16.1 Patterns in Texts

What would such a pattern look like? Chapters 11 and 15 have talked about sentence patterns and discourse patterns above the sentence, text types, and collocation. It is possible to associate some sentence patterns with text types or authors. For example, in scientific lab reports we can recognize extensive use of the passive voice construction in sentences (“a color change was observed after adding the reagent”). It is often noted in literary circles that Hemingway uses short sentences, and Faulkner uses long sentences. These are qualitative statements, impressions after reading that do not actually count the number of passive-voice sentences in scientific reports or calculate the mean length, or tabulate the number of sentences of a given length, in Hemingway and Faulkner. It might be possible to count them, but it would be difficult to account for all of the variation, all of the optional elements in such patterns. There is nothing wrong with this kind of impressionistic statement, but it does not really establish what is going on in the reports or writing by the authors.

The complex system of speech gives us another way to work with text types, but we have to be careful about how we use frequency profiles to describe the text type(s) they come from. If we calculate the A-curve for the frequency of words or collocates in a text type, that is good information that does establish something about the text type, but it still does not tell us what we should pay attention to in the A-curve. For example, if we make an A-curve for the Brown Corpus (about one million words, 500 texts selected from 15 different text types of American English published in 1961, as described in the last chapter), the top-ranked words are:

This is nice information to have, but we do not know whether any of the information is useful as part of a pattern that describes American English, which is what the Brown Corpus is used for. The significance of patterns is not apparent just from observing a text, whether we notice the patterns qualitatively or quantitatively.

In order to begin to see the significance, we need to make a comparison. If we are observing the writing of an author, it is true that what we observe does belong to the author’s style, but we will not know what characteristics belong to that author and not to other writers unless we compare them. So, what are the most common words in Hemingway’s A Farewell to Arms?

Many of these are the same as what we saw in the Brown Corpus: the, a, and, to, in. The different ones are I (Brown 24), you (Brown 44), it (Brown 15), was (Brown 12), said (Brown 69). We might be tempted to say that these words are part of Hemingway’s style, except that they are all from the top 70 words in the Brown list, all really common words. The fact that I, you, and said are more common for Hemingway suggests that he is writing something more personal than the general run of writing in the Brown Corpus. Douglas Biber (Biber, Conrad, and Reppen Reference Biber, Conrad and Reppen1998: 148) reports that first- and second-person pronouns indicate a more involved and less informational register. A past tense verb, like said, may indicate a more narrative and less non-narrative register. What Biber means by register is a level of discourse lower than top-level text types like “writing” but higher than any specific text type like “novel” or “report.” A comparison of the top ten words from the Brown Corpus and Hemingway, therefore, does not tell us much about Hemingway’s personal style that we can recognize; however, it does begin to tell us that Hemingway is not writing just any old kind of English, but instead a story about people, as opposed to a non-narrative informational piece.

We can learn more about the novel by considering those words that are higher on the A-curve in the novel than they are on the Brown A-curve. Here are 50 words that are much more frequent in A Farewell to Arms than they are in the Brown Corpus:

This list takes account of the differing number of words overall in each source, a fair way of finding words that are higher up the A-curve in one source than in another. The list fairly jumps off the page for anyone familiar with the novel, which, for those who have not read it, is a love story about ambulance drivers in rainy Italy. The names of characters stand out, and aspects of the story like the road, the rain, and doctor and nurse as medical personnel. The list tells us pretty clearly what the novel is about. Here again, we see pronouns and simple past tense verbs, evidence of a narrative about people. Finally, we see lots of short words (only two words are as long as seven letters, darling and details), the kind of simple vocabulary for which Hemingway is well known.

Just to confirm the point, here is a comparison of words from Mark Twain’s Huck Finn with the Brown Corpus, showing 50 words that are more common in Huck Finn than in the Brown Corpus:

Again we see personal pronouns and simple past tense verbs that fit the personal narrative, and we see character names, the raft (river is a few words further down the list), and some of the dialect words that this novel is famous for. Again, differences in frequency on the A-curve in a comparison of two sources has told us about the novel. In a text analysis, then, we can use the properties of the complex system of language to help us find out about texts. Now, however, we are observing the style of each novel. The words that are used more often in the novels than they are in a general corpus belong to the special pattern of each novel. They describe what is different about it. When we as readers encounter such differences in usage when we read a novel, we build up a mental picture that we may not remember as precisely, but which still allows us to tell the differences between, say, a Hemingway novel and a Twain novel, or between either of them and the language that we regard as normal in our own experience with language. Making counts and lists just formalizes what each reader does silently, while reading.

The reason that we can do this is the complex system of speech, and the dependable A-curve frequency profiles that come from the interactions of speakers and writers. Once we know that there will be an A-curve frequency profile of the words in any text type, large or small, we can compare the profiles in order to find out what makes any particular text type special. That is, if we know that one text type has a different profile from another text type – and it will, whether we are talking about American English in the COCA vs. British English in the BNC, or about a Hemingway novel vs. a Mark Twain novel – we can use the differences to define how they are different. The shape of the profile will be the same each time – an A-curve – so what we have to do is find out what particular words or collocates are higher on the profile for one text type than they are for another.

16.2 Author Style

If we want to consider the style of an author, not just the style of a particular novel, we will need to compare an entire collection of an author’s work with a general corpus like the Brown Corpus, or the BNC, or the COCA. The following list again contains 50 words that are particularly more common in Mark Twain’s collected works (as assembled by Project Gutenberg, www.gutenberg.org) than they are in the Brown Corpus:

This list no longer contains the I, you, and said that marked the novels as involved narratives. It has only one character name, Tom, but here the name can come from any of the 60 works collected by Project Gutenberg. We no longer see any of the dialect words on the comparison list from Huck Finn. We do see am, a verb that goes only with I, which suggests that Twain overall writes “involved” works in Biber’s sense. We also see simple past tense verbs like took, done, saw, told, seen, began, and knew that suggest that Twain overall writes “narrative” works in Biber’s sense. We can tell that Twain cared about mental and emotional states, from think, head, mind, heart. He was interested in the young and poor, but another of his especially common words is king (probably not just from A Connecticut Yankee in King Arthur’s Court). Twain often wrote about existence, from his common use of thing, nothing, things, matter, anything. It is less clear how we should interpret other words on the list, but nonetheless they are Twain’s words, used by him more often than they were used in a general resource like the Brown Corpus. Since Mark Twain died in 1910, about 50 years before the year from which the texts in the Brown Corpus were written, some historical change in the language may exist in the list (if we were to make a late-nineteenth-century general corpus, we could test that idea), but we can certainly begin to see Mark Twain’s style here.

If we want to examine how Twain used particular words, we can look for what words he used together, his collocates for any given word. For example, if we consider the word mark from his list of common words, we find that it most frequently collocates with twain (877 times!), which tells us more about how often the author’s own name occurs in the Project Gutenberg collection than about his style. For a more stylistic example, if we consider the word poor from the list of Twain’s common words, we find that he most often combines it with words like chap, friendless, lad, creatures, shabby, devil, fellows, fellow, cheap, creature, ignorant, pity, soul, boy, child, girl, ah, poor, brother, old. These words suggest that Twain was not so much interested in the plight of the poor, like Dickens was, as he was in pointing them out. If we consider the word book from the list, we find that Twain pairs it with memorandum, hymn, scrap, co, spelling, text, and guide, each of which makes a common compound (memorandum book, hymn book, scrap book, book co[mpany], spelling book, text book, guide book). He also pairs book with publish, published, and author, which addresses his own role as a writer. If we consider the word mind, we see unusual collocates like mortal and healing that come from Twain’s commentary on Christian Science as a religion, and then see that body is a frequent collocate more broadly in phrases like body and mind/mind and body, more a stylistic attribute of his writing. Finally, if we consider the word face, we find that lighted and lit are frequent collocates, surely a stylistic trait of Twain’s, as in her face lighted joyously from his Personal Recollections of Joan of Arc. In each of these cases (except for Mark Twain!), the particular combination of words marks Twain’s style.

Some collocates that Twain used frequently are also common in general use. For the word mind, collocates like peace, never, and presence come from the common phrases peace of mind, never mind, and presence of mind. The compounds hymn book, text book, and guide book are also common in general use. Still, because they rise to become among the most frequent collocates in Twain’s writing, we are entitled to count them as aspects of his style.

Analysis of collocates, like the analysis of word lists, requires comparison with a general resource before we can decide whether any collocate is especially common in an author’s writing. Collocates have a frequency profile just like words do; we are always interested in which collocates are higher on the A-curve for a particular author than they are in general use. The comparison makes the case that we are entitled to say that a word or collocate belongs to an author’s style because the author has used it more often than it occurs in general use. Again, the complex system of speech is what makes it possible for us to conduct such comparisons. We know that the A-curve will always be there for us to find, and then compare.

16.3 Other Elements of Style

Of course there are other things besides words and collocates that we can count in order to find especially common elements that make up an author’s style. We could measure the number of times an author uses the passive voice in sentences, for instance. Or we could measure sentence length. Or we could count the number of subordinate clauses in order to create a measure of complication in an author’s sentences. These counts, however, are more complicated to carry out than the simple word counts illustrated in this chapter. The important thing for the more complicated measurements will still be a comparison. We still need to know, not just that an author used a grammatical construction a certain number of times, but how an author’s number of constructions compares to the number of similar constructions in a general corpus before we can claim that an author’s usage is part of the author’s style. We are interested in a pattern of writing in order to talk about style, and we want the pattern we find in an author’s writing to be different from what we find more generally.

16.4 Looking Ahead

We’ve seen that text types and authors’ use of particular words and constructions can be evaluated quantitatively in order to identify emergent patterns that make up what we think of as style. The idea that a “way of doing things” is an emergent pattern is one that will serve us well as we dive into sociolinguistics in the next chapter, where we’ll be talking more about variation in language and how that relates to the speech of groups of speakers.