Style, Precept, Personality: A Test Case (Thomas Sprat, 1635-1713)* Robert Cluett For 300 years, the prose style of Thomas Sprat has been admired by devotees of Restoration prose (Evelyn, Bumet, Macaulay), especially by those who would establish a relationship between the "scientific" state of mind and a "plain" style (Glanvill, Mitchell, Spingarn, Jones), and scholars since Spingarn have often tended to regard Sprat's attacks on eloquence et at as part of a conscious and successful attempt to alter English prose style. Despite an occasional dissent (e.g., Williamson, Adolph), there is a widespread impression that Sprat was influential in his practice of a "plain" or "practical" or "unadorned and unmetaphorical" style, eschewing eloquence in favor of that "primitive purity and shortness," that "dose, naked, natural way of speaking" (History) 1 in conformity with dictates of the new science. Piqued by the idea of a man who could follow anyone's precepts in the writing of English prose, I undertook in 1967-68 an extended study of Sprat's style in relation to both the precepts and the intellectual attitudes set forth by him and his chief mentors: Francis Bacon, Thomas Hobbes, and John Wilkins. The study was a traditional one in its assumptions and objectives; yet it employed and developed some of the computer techniques devised by Louis T. Milic for the analysis of style. It employed these techniques not for their methodological value as approach (Milic) nor for their possible incisiveness as attributive instruments (K6ster), but for their economy in surveying a plenitude of examples and for the protection they seemed to offer against the normal biases of any literary critic as he selects and describes the necessarily small portion of an author's corpus that becomes the demonstrative material of a stylistic study. Though I shall mention the larger aspects of the study-the relationship between the new science and prose style and, within one writer, between precept and practice-I shall confine this article largely to aspects of the study that are of moment to readers of this journal. Sprat himself is an interesting historical personage: first historian of the Royal Society, Chaplain-in-ordinary to Charles II, Dean of Westminster, and Bishop of Rochester (1684-1713); he is best remembered by *This article is an account of a larger study. Necessarily, data has had to be compressed. Wherever I have given less than full information, I have indicated with a n asterisk that full information is available on request from the author. 1The following abbreviations are used throughout: History for History o f The Royal Society, Cowley for Account o f The Life and Writings o f Cowley.
Robert Cluett is an assistant professor in the Department o f English, York University. Computers and the Humanities/Vol. 5/No. S/May 1971
257
Cluett: Style, Precept, Personality political historians for his active, expedient, and persistent High Church Toryism and for being the most eminent divine in the English Church to read James II's second Declaration of Indulgence to his parishioners. He is remembered by literary and cultural historians as the putative formulator of the Royal Society's "stylistic ideals" (Jones, Spingarn), and he is mentioned as a simplifying stylistic influence in more than one literary encyclopedia. The literary aspect of his reputation is the consequence of over-reliance on a few passages (Spingarn, vol. 2) from his History of the Royal Society. In fact, he never prescribed a totally unadorned style except in Natural Philosophy, and his stated critical position (Cowley) admitted a wide range of styles according to the traditional classical prose levels, even though he did once seem to favor a "naked, natural, way of Speaking" (History). The main distinctions of Sprat's theoretical position will be found not in any plain-fancy antithesis, but in his career-long attacks on "precepts," abstractions, and poetry, 2 and on all activitiesreligious, political, literary-identified with the Fancy, i.e. with what we would call the imagination. We should not be misled by his attack on "specious Tropes and Figures," among other things (History), into thinking that Sprat disapproved of rhetoric; what he disapproved of was use of the verbal imagination in making statements not empirically or sensibly true, and its use towards purposes other than brief displays for "delight." He explicitly admired eloquence in some of his contemporaries and urged its cultivation upon his clergy (1695), although the deep distrust of words that he shared with many of his co-Baconians, notably Wilkins, was but one aspect of a pragmatic and anti-intellectual cast of mind. Hence there was a conflict between Sprat's rhetorical training, gifts, and tastes, and his distrust of the verbal imagination. This conflict produced a wide range of seemingly contradictory attitudes in his own statements about writing. He could despise metaphor (History, p. 112)in the same volume in which he praised Bacon's style for "comparisons fetch'd out of the way" (History, p. 36); he could despise "Ornaments of Speaking" (History, p. 111) soon after suggesting that what the English language needs is a dose of "Wit, and Variety, and Elegance of Language" (History, p. 41). In addition, both the conflict and Sprat's attempts at resolution were manifested in a number of areas of distinctive synthesis in his style. Similitude was one such area: to be eloquent, Sprat-both apologist and preacher-admittedly needed comparison; yet, to speak truth as he saw it, he had to avoid trope. A catalog of Sprat's elaborated analogies in prose from 1659 to 1682" reveals a consistent compromise between these apparently conflicting standards. Metaphors and other comparisons abound, especially in peroration; yet they are anti-fanciful (Adolph, p. 210 ff., gives numerous examples), being repetitive of things stated proposi2In poetry Sprat preferred the Cowleyan Pindaric stanza above all others "for its near affinity w i t h prose, which is the style o f all business and conversation."
(Cowley)
258
Computers and the Humanities/Vol. S/No. 5/May 1971
Cluett: Style, Precept, Personality tionally elsewhere, and they are carefully limited and explained by the author. A similar synthesis was revealed by manual counts of doublets and triplets 3. in Sprat's prose: where eloquence was called for, especially in sermons, doublets and triplets are numerous (280 per 10,000 words in sermons, vs. 260 for Burnet, 220 for Tillotson; 200 per 10,000 in his "apologetic" prose, vs. 180 for Hobbes, 160 for Glanvill, 155 for Dryden). Here again, we have a synthesis: a device of eloquence that does not violate empirical, sensible truth. The competing claims of truth and eloquence-together with those of "Art" and an "easie," "natural way of speaking"-were markedly revealed in the computerized portion of my examination of Sprat's style. The techniques of this portion of the study were essentially those of Milic, with modifications as described below. These techniques were consistently and deliberately subordinated to my own critical intuition about Sprat as a stylist. The "D" (Number of different 3-Class patterns in a sample) and 'T' (Number of unique 3-Class patterns in a sample) values used by Mille and K6ster seemed less important in Sprat's case than did other features of his prose: notably, his apparent fondness for frontal parallelism and other kinds of repetition, his apparent dependence on syndeton, his linear techniques (i.e. non-subordinating, non-embedding) of elaborating a sentence which was, by standards of the period, about 20 percent shorter than average (cf. Milic), and his preference for systems of modification over those involving finite predication. While not every test revealed findings that seemed new and worthwhile, several did, and these are reported below. An interesting by-product of the study was a surprising compromise between the "genre hypothesis" that a writer (notably Sprat) will alter his style to his occasions, adopted by me, and the "authorship hypothesis," adopted by Mille (and later by K6ster), that an experienced writer will show distinctive consistency in his solutions to stylistic problems. Modifications to Milic's techniques were as follows. In his version of the Fries word-class system, we removed the infinitive signal (to) from the pattern-maker class, 61, to class 62 by itself, and we isolated as class 63 the contractions 'tis, ?was, "twould, and their permutations. This list is given below:
Parts of Speech 01 02 03 04
Noun Verb Descriptive Adjective Descriptive Adverb
05 06 07 08
Infinitive Participle Gerund Miscellaneous substantive
3Doublet: a word-pair modified by nothing other than single-word modifiers; internal modifying phrases disqualify a pair. For example, "strength and quickness of invention" would be counted; "strength of invention and quickness of w i t " would n o t be counted. "Bathetic depths and amorous heights" is counted; "depths of bathos and heights.. 2' is n o t . Triplet: a word trio with the same disqualifying criteria.
Computers and the Humanities/Vol. 5/No. 5/May 1971
259
Cluett: Style, Precept, Personality Funclion Words 11 Pronoun 21 Auxiliary Verb (plus be) 31 Determiner, limiting adjective 32 Postposition 33 Intensifier (plus not) 34 Function adverb (plus only) 41 Coordinator 42 Subordinator 43 Relative
44 45 51 61 62 63 71
Interrogative Correlative True Preposition Pattern marker (there, it) Infinitive signal (to) Contraction Interjection
81 Numeral 91 Sentence Connector
We used two kinds of sample: eleven samples of text of full sentences (approximately 95 encoded sentences per sample), and 21 samples of sentence openings (two encoded words per sentence), with a minimum of 150 consecutive sentences per sample. Full text samples were not confined to those selected by random number table (Hoel); we also used stratified samples (every 4th sentence, every 9th page, etc.) and contn'ved samples picked from obvious "high style" (peroratio) and "low style" (narratio) areas of the History of the Royal Society, these to test the hypothesis of Sprat's wilful variation of his style. An average sample size of full text of 3,300 words and 96 IBM cards was close to Milic's (3,500 words, 100 IBM cards). We used eight Sprat samples and three from control writers as listed below. As a measure of the internal homogeneity of each sample, the coefficient of determination, r 2, was computed by Pearson's formula for each encoded sample for the 50 or so odd cards vs. the 50 or so even cards. PREFIX
AUTHOR
TITLE//SAMPLING TECHNIQUE//SIZE IN WORDS
0101 Sprat
deSorbikre's Voyage (London, 1665); regular recurrence
0102 Sprat
History of the Royal Society (St. Louis, 1958); random
260
(every 9th page); 2,628 words, r 2 = 0.990. This sample, with 0103, 0207, and 0410, fell short of Milic's optimum of 3,500, but because 0101 and 0410 had high internal reliability (split half), we did not cut, Xerox, and translate further samples from these books. Cowley always produces a short stratified sample, since it is only 9,800 words total. 0207 was short because the two perorations involved were only about 1,500 words apiece. sample (Hoel) (pages 9, 11, 19, 58, 68, 102, 111,116, 328, 359, 395, 396, 418. Pages 133-320 were omitted because most of them were not written by Sprat); 3,498 words, r 2 = 0.996. Computers and the Humanities/Vol. S/No. 5/May 1971
Cluett: Style, Precept, Personality 0103 Sprat
Account of the Life and Writings of Cowley (New York,
0104 Sprat
Sermons, 1676-1682 (London, 1722); regular recurrence
0105 Sprat
Visitation Discourse, 1695 (London, 1696); regular recurrence (every 4th sentence); 3,542 words, r: = 0.988. "Oral" sample.
0109 Sprat
Sermons, 1690-1694 (London, 1722); regular recurrence
0206 Sprat
Ice,tory of the Royal Society; contrived low style sample
0207 Sprat
History of the Royal Society; contrived high style sample
0410 Glanvill
Plus Ultra (1668 edition); random sample (pages 4, 19, 36,
0411 Hobbes
Leviathan (Meridian edition, Plamenatz ed.); random sample
0412 Dryden
Essays (Bredvold's The Best of Dryden); regular recurrence
1907); regular recurrence (every 3rd sentence); 3,109 words. r 2 = 0.992. (every 10th page in 1722 edition, skipping blanks, title pages, dedications, etc.); 3,606 words, r ~ = 0.988. "Oral" sample.
(every 5th sentence in 1722 edition); 3,524 words, r 2 = 0.990. "Oral" sample. (pages 5-1 ] and 53-59); 3,499 words, r 2 = 0.986. (pages 46-51 and 413-417), 3,2,57 words, r 2 = 0.984. 40, 47, 51, 62, 72, 78, 85, 95, 98, 119, 133, 144); 3,096 words, r 2 = 0.982. (pages 61, 69, 71, 80, 88, 175, 241,279, 305, 326); 3,493 words, r 2 = 0.992. (first sentence each page, 431-512); 3,516 words, r 2 =0.988.
These samples were each translated twice from Xeroxed text into the Fries-Milic code (once by me, once by a research assistant), compared and reconciled, punched onto cards, and computer-analyzed on the York University computer (IBM 360/40, OS, memory 256k) in accordance with programs devised by me and written in FORTRAN IV for the G compiler by Lt. Col. G. J. Carpenter, Canadian Defense Forces. These programs called for word-class distributions and for certain linked prepositional phrase strings. In addition we encoded sequentially the first two words of each sentence in the following passages: pages 1-128 and 314-428 of the History of the Royal Society, all sentences of Sprat's Cowley, and all sentences in Sprat's first four published sermons (1676-1682, the same as comprise full text sample 0104). As controls we used four overlapping 200-sentence samples from Plus Ultra (I-200,101-300, etc., to the 500th sentence, some 35 sentences from the end of the book), six overlapping Computers and the Humanities/Vol. 5/No. 5/May 1971
261
Cluett: Style, Precept, Personality 200-sentence samples from Dryden's essays (samples taken as for Glanvill, starting at top of page 431 in the The Best o f Dryden and continuing for 700 sentences), and all the sentences in three Tillotson and two Bumet sermons. 4 An analysis of all these two-word clusters according to a computerized frequency distribution* revealed several facts about the samples. Burnet had a strong predisposition for broad-reference pronouns, his most frequent opening being 11 21 (16.7% of the sentences samples); Tillotson showed a pronounced tendency to open with coordinators, as class 41 showed in four of his eight most frequent openings, totalling 24.3% of the sentences in his sample; Glanvill showed as the least various of the writers; the samples of Dryden and Sprat, especially Sprat, were distinguished by the authors' apparent reliance on frontal pattern-markers; Dryden was the only author to open a considerable number (1.4%) of his sentences with infinitives; Sprat was the most abundant of the five in sentences opening with adjective constructions (three of his ten most frequent openings). The frequency distribution enabled us to see also the amount of initial syndeton (classes 41, 42, 91) in each sample. The five writers all having been born within a decade, all being subject to similar educations, and all having had extensive exposure to the homiletic training and practice of the day, there was an expected overlap in their percentages of initial syndeton, the range of occurrence of which is charted below in Table 1. Table 1. Range of Initial Syndeton as Percent of Sentences
Name GlanviU (4 samples) Tillotson (3 samples) Dryden (6 samples) Sprat (6 samples) Burnett (2 samples)
Lowest %
Highest %
33.5 39.1 21.0 30.5 17.9
55.0 52.9 43.0 35.5 35.7
We thought it possible that more accurate registration of apparent effect might be achieved if we computed initial syndeton as a percentage of whole text rather than as a percentage of sentences (e.g., a South sermon, with 60-word sentences, seems less syndetic with 50% initial connection than does a Sprat sermon, in 37-word sentences, with 35%). 4Burnet, Sermon on the Fast-Day, March 12, 1690, and Sermon f o r The Protestants o f Orange, n.d., London, 1704; Tillotson, sermons preached and published in London, November 5, 1678, April 4, 1679, and April 2, 1680. Though t h e s e men were n o t c o n t r o l writers in the computer study described below, they were included and Hobbes was dropped for this study for two reasons: they o f f e r e d generic controls for Sprat's sermons, and the distinctiveness of Hobbes's sentence openings (vs. Glanvifl, Dryden, Sprat) is so o b v i o u s t o t h e naked eye as to make a study of them useful only for statistical purposes. The overlap in the Glanvfll and Dryden passages, introduced to give a number o f samples over 150 sentences in length from the material sampled in computer samples 0410 and 0412 (below), was eliminated for purposes o f t h e frequency count. All references to sermons are by author and year: "Burnet (1690)."
262
Computers and the Humanities/Vol. 5/No. 5/May 1971
Cluett: Style, Precept, Personality Such figures were computed (as ICW, initial connectors as a percentage of total text). While they offered no radical reordering of Sprat and the four controls in Table 1, they were comparable to initial repetition as a percentage of total text, a statistic suggested by Sprat's obvious reliance on anaphora and other kinds of initial repetition. The figure for initial repetition (RW) was derived by a manual count of the following clusters of encoded sentence openings in sequence: ABAA, AABA, A B B A , AA, A A A , AAAA, plus any other group of four or more sentences in which the sentence openings fell into only two different two-class patterns. The configuration AA (e.g. 61 21//61 21 or 11 02//11 02 was counted as two repeated patterns; AAABBA was counted as six. The RW figure was derived by taking the total number of words in the repeated patterns and dividing that figure by the number of words in the sample. The results are below, in Table 2. Table 2. RW and ICW: Repetition and Syndeton as % of Text
Name GlanviU Glanvill Glanvill Glanvill Sprat Sprat Sprat Sprat Sprat Sprat
1 2 3 4
1 (Cowley) 2 (History) 3(S1676) 4(S1678) 5(S1678) 6(S1682)
RW
ICW
.415 .847 .682 .976
1.151 .767 .758 .898
1.218 1.069 1.667 1.153 1.333 1.057
1.156 .848 .971 .849 1.061 .908
Name
RW
ICW
.669 .751 .676 .400 .601 .766
.876 .909 .707 .610 .436 .473
Tillotson (1678) .377 Tillotson (1679) .188 Tillotson(1680) .536
.981 1.377 1.342
Dryden Dryden Dryden Dryden Dryden Dryden
1 2 3 4 5 6
Burnet (1690) Burnet (1704)
.535 .793
.667 .383
The differences above, together with the general consistency of each writer, can be better seen in graphic form (Table 3). Tables 2 and 3 suggest two things: first, that each of the five writers was consistent in his habits of opening sentences; second, that Sprat and Tillotson were each entirely distinctive. Tillotson's syndetic and antirepetitious technique can be seen in any of his sermons. Sprat's techniques, both highly repetitious and highly syndetic, can be seen in any extensive sample of his work, but especially in his History of the Royal Society, pages 49-50 and 149-150, and in his Sermons (1722), pages 76-78. The anaphora-syndeton combination for Sprat was a compromise among several values: syndeton for coherence and ease of communication-important to all Wilkensian preachers; anaphora for both coherence and a high eloquence that avoided sinking into the "specious Tropes" and flights of Fancy that Sprat so despised. Computers and the Humanities/Vol. S/No. 5/May 1971
263
Cluett: Style, Precept, Personality Table 3. Syndeton (ICW) vs. Anaphora (RW) in Graph Form 1.5%
1679
~
1680
A
1.o% z O j-
\,Y"
.~
I
~
f G 4 h--~----"~
S2 S4
iii
z >. o3
D4
en
.s%
D5 \ 1704
o
I
.s%
I
1.o% ANAPHORA (RW)
I
1.5%
The larger computer study, that of the full text, which assessed full word-class distributions (with standard deviations) and certain phrasal strings, revealed other areas where Sprat seemed distinctive not only to the literary eye but in the numerical test. In these areas we shall first consider the evident stability of Sprat's style and second its prominent features. I should note here that in both the selection of Sprat samples (different sampling techniques, deliberate quest for variance) and the selection of controls (all three identified with the seventeenth-century reform of prose, none so obviously distinctive as Gibbon), the study was designed for maximum chance to refute the Milic hypothesis that a writer's style is consistent and that it will show greater numerical homogeneity than a group of outside controls. Table 4 presents the word-class distributions and Table 5 presents the standard deviations and the coefficients of variation (CV = 100 x S/X, where S = Standard Deviation from Mean and X = Mean). As can be noted from Table 5, the style of the Sprat samples is surprisingly uniform in view of the method of selection. The uniformity is suggested by the Standard Deviations and Coefficients of Variation. Sprat's SDs are lower than those of the controls with a very few exceptions (Class 02 oral, 04 oral and consolidated, 05 written, 11 written, 34 written and consolidated, 43 oral and consolidated, 45 written). Most of the unusual cases where a group of Sprat samples is less homogeneous 264
Computers and the Humanities/Vol. 5/No. 5/May 1971
Cluett: Style, Precept, Personality than those of the controls can be explained. Sprat's mean figure for class 02 (verb) is below the mean for the controls; the larger standard deviation produced by the oral decks for class 02 is the result of distortion of his style in a direction in which it already was distorted. Class 04 is a small class, subject to both strong conscious control s and high variability; it was not large enough in Milic's study of Swift to show in every case as part of a normal distribution. The wide dispersion of class 05 (written) cannot be explained o n literary grounds; perhaps the small size of the class and the small number of samples accounts for the dispersion. The large swing in class 11 is produced entirely by the peculiar sample from deSorbibre's Voyage, in which the pronoun he is a constantly repeated key word (in Cowley Sprat somehow managed the problem more deftly). The variability in class 34 is parallel to a phenomenon that Milic found, which he attributes to the fact that "34 is more miscellaneous than the other classes." Large variance in class 43 is attributable entirely to the Visitation Discourse, 1695, which was Sprat's last work, delivered and published under very different circumstances than either his sermons or his other prose. 6 In the .use of class 45 (correlatives) there is even more conscious control than there is in the use of 04 (adverbs), and this class, like 04, did not show as part of a normal distribution in Milic's study. Despite the use of eight Sprat samples (vs. only three control samples, all of which were from written discourse) and despite the biases of the sampling technique, these ten exceptions (out of 208 individual word-class samples and 64 standard deviations computed for Sprat_) were the only cases in which the Sprat samples were less homogeneous than those of the controls. And nine of these ten, as we have seen, are easily accounted for. Despite whatever lack of precision may exist in a standard deviation test, 7 the data in Table 5 demonstrate strong consistencies in all of Sprat's writing. Within this stability we can infer some stable habits of sentence construction when we add to Table 4 the information from Table 3, plus information derived from manual analysis of subject-verb placement and separation in the first independent clause of 581 Sprat sentences and of 647 sentences from the three control writers,* presented as Table 6. SUnlike a subordinator (42) or relative (43), an added adverb does not change the entire cast of a sentence. 6Sprat's homiletic prose was delivered aloud to churchgoers, taken down by a stenographer, then published within ten days; his other prose was fully written out by him. The Visitation Discourse was delivered aloud to clergymen only, in the fall of 1695, but was not published until the following winter, 1696. Though the word-class distributions are very similar to those of the late sermons (0104), there are other aspects of the Discourse (sentence length and punctuation) which suggest an alien editorial hand. There is no extant MS. 7A detective seeking Sprat's stylistic fingerprint could probably find it through multiple regression or multiple discriminant analysis, but a fingerprint was not the object of the present study; what was wanted was quantitative demonstration of qualities visible to the eye of the literary reader. Computers and the Humanities/Vol. S/No. 5/May 1971
265
Ctuett: Style, Precept, Personality
de Seth 1665
Hist 1667
Cowley 1667
Sermons 1676-82
ViD 1695
Fhst Low
Fgist High
Sermons 1690-94
Glan Plus Ultra
Hobbes Leviath
D~yden Essays
0101
0102
0103
0104
0105
0206
0207
0109
0410
0411
0412
Noun
01
18.67
19.35
19.66
20.52
18.04
21.13
17.63
18.79
21.87
21.51
17.75
Verb
02
7.44
7.09
6.60
6.49
5.03
6.35
7.08
5.51
7.62
6.39
7.~8
Adj
03
7.44
8.00
7.18
9.93
11.07
8.72
8,15
8.34
8.79
4.92
6,77
Adverb
04
.88
1.17
1.19
1.47
2.57
.74
2.26
1.70
1.23
1.00
1.28
lnfin.
05
1.47
2.20
1.42
1.80
1.47
2.12
1.96
1.90
1.20
1.41
1.99
Partielple
06
.56
.69
1.13
.63
1.30
.83
.75
1.56
.58
1.65
1.25
Gerund
07
.48
.77
.58
.60
1.16
.92
.72
1.33
.68
.53
.54
M1sc,
08
.36
.03
Pronoun
ii
9.99
6.83
Auxiliary
21
7.68
8.69
6.95
6.27
6.64
6.09
7.98
5.93
6.10
7.31
8.25
Determiner
31
15.25
13.32
15.38
14.59
13.33
14.41
13.08
15.04
14.70
15.14
12.63
.33 7.92
6.85
6.30
6.21
9.06
7.63
1.00
-
6.27
6.75
.54 8.99
Postposition
32
.44
.57
.42
.17
.37
.66
.26
.45
.23
.32
.51
lntensltier
33
1.83
2.09
1.83
2.47
2.71
1.23
2.65
2.24
1.81
1.53
2.67
1.87
2.03
2.22
1.41
2.51
3.15
1.86
2.16
1.03
1.71
1.99
Function 34 Adverb Coordmator
41
3.54
4.60
4.86
4.44
6.21
5.09
5.20
5.39
6.46
5.83
3.98
Suborth- 42 hater
2.43
2.40
1.93
2.27
2.68
2.12
2.42
2.10
1.65
2.30
2.96
Relative
43
1.67
2.12
1.93
1.80
.99
1.74
2.03
1.56
2.16
2.68
2.36
44
.48
.26
.03
.22
.03
.06
.29
.37
.06
.06
.06
Correl- t~5 atlve
.52
1,63
1.13
.78
i.i0
1.23
1.37
i.ii
.84
1.71
1.73 12.09
Int
True Prep.
51
12.70
12.49
13.61
13.73
13.04
13.38
11.87
13.34
13.60
13.55
P.M
61
.92
.62
.68
.42
.54
.63
.85
.37
.19
.91
.85
Infin. S*g.
62
1.59
2.23
2.06
1.80
1.81
2.06
1.90
1.87
1.07
1.59
2.10
Contr.
63
.04
.03
.03
.14
-
.09
.07
.06
.17
Interjee-7] tlon
.76
.II
.13
.14
.17
.i0
.17
.06
,06
.09
Number 81 Senten~ Conn. 91
.17
.64
.20
.45
.22
.ii
.34
.07
.09
.26
.15
.63
.36
.46
.71
.78
,85
.71
.36
.99
.39
.97
.54
Me&tiers(M) 12.02
13.29
12.42
15.28
18.86
13.84
14.92
14.44
12.86
9.16
12.71
M f f i 0 3 + 0 4 + 3 3 + 34
Table 4. Class Percentages per Sample
266
Computers and the Humanities/VoL 5/No. 5/May 1971
9
0
<
C
0
0
<
0
~9
m m.
0
E
1.96
51
62
.332
.595
.556
.226
.280
.899
.175
.150
1.153
.874
1.603
.437
.174
.420
.422
16.94
4.60
51.01
11.83
12.44
16.14
8.57
7.81
7.87
11)24
19.43
25.71
16.11
5.57
5.99
1.83
13.37
1.00
1.45
2.35
5.33
2.03
2.47
14.32
6.28
6.93
1.72
1.91
9.78
5.68
19.11
.038
.346
.188
.416
.298
.886
.562
.235
.886
.335
.668
.228
.580
1.372
.744
1.271
s
CV 6.65
2.08
2.59
18.80
28.69
12.68
16.56
27.68
9.51
6.19
5,33
9.64
13.26
30.37
14.03
13.10
3 Samples. Sprat oral (0104. 0105,0109)
CV = 190 x S/X where S = S t a n d a r d Dev~atlon fro m Mean = Mean
1.09
12.93
45
2.28
1.91
4.33
41
43
2.04
34
42
1.92
33
7.77
14.65
21
1.70
05
31
1.08
04
8.25
7.54
03
ii
7.04
02
U 01
3 Samples, Sprat Written (0101,0102,0103) x s cv 19.23 .516 2,68
1.89
13.15
1.05
1.68
2.30
4.84
2.03
2.19
14.48
7.02
7.59
1.71
1.49
8.66
6.36
.223
.497
.375
.390
.264
.904
,372
.352
.938
1.011
1.316
.309
.596
1.525
.924
6 Samples, Sprat written and oral (0!01-0105,0109) X S 19.17 .868
11.80
3.78
35.71
23.21
11.48
18.68
18.33
16.07
6.47
14.40
17.34
18.07
40.00
17.61
14.53
CV 4.53
1.92
13.02
1.11
1.73
2.29
4,92
2,15
2,13
14.30
7.03
7.60
1.79
1.50
8.60
6.45
.197
.631
.340
.352
.238
.778
.515
.496
.934
.994
1.348
.307
.647
1,301
.821
10.26
4.85
30.63
20.35
10.39
15,81
23.95
23.29
6.53
14.14
17.74
17.15
43.13
15.13
12.73
8 Samples. Sprat (all Ol . . . . all 02 . . . ) S CV 19.22 1.191 6.10
1.59
13.08
1.43
2.40
2.30
5.42
1.58
2.00
14.16
7.22
7.34
1.53
1.17
6.83
7.1
20.38
.515
.858
.508
.262
.655
1.289
.494
.594
1.340
1.078
1.452
.409
.149
1.935
.635
2.282
s
32.39
6.56
35,52
10,92
28.48
23.78
31.27
29.70
9,46
14.93
19.78
26.73
12.73
28.33
8.94
cv 11.20
3 Samples, Controls (all 04 . . . )
.46
1.32
N.F.
.54
.57
.82
.41
.75
1.98
l.lO
2.40
.47
N.F.
1.30
1.14
8 Samples Mlllc' s Cntrls. S 2.6
q
q
Cluett." Style, Precept, Personality Table 6. Subject Hacement 4
5
I
6
I
7
I
8
l
9
I
10
I
11
I
12
|
I
Sprat=(581 sentences) I Drydent(267 Sentences) ,,Glanvillt (230 Sentences) i Hobbes ( 150 Sentences) i Verb Placement 5
6
I'
7
I
8
I
,
9
I
10
I
I1
I
12
I
13
~
I
14 I
15
I
Sprat, Dryden~
L
Hobbes,
|,
Subject- Verb Separation 2
3
!
4 t
!
, i
Sprat, ~,Dryden, ~Gvill=
I
5 I
I
6 l
Hobbes,
I
Expletives & Inversions 2 89
5
I
71/2
10
I
I
I
15
20
I
Sprat,
I
1
I Dryden ~ ~Gvill~
I
Hobbesj
4
$
Late Subjects 2%
= ,
~
5
10
I
"1"
15 I
20 I
Sprat~ Dryden D iGvill I
Hobbesp
Sentence Length in Words, Ave. per 100 Sentences 25
35
I
45
I
i
Dryde~ ,
268
I
I
Spratl
Mean Number is designated by arrow
65
55
I
/~Glanvill, 9
4
Hobbes
Computers and the Humanities/Vol. 5/No. 5/May 1971
Cluett: Style, Precept, Personality In Table 6, Sprat, though not conclusively differentiated from his colleagues Glanvill and Dryden in all respects, shows clear tendencies, especially towards shorter sentences, towards frequent use of expletive and inverted constructions 8 in main clauses, and away from late disclosure of subject (16th word or later). Once Sprat was into his sentence, past the quickly disclosed first subject and first verb, he tended towards linear and non-clausal modes of development. Such an inference is suggested not only by intuition but by the relatively small number of classes 01 (noun) and 02 (verb) in all the Sprat samples in Table 4. The inference is reinforced by a consideration of his subordination, his coordination, his use of doublets and triplets, and his use of certain phrasal strings. Of these, the first three can be extruded from the word-class distributions in Table 4. Density of subordination can be approximated simply by the addition of classes 42 and 43, as has been done in Table 7. Table 7. Subordination I
CONTROLS
"!I
0410
0411
0412
Glanvill
Hobbes
Dryden
2.30 2.68 4.98
2.96 2.36 5.32
Subordinators(42) 1.65 Relatives(43) 2.16 TOTAL 3.81
.....
SPRAT9
I
0102
0103 0104
2.40 2.12 4.52
1.93 2.17 1.93 1.80 3.86 3.97
The Table shows Sprat a sparse subordinator, together with Glanvill, and shows Dryden and Hobbes relatively heavy subordinators (1% or so may seem small, but it amounts to a subordinate clause every other sentence, an amount detectable even to the naked eye of a casual reader). The assessment of coordination is not so easy. However, one aspect of coordination of clauses and long phrases can be derived by consideration of the total number of coordinators (41) in each of the six samples in Table 7, and subtracting from the total the coordinators used for doublets and triplets. This is done in Table 8.
Sin this test there were 7 Sprat samples, 8 control samples. The inverted and p a t t e r n - m a r k e d const ructions in the Sprat samples were, per l e o sentences: 20, 17, 15, 14, 14, 14, 6; the range on the graph is deceptively large: there is a me a n of 14.28, a Stan dard Deviation of 3.95. 9The three Sprat samples were chosen as the " c o r e " of the stylistic argument because nearly all that has been w r i t t e n in the t w e n t i e t h c e nt ury a b o u t the " s i m p l i c i t y " of Sprat's style and its putative influence has rested on Cowley, History, and the early sermons. C o m p u t e r s and the Humanities/Vol. S/No. 5/May 1971
269
Cluett: Style, Precept, Personality Table 8. Coordination
j
Controls ,f - 0410 0411 0412 0102 Glanvill Hobbes Dryden
Sprat 0103
0104
Total Coordinators
200
198
140
161
151
160
Coordinators as %
6,46
5.83
3.98
4.60
4.86
4.44
2-lets & 3-lets
52
58
54
78
66
103
148
140
86
83
85
57
4.93%
4.10%
2.44%
2.33%
2.63%
1.55%
NET (coordinators for long phrases & clauses) NET as %
The table suggests that Sprat, though the most fecund seriator of the four in words and brief phrases, is a sparse coordinator of larger sentence components. Tables 7 and 8 together give each man a distinct pattern: Hobbes, highly dependent on clausal expression 1~ abundant in both subordinated and coordinated clausal units; Dryden, eschewing coordination of all kinds, yet using a highly subordinated sentence architecture; Glanvill, also dependent on clausal expression, but setting out the clauses in an amorphous coordinated string; Sprat, suppressing clauses of both kinds, developing his sentences in phrasal patterns that tend towards distribution and modification rather than finite predication. A biaxial treatment of Tables 7 and 8 together is revealing; it is below as Table 9. Sprat's tendencies towards distribution and modification cannot be exhaustively plumbed within the limits of the Fries-Milic coding system, but two tests based on that system are highly suggestive. The first of these is a test for strung prepositional phrases. Colonel Carpenter and I devised a program to extract the following phrase patterns from the six samples in Tables 7 and 8. A 510151 - in houses at E 51030151 - for young men in B 510751 - in going to F 51070151 - by throwing balls to C 511151 - to those of G 5131030151 - at the big tree on D 51310151-onthetopof H 5 1 3 1 0 7 5 1 - in the going to 10Hobbes, s main verb total (02) is low, but he did a great deal of predicating with be as a main verb, and in this grammar be is indistinguishable from the auxiliaries; the subordination figures (classes 42 and 43) plus those for correlatives (45) verify what the naked eye perceives: a clausally elaborated style. This no doubt refieets a weakness in the grammar itself, which could be extensively enlarged without any damage to its integrity. The author is currently experimenting with an enlarged form (CHum 5, ii (Nov. 1970): DSA, L398).
270
Computers and the Humanities/Vol. 5/No. 5/May 1971
67uett: Style, Precept, Personality I 5101410151 - with boys and girls at J 51310341030151 - in the pink and green garden of K 51071151 - in teaching them at L 5131070151 - in the worshipping gods of The major limitation of this test-its inability to discriminate between a
string like Sprat's "'of his life to his curiosity," picked up by 51310151, and the antepenultimate members of a parallel series like Bacon's for delight, for ornament, picked up by 510151-was deemed an inconsequential factor in these four writers, even though it no doubt would have a distorting effect in examination of a phrasal seriator like Bacon, Donne, or Jeremy Taylor. The string totals for six samples are given in Table 10. Table 9. Subordination and Coordination
Q
z O
Dryden 0412
5
Q
.
Hobbes 0411
0102-S
z
0 110
m4
)104-S
0103-S
Glanvill Q 0410
I
3
,,,
t,,
4
~L
I
5
6
COORDI NATION
In view of the shortness of the Sprat sentence (Table 6) and in view of his heavy use of pairs and triplets (Table 8), his fecundity in strung phrases must be considered remarkable. We should also note, as we did in his sentence openings, his marked propensity for adjectival constructions (Patterns E, G, and J), a propensity we shall deal with after considering the information from Tables 7, 8, and 10 all together as Table 11. The obvious Sprat pattern requires no reiteration. Perhaps the most revealing test of Sprat's propensity for modification can be constructed from other information already disclosed in Table 4, the word-class distributions. Table 12, below, takes all of my samples Computers and the Humanities/Vol. 5/No. 5/May 1971
271
Cluett: Style, Precept, Personality Table 10. Strings per 3,500 words (normalized) -Controls
A B C D E F G H I J K L
,
0410 0411 Glanvill Hobbes 20 28 0 0 8 10 50 60 5 2 1 0 13 4 0 2 0 2 0 0 1 1 0 0 98 109
510151 510751 511151 51310151 51030151 51070151 5131030151 51310751 5101410151 51310341030151 51071151 5131070151 TOTAL
,--Sprat-
0412 Dryden 15 3 7 37 2 0 10 1 4 0 1 0 80
0102
,
0103 0104
15 24 3 1 4 5 49 47 2 3 1 0 21 24 0 0 2 0 1 0 1 1 1 0 101 106
28 1 13 57 6 1 28 0 5 0 0 0 139
and all of Milic's and measures the use of main, predicative words (nouns, verbs, auxiliaries: 01, 02, 21 respectively) in an inverse relationship with modifying words (adjectives 03, adverbs 04, intensifiers 33, and function adverbs 3 4 - t h e words totalled as "M" in Table 4). Table 11. Subordination, Coordination, Strings SUBORDINATION
3.5
4.0
4.5
:
CONTROLS
0410
.
NET COORDINATION
5.0 ,
,a
:
0411 0412
,SPRAT
0103 0104
:
a
2.5
3.5
r
CONTROLS.
0412-D
i SPRATI
0102
60
70 80
90
CONTROLS
SPRAT,
0104:
272
0102 '0103
:
0411-H 041043
I
100
STRINGS 80
100
I CONTROLS
0410 0412 0411
~.;5
0104 0102 0103
DOUBLETS & TRIPLETS 50
4.5
0412
0410 i
0102
120
i
0411 I SPRAT Old 0103
Computers and the Humanities/Vol. 5/No. 5/May 1971
67uett: Style, Precept, Personality Table 12. Main Words and Modifying Words
29 A
31 .ca X
<
N 0
33
G-
Hobbes Glanvill Dryden
J-
Johnson
GiM-
Gibbon Macaulay
H-
:~ 35
|
Addison A
g 37
0
z
Gi Sprat 1 - Written Prose 0102, 0103 0206, 0207
r 39 r162 0 z <
Sprat 2 - Oral Prose
Gi
41
0104, 0105 0109
43
I
6
8
I
I
I
I
I
10
12
14
16
18
I
20
MODIFYING WORDS: Adjectives (03), Adverbs (04), Intensifiers (33), Function Adverbs (34)
The pattern of Sprat's distribution of meaning away from nouns and verbs and into modifiers is given convincing demonstration by the table. A final, quantitative question relating to Sprat is that of stylistic level: Could he alter his style to suit his occasions? Despite the homogeneity revealed by the standard deviation tests (Table 5), Sprat did bend his basic style, which consisted primarily of repetition (both structural and informational), pattern-marking, distribution, modification, and clausal suppression. The bending occurred along two axes, most properly termed high-low and oral-written. Let us briefly consider some of the variations in Table 4 among the following samples: Cowley 0103 (low written), History 0102 (middle-high written), History 0206 (middle written), History 0207 (high written), and early Sermons 0104 (oral). The written-oral transformation in Sprat involved mainly an intensification of his ordinary tendencies: more adjectives (from 7.5% to 9.8%); a substantial increase in total modifiers (from an "M" quotient of 12.8 to one of 15.3); a sharp reduction in verb forms of all kinds (verbs, Computers and the Humanities/Vol. 5/No. 5/May 1971
273
Cluett: Style, Precept, Personality auxiliaries, and verbals drop from over 18% to under 16%); a greater-thannormal suppression of clauses (see Tables 7, 8, and 11); a pronounced extension of the tendency to string phrases (Table 10). All of these variations are corroborated by a comparison of the other oral samples 11 with the written samples. When Sprat rose in the pulpit to speak, he spoke just as he wrote, only more so. The low-high transformation was as much rhetorical as purely linguistic, but it is still detectable from the purely linguistic data of Table 4. The concordance of Sprat's similitudes* revealed in peroration roughly ten times the density of comparison that there was in narrative passages. The figures in Table 4 for interrogatives, class 44, in 0206 and 0207 show an analogous increase in the use of rhetorical question: though the difference looks small as 0.06 vs. 0.29, it seems quite large as two interrogatives vs. 10, even more so in a manual count that reveals two interrogative sentences in all of 0206 and 13 in 0207, there being 90 + sentences in each sample. A similar difference exists between Cowley (0103) and the History (0102). Another low-high change was a tendency towards more elaboration of sentence and a diminution of Sprat's tendency to suppress clauses. From low (0206) to high (0207) subordinating words (42, 43) increase from 3.85% to 4.45%, and main verbs and auxiliaries increase from 12.44% to 15.06%. From Cowley 0103 to the History 0102 the pattern is parallel, with subordinating words increasing from 3.86 to 4.52, and verbs and auxiliaries moving from 14.13 to 16.69. Not surprisingly, both the number of function adverbs (34) and the incidence of initial connection decrease as Sprat moves from narrative to exhortation: from deck 0206 to 0207 function adverbs decline from 3.15% to 1.86%, with a parallel though less pronounced decline from Cowley to History; and percentage of initial connection declines from 41.3% of sentences to 25.3%. Initial repetition, along with formal anaphora, increases perceptibly when Sprat turns to exhortation: 13% of the sentences in 0206 have repeated openings, 23% in 0207; 17% have repeated openings in Cowley, 20.1% in the History, and this despite the "funeral oration" effect that Samuel Johnson (.Lives of the Poets] accurately noted in the last third of the Cowley. Again not surprisingly, as style is very much elevated, so even more is the number of intensifiers (33). In general, the overall pattern of movement discussed above is one away from conversational devices and towards those identified with artifice and contrivance, away from a "natural way of speaking" towards "Wit and elegance of language," and we should remember that Sprat advocated both. Still, for all the bending in obedience to convention, Sprat is still Sprat, for both the highest and lowest Sprat styles are more homogeneous with other Sprat styles (Table 5) than are our carefully selected controls with each other. 110105 produced 143 strings per 3,500 words, 0109 produced 124.
274
Computers and the Humanities/Vol. 5/No. 5/May 1971
Cluett: Style, Precept, Personality Conclusions It is impossible to say that a writer "wills" certain traits into being in his own style. It is, however, possible to say that of his given resources and tendencies a writer can emphasize those that are most reflective of his temperament and predispositions and suppress those that are least so. For example, Sprat's consistent avoidance of dictional specificity (Adolph), can be seen as an undoubted reflection of his aversion to the rhetorical climate of the Commonwealth years (Cope), even though we cannot say anything so firm as "Sprat decided on a general level of diction because . . . . " The process, rather, is permissive and acquiescent-as any writing teacher knows-rather than prescriptive and positivistic. The history of writers' styles in relation to their most firmly enunciated precepts is a dismal one, from which a pessimist might reconclude that hypocrisy is an endemic aspect of the human condition: Swift on monosyllables (Milic), Orwell on the passive voice (Orwell), E.B. White on manly plain directness (Strunk and White), and Thomas Sprat on words and things (Cluett) form but a beginning list of writers who have failed notoriously to follow their most renowned prescriptions. The failure is easily explained. Part of it is traceable to the permissive rather than positive transfer between rhetorical concepts and stylistic facts. Another part derives from the fact that stylists and rhetoricians always underestimate the language content of style and overestimate the power of their own personalities; for example, it seems nearly impossible'to write English with less than one word in six or more than one word in four being a noun, and even near these frequencies you are likely to sound as idiosyncratic as Dryden or Burton or Gibbon; ~2 yet, Strunk and White tel1 us always prefer the noun to the modifier. A third part derives from the nature of stylistic percepts: in the mind of the maker they are hedged about with qualifiers, but preceptorial convention demands a pithy, aphoristic command. And there is no enemy of pith like a chain of qualifiers. Still another part derives from the fact that statements shed their contexts with time; the precept in its proper place is always less tendentious than it seems when it has been lifted by the literary historian and placed in an improved setting. And finally, the precept reveals but one side of the preceptor; there is in each of us a Seneca and a Cicero, a Montaigne and a Cardinal Bembo, an oral and an anal literary persona; one of these, to be sure, will dominate, but never to the entire exclusion of the other, and in the precept one persona speaks alone, usually, though not always, the dominant one. Sprat's repetitious, distributed style was surely a violation of his most frequently quoted dogmatic statements, statements he did not intend nearly so categorically as has sometimes been inferred. On the other hand, 12Burton and Gibbon, one noun every four words; Dryden, one n o u n every six. Figures on Gib bon are from Milie, on Dryden from table 4, on B urt on from a 10,500 wo rd r a n d o m sample from his Anatomy. Comp uters and the Humanities/Vol. 5/No; 5/May 1971
275
67uett: Style, Precept, Personality that style does reflect in its chastened metaphors and systems of modification his standards of truth and his aversion to Fancy; in its anaphoras, seriations, and rhetorical questions, his love of eloquence; in its control of level, his traditional rhetorical training and his conventional vision of prose genre; in its linearity and syndeton, his striving for the "easie" and "natural"; in its diction his preference for politeness in controversy. It also reflects in its ease and facility the facility, speed, and shallowness of the mind that produced it. There are in Sprat's case connections between taste, personality, and thought on the one hand, and, on the other hand, the way in which he wrote. But they are connections whose precise character cannot be readily described with the present literary vocabulary, and how conscious some of them were in their growth is beyond knowing. Of no writer can we yet say, "There was this kind of personality; therefore the style was thus-and-so." Rather, we must proceed backward from the phenomena to probable and possible cause. There are innumerable pitfalls waiting for the critic who so proceeds, 13 and we must exercise constant care and discretion to insure that our perception of things is detached and accurate. In many respects, the computer (with Milic's technique) offers to stylistic critics an unprecedented opportunity to examine minutely large quantities of text and to protect themselves from the biases and predispositions that are inherent in the critical process. And it offers this opportunity not merely to the statistically oriented but to critics like myself, who bring to their task a traditionally humanistic set of assumptions. Such an opportunity is long overdue: for all the infinity of words to have been expended on English prose style, 1600-1700, we have yet to see a satisfactory full-length account of so important a figure as Dryden. Despite the need for an even greater elaboration of the grammar (see note 10) the technique as it stands is an extremely incisive instrument. Whatever it was that the New Science did for or to prose style cannot be said without far more specific information about individual styles than we have now, though valiant attempts have been made. We do know that concomitant with the New Science there arose new, Baconian standards of truth, together with a new, crustier attitude towards strictly personal experience, and a renewed preference for "matter" over "words." However, beyond what Jones, Williamson, and Adolph have already said about the effects of these changes on such conscious elements as metaphor, rhetorical training, and the writer's conception of his task, not a great deal can be added on the basis of present facts. We need accounts of many more individual styles than we have now. 13Se e Milic's Stylists on Style, p. 10.
276
Computers and the Humanities/VoL 5/No. 5/May 1971
MeeHngs Announced
References Adolph, R. The Rise of Modern Prose Style. Cambridge, Mass.: 1968. Burnet, G. Bishop Burnet's History of His lime, Vol. I. London: 1724. Cluett, R. These Seeming Mysteries: The Mind and Style of Thomas Sprat (DA 70-6950). Cope, J. Introduction to Sprat's History of the Royal Society. St. Louis: 1958. Evelyn, J. Diary, edited by E.S. deBeer, London: 1959. Glanvill, J. Plus Ultra. London: 1668. Hoel, P. G. Elementary Statistics. New York: 1960. Jones, R. F. Articles in Schelling, ed., Essential Articles. Hamden, Conn.: 1963. K6ster, P. "Words and Numbers..." CHum, 4, v. (May 1970). Macaulay, T. B. History of England. 5 vols. Philadelphia: 1881. Milic, L. T. A Quantitative Approach to the Style o f Swi[t. The Hague: 1967. Mitchell, W. F. English Pulpit Oratory. London: 1931. Orwell, G. "'Politicsand the English Language," In Essays. Garden City: 1956. Spingarn, J. E. 17th Century Critical Essays. 3 vols. New York: 1907. Strunk, W., and E. B. White. The Elements of Style. New York: 1959. Williamson, G. The Senecan Amble. Chicago: 1952.
Meetings Announced 11-12 June 1971 A SYMPOSIUM ON "QUESTIONS OF MACHINE PROCESSING OF
MIDDLE HIGH GERMAN TEXTS" will be held at the Institut flit Deutsche Sprache in Mannheim, W. Germany. Papers will be presented by G. Kochend6rfer (Freiburg), Gerd Simon (Tiibingen), Rudolf Hirschmann (Los Angeles), Erich Strafiner (Erlangen), Jiirgen Sch6nhut (Eflangen), Ilpo Piirainen (Helsinki), Monika R6ssing-Hager (Marburg), Hans Dieter Lutz (Bonn), Stanley N. Werbow (Austin), Roy A. Boggs (Pittsburgh), Ute Krumnack (Bonn), Winfried Lenders (Bonn), Roy A. Wisbey (Cambridge), Paul Sappler (Tiibingen), Harald Scholler (Ann Arbor), and Bernward Plate (Regensburg). Additional information is available from Ute Krumnack or Dr. Winfried Lenders, Institut f'tir Deutsche Sprache, Forschungsstellen Bonn, D53, Adenauerallee 96, W. Germany. 23-25 June 1971 THE CONFERENCE ON COMPUTERS IN THE UNDERGRADUATE CURRICULA will be held at Dartmouth College. Two sessions are of special interest. One, chaired by Sally Y. Sedelow on 23 June, includes papers by Ben R. Schneider, Jr., "Building a Machine-Readable Text Library"; R. N. Killam, "Computer-Aided Analysis of Music Used in Undergraduate Music Theory"; and Floyd R. Horowitz, "Training the Humanist for his Role in Developing Undergraduate Computer Oriented Computers and the Humanities/Vol. S/No. S/May 1971
277