrease in mean score of Judge 1, Group B globally has a higher score after
the training (from 3.37 to 3.67). The evolution of +8% confirms Hypothesis 3 #2, and the
prosodic training has helped the learners improve their read production skills.
Although both groups increased from the pre-training recordings to the post-training
recordings, the rates of evolution of the two groups are quite different, which is why a
comparison of the two evolutions is relevant.
Overall evolutions of Group A vs. Group B (Hypothesis 4 #1)
Analyzing which of the two groups has evolved more than the other after the trainings
enables us to get further insight into whether prosodic features can actually help learners
improve their production skills at the segmental level and not only at the suprasegmental
level, as is claimed in Birdsong (2003), for example. In fact, a close examination of the mean
evolution of each group clearly invalidates Hypothesis 4 #1, at least concerning read speech
capacities of French learners. The pre-training to post-training evolution of the segmental
group is stronger than that of the prosodic group. The former evinces a significant overall
evolution of +23.5%, whereas the latter only has an evolution of +8%.
According to the mean evolutions worked out from Judge 1’s scores, the gap between the
two groups is very pronounced. While Group A has increased by 29%, Group B has
decreased by 1.5%. This extreme case is completely at odds with our claims that a
suprasegmental training has a better effect on L2 learners’ production skills than a segmental
training, and that the group should have higher scores after the training.
The evolution calculated from the mean scores of Judge 2 also invalidates Hypothesis 4 #1,
even though the gap between Group A and Group B is here smaller. The increase by 27% of
the segmental group is in keeping with that of Judge 1, and it is evidence of an advantage
over the prosodic group and its 11.9% increase.
With Judge 3’s scores, the gap is even smaller, and the evolutions of the two groups are
very similar this time. The mean score of Group A has increased by 14.5%, and the score of
Group B, by 13.8%. On the one hand, the difference is inferior to 1%, and it seems that the
two groups have evolved in a similar way. On the other hand, the hypothesis is not
completely confirmed according to Judge 3, and it is the segmental group that evolved the
more after the trainings.
Evolutions of Group A vs. Group B for phrases and words (Hypothesis 4 #2)
As far as the productions of read phrases are concerned, we have hypothesized that Group B
would evince a stronger evolution than Group A, mainly because Training B included oral
practice at the level of the entire phrase. Yet, the opposite is actually found; the segmental
group showed a greater evolution from pre-training to post-training production than the
prosodic group. In Table 6 below, the evolutions of both groups according to the three judges’
mean scores are presented, following the schema pre-training → post-training = rate of
evolution (whether increase + or decrease -, expressed in percentage):
Table 6: Pre-training to post-training evolutions for phrases
For Judge 1, the better evolution of Group A is clear, and the gap between the two groups
is large; Group A has increased by 31.2%, but Group B has decreased by 0.56%. This result is
all the more unexpected as the evolution of Group A is very strong. On the contrary, the
mean evolution of Group B is close to that of Group A if one follows Judge 2’s scores.
Although still higher, the evolution of +15.7% of the segmental group is almost on a par with
the 15.3% increase of the prosodic group. As regards Judge 3, yet another pattern can be
observed, since Hypothesis 4 #2 for phrases is here validated. The increase of Group B is
greater than that of Group A, with +20.6% and only +7.74%, respectively.
Despite the latter result, the general result denies the hypothesis concerning phrases. The
mean evolution of Group A, i.e. +18.15%, is above the mean evolution of Group B, which is
only +11.2%.
The pre-training to post-training evolution for isolated words may lead to a twofold
hypothesis. On the one hand, if prosodic features have more importance than segmentals,
and Birdsong’s (2003) idea proves to be true, then the prosodic group will have evolved more
that the segmental group. This would imply that a suprasegmental training help learners
with segmental production, as well. On the other hand, considering that a segmental training
is based on phonemes and oral practice of individual sounds and words, it is just as
conceivable that the mean score of Group A for words will have increased more than that of
Group B.
As is shown in Table 7, we may conclude that the second hypothesis is true; the segmental
group has a stronger pre- to post-training evolution than Group B.
Table 7: Pre-training to post-training evolutions for words
Once again, Judge 1’s scores and the calculation of the evolutions foreground a significant
gap between the 26.6% increase of Group A and the 2.53% decrease of Group B. Even though
it is confirmed that the segmental group has evolved in word reading more than the other
group, it is hardly comprehensible that Group B should have thus decreased. For Judge 2,
both groups have increased in the word reading task, but the increase of Group A is much
stronger than Group B, with +39.4% compared with +8.67%. Finally, the increase of Group A
is slightly weaker with Judge 3, i.e. +21.3%, but it is still above Group B, with +8.12%.
On the whole, the evolution of word production of Group B is rather weak, as the mean
score has changed from 3.42 to 3.6 (+5.26%). As could be expected, however, the segmental
group showed significant improvement in word reading, going from a mean score of 2.84 to
3.68 (+29.6%).