Clusters in a Japanese FCE essay
4.4 Results and Discussion: Vocabulary Breadth
4.4.3 Lexical frequency profiling: ‘other’ words
4.4.4.1 Lexical errors in complete essays
Figure 4.24 and Table 4.16 below show the raw frequencies of lexical errors in the two groups of learners as they progress through the levels. The error rates for each essay were then calculated by expressing the number of errors per essay as a percentage of the total number of words. The mean for each level was then calculated to give the lexical error rate graph.
170
Figure 4.23: Graph showing lexical error frequencies
Figure 4.24: Graph showing lexical error rates 0
171
Table 4.16 Table showing total lexical errors and average lexical error rates at each level
The first graph shows raw frequencies of error. The increase up to level CAE in both languages is to be expected as the total word counts are also increasing, but the decrease between levels CAE and CPE is perhaps more surprising, especially given the high word count. However, such a finding is in line with the fact that CPE is the most advanced level, and students are expected to write at near-native levels. However, an interesting difference between the language backgrounds can be observed in the error rate graph. While the Japanese error rate decreases steadily as the levels progress, the French error rates follow a slightly more erratic path, with an increase between levels PET and FCE (putting their error rate in excess of the Japanese rate, where it remains up to level CAE) and an even sharper
172
decrease between levels CAE and CPE, by which time it has dropped below the Japanese rate. Such findings run somewhat counter to the conclusions drawn in Littlemore et al.
(2014) that at level FCE, learners are being more creative with their language use, which leads to an increased error rate. This is perhaps the case for the French learners, but no indication of this is observed for the Japanese learners. Littlemore et al. (2014) also found that error rates involving metaphor were higher than those that did not, but interestingly, the French learners’ spike in error rate between levels PET and FCE is accompanied by a drop in metaphoric density (see Chapter 3) while the Japanese learners’ increased metaphoric density was not accompanied by an increased error rate. The differences observed in the two studies could perhaps be explained by the fact that this analysis is taking into account orthographic errors only, but nonetheless, language background does seem to have a significant effect on error rates. It could be that the Japanese learners are more ‘cautious’ at the intermediate levels, although this was not borne out by the LFP analysis: Japanese learners use more band 3 words than the French at KET level, and more
‘beyond 3000’ words at both KET and PET. At level FCE, too, they use more band 3 words, although the French learners use more ‘beyond 3000’ words. It is also possible that their higher error rates in the earlier levels give them a more solid foundation as they progress to the intermediate and advanced stages. However, this is speculative and further research would have to be conducted to reach a firm conclusion.
Again, it is unsurprising to note the high levels of individual variation in lexical error rates.
As above, the graphs show the mean error rate percentages along with the range and quartiles, while the tables include statistics calculated from raw frequencies:
173
Figure 4.25 graphs showing variation in lexical error rates in the writing produced by Japanese and French learners
Raw
frequencies Japanese KET Japanese PET Japanese FCE Japanese CAE Japanese CPE French KET French PET French FCE French CAE French CPE
Mean 0.52 0.90 1.76 2.38 2.00 0.62 0.76 2.29 3.24 1.67
SD 0.68 1.55 2.07 2.04 2.19 1.07 0.83 2.10 3.21 1.39
Minimum 0 0 0 0 0 0 0 0 0 0
Maximum 2 6 8 9 7 4 3 7 10 5
Table 4.17 Table showing statistics relating to raw frequencies of lexical errors
Mean and SDs for %
rates Japanese KET Japanese PET Japanese FCE Japanese CAE Japanese CPE French KET French PET French FCE French CAE French CPE
Mean 1.67 1.17 1.07 0.82 0.63 0.97 0.88 1.14 1.08 0.50
SD 2.22 2.26 1.47 0.71 0.68 1.36 1.04 1.03 1.06 0.45
Minimum 0 0 0 0 0 0 0 0 0 0
Maximum 7.41 9.52 6.35 3.31 2.14 3.60 4.00 3.72 3.39 1.74
Table 4.18 Table showing statistics relating to lexical error rates
These graphs show high levels of variation within each level, with some essays showing relatively high error rates (a maximum of 9.52% in Japanese PET) and others containing no lexical errors. However, it is interesting to note that there is substantially more variation in the learners’ error rates from levels KET-PET, particularly in the Japanese data. These levels
0.00
174
therefore seem to be somewhat more chaotic, reflecting the developmental work the learners are undertaking before stabilising somewhat at FCE level. In dynamic systems terms, such ‘chaotic’ data is particularly interesting as it can be an indicator of points where the system is about to change (Larsen-Freeman and Cameron, 2008b), and it has already been noted that FCE seems to be a threshold between the beginner levels and the more advanced stages.
Finally, it is worth investigating the impact of average LFP scores on error rates, to see if learners are more likely to make more lexical errors if they use words from the higher frequency bands. The correlation was run using the average LFP scores of the corrected essays and the lexical error rates given above. A Spearman’s rho test showed no significant relationship between LFP score and error rate, rs = .089, p = .100.