Method | Style

As an addition to the classical toolkit of philology, the study of style offers certain possibilities.

There is what you say, the content of your message. And there is how you say it. The latter is what we call style. Stylistics has many aspects, but in perhaps its purest form, it consists of the study, not of content words, the message carriers, but of function words or connectives, the words which articulate the content words. It is this top end of the lexicon, the high-frequency connectives or particles, with which we are here concerned.

Mary McCarthy once said, of her enemy Lillian Hellman, Every word she writes is a lie, including "and" and "the."

This is a witty remark, and the wit consists in the fact that "and" and "the" are precisely the sort of high frequency connectives and function words we have been talking about. These words articulate the message, but they do not carry the message. They cannot lie, for the same reason that they cannot tell the truth. They operate below the level at which statements of any kind are made, where judgements of "true" or "false" can be made about those statements.

Bias. Personal judgements of style are open to the charge of bias. And people do differ in their impressions of the style of a text. The advantage of an objective measure is that it has no bias. Anyone who does the counts and runs the arithmetic (by hand, by computer, it does not matter), will get the same result. That result will be interpreted by a human investigator, but the result itself is objective.

Our test, called BIRD (the Brooks Index of Rhetorical Difference), was developed over many years by Bruce and Taeko Brooks, first for Queen Anne's English (with amusing results for some writings of Jonathan Swift), then for literary Chinese. It has recently been adapted by Keith L Yoder for Biblical and Homeric Greek, and for Biblical Hebrew.

Technica. How does it work? Most modern stylistic methods assign a number to each of several texts, and then compare the numbers. The BIRD test instead takes two passages at a time, and considers, for each, the degree of departure of 14 high frequency test words from what is expected on general principles. It then compares the resulting stylistic profiles of the two texts to get D (for Difference), a measure of the degree to which they depart in the same way from that expected norm. If one text zigs where the other zigs, and zags where it zags, then they will have a low D number, and will be considered to be closely similar.

Interpretation. In all languages so far examined, BIRD results define the following areas:

Extreme: D = 1.00 or more. The two passages cannot be from the same author or source.

High: D = 0.75 ~ 0.99. A different author (or source), or the same author in a disturbed mood, like Paul in passages where he defends himself against some criticism.

Normal: D = 0.51 ~ 0.76. The usual thing, the boilerplate of that text. Indeterminate as to author or source.

Low: D = 0.50 or less. Significantly close; often occurs with segments of a continuous narrative, or when one passage has the other "in mind," as with intentional (or sometimes, involuntary) imitations.

An advantage of this test is that, unlike some others, which require samples of 10,000 or more words, BIRD works with much smaller passages, the zone where most questions of scholarly interest are located. As sample size decreases, the test words gradually lose their power of discrimination. The minimum sample size varies with the language in question. For classical Chinese, the lower limit is 218 words; for Biblical Hebrew 88 words; for Biblical Greek 105 words; and for Homeric Greek 90 words. As those lower limits are approached, the risk of false positives increases; below them, the test becomes virtually inoperative, and one continues at one's peril.

Proof. Does it work? And how would we know? We suggest that if the results of a style test tend to agree, over time, with what seem to be competent human opinions, then the test itself gains credibility, and its results may be considered in adjudicating differences of scholarly opinion, on in the preliminary assessment of a less studied text. Results so far suggest that the test results are meaningful - that the test, at least to some extent, is seeing what human investigators have seen, over the centuries. Sample confirmations include:

Homeric Greek. Already in antiquity, Iliad 10, the Doloneia, was suspected as an intrusion. The BIRD test reports that it is so unlike the preceding Iliad 9 (the Embassy to Achilles) as to preclude its coming from the same author, or even the same school of rhapsodes. It is indeed intrusive. The Ancients were right.
Biblical Greek. Origen had suspected that Revelation was wrongly associated with the Johannine group (the Gospel and three Epistles). The BIRD results show that not only is this true, but that Revelation resembles no other text in the NT canon, probably because it is written in an archaizing, Scriptural style. Origen was right.

For further examples, including some from classical China, see the primary publication by Bruce and Taeko Brooks, for the second time (the first was at Leiden University in 2003) formally introducing the test to the scholarly public.

We conclude with deep thanks to Keith Yoder, who has made the BIRD test available for wider applications; and with our best wishes to any investigators who may be disposed to include such results in their own future researches. May our BIRD, like Noah's dove in an earlier day, help to find where the firm ground lies, in the uncertain world outside.