Summary of Possible Positive and Negative Correlations

The possible correlations described above (assuming that A was written first) are summarized here:

As previously mentioned, these results are reversible. In other words, if we had originally assumed that B was written first, and that aA copied and edited from B, then the above relationships would be reversed, as follows:

It should be noted that six of the tests in the two tables above have the same possible results after reversing the initial assumption, indicating that these tests cannot be used to determine directionality. Therefore, by combining the above tables we get the following possible profile correlations that vary according to whether A or B was first:

Implications

The above are key results, since where correlations (either positive or negative) exist, they provide us with a way of determining whether aB used text from A, or aA used text from B. We have two possible ways of testing which of these is more likely:

1.     If A is more homogeneous than B it is likely to be more original than B, and vice versa:

• If p20 and p2N have a more positive correlation than p02 and pN2, then aB is likely to have used A;

• If p22 and p2A have a more positive correlation than p22 and pA2, then aB is likely to have used A;

• If p02 and pN2 have a more positive correlation than p20 and p2N, then aA is likely to have used B;

• If p22 and pA2 have a more positive correlation than p22 and p2A, then aA is likely to have used B.

2.     Checking other correlations involving p22 can indicate whether it originally came from A or B. In particular, if p22:

• has a positive correlation with p20 and does not correlate with p02, then the words in 22 came from A, i.e. A came first;

• has a positive correlation with p02 and does not correlate with p20, then the words in 22 came from B, i.e. B came first.

However, it is possible that the styles of the various categories may not correlate with each other, or perhaps that the correlations are mixed and provide inconclusive results. For example:

• • If the number of words in c21, c22, or c12 is small (e.g. the second author only copied or edited a small number of the first author’s words), then there may not be enough words in these categories to be able to determine any relationships.

• If the second author copied a very ‘eclectic’ selection of the first author’s words, then p22 could be unique.

• c22 could be common to both texts because it is taken from some other source, and not because it was written by either aA or aB.

• If the first text was actually created by editing from earlier sources, then the starting assumption was incorrect, and A or B (whichever is first) may contain a mixture of profiles from other sources. If so, then c20, c22, and c02 may have originally come from different sources, and p22 may show no correlation with either p20 or p02.

• aB may not have edited A directly, but perhaps edited an intermediate text that was itself edited from A.

• If the ‘gestation’ periods of A and B overlapped, then some portions of A may have been influenced by B, while some portions of B may have been influenced by A.

This means that this technique (comparing the profiles of the different categories) may produce an indeterminate result, and in this case we cannot tell whether A or B came first without further analysis.

There are also some possible profile correlations that do not depend on whether A or B was first, and therefore cannot be used to determine directionality, although they may tell us something about the subject matter and the method of editing employed by the second author:

Next: Three Authors