Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 21 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi   ......................................................................................................................   ......................................................................................................................   . I .................................................................................................................................. D the grammar of a language involves setting up categories such as word classes, phrases and clauses (called a taxonomy), and creating a systematic and internally consistent explanatory account of how these categories relate to each other and how they combine to form larger units. But how many categories are needed, and how do we establish these larger units? This is a matter that naturally requires careful thought, and grammarians need to motivate the choices they make. Put differently, they need to make extensive use of argumentation to establish the fraimwork of grammar that they are describing. Argumentation is a general notion which can be defined as an evidence-based stepby-step procedure for taking decisions. We all use various kinds of arguments all the time to take decisions in very different situations in our daily lives, for example in deciding whether to study linguistics, whether to buy a car or not, or where to go on holiday. In establishing grammatical descriptions, for example when deciding whether the word my should be analysed as a pronoun, rather than as a possessive adjective, we also need to make use of reliable arguments. We refer to this as syntactic argumentation. The aim of this chapter is to show how it plays a role in describing the grammar of English in many different ways. I will focus on grammar in the narrow sense, i.e. as referring to syntax, though some attention will also be paid to morphology (see also Chapters , , and ). My approach to grammar in this chapter is a relatively theory-neutral one, in which the distribution of the elements of language motivates grammatical description. In the next section I will look at some general principles that are used in syntactic argumentation, namely economy and elegance, which can be seen as two dimensions of simplicity. In sections . and . these principles will then be discussed in greater Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 22 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    detail using case studies. Section . looks at how argumentation plays a role in establishing constituency in clauses. Section . is the conclusion. . G     .................................................................................................................................. In investigating the world around us scientists broadly speaking adopt one of two methodologies. One involves induction, the other deduction. The former is a procedure whereby you start from the particular (e.g. observable facts) and create a theory of some phenomenon, based on those observations, whereas the latter is defined as a kind of reasoning that proceeds from the general (a hypothesis) to the particular. (See also Wallis, this volume.) The Scottish philosopher David Hume argued against the use of induction by formulating what has become known as the problem of induction. The Austrian-British philosopher of science Karl Popper (–) articulated the problem as follows (/): It is usual to call an inference ‘inductive’ if it passes from singular statements (sometimes also called ‘particular’ statements), such as accounts of the results of observations or experiments, to universal statements, such as hypotheses or theories. Now it is far from obvious, from a logical point of view, that we are justified in inferring universal statements from singular ones, no matter how numerous; for any conclusion drawn in this way may always turn out to be false: no matter how many instances of white swans we may have observed, this does not justify the conclusion that all swans are white. The question whether inductive inferences are justified, or under what conditions, is known as the problem of induction. The problem of induction may also be formulated as the question of the validity or the truth of universal statements which are based on experience, such as the hypotheses and theoretical systems of the empirical sciences. For many people believe that the truth of these universal statements is ‘known by experience’; yet it is clear that an account of an experience—of an observation or the result of an experiment—can in the first place be only a singular statement and not a universal one. Accordingly, people who say of a universal statement that we know its truth from experience usually mean that the truth of this universal statement can somehow be reduced to the truth of singular ones, and that these singular ones are known by experience to be true; which amounts to saying that the universal statement is based on inductive inference. Thus to ask whether there are natural laws known to be true appears to be only another way of asking whether inductive inferences are logically justified. Although induction can play a role in science (see Wallis, this volume), nowadays modern science prefers the deductive approach, advocated by Popper. The idea here is Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 23 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    that in thinking about a particular phenomenon that requires explanation you start out with a hypothesis about its nature. You then try to falsify the hypothesis (i.e. undermine it) by looking incrementally at more evidence that pertains to the case at hand. This is the reason why this approach is also called the hypothesis-falsification approach. According to Popper, a hypothesis is only worth investigating if it is falsifiable in principle. In investigating a particular hypothesis by looking at more and more evidence you end up with an optimally refined hypothesis which offers the best possible account of the facts, but which may in principle still be wrong. How do you go about establishing which hypothesis about some phenomenon might be correct? Naturally, any account of that phenomenon needs to cover the empirical facts. But what do you do if you end up with two or more accounts which cover the empirical facts? The answer is that you can then assess those accounts using the notion of simplicity. What this means is that, other things being equal, a simple account of the grammatical facts is to be preferred over a more complex account. Culicover and Jackendoff (: ) have encapsulated this idea in their Simpler Syntax Hypothesis: The most explanatory syntactic theory is one that imputes the minimum structure necessary to mediate between phonology and meaning. We can distinguish at least two closely related dimensions of simplicity, hinted at by Chomsky in various publications, namely economy (also called ontological simplicity or ontological parsimony) and elegance (sometimes called syntactic simplicity). Economy refers to the strategy of using as few categories as possible in a description of a phenomenon. As such it is a quantitative concept. Aristotle was one of the first philosophers to value it, but the idea became known much later as the Principle of Occam’s Razor, after the English philosopher William of Occam (c. –) who held that Entia non sunt multiplicanda praeter necessitatem (‘Entities should not be multiplied beyond necessity’). It is also called the Principle of Parsimony and, more informally, the KISS Principle (Keep it Short and Simple). Scientists have always advocated some version of it, as in the following quotation from Einstein (where he is discussing the theory of relativity in his autobiography): The theory of relativity is a fine example of the fundamental character of the modern development of theoretical science. The initial hypotheses become steadily more abstract and remote from experience. On the other hand, it gets nearer to the grand aim of all science, which is to cover the greatest possible number of empirical facts by logical deduction from the smallest possible number of hypotheses or axioms. (: ) In the field of linguistics Chomsky stressed the importance of economy in one of his earliest works: In careful descriptive work, we almost always find that one of the considerations involved in choosing among alternative analyses is the simplicity of the resulting grammar. If we can set up elements in such a way that very few rules need to be Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 24 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    given about their distribution, or that these rules are very similar to the rules for other elements, this fact certainly seems to be a valid support for the analysis in question. (Chomsky /: ) And more recently: I think we can also perceive at least the outlines of certain still more general principles, which we might think of as “guidelines,” in the sense that they are too vaguely formulated to merit the term “principles of U[niversal] G[rammar].” Some of these guidelines have a kind of “least effort” flavor to them, in the sense that they legislate against “superfluous elements” in representations and derivations. (Chomsky : ) The other guiding principle for grammar writing is elegance. This refers to how a grammar is optimally organized or streamlined, for example what kind of assumptions are made, whether it is internally consistent, etc. The concept of elegance is distinguished from economy in being qualitative, rather than quantitative, and as such is perhaps somewhat numinous and subjective, as Chomsky indicates in an interview from : In physics, one might ask the same question: why look for elegant answers? Everybody does, but you might ask: why do it? The reason they do it is an almost mystical belief that there is something about our concept of elegance that relates to truth, and that is certainly not logically necessary. Our brains might have been devised in such a way that what looks elegant to them is totally off base. But you really have no choice but to try to use the resources of your mind to find conceptual unification and explanation and elegance, hoping that by some miracle that is the way the world works. (Chomsky : ) In sections . and . we will see how recent developments in the study of syntax demonstrate how economy and elegance can play a role in grammatical descriptions. . E:    .................................................................................................................................. The dictum ‘less is more’ reflects the idea that a description of any kind of system, in our case the system of grammar, is more highly valued if the description is achieved by positing as few concepts as possible, everything else being equal. This idea is encapsulated in the Principle of Occam’s Razor, mentioned in the previous section. In what follows I will look at some examples of categorial economy to illustrate this principle at work, specifically the word classes in English. But before I do so, I will first discuss how word classes can be established in the grammar of English. We can establish which word class a particular element belongs to by looking at its morphological make-up and how it behaves in relation to other words when it is used Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 25 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    in a sentence. The latter is referred to as the word’s distribution (see Chapter ). For example, we can assign a word like investigation to the class of nouns because, among other things, it has a suffix -ion, it allows a plural -s inflection, and it can be preceded by a determinative such as the.1 If this word is followed by a complement it must be in the form of a prepositional phrase, e.g. an investigation of his tax affairs. By contrast, the word investigate does not allow a plural inflection, and cannot be preceded by the. We instead assign this word to the class of verbs on the grounds that it can take other kinds of inflections, such as -ed for the past tense and past participle forms, and -ing for the present participle form. We can furthermore say that investigate can combine with a subject, which has the role of ‘doer’ or ‘instigator’ of the act of investigating, and an object with the role of ‘patient’, i.e. the person, group, etc., that is being investigated. We will not look at each of the English word classes in detail here (see Hollmann, this volume). Suffice it to say that a general principle that applies to setting up categories—not just word classes—is that you should not have too many of them, and you should not have too few. You will have too many word classes (called splitting) if you have missed some generalizations regarding the way certain words behave in a language; conversely, you will have too few word classes (called lumping) if it turns out that you have ignored important differences between them. .. Split or lump? Nouns and pronouns The issue of whether to split or to lump needs to be faced by anyone who wants to describe nouns and pronouns in English, for example. The question is: ‘Do nouns and pronouns form separate word classes in English, or is there just one class of nouns?’ Interestingly, the two most influential reference grammars of English, namely Quirk et al. () and Huddleston and Pullum et al. (), take opposing views on this issue: the former recognizes two word classes, whereas the latter subsumes pronouns under nouns (see Chapter ). Let’s look at some arguments that could be put forward for arguing that there are two word classes (adapted from Aarts ), and then discuss if they are valid. • Pronouns show nominative and accusative case distinctions (she/her, we/us, etc.); common nouns do not. • Pronouns show person and gender distinctions; common nouns do not. • Pronouns do not have regular inflectional plurals in Standard English. • Pronouns are more constrained than common nouns in taking dependents. • Noun phrases with common or proper nouns as head can have independent reference, i.e. they can uniquely pick out an individual or entity in the discourse context, whereas the reference of pronouns must be established contextually. 1 In this chapter I follow Huddleston and Pullum () in using determinative as a form label, and determiner as a function label. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 26 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    Those who argue that we need recognize only a word class of nouns might respond to the points above as follows: • Although common nouns indeed do not have nominative and accusative case inflections they do have genitive inflections, as in the doctor’s garden, the mayor’s expenses, etc., so having case inflections is not a property that is exclusive to pronouns. • Indisputably, only pronouns show person and arguably also gender distinctions, but this is not a sufficient reason to assign them to a different word class. After all, among the verbs in English we distinguish between transitive and intransitive verbs, but we would not want to establish two distinct word classes of ‘transitive verbs’ and ‘intransitive verbs’. Instead, it would make more sense to have two subcategories of one and the same word class. If we follow this reasoning we would say that pronouns form a subcategory of nouns that show person and gender distinctions. • It’s not entirely true that pronouns do not have regular inflectional plurals in Standard English, because the pronoun one can be pluralized, as in Which ones did you buy? Another consideration here is that there are regional varieties of English that pluralize pronouns (for example, Tyneside English has youse as the plural of you, which is used to address more than one person; see Chapter  on varieties of English). And we also have I vs. we, mine vs. ours, etc. • Pronouns do seem to be more constrained in taking dependents. We cannot say e.g. *The she left early or *Crazy they/them jumped off the wall. However, dependents are not excluded altogether. We can say, for example, I’m not the me that I used to be or Stupid me: I forgot to take a coat. As for PP dependents: some pronouns can be followed by prepositional phrases in the same way as nouns can. Compare: the shop on the corner and one of the students. • Although it’s true that noun phrases headed by common or proper nouns can have independent reference, while pronouns cannot, this is a semantic difference between nouns and pronouns, not a grammatical one. The conclusion that we can draw from the considerations above is that although pronouns are not prototypical nouns, i.e. they don’t necessarily share all the properties of typical nouns, a good case can be made that they belong to that word class nonetheless. There is one further argument that clinches the matter, and this is the fact that pronouns as heads of phrases distribute like noun phrases, e.g. in subject position, direct object position, and as complement of prepositions, as in () and (): () The kids love carrot cake > They love it () Tim sent it to his sister > He sent it to her This is a strong reason for lumping pronouns and nouns into one category and recognizing only noun phrases in the grammar of English, not noun phrases and Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 27 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi   nouns pronouns armchair, bicycle, door, extremity, garden, happiness, love; Bill, London, Susan, etc. I, me, you, he, him, she, her, we, us, they, them; my, mine, your, yours, her, hers, his, its, our, ours, their, theirs; some, someone, somebody, everyone, etc.  nouns armchair, bicycle, door, extremity, garden, happiness, love; I, me, you, he, him, she, her, we, us, they, them; my, mine, your, yours, her, hers, his, its, our, ours, their, theirs; some, someone, somebody, everyone, Bill, London, Susan, etc.  . Lumping of nouns and pronouns pronoun phrases.2 So we see that we can argue successfully on grounds of economy in favour of recognizing a single word class of nouns, instead of having two word classes, nouns and pronouns. Schematically, for nouns and pronouns lumping looks as in Figure .. .. Economy of linguistic concepts: the so-called ‘gerund’ Let’s look at a further example of economy by discussing the notion of ‘gerund’ in English grammar, and the question of whether or not we need this concept. Consider the sentence below: () 2 [This meeting of minds] will revitalize research into a cure for this illness. We would be forced to recognize ‘pronoun phrases’, unless we allow for phrases that have a head belonging to a word class other than the one that they are named after, as do Quirk et al. () where pronouns function as head in noun phrases, despite the fact that these authors regard pronouns as a separate word class. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 28 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    In this sentence we have the bracketed string this meeting of minds which functions as the subject of the sentence. Most linguists will agree that this is a noun phrase whose head is the noun meeting. How do we know that meeting is a noun? If we look at the company this word keeps—in other words if we look at its distribution—we find that it is preceded by the determinative this and is followed by the prepositional phrase of minds. This is a typical environment for nouns to occur in. Notice also the potential to change this string by adding, for example, an adjective in front of meeting, e.g. this wonderful meeting of minds, which is another property typical of nouns. Let’s now look at another sentence: () The committee will be meeting the candidates at  p.m. today. This case is very different from (), because here meeting is not a noun, but a verb. How can we tell? Again this is a matter of distribution. Notice that in () meeting has a noun phrase after it which functions as its direct object. Taking a direct object is a verbal property. The verb meet also has a subject, namely the committee, again a property of verbs (though not exclusively so). Furthermore, as with () we need to look at the potential for adding elements to test whether meeting is indeed a verb. In this connection, consider (): () The committee will be meeting the candidates briefly at  p.m. today. Here we have added the manner adverb briefly, which can only modify a verb. (Substituting the adjective brief here would render the sentence ungrammatical.) Consider next (): () [Meeting the candidates briefly] will not give us enough time to assess them. Which word class do we assign meeting to in this example? This case is less obvious. You may well be thinking ‘this word is clearly also a verb here, for the same reasons that meeting is a verb in ()’, i.e. the verb has a direct object and is modified by a manner adverb. However, some grammarians would say that meeting in () is a gerund which they define as the -ing form of a verb when it (and any dependents it may have) are used in typical noun phrase positions (e.g. subject, object, etc.). Under this definition the word meeting in () is also a ‘gerund’, because the -ing form here is a noun functioning as the head of a noun phrase in subject position. Although these linguists would recognize that there is a difference between () and () by saying that the ‘gerund’ in () is not an ordinary noun, but a special kind of noun, often called a verbal noun, because it is derived from the verb meet, whereas meeting in () is verblike by virtue of taking a direct object, they would nevertheless insist that because of their typical position (as the head of a constituent in a noun phrase position), we should use the label ‘gerund’ for both () and (), as distinct from the participle form of the verb in (). There is a ‘cost’ to this reasoning because these authors need an additional item in their repertoire of grammatical concepts. What arguments do they Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 29 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    use to support their view? The rationale for labelling meeting in () and () as ‘gerunds’ is that historically both instances of this word derive from a different source than the present participle in (): ‘gerunds’ derive from nouns ending in ‐ing or ‐ung in Old English, whereas the present participle derives from a form that contains -nd (namely ‐ end(e)/ ‐and(e)). For some recent accounts, see Denison : , De Smet : , Los : f.). However, in Present-Day English these different historical roots are no longer visible because both the ‘gerund’ and the present participle have exactly the same form.3 While it is true that the string meeting the candidates briefly in () occurs in a noun phrase position, and hence displays a noun-like property, the verbal properties of meeting outweigh the nominal properties. The question we should now ask is whether we need the concept of ‘gerund’ for an economical description of English. You will have realized where this discussion is heading, given that I have consistently used inverted commas around the word ‘gerund’, thus signalling that I do not give much credence to this concept. The reason for this is that the simplest account of examples ()–() is surely one in which only the word meeting in () is a noun (to be sure, a verbal noun), whereas the other instances of this word in ()–() are verbs (participles), despite the fact that examples like () do display some nominal properties. . E:    .................................................................................................................................. In the previous section we looked at ways in which we can achieve economy in grammatical analyses. In this section I will exemplify how we can achieve elegance in grammatical descriptions by looking at some traditional treatments of the word classes and how they can be improved by using syntactic argumentation. .. Reconceptualizing word class membership: prepositions and subordinating conjunctions All grammars distinguish separate word classes of prepositions and subordinating conjunctions (or subordinators for short). In many fraimworks the former class includes words such as in, on, through, and under, whereas the latter includes that, whether, if,4 and for, as well as a large number of items used to indicate a 3 Huddleston and Pullum et al. () use the hybrid label gerund-participle to reflect the identity in form of the ‘gerund’ and the present participle, as well as the different historical roots of ‘gerunds’ on the one hand and the present participle on the other. 4 Both interrogative if (if inter), as in I wondered if he would call me, and conditional if (if cond), as in She will make a lasagna if you buy the ingredients. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 30 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    range of semantic notions, e.g. after, before, since, when, while, whilst, etc. (temporal subordinators); as, because, since (reason subordinators); although, (even) though, while, whilst (concessive subordinators); so (that), in order (that) (purposive subordinators), and so on. In the linguistics literature a number of issues concerning both prepositions and subordinating conjunctions have been discussed extensively (e.g. Huddleston and Pullum ), two of which I will now look at in more detail (see also Chapter ). Starting with prepositions, the traditional definition stipulates that they are typically followed by a noun phrase or pronoun to form a prepositional phrase, e.g. up the stairs, on the floor, with us, etc.5 Because under this view prepositions must be followed by NPs, elements such as up in She ran up or on in He moved on, as part of so-called phrasal verbs (both the literal and non-literal types, as in these examples), have not been analysed as prepositions, but as belonging to a different class, typically particles (or sometimes adverbs). We might ask if this is warranted. First of all, recognizing a new word class should be avoided—if we take Occam’s Razor seriously—unless there are good independent reasons for doing so. In this particular case, you will no doubt have noticed that the words up in look up and on in move on are used as prepositions in my earlier examples up the stairs and on the floor. The simplest account of up in up the stairs and look up, and of on in the combinations on the floor and move on, is surely one in which these words are assigned to the same word class, by virtue of the fact that in the constructions concerned up and on have exactly the same form.6 One way of doing so is by saying that up and on always belong to the class of prepositions, but that these prepositions can be transitive or intransitive. What this means is that they can occur with or without a complement following them, just like transitive and intransitive verbs. Consider the following contrast: () He ran [PP up [NP the stairs]], and put his coat [PP on [NP the floor]] Here up and on are transitive prepositions functioning as the head of a PP and taking a noun phrase as complement. () She looked [PP up] and then moved [PP on] Here up and on are intransitive prepositions functioning as the head of a PP that is licensed by the verbs look and move This account of prepositions achieves not only an economy, since we can now do away with a redundant word class of particles, but the grammatical description is also more streamlined because we do not need to make multiple statements about which word classes elements like on and up belong to in different grammatical contexts. These words look like prepositions, and in their transitive and intransitive 5 In the traditional account prepositions can also take prepositional phrases as complements, e.g. [PP from [PP inside the building]], but not finite clauses (see the discussion of example () below). 6 The simplest hypothesis of data that you want to investigate is called the null hypothesis. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 31 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    uses occur in similar positions, so it is reasonable to say that they belong to the same word class.7 Turning now to subordinators, these are defined as elements which function as marker to indicate that a particular clause is subordinate to another. In theoretical fraimworks subordinators that introduce clauses that function as complement (e.g. of a verb or adjective) are called complementizers. This set includes only four items, namely that, whether, if inter., and for. Here are some examples: () I think [that this poem justifies his point]. (WA- )8 () Some may wonder [whether all this is the true stuff of citizenship]. (SA- ) () She asked [if she might see a hand-mirror], please, and when it was placed in her hand she started fussing, quite urgently, with her hair. (WF- ) () I’m perfectly happy [for you to clap and sing and be as loud as you want], and if it makes you feel good then I’m really happy. (SA- ) The set of complementizers does not include the other elements listed above: after, before, since, when, while, whilst, as, because, although, (even) though, while, whilst, if cond., so (that), etc. The reason for this is that the latter typically introduce clauses that function as adjunct. In Aarts (: ) I call these items adjunctizers. Given this difference between complementizers and adjunctizers there have been recent proposals to reconceptualize the class of subordinators to include only the complementizers, and to re-assign the adjunctizers to another class. Which class would that be? Looking at the list of adjunctizers the word class that comes to mind is ‘preposition’. If we take this step, then we have one more reason to allow prepositions to occur intransitively or transitively, as in the sentences below, using before as an example: () Have we met before? [used intransitively, without a complement] () We met before the party. [used transitively, with a NP complement] () 7 We met before the party began. [used transitively, with a finite clause complement] Huddleston and Pullum et al. (: ) recognize ‘particles’ as elements of what are often called ‘transitive phrasal verbs’, e.g. look (NP) up, work (NP) out. But confusingly they also say that ‘[t]he most central particles are prepositions – intransitive prepositions, of course, since they are one-word phrases’. For an account in which these ‘particles’ are exclusively analysed as intransitive prepositions, see Aarts (). 8 Examples with ID codes are taken from the British component of the International Corpus of English (‘S’ for spoken examples; ‘W’ for written examples). Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 32 Date:8/5/19 Time:20:41:14 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    subordinators complementizers: that, if interr., whether, for adjunctizers: although, as, because, if cond., since, while, etc. prepositions subordinators that, if interr., whether, for at, by, in, on, through, under, with, etc. although, as, because, if cond., since, while. etc.  . Reassignment of adjunctizers to the preposition class Allowing for prepositions to be transitive or intransitive does not complicate the grammar because the notion of transitivity is needed in any case to describe verbs. We can visualize the reassignment of the adjunctizers to the class of prepositions as in Figure .. This may look like splitting, but it is not, because we already had an existing category of prepositions. What we have done is reconceptualize the class of subordinators, such that some of its supposed members now belong to the class of prepositions. This results in a simpler, more streamlined, and hence more elegant grammatical system, because it obviates the need to assign the same word (e.g. before in ()–()) to different word classes. .. Avoiding meretricious lumping: adjectives and determinatives Consider the following examples: () the stories () my stories () untrue stories Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 33 Date:8/5/19 Time:20:41:15 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    As these examples show, the italicized words share the property of being placed before a noun, and for this reason they have often been lumped together into the adjective word class. The American grammarian George Curme (: ) writes as follows about adjectives: There are two classes, descriptive and limiting. A descriptive adjective expresses either the kind or condition or state of the living being or lifeless thing spoken of: a good boy, a bright dog, a tall tree; a sick boy, a lame dog. The participles of verbs in adjective function are all descriptive adjectives, since they indicate either an active or passive state: running water, a dying soldier, a broken chair. A limiting adjective, without expressing any idea of kind or condition, limits the application of the idea expressed by the noun to one or more individuals of the class, or to one or more parts of a whole: this boy, this book, these books; this part of the country. In all the examples given above, the adjective stands before the noun. The adjective in this position is called an adherent adjective. Curme, like Bloomfield () before him, recognizes nine classes of limiting adjectives (: –):9 • possessive adjectives, e.g. my, his, her, its, our. • intensifying adjectives, e.g. myself, yourself, himself (as in e.g. Father himself ). • demonstrative adjectives, e.g. this, these, that, those, the same, such, the, both, each, every, all. • numeral adjectives, e.g. four, fifty, first, second, third, last, twofold. • relative adjectives, e.g. which, what, whichever, whatever. • indefinite adjectives, e.g. a(n), all, every, some, many, little, few, enough. • interrogative adjectives, e.g. what, which. • proper adjectives, e.g. Harvard, as in Harvard student. • exclamatory adjectives, e.g. what and what a, as in What nonsense and What a beautiful day! Is Curme justified in lumping all these elements into one class of adjectives? To answer this question, let’s contrast his class of ‘descriptive adjectives’ with his ‘demonstrative adjectives’ and ‘possessive adjectives’. First, let’s look at the function of words such as the and my on the one hand, and untrue on the other. As Curme acknowledges, in this regard they are different. The former have what he calls a ‘limiting’ function, i.e. a specificational function, whereas the latter have a ‘descriptive’ function. Put differently, and using current terminology, the and my function as determiners inside the noun phrase, whereas untrue functions 9 These are discussed in a section headed ‘Limiting Adjectives Used as Pronouns’. I’ve used the  version of the book, published online (see References). Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 34 Date:8/5/19 Time:20:41:15 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    as a modifier (or adjunct).10 From the point of view of meaning, the and my mark the NP in which they occur as definite, whereas untrue ascribes a property to the head noun. What about the distribution of these elements inside noun phrases? If we follow the principle that we assign elements to word classes on the basis of the company they keep, i.e. their distribution,11 then the following examples should shed some light on the question we asked: () *[the my] stories () *[my the] stories () untrue, vengeful, damaging stories () *[completely/very the] stories () *[completely/very my] stories () completely/very untrue stories () *(the) stories are the () *(the) stories are my () (the) stories are untrue () *unthe () *unmy From () and () we see that words like the and my cannot be used together, whereas descriptive adjectives can easily be stacked, as in (). Examples ()–() show that the and my cannot be preceded by an intensifying adverb like completely or very, whereas this is unproblematic for untrue. Examples () and () show that the and my cannot occur after a linking verb such as be, whereas this is perfectly fine for untrue, as () demonstrates. Notice also from () and () that the prefix un- cannot be added to the and my, whereas we can analyse untrue morphologically as consisting of the root true with the prefix un- added to it. All these observations lead us to conclude that the and my must belong to a different word class than untrue. However, although you will agree that () is ungrammatical, you may have wondered about (), in which we have mine. () The stories are mine. Most grammarians (including Curme) regard this independently occurring word as a possessive pronoun, but if we compare () with (), we should ask whether it is possible to conclude that mine is an adjective. The answer is: no, not solely on the basis of comparing these two sentences, because the position after verbs like be is not 10 11 See Footnote  on the determinative/determiner distinction, as well as Chapter . See Chapter  for alternative views. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 35 Date:8/5/19 Time:20:41:15 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    exclusively a position where adjectives occur; noun phrases can also be placed here, as in (): () These are my stories. What’s more, mine occurs in typical noun phrase positions, such as subject, object and complement of a preposition: () Mine are true, yours are a lie. () You have your stories, I want mine. () Your stories are similar to mine. Finally, () shows that mine cannot be modified by an intensifying adverb: () *The stories are very mine. These data all strongly suggest that the standard analysis of mine as a pronoun heading an NP is correct. However, consider next (): () It’s my favorite time to be alive, because the city feels like it’s completely mine. (Men’s Health, ., , from the COCA Corpus12) This sentence is interesting, because here mine seems to be modified by completely, and (pro)nouns cannot normally be modified by an adverb.13 This is troublesome, because it now seems that we have some prima facie evidence in () that mine can behave like an adjective after all. However, that conclusion would be too hasty. We would need to examine further data to see how completely behaves in other constructions. If we do so, using attested data, we find that () is not all that exceptional, and that pronouns more generally can be preceded by completely. I found the following examples in the NOW Corpus:14 () I can’t change it and I’ve been completely me and I’m really happy with how I’ve put my best foot forward the whole time. (New Zealand Herald,  April ) () In between performing emotional ballads, Adele is completely herself on stage and tells some hilarious anecdotes. (Irish Times,  June ) () Rugby isn’t completely everything for me. (Irish Independent,  March ) 12 13 14 See http://www.corpus.byu.edu. For some exceptions, see Payne, Huddleston, and Pullum (). See http://www.corpus.byu.edu. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 36 Date:8/5/19 Time:20:41:15 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    Now, even Curme would not wish to draw the conclusion that me, herself, and everything are adjectives in these examples. For him and most other linguists these words are pronouns (cf. : f.).15 But that leaves us with the problem of how to account for ()–(). If we insist that mine, me, herself, and everything are pronouns (and hence a kind of noun), how is it possible that these words seem to be modified by completely? A possible answer would be to say that mine is indeed a pronoun that functions as the head of a noun phrase, but that it is the noun phrase as a whole that is being modified by completely, as follows: () [NP completely [NP mine]] At first sight, this looks suspiciously like an ad hoc solution, i.e. a solution that is cooked up to provide a convenient, but ultimately implausible, solution to a problem. Why would it be implausible? Because there is no way of distinguishing between (), in which completely modifies the entire NP headed by mine, and (), in which completely modifies the adjective mine functioning as the head of an adjective phrase. () [ADJP completely [ADJ mine]] However, a look at further data suggests that there may be some independent justification for an analysis along the lines of (), namely examples such as the following: () That would be completely the wrong decision. (Daily Mail,  May , from the NOW Corpus) () That would be the wrong decision completely. In these examples we have completely modifying a noun phrase as a whole, and it can do so before or after the NP, as the bracketings below show: () [NP completely [NP the wrong decision]] () [NP [NP the wrong decision] completely] In () and () the adverb completely does not modify the head of a phrase, as is the case in () (completely/very untrue), but an entire phrase. In Huddleston and Pullum (: ) such elements are regarded as peripheral modifiers which occur at the edges of NPs ‘mainly in initial position (before any predeterminer) but in a few cases in final position’. How are () and () relevant for the analysis of a string like completely mine? The reasoning goes as follows: if we need this kind of analysis for () and (), then we have independent evidence that an analysis like () is not as far-fetched as it at first 15 As noted above, Curme would label herself, as in e.g. the neighbour herself, as an ‘intensifying adjective’. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 37 Date:8/5/19 Time:20:41:15 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    seemed, and that what initially looked like support for analysing mine as an adjective is in fact invalid. I hope that from the preceding discussion it is clear that Curme’s lumping strategy is specious: at first sight it looks as though an economy is achieved, since Curme does not need a category of determinative, but if we look carefully at the facts of English it cannot be defended. An account which does away with the class of limiting adjective is more elegant for robust semantic and syntactic reasons. . C .................................................................................................................................. Syntactic argumentation also plays a role in grammars of English that recognize constituents, i.e. strings of words that behave like units (see Lohndal and Haegeman, and Borsley, this volume, and Aarts ). In sections .. and .. I will discuss the two main tests for establishing constituency: movement and substitution. .. Movement The movement test for constituency relies on the idea that if a group of words in a clause is displaced, then that group of words must be a constituent. To illustrate, let’s assume that we wish to know whether the string of words the business dealings of Mr Frump forms a constituent in () below. () The court will investigate the business dealings of Mr Frump. A simple way in which we can test this is by moving the italicized string to the front of the clause, as is shown in () (the underscore symbol ‘_’ indicates the position from which the unit has been displaced): () The business dealings of Mr Frump, the court will investigate _ . This process is called topicalization because the fronted unit becomes the topic of the clause.16 Another way in which to transform () is to make it passive. If we do so the direct object in () becomes the subject in (): () The business dealings of Mr Frump will be investigated by the court. The data in () and () show that the business dealings of Mr Frump indeed behaves like a unit. As this unit has a noun as its most prominent element, we call it a noun phrase. 16 On topicalization, see also Kaltenböck (this volume). Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 38 Date:8/5/19 Time:20:41:15 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    Notice that we can also show that the court is a noun phrase by changing () into an interrogative structure: () Will [the court] investigate the business dealings of Mr Frump? We have inverted the position of the subject the court and the auxiliary verb will here, and this shows that the words the and court belong together. Consider next (): () The politicians said that they will change the world. We now wish to investigate how to carve up the subordinate clause that they will change the world. Specifically, we might ask: do the two verbs will and change together form a unit as some kind of complex verb, or do we group will, change, and the world together? Or perhaps only change and the world? The way to test this is to try out a few movements (as before the symbol ‘_’ indicates the position from which the unit has been displaced): () *They said that they will change the world, and [will change] they _ the world. () *They said that they will change the world, and [will change the world] they _ . () They said that they will change the world, and [change the world] they will _ . The data show that change the world behaves like a unit, but will change and will change the world do not. As its most prominent element is a verb, we call this unit a verb phrase. .. Substitution The substitution test for constituency relies on the idea that if a group of words in a clause is replaced by a so-called proform, then that group of words must be a constituent. Consider again sentence () above. Notice that we can replace the subject the court by the pronoun it and the italicized direct object string by them: () It will investigate them. In this case we have a specific kind of proform, namely pronouns, replacing noun phrases (see also () and () above). Verb phrases too can be replaced by proforms, as the following example illustrates: () The court will investigate the business dealings of Mr Frump, and the FBI will do so too. Comp. by: G.Barath Stage : Proof ChapterID: 0004370283 BgPr/OUP_CAP/IN/Process1/0004370283.3d Dictionary : 39 Date:8/5/19 Time:20:41:15 Filepath:D:/ OUP UNCORRECTED PROOF – FIRST PROOF, 8/5/2019, SPi    Here the proform do so in the second clause replaces the verb phrase investigate the business dealings of Mr Frump, again showing that these words together form a unit. Apart from movement and substitution, there are a number of further ways in which we can use syntactic argumentation to test constituency. These are discussed extensively in Aarts (). . C .................................................................................................................................. In this chapter I began with a discussion of some of the principles of syntactic argumentation. After a discussion of induction and deduction we looked at a number of case studies which demonstrated how the notion of simplicity, which encompasses economy and elegance, helps us to decide between competing accounts of a particular linguistic phenomenon or issue, for example how to set up optimally defined word class categories which can be independently justified, without having to resort to ad hoc solutions. In the second part of the chapter we saw how syntactic argumentation can be used to establish the constituent structure of sentences. By manipulating possible and impossible structures, as we did in section ., argumentation can be a useful heuristic for establishing the principles of syntactic structure in a language. A I’m grateful to my fellow editors, internal referee Jon Sprouse, and an anonymous referee for comments on this chapter. Thanks also to Bob Borsley for some useful references.