16 Aug 2007

What's in a phrase?

Submitted by Karl Hagen
This is part of a continuing series documenting the shortcomings of the College Board's view of grammar, particularly as exemplified in its explanations for multiple-choice writing questions. (Part 1, part 2, and part 3 here). Please see part 1 for an explanation of how I reference individual questions.

To this point, I have focused my attention on a large number of relatively small points: individual errors of analysis and small points of inconsistency. Taken cumulatively they demonstrate, at a minimum, an error rate that reflects unprofessional sloppiness and a failure to follow a consistent grammatical framework.

At this point I want to examine more fundamental flaws in the explanations. These flaws can be understood as inconsistencies, since practice is not uniform throughout the explanations, but a large measure of the inconsistency derives directly from the tradition in which the College Board situates itself: the archaic and inaccurate schoolbook grammars exemplified by Warriner and Griffith (1965).


Confusion about basic issues of constituency pervade the explanations. Constituency—the way that sentences are composed of hierarchically arranged parts—is fundamental to all varieties of syntactic analysis. It would be reasonable to expect the experts who create the questions would understand what a phrase is. Unfortunately, the term is rarely used properly in the explanations.

In ordinary, nontechnical language, phrase can simply mean a group of words. In syntax, however, it has a more precise meaning. A phrase is not just any string of words. Even in very traditional accounts of phrases, the words that form a phrase must function as a single constituent within a sentence:

Words in a sentence act not only individually but also in groups. The grouped words act together as a unit which may function as a modifier, a subject, a verb, an object, or a predicate nominative. The most common group of related words is the phrase. (Warriner & Griffith, 1965, p. 35).

This is not a particularly precise explanation, and the later, more formal definition is no more helpful, but it does convey the central notion of constituency: the words must act as a single unit. And yet, over and over, the College Board uses phrase to refer to linear strings of words that are not constituents under any plausible analysis. For example,

"and that was when" (SG-1.7.1)
"in the way of having a"; "also having the" (SG-1.7.5)
"in contrast to" (SG-1.7.8)
"By considering that there" (SG-1.7.11)
"speak about" (SG-1.7.14)
"the dense" (SG-2.7.12)

In other places, the term is used for what I would call a "surface phrase," i.e., for a string of words which, taken out of context, could be a phrase but which does not actually match the real phrase boundary for the particular sentence.

For example, in SG-2.7.15: "may be" is called a verb phrase, when the complete verb phrase is "may be recovered."

Similarly, in SG-3.6.4, "increase profits" is also called a verb phrase, omitting the prepositional phrase "for landlords," which is also part of the same phrase. See also SG-4.6.3

These nontechnical uses of phrase are wholly unnecessary. In the context of a grammatical explanation, the term, when it is used, should be used precisely. Such use would not make the explanations more complex. All that is necessary is not to use the term when the string of words is not a true phrase, or to alter the quoted text to reference the entire phrase. Some explanations already do just this.

The term phrase is also used inconsistently in that sometimes it reflects the definitions given in traditional grammars and other times it reflects the definitions used by linguistically influenced grammars. This inconsistency is compounded by the fact that the traditional use of the term phrase is rife with internal inconsistencies.

As Warriner and Griffith (1965) describe them, the traditional phrasal categories look like this:

A verb phrase is a lexical verb verb along with any associated auxiliary verbs (Warriner & Griffith, 1965, p. 12). For example, in the following sentence, the underlined words, and those words only, constitute the verb phrase:

Marissa could have completed her project rapidly.

This definition implies that the direct object ("her project") and the adverb ("rapidly") are not part of the verb phrase. By contrast, more recent accounts of phrase structure recognize that these other elements are part of the verb phrase too.

One peculiar thing about this use of verb phrase is that it is inconsistent with the definitions of phrases formed with "verbals" (i.e., participles, gerunds, and infinitives), which do include complements and modifiers (Warriner & Griffith, 1965, pp. 42, 46, and 48). And muddying the water even further, Warriner and Griffith also use the term verb as a synonym for their use of verb phrase (Warriner & Griffith, 1965, p. 24).

The term prepositional phrase is defined with extreme vagueness (Warriner & Griffith, 1965, p. 36), but in practice amounts to what linguists mean by a prepositional phrase.

The terms adjective phrase and adverb phrase, however, are used very differently from the way contemporary linguists use them. According to Warriner and Griffith, an adjective phrase refers to a prepositional phrase that modifies a noun, and an adverb phrase refers to a prepositional phrase that modifies a verb, adjective or adverb (Warriner & Griffith, 1965, pp. 36 and 38).

In contrast, more recent accounts reserve adjective phrase and adverb phrase for constituents that are headed by adjectives and adverbs respectively, for example "fond of chocolate," or "less quickly." Warriner and Griffith have no term for these constituents.

The term noun phrase does not appear at all in Warriner and Griffith, although the Reed-Kellogg diagrams in the book implicitly treat nouns and their modifiers as constituents.

In contrast to this conceptual mess, works that have been influenced by linguistics treat phrase structure in a much more consistent and sensible manner. Phrases are named according to the category of the head word, and form is clearly differentiated from function. Thus a prepositional phrase remains a prepositional phrase whether it modifies a noun (e.g., "the boy with whooping cough") or a verb (e.g., "He infected his sister with whooping cough.") In some works that use a linguistically inspired account of phrase structure, the terms adjectival and adverbial are used to describe the functional role that modifiers play, that is, whether they modify nouns or other words. These terms are distinct from the part of speech labels to indicate that the prepositional phrases do not become a part of speech. Rather, they have a function that matches that of adjective and adverb phrases.

There are two key points to note about these different views of phrase structure. First, they are incompatible. The terms adjective phrase, adverb phrase, and verb phrase all mean completely different things. Second, the traditional version is, quite simply, inadequate. It fails to meet the minimum standards for accuracy or explanatory adequacy. Moreover, its weaknesses have long been known to experts. It therefore has no basis being used in any current standardized test, which should be based on the best current understanding of the state of the subject.

For an example of the incoherence that the traditional account creates, consider question SG-4.10.4, choice C. The sentence in question (with the text of C substituted, reads as follows):

The first world computer chess championship, held in Stockholm, was won by a Russian computer program called Kaissa, having four victories and with no defeats.

And here is the explanation:

Choice (C) results in a lack of parallelism. The expression "having four victories" is an adjective phrase, while "with no defeats" is a prepositional phrase.

It is true that this sentence is not parallel, and that "having four victories" does not coordinate well with "with no defeats." But the remember that adjective phrase is a functional description. Traditionally, a prepositional phrase is also called an adjective phrase if it modifies a noun. So we could with equal justice describe both as adjective phrases, in which case we have explained absolutely nothing as to why the sentence lacks parallelism. A better explanation would invoke the formal phrase times for each phrase and call "having four victories" a participial phrase.

To measure the mixture of traditional vs. contemporary uses of phrase terminology, I analyzed all instances of the terms verb phrase, adjective phrase, and adverb phrase in the explanations. Excluding instances where it cannot be determined which usage is intended, the traditional use trumps the modern use at a rate of more than three to one.[1] Note that this count understates traditional usage, because many explanations use the term verb as a synonym for the traditional definition of verb phrase.

In at least one case, traditional and contemporary understandings of phrase structure appear in explanation for the same question.

Because they must compete with a large chain of super-stores that can afford to charge very low rates for certain items, the owners of small hardware stores know that you are unlikely to make much profit and may, in fact, go bankrupt. (OC-1.7.17)

The explanation for choice (A) uses the traditional understanding of a verb phrase ("must compete" is labeled the phrase when there is a prepositional phrase "with...items" that, in contemporary terms is also part of the VP). In the explanation for choice (D), however, a modern understanding of verb phrases is employed:

The verb phrase "may... go bankrupt," introduced by the word "and," is properly parallel with the preceding verb phrase "are unlikely... profit." (ellipses in original)

In explanations that use the traditional notion of verb phrases, ellipses are sometimes used to indicate that adverbs, etc. are not part of the verb phrase, but in each case here, the first and last quoted words correspond to the complete phrase boundary, and the final words are in neither case verbs.

This conflict between incompatible terminology does not exhaust the inconsistencies. There are also instances clearly intended to follow the traditional usage but which do not work by this or by modern usage.

For example, an explanation in SG-4.6.30 reads, "Choice (D) is unsatisfactory because it creates a sentence fragment, using an adjective phrase, 'Having been unable,' instead of a subject and verb."

The full phrase is actually "Having been unable, usually, to locate its source." Further, such phrases are, in the traditional scheme, only adjective phrases by virtue of modifying a noun. There is no noun in this sentence for the phrase to modify, so it would be inaccurate to call it an adjective phrase.

Or to take another example, the infinitive phrase in the following sentence is called an adjective phrase:

Only after the floodwaters had risen two feet was the mayor willing to order the evacuation of some homes. (OC-3.4.13)

But the phrase modifies willing, not mayor, and by the traditional scheme should be called an adverb phrase.

Adverbs that occur between auxiliary verbs and the main verb are sometimes called part of the verb phrase, for example, in SG-6.4.1 or OC-2.10.12

In such cases, verb phrase appears to mean little more than "a string of words that contains a verb." Another instance of such imprecise use occurs in OC-2.6.1, where we are told that "'A stranger,' the object of the verb phrase 'were surprised to see,' is placed before the subject, 'the students.'" (The original sentence reads, "A stranger, the students were surprised to see him enter the classroom carrying a bowling ball.") The quoted words, though, are not a single phrase. "A stranger" is the subject of were only, surprised is an adjective (a subject complement), and the infinitive "to see him" is a complement of surprised. A stranger is the (repetitious) direct object of see, not of the whole string of words.

Or consider this problem:

Experts disagree about how to define and measure intelligence. (OC-p.10.4, text of C substituted for original)


Choice (C) is correct. It avoids the error of the original by providing parallel verbs ("define" and "measure") to complete the verb phrase "how to" and modify "intelligence."

The analysis of constituency is strange. By itself, "how to" isn't any sort of verb phrase. This might just be bad wording and the intended phrase would be "how to define and measure," but (1) the verb coordination doesn't modify intelligence, intelligence is the direct object of the verbs. And (2), how isn't part of the verb phrase (i.e., the infinitive). In traditional terms, it's a relative adverb.

In OC-1.7.20, "is helpful" is called a verb phrase. Helpful is an adjective, so this isn't a traditional definition of the verb phrase. But in modern terms, this isn't the complete verb phrase either, as helpful has its own complement.

To summarize, the College Board's use of phrase is deeply problematic. Not only are there numerous internal inconsistencies and plain errors of analysis, but the predominant analytical scheme used is hopelessly outdated, demonstrably incorrect, and not particularly useful in explaining things.

In the next installment, I will continue exploring the College Board's flawed notion of constituency, particularly the way that form and function are regularly confused, and the consequences such confusion has for significant misunderstanding.

[4] 64 Traditional: SG-1.7.25, SG-1.7.29, SG-2.7.13, SG-2.7.28, SG-2.10.4, SG-3.6.18, SG-3.6.19, SG-3.6.20, SG-3.6.21, SG-3.6.22, SG-3.6.24, SG-3.6.27, SG-3.6.28, SG-3.10.13, SG-4.6.21, SG-4.6.22, SG-4.10.12, SG-5.4.7, SG-5.4.16, SG-5.4.21, SG-5.4.24, SG-5.4.25, SG-5.4.28, SG-6.4.13, SG-6.4.15, SG-6.4.16, SG-6.4.23, SG-6.4.27, SG-6.4.33, SG-6.10.1, SG-6.10.8, SG-7.3.5, SG-7.3.12, SG-7.3.13, SG-7.3.16, SG-7.3.17, SG-7.3.19, SG-7.3.21, SG-7.3.24, SG-7.3.25, SG-7.3.29, SG-8.3.13, SG-8.3.22, SG-8.10.10, OC-1.7.8, OC-1.7.12, OC-1.7.13, OC-1.7.25, OC-1.10.1, OC-1.10.10, OC-2.6.9, OC-2.6.11, OC-2.6.13, OC-2.6.21, OC-2.6.26, OC-2.10.6, OC-3.4.5, OC-3.4.12, OC-3.4.17, OC-3.4.25, OC-3.4.26, OC-3.10.6, OC-4.5.9, OC-4.5.26

20 Modern: SG-1.7.18, SG-1.7.20, SG-1.7.24, SG-2.10.2, SG-2.10.3, SG-2.10.14, SG-4.6.2, SG-4.6.7, SG-4.10.5, SG-5.4.2, SG-5.4.17, SG-5.10.11, SG-7.3.22, SG-7.3.23, OC-p.5.8, OC-p.10.10, OC-5.6.13, OC-5.6.27, OC-5.10.10, OC-6.5.4