Clusters in a Japanese FCE essay
4.4 Results and Discussion: Vocabulary Breadth
4.4.1 Piloting the LFP methodology and gaining an overall picture of vocabulary development: LFP analysis of whole essays with spelling errors corrected development: LFP analysis of whole essays with spelling errors corrected
4.4.1.1 Variation in frequency scores among learners of the same level and language background
So far, the analysis has focused on the average frequency scores observed at each level.
However, given the variation in metaphoric density observed among learners of the same level and language background in Chapter Three, it was also considered important to explore the range of frequency scores and their standard deviations observed in the data.
To do this, box and whisker plots were constructed for the first three frequency bands and the ‘beyond 3000’ words. The plots (Figure 4.8 - Figure 4.11) express the percentages of words coded at each frequency band, with the tables below them (Table 4.4 -Table 4.11) giving the raw frequencies from which these percentages were calculated and the
0 2 4 6 8 10
KET PET FCE CAE CPE
Average % of words
CEFR level
%s of words coded at band 2, 3 and 'beyond 3000' in French learners' writing
Band 2 Band 3 Beyond 3000
151
mean/standard deviations for the percentages themselves. For example, at French KET, the average percentage density of words found in frequency band one was 90.26% (50.86 words). The minimum percentage coding at band one at this level was 71.21% (26 words), and the maximum was 98.36% (122 words).
Figure 4.8: Box-and-whisker plots showing variation in percentages of Band 1 words for writing produced by French and Japanese learners
Table 4.4 Table showing raw frequencies of Band 1 words in complete essays 50
Variation in %s of Band 1 words - Japanese learners' writing
Variation in %s of Band 1 words - French learners' writing
152 Table 4.5 Table showing percentages of Band 1 words in complete essays
The statistics for Level 1 words show a moderately large degree of variation within languages, with the means decreasing slightly as the CEFR level increases and words from the lower-frequency bands occur more often.
Figure 4.9 Box-and-whisker plots showing variation in percentages of Band 2 words for writing produced by French and Japanese learners
Band 2 Raw
Frequencies Japanese KET Japanese PET Japanese FCE Japanese CAE Japanese CPE French KET French PET French FCE French CAE French CPE
Mean 0.62 2.71 13.95 23.48 27.52 1.76 3.24 17.71 20.10 24.90
SD 0.50 2.63 6.95 8.69 10.83 1.76 2.17 6.38 6.32 6.85
Minimum 0 0 3 4 11 0 0 6 8 12
Maximum 1 11 27 44 59 7 8 29 34 41
Table 4.6 Table showing raw frequencies of Band 2 words in complete essays 0
153 Table 4.7 Table showing percentages of Band 2 words in complete essays
The graph for the band 2 words clearly shows the increase in words from this frequency band at level FCE for both languages, as noted above. However, high levels of variation are also noted, with one learner at Japanese PET level using 11 words from band two while the mean was 2.71.
Figure 4.10 Box-and-whisker plots showing variation in percentages of Band 3 words
Band 3 Raw
Frequencies Japanese KET Japanese PET Japanese FCE Japanese CAE Japanese CPE French KET French PET French FCE French CAE French CPE
Mean 0.62 0.33 4.14 9.29 15.95 0.62 0.62 4.43 7.43 15.76
SD 0.97 0.66 3.42 4.53 6.98 0.97 0.86 3.37 5.31 5.54
Minimum 0 0 0 2 7 0 0 0 1 4
Maximum 3 2 12 23 31 3 3 12 20 24
Table 4.8: Table showing raw frequencies of Band 3 words in complete essays 0
Variation in %s of Band 3 words - Japanese learners' writing
Variation in %s of Band 3 words - French learners' writing
154 Table 4.9 : Table showing percentages of Band 3 words in complete essays
Again, the ‘jump’ at level FCE can be noted for the level three words, although it is not as pronounced.
Mean 0.38 1.71 2.95 6.29 6.10 0.10 0.33 1.00 1.43 4.10
SD 0.59 1.82 2.31 4.69 3.62 0.30 0.58 1.18 1.33 4.12
Minimum 0 0 0 0 1 0 0 0 0 0
Maximum 2 6 9 18 17 1 2 4 4 18
Table 4.10 Table showing raw frequencies of ‘Beyond 3000’ words in complete essays 0
Variation in %s of 'Beyond 3000' words - Japanese learners'
Variation in %s of 'Beyond 3000' words - French learners' writing
155 Beyond
3000 %s Japanese KET Japanese PET Japanese FCE Japanese CAE Japanese CPE French KET French PET French FCE French CAE French CPE
Mean 1.14 2.66 2.30 1.59 0.70 0.29 0.49 0.54 0.71 5.00
SD 1.54 2.70 1.72 0.80 1.23 0.57 0.67 0.42 0.73 4.77
Minimum 0.00 0.00 0.00 0.36 0.00 0.00 0.00 0.00 0.00 0.75
Maximum 4.76 11.54 5.59 3.71 4.84 2.00 2.13 1.49 3.09 20.00
Table 4.11 Table showing percentages of 'Beyond 3000' words in complete essays
Analysis of variation has made it possible to begin attaining a better insight into the nuances in the data, obscured in the first analysis which was based solely on averages. The degree of individual variation is shown to be high, which was to be expected given the variation observed in previous chapters and in the literature. It also provides further evidence that while it is useful to examine overall patterns of development, such analyses obscure the variation which is such an important feature of the data.
This preliminary analysis is significant for three main reasons. First, it suggests that the LFP procedure using the BNC-COCA word lists is an adequate methodology: development proceeds in both languages as expected, with the majority of words coming from band 1 and words from the later bands appearing in more advanced levels. Second, the fact that some variation can be observed in developmental patterns between French and Japanese learners implies that there will also be a similarly observable level of variation in use of metaphor from different frequency bands. Third, it provides a way to chunk the data into the first three frequency bands and ‘beyond 3000’. Now that an adequate methodology has been developed and an overall picture of general vocabulary development attained, the focus can be turned to distinguishing between the open-class MRWs on the one hand, and the non-open class MRWs, non-MRWs and possible personification on the other: the ‘other words’ category.
156
4.4.2 Results of LFP analysis on corrected open-class MRWs (direct and indirect) Having attained a general picture of lexical development and how it varies in the data from the different language backgrounds, it is now possible to hone in on metaphor in particular.
Again, this part of the analysis focusses on metaphor where the spelling mistakes have been corrected. Table 4.12 shows average coding density percentages for open-class metaphor for the frequency bands identified in Section 4.1. Essays containing no open-class metaphors were excluded from this part of the analysis, which led to the exclusion of twelve essays from Japanese KET, ten from French KET, seven from Japanese PET and four from French PET. Two metaphors were discarded from the analysis from Japanese CPE and one from French CPE for the same reasons as in Section 4 above.
Average no. of
CPE 27.33 70.26 15.55 7.84 6.19
Table 4.12 Table showing percentages of Bands 1, 2, 3 and 'Beyond 3000' words, in corrected OC MRWs
The composition of the ‘Beyond 3000’ band is as follows:
157
Table 4.13 Table showing coding percentages of Bands 4-17 in corrected OC MRWs
158
Figure 4.12 Graph showing the percentages of corrected OC MRWs coded at bands 1, 2 3 and 'beyond 3000' in Japanese learners’ writing
Figure 4.13: Graph showing the percentages of corrected OC MRWs coded at bands 1, 2 3 and 'beyond 3000' in French learners’ writing
159
Figure 4.14: Graph showing %s of words coded at bands 2, 3 and 'beyond 3000' in Japanese learners' writing
Figure 4.15: Graph showing %s of OC MRWs coded at bands 2, 3 and 'beyond 3000' in French learners' writing
These graphs provide further insights into some of the preliminary findings reported in Chapter Three. In Chapter Three, it was observed that open-class metaphor use overtakes that of closed-class metaphor at different points depending on the language background.
160
learners produced this crossover one stage later, between levels PET and FCE. The frequency profile analysis results shown here further corroborate this finding. At those points where open-class metaphor use was shown to overtake closed-class use, the learners also begin to use metaphor from bands two, three and beyond, suggesting that the increases in open-class use coincide with a greater level of sophistication. The Japanese learners, for example, do not use metaphor from bands two and above until level FCE, although it was seen in Section 4.4.1 that they do use non-MRWs from these bands at the same earlier levels as their French counterparts, but usually not as frequently. This difference may suggest a certain tendency on the part of the Japanese learners not to use metaphor from the lower frequency bands until they are at level FCE, when their use corresponds with the significant increases in open-class metaphor density.
At this point, it is possible to begin comparing the developmental patterns of metaphor to the patterns observed in the learners’ lexical development more generally. The most notable difference is that LFP scores for metaphor are somewhat lower than overall LFP rates. An examination of the frequency bands represented in the ‘Beyond 3000’ section revealed that while band 17 was the highest frequency band represented, but with only one word. Band 15 was the highest frequency band with more than one word coded. However, when looking specifically at open-class MRWs, bands 15 and 11 each have a single metaphor, and band 7 is the highest frequency band with more than one word representing it. There were no metaphors from bands 8-10 or 12-14. There is also evidence of a difference according to language background, with the French learners venturing further into the higher frequency bands and producing higher densities.
161
These findings could be taken to suggest that learners may, on the whole, be less confident to use open-class MRWs. However, this could equally be due to the high frequency of metaphor in discourse as discussed in Chapter One: metaphorically-used words occur frequently in discourse to give the high metaphoric densities observed in the literature, so metaphors will often be found in the higher frequency word lists. Furthermore, if a word has the potential to be metaphorically used, it is by definition polysemous. The wordlists cannot make a distinction between senses of a word in LFP analysis, so logically, a word with more senses and thus the potential to be used in multiple contexts will be more frequent than more specialised words with less senses (Crossley et al., 2010).