The within-groups analysis of the pre- and post-training scores was supposed to show that
both groups had better scores after their respective trainings than before the trainings
(Hypothesis 3), even if this might have seemed to go without saying. Indeed Group A
increased by 23.5%, and Group B increased by 8%, as was reported in the previous section.
Hypothesis 3 was confirmed.
For the sake of a comparison between the importance of L2 prosody and L2 segments, the
within-groups evolutions were compared with each other in order to determine whether the
prosodic group evinced a stronger evolution than the segmental group (Hypothesis 4). If that
was the case, then the importance of suprasegmental features vis-à-vis segmental features
would be confirmed. Nevertheless, the results of the experiment bring evidence of the
opposite conclusion. The global evolution of Group A, i.e. +23.5%, is far superior to the
evolution of Group B, which is +8% only, thus giving more strength and efficiency on
learners’ skills to the segmental training. But apart from the possibility that a segment-based
training is more efficient than a prosodic training, several explanations to this observation
can be considered.
First of all, given the intentional resemblance between our trainings and typical
pronunciation lessons as given by an EFL teacher in France, it is possible that the Frenchspeaking
subjects regarded the trainings as serious lessons. As a consequence, French
learners being more familiar with phonemes, the participants had more difficulty in
understanding the training on prosodic features of English, which may have been something
completely new to them. For that matter, we noticed that several subjects already mastered
some basic elements of English phonemic contrasts before the trainings (e.g. the /i?/-/?/
distinction, the realization of /??/), while none of them showed any knowledge of English
prosodic features(2), including lexical stress, judging by their productions. If the
suprasegmental training seemed harder to the French speakers who attended it, then it may
have caused them difficulty in applying what was taught during the session, even though
oral practice was done. As we have seen in the introduction and Chapter 1, the differences
between the French and the English prosodic systems and rhythmic structures are so
pronounced that a French EFL learner may not even realize that work on and practice of the
L2 prosody are crucial.
Concerning Subject 7 of Group B whose mean score surprisingly decreased after the
training (see 4.2.2.), it has been noticed that the cause for his thus decreasing is that he largely
exaggerated lexical stress and nucleus realizations. Together with the observations from our
own teaching experience during the prosodic trainings to five French speakers, it shows that
the latter have real difficulties in understanding, and thereby learning and producing
English rhythm and prosodic system. Obviously, a training that lasted a few hours could not
make learners become experts in English prosody.
Finally, the strong differences of evolutions among the three judges of the experiments are
sometimes noteworthy, as they may explain why the prosodic group globally evolved less
than the segmental group. The most striking instance of discrepancy among the judges is the
mean evolutions calculated from the scores of Judge 1 and Judge 3. According to the former,
Group A increased by 29%, and Group B decreased by 1.5%. On the contrary, Judge 3’s scores
show that both groups have increased in a very similar way: +14.5% for Group A, and +13.8%
for Group B. If one had to follow one result rather than the other, the conclusion would be
completely different. According to Judge 1, a prosodic training does not help French EFL
learners at all, whereas with Judge 3, both a prosodic training and a segmental training have
similar effects on the learners’ pronunciation skills. Now, one may wonder if the origin of the
listener plays a role in this sharp contrast, just as Flege, Bohn and Jang (1997: 451)
emphasized that listeners’ judgements might not be a hundred percent reliable (cf. subsection
3.3.). The scores that were given by the British speaker often evince a decrease of the prosodic
group members after the trainings, while the scores of the expert in English phonology does
not. Even if one is tempted to conclude that segments are more important to British speakers’
ears, the validity of this argument is not safe as only one British speaker did the rating task.
In addition, that Judge 1 unconsciously focused on the learners’ realizations of phonemes
despite the instructions (Appendix D) is another possibility, just as a teacher and expert in
phonology may unconsciously keep in mind the importance and realization of prosody.
Since our results do not support the claim that suprasegmentals make learners’ read
productions better than segmentals do, the existence of a link between the good production
of L2 suprasegmental features and the improvement at the segmental level is also
challenged. In Birdsong’s (2003) experiment on the production of French by English learners,
the findings are similar to ours. The issue of finding out whether native-like production of
suprasegmental features could predict native-like production of segmentals was raised. The
results did not confirm the author’s hypothesis, and the French productions by the English
speakers did not bring evidence of the link between L2 prosody and segments. As was
pointed out by the author, though, more research is needed, and the claim is maintained that
prosody has more weight in communication than segments. Still, considering the similarity
between Birdsong’s and our own results, it seems that the two phonological aspects of an L2
are at least of equal importance in pronunciation to native speakers’ ears, even though
Birdsong’s experiment focused on productions of French, and ours did not investigate
spontaneous production.