Overall post-training scores (Hypothesis 1)
Interestingly enough, the overall comparison of the post-training scores (marked out of
seven: cf. 3.2.3.) obtained by the two groups reveals that neither group is better than the other
after the trainings. As a consequence, neither training has had a better effect on the French
EFL learners’ read production capacities than the other. The detailed scores that were given
to each speaker by each judge are listed in the tables of Appendix E.
The following table (Table 1) shows the mean post-training scores of Group A and Group B
given by the three listener-judges, as well as the general mean score of each group and the
conclusion that can be drawn from that (in the “result” column):
Table 1: Overall post-training scores
Concerning the mean scores of Judge 1 – the British English speaker –, Group A has
obtained 3.37 out of 7, and Group B has got 3.30 out of 7. Even though Group A, i.e. the
segmental group, has a slight advantage of 0.07 over Group B, it is impossible to draw
significant conclusions from such results. The gap between the two scores is far too small
(inferior to 0.1), and it is therefore non-significant. Judging by the scores given by the British
English speaker, neither a prosodic training, nor a segmental training helps French EFL
learners improve their (read) production skills more than the other.
Although higher than the scores given by Judge 1, the mean scores given by Judge 2 – the
American English speaker – are just as inconclusive. While Group B has a mean score of 4.24
out of 7, the slightly higher score obtained by Group A, that is, 4.29, cannot be said to
provide evidence that a segmental training has a better effect on production skills and global
intelligibility than a prosody-centred training. Again, the difference from 4.24 to 4.29 is too
little, i.e. 0.05. The two groups may be considered to be at the same level.
As for the mean scores that Judge 3 – the academic expert in English phonology – has
given to each group, they seem to point to a slight advantage of Group B over Group A. The
former has obtained a mean score of 3.37, whereas the latter has obtained 3.24. The difference
between the two is superior to 0.10, i.e. 0.13. Still, it is too little to ascertain that Group B is
better than Group A according to Judge 3.
When the mean score of each group is worked out from the three judges’ mean scores, the
closeness between the general levels of the two groups is even more striking. As Group A
obtains 3.63, and Group B obtains 3.64, the surprisingly tiny difference of 0.01 is of even less
significance than it is with the mean scores of each judge separately. As a conclusion, the two
groups can be said to be at the same level after their respective trainings. Neither the
prosodic training, nor the segmental training has helped the French EFL learners more than
the other. Consequently, it seems that as far as read speech skills are concerned, prosodic
features are not more important than individual sounds. Instead, they have the same weight
in intelligibility, indeed foreign-accentedness.
Albeit it is still very risky to jump to conclusions, it is worth noticing is that the segmental
group has obtained slightly better scores than the prosodic group in the two native speakers’
judgements (Judges 1 and 2), and only to the expert’s ears (Judge 3) is the prosodic group
better. One may thus wonder if the native speakers’ judgements should not be followed, and
segments are actually slightly more important than prosody. Nevertheless, such a conclusion
is far too hasty, since the experiment only tested read speech capacities of French learners,
and the gaps between the scores of Group A and Group B according to both native speakers
are, once again, too insignificant (i.e. 0.07 and 0.05). Further research needs to investigate the
spontaneous speech capacities and the perception skills of French EFL learners after either a
prosodic training or a segmental training.
Even though the two groups turn out to have similar capacities of read production after
their trainings, it is interesting to have a look at the scores of each group in the production of
the words only, and then the productions of the phrases only.
Scores for the words (Hypothesis 2 #1)
Given that Group A received a training at the phoneme-level and the word-level, the
participants of this group should have better scores at individual words than those in Group
B, who received a training that was more centred on the whole phrase. Yet, if Birdsong’s
(2003) claim about the link between prosody and segmentals is true, then one should not
dismiss the idea that the prosodic training has helped the learners improve their
pronunciation at the segmental level, as well. In that case, Group B may have higher scores
than Group A.
As a matter of fact, just like the general post-training scores, neither group was better
evaluated than the other for the reading of isolated words. Table 2 below shows the mean
scores of the groups as given by each judge for the words only:
Table 2: Post-training scores for words
For Judge 1, the two groups have similar levels regarding the read production of words.
The slight advantage of Group B – 3.08 – over Group A – 3.04 – is of no great significance, as
the gap is only 0.04. However slight the advantage of Group B over Group A may be, it
actually runs counter to the hypothesis that the segmental group would be better at words
than the suprasegmental group. A little tendency towards the opposite can be observed,
pointing to the other possibility that prosodic training may improve the learner’s
pronunciation at the segmental level.
On the contrary, an examination of Judge 2’s mean scores confirms the claim that the
segmental group should have better scores for words than the suprasegmental group. The
difference between the two is superior to 0.30, i.e. 0.34, with 4.60 out of 7 for Group A, and
4.26 for Group B. It must be borne in mind that such a difference is small all the same.
Judge 3’s mean scores for words are close to Judge 1’s, in so far as Group B is slightly
above Group A, with 3.46 and 3.42 respectively. Once again, the difference of 0.04 is nonsignificant,
even though it points to the invalidation of the hypothesis that Group A should
obtain better scores for words. However, the hypothesis that a prosodic training prevents
segmental errors is not safely validated, either.
On the whole, it cannot be said that Hypothesis 2 #1 is confirmed. The mean score of
Group A is 3.68, which is more than Group B’s 3.60, but still insufficient to draw safe
conclusions and generalize. In other words, neither a segmental training, nor a
suprasegmental training has a better impact on French EFL learners’ production of isolated
words than the other.
Scores for the phrases (Hypothesis 2 #2)
As Training B was based on prosodic features and thereby implied oral practice at the
sentence level, Group B should have obtained better scores for phrase reading than Group A.
The analysis of the mean post-training scores for phrases, detailed in Table 3 below, indeed
seems to confirm Hypothesis 2 #2, contrary to the mean scores for words.
Table 3: Post-training scores for phrases
On the one hand, the mean post-training scores given by Judge 1 for the phrases are 3.70
for Group A, and 3.52 for Group B. The hypothesis is invalidated because it is the segmental
group that has got the better mean score. The difference of 0.18 is superior to 0.10, but the
question of the significance of such a difference may be raised.
On the other hand, if one follows the scores that were given by Judge 2, Hypothesis 2 #2
appears to be confirmed. Group A has obtained 3.98, and Group B, 4.22. The (small)
difference of 0.24 enables us to note that a suprasegmental training may help French learners
in phrase reading more than a segmental training.
Similarly, Group B is better at phrase reading than Group A according to Judge 3’s scores,
with 3.28 against 3.06 respectively, and a difference of 0.22. This advantage of the effects of a
prosodic training over a segmental training confirms the hypothesis.
As regards the mean score of each group, it is Group B, i.e. the prosodic group, that has
obtained the better score (3.67). With 3.58 out of 7, Group A has 0.09 less than Group B. The
hypothesis is therefore confirmed as far as phrase reading is concerned, yet the difference
between the two groups is so slight that generalizing about the importance of prosody vis-àvis
segments is too risky. Finally, what is noteworthy is that among the one hundred phrases
that were recorded after the trainings, only one phrase has obtained the full score (7 out of 7,
i.e. native-likeness). It was given by Judge 2 to the reading of Phrase 5 by Subject 9, that is,
one participant of the prosodic group. If the prosodic training is the cause for that, then our
central hypothesis may be close to the truth. The analysis of the evolutions of the groups
(within-groups design) will provide more answers and confirmations – or invalidations.