Chaucer's tell-tale lexicon: romancing seinte cecyle - Geoffrey Chaucer

Style, Fall, 1997 by Judith A. Weise

In the second step, I produced confidence intervals of varying probabilities for all five parts of the Prologue and Tale. It would, of course, be easier to use the mean ([Mu]) as a tool in the comparison of the percentages of Romance words in the five parts. However, in Statistics in Language Studies, Anthony Woods, Paul Fletcher and Arthur Hughes explain that a confidence interval is more accurate than a point estimator, e.g., [Mu], which "is only a single number calculated from a sample of data and used to estimate a population parameter," and "is of limited usefulness by itself. It is therefore preferable to provide estimates which take into account explicitly this sampling variability and state a likely range within which the population value may lie. This is the motivation behind the idea of a confidence interval" (96).

Confidence intervals can be calculated at any percentage of confidence or probability, but analyses with the highest possible probability are, of course, the strongest. In an ideal statistical sample with a large amount of data all distributed symmetrically around the mean, various percentages of confidence are defined as follows: in a 68% confidence interval, 68% of all data lies in the interval (or range of numbers) of [Mu]-SD to [Mu] SD; in a 95% confidence interval, 95% of all data lies in the interval [Mu]-2SD to [Mu] 2SD, and in a 99% confidence interval, 99% of all data lies in the interval [Mu]-3SD to [Mu] 3SD. Stated another way, in 68 cases out of 100, or 95 out of 100, or 99 out of 100, the data will fall within the respective intervals. These formulas show why smaller standard deviations reflect data located in a relatively narrow interval (or range) centered around the mean. For a non-ideal statistical sample such as this, the confidence interval also depends on the size of raw data available. The tables in Figures 1-3 show the confidence intervals of percentages of Romance usage in the five parts of the Tale with 90%, 70% or 60% confidence, that is, in nine, seven, or six cases out of ten. In the first part of the Prologue, lines 1-28, for example, in nine cases out of ten the percentages of Romance vocabulary will fall in the interval between 4.5 and 24.2%.

Mersand finds an 11.1 percent Romance usage in the Second Nun's Prologue and Tale (98) and points out the similarity in the number of Romance words per line of the Second Nun's Tale and The Parlement of Foules, an acknowledged early work (97). He does not distinguish, however, between the two parts of the Tale, even though he is aware that they have two different sources (48). Mersand also acknowledges Carleton Brown's conclusion that at least part of the Prologue was written later than the Tale (71), but he ignores the poem's internal divisions. When the percentages of Romance vocabulary in all the parts of the Prologue and the Tale are analyzed separately, as I have done, great differences among the percentages emerge. Figures 1-3 include graphs showing these intervals.