We thank Martin Haspelmath for allowing us to reply to his review of our book. We have divided our reply in two parts. In the first part, we make explicit what our choices have been in writing this textbook and why we made them. We believe that quite a few of Martin’s criticisms relate to these, often didactic, choices. In the second part we reply to some of the more detailed comments about the first four chapters.
In writing Introducing Syntax, we formulated the following central objective for ourselves: To inform students about syntactic theory, in a way that makes them enthusiastic enough to pursue syntax/linguistics further. In an attempt to achieve this goal, we made a number of choices. The fact that one has to make choices when writing a textbook is inevitable, the particular choices we made are not. Below, we will explicate the choices we made with a rationale for why we made them. Please realize that these choices are hardly unique to our textbook, but it is helpful to make them explicit and state why we stand by them.
Choice 1: The book is a didactic text intended for beginning students.
Although this may be evident for an introduction, it is important to make explicit that our primary goal was not to provide a state of the art for fellow linguists (although we are glad to hear that Martin found it very useful as an update). This choice has far-reaching consequences. It does not only entail that the materials are simplified here and there, it also means that they are presented in a way that serves the overall narrative, rather than in a way that is topically or historically correct. To give one example, we introduce the Elsewhere Principle as a way to understand the competition between pronouns and reflexives in Chapter 5.3, only to note in Chapter 8 that the concept is also useful in morphology. The choice also entails that many questions that come up for an average or advanced linguist are not addressed, simply because it is unlikely for a beginning student to come up with them (and if one does, good for him/her!), such as the difference between features and feature values.
Choice 2: The book restricts itself to presenting a generative/minimalist approach.
Since the book teaches syntactic theory, it is important that the student understands the nature of theorizing. What is a theory? How does one provide arguments for a theory, conceptually and/or empirically? What do you do when the theory does not work: start over or solve the problem (both avenues are shown throughout the book)? The book dedicates quite some attention to these issues. We also make it a point that a theory is never finished. In the afterword, we present four major problems for the theory in quite some detail, and we could have presented other or more problems.
Since a beginning student is not very familiar with theorizing (let alone syntactic theorizing), it is important that (s)he gets to see what syntactic theory can do, where it can get you, how one hypothesis provides the scaffolding for the next one, etcetera. In order to get a feel for this, the student needs to see some results, and not get stuck somewhere halfway. One result, very central in our book and the motor of the narrative, is the idea that very distinct and complicated data can be reduced to a small set of principles, the general goal of science. This goal is relatively easy to reach if one focuses on one particular theoretical approach, and show on what grounds each step in specifically that approach has been taken. It also entails that one chooses a specific syntactic framework, and not remain agnostic about possible choices. Note that, although we build up the theory using technicalities that are adopted from current minimalism, we do not adhere to or to try to persuade students about most of its philosophical or biological underpinnings, such as innateness claims or conjectures about the biological function of language. In addition, several of the technicalities we use are fully compatible with, or straightforwardly translatable into, assumptions made in other linguistic frameworks.
In short, the impetus for the choice we made is didactic, namely the desire to show a product of extensive theorizing. This gives the student a good idea of what to aspire to.
Choice 3: The book focuses on a single language, namely English.
The third choice was not informed by didactics and is therefore more arbitrary. One could imagine an introduction that chooses a particular framework and then looks at different languages, or language families. Instead, we have decided to focus on English and only make some comparative remarks when we felt this contributed to the understand of a particular issue for the student (not the linguist). The fact that we focus on English leads Martin to object that, although we only consider one language, we at the same time claim that the principles underlying English syntax hold for other languages as well. This criticism is fully understandable, but nevertheless we had good reasons to pursue out chosen approach.
In the beginning (in the section ‘About this book’) we say that ‘syntax (and especially its level of formalisation and abstraction) can quite often be a surprise for students, especially those enrolled in an (English) language and literature/culture programme. The challenge is to make this formal theory accessible and interesting without oversimplifying.’ In our introduction we say that ‘we are going to show you what the study of the English language can contribute to this bigger enterprise, understanding the human brain.’ We explicitly chose, then, to see how much can be said about English syntax before comparing it to other languages. By first providing a detailed theoretical analysis of English, students become well equipped for future courses in which they can compare what they have learned with what other languages seem to show. An alternative would have been to first compare a number of languages and then say what can be shown theoretically. For a syntax course in a general linguistics department that might perhaps make more sense, and there are also various textbooks doing this. For students of English it might make things much harder and less accessible, though.
As said, we made a choice and other choices could have been made as well. However, we seriously wonder whether any other course material that can be used in a first-year course for students with no background in formal theory can reach our own objectives any better.
Choice 4: Simplicity and elegance are very, very important.
Related to the previous choice, focusing on one language allows us to dive into a number of fairly complicated grammatical constructions, which would be harder to do under a cross-linguistic perspective. In the endeavor to analyse the properties of these constructions and come to some overall synthesis, various abstract notions are introduced. Martin wonders what the purpose of this additional abstractness is: ’Do we need it to understand which English sentences are well-formed and which ones are not? Not really. Instead, the abstract analyses are justified by the uniformity of mechanisms.’ Here we disagree. One can describe all rules of English without abstract notions, but that is not the same as understanding these rules of English, or making progress with this. Uniformity leads to better understanding if what appear to be completely different phenomena turn out be variations on the same theme. Then only that theme is what needs to be understood. In other words, achieving uniformity is not a goal separate from understanding the (un)wellformedness of sentences but a a way towards this understanding. In short, what we want to show students is how theories are built up and what makes theories good, and simplicity, uniformity and elegance are hallmarks of any good scientific practice
Apart from teaching syntax, the book is also intended to make students enthusiastic about it. We think that the choices we made are helpful in this endeavor. If we now compare our choices in the light of Martin’s review and his vision on the teaching of linguistics, we feel that our idea of a beginner’s textbook fundamentally clashes with what Martin would want to see in an introductory course. He argues that it would be better to teach students those ideas and concepts that most linguists, irrespective of their background, take for granted, an approach that is essentially less speculative. We believe that the chances of achieving our objective stated at the beginning of this section would be seriously reduced and partly undermined, though maybe Martin does not share our goals or has different students in mind.
Two remarks are in order, however. First of all, a slower start may turn students away that would have become more enthusiastic about the field after more headway has been made, even if some of that headway is speculative. Second, for many academic major programmes, an introductory course in syntax is all there is, and this is your chance to grab them. Students that are not enthusiastic you won’t see back, whereas students that are enthusiastic can at the next stage be shown anything that the teacher finds important: the speculative nature of some of the conclusions, other frameworks, other languages, relations to other subfields, such as psycholinguistics or neurolinguistics, empirical methods, etcetera. It is for this reason that Introducing syntax is not only trying to be argumentative, but sometimes also persuasive. But every theoretical choice in the book is argued for, and the reasons for drawing certain conclusions are always made explicit, thereby keeping the number of unmotivated assumptions to a minimum.
Detailed points about the first four chapters
Apart from the more general criticisms that we addressed above, Martin also lists a number of things he takes issue with in the first four chapters. It would stretch too far to address each and every point, but below we make explicit why we made certain choices and how, in our eyes, they serve the general purpose of the book that we outlined before.
The first criticism on chapter 1 is that ‘[t]he text makes it appear that it is “interesting and striking” that every word belongs to a category, but this is so by definition, not because we discovered anything.’ That is not true. It has been discovered that there are words whose distribution is fully similar to other words. Proper names are already an example. The distribution of Mary is the same as that of Susanne. What we show in the book is that this opens up the possibility of formalizing rules in terms of categories (i.e. groups of words with the same distribution) and that this leads to (i) a theory that is simpler than having to spell all syntactic rules per word; and (ii) a theory that allows is to make testable predictions. One such prediction is that every lexical item should belong to some category; another prediction is that differences that exist between elements that belong to the same category should follow from additional (sub)categorial differences.
Our primary goal here is to teach students that by assigning an abstract, formal property to words, one can make a stronger predictive theory about language than without doing so. Such abstract, formal properties we refer to as ‘features’. Students mostly enter class already knowing the ‘features’ [Noun] and [Verb] but mostly without realizing that they are abstractions and without knowing how we got to them. A detailed discussion of these, then, sets the stage for more abstractions to come, and how one can justify them.
Martin states what we take to be features are not features in the strict definition of Corbett: ‘[b]ut it turns out that “feature” actually means “value” of an attribute-value pair (and is thus a value, not a feature in Corbett’s 2012 sense). So saying that car has the “feature” [N] means that its attribute [±N] is set to the value [+N] – in other words, there is no “syntactic category” as an attribute that could take on multiple values.’ For Martin this leads to more rather than less confusion.
As unhappy we are about triggering confusion with a well-established and respected scholar, we would be much unhappier if at this stage students, in their first or perhaps second syntax class, would be totally confused. Of course, the notion of a feature, as virtually every technical notion we introduce, is more complex and requires more discussion than we provide in the text. But students at this stage still lack the necessary knowledge and skills to fully grasp these differences. Here it suffices that syntax works on the basis of abstract, formal properties. The exact nature of these properties is to be discussed later (and, for instance, in chapter 8 we talk about the differences between features and feature values), when the theory developed so far requires such refinements. The same applies to the claim that proper names carry both [D] and [N] features. Later on in the book, when functional heads are introduced, we separate DPs form NPs. That would be a place where the syntactic distinction between proper names and other nouns could be further disentangled, not in chapter 1.
In the discussion on chapter 2 Martin accuses us of arguing ‘against a strawman alternative that does not use hierarchical structures and that nobody has ever advocated’. He refers to the syntactic operation Glue, which glues together categories in a very rudimentary left-to-right fashion: the structure of cheap red wine would be [[A]+[A]+[N]]. Even if Martin were correct about this being a strawman, there is still a good reason for including a discussion of it. The Glue operation looks simpler than the Merge operation, certainly for a beginning student. It is not self-explanatory that sentences must be hierarchically ordered. Prima facie, one may wonder why grammar rules should not be formulated in terms of linear orders. It is therefore important to show that it is wrong, and that phrase- and sentence-internal constituency exists, to which the grammar is sensitive.
At the same time, Glue sets the stage for what we call List. An important problem for the operation Glue (namely, [A] can be glued to another [A], whereas cheap red is not a grammatical constituent in English) can be circumvented by letting the grammar list the grammatical and ungrammatical strings: [A]+[N] is possible, whereas [A]+[A] is not. This conception of syntactic structures is very close to the schemas used in Construction Grammar. More generally, it is not true that proposals claiming that syntax goes without hierarchy are not defended. For instance, Frank et al. in a press release on their 2012 paper claimed that they have discovered that sentence structures are not hierarchical (though, admittedly, their paper actually never corroborates this claim). Also, among neurolinguists, there is no consensus either that sentence structures must be hierarchically structured. In other words, we don’t think that we introduce a strawman.
Another criticism concerns the claim that every constituent is endocentric. Martin does not deny the existence of endocentricity (even though various syntacticians in MGG nowadays question endocentricity, including Chomsky), but raises the possibility that certain constituents are not endocentric, e.g. clauses or coordinates expressions. Indeed, both endocentric and non-endocentric analyses of clauses or coordinated expressions have been formulated. We do not discuss coordinated expressions in our book (as we needed to restrict ourselves), but in chapter 4 we argue indeed that clauses are best analysed in endocentric terms. This, we think, is both conceptually and didactically advantageous, as the way we presented Merge and the notion of constituency takes every constituent to be endocentric. Of course, this may be wrong, but then this has to be shown. This would require first spelling out the exact predictions that an endocentric approach to clauses makes and subsequently showing that these predictions are not met. It only makes sense to present alternatives accounts if the data analysed so far give good reasons for that. At this stage in the narrative, this is not the case, however, and therefore we deliberately refrained from doing this.
One of Martin’s criticisms raised about chapter 3 is that a ‘key question was not answered for me: Is the chapter about semantic restrictions or about syntactic restrictions?’ In this chapter, we say indeed that it is the meaning of a verb that determines the kind and number of theta-roles that it can assign, and later on in the book (e.g., in chapter 5) we make more explicit that this is actually a semantic restriction. Of course, one may wonder why such semantic restrictions should be incorporated in a syntax textbook, but showing what syntactic theories should explain also amounts to showing what they do not need to explain.
This, furthermore, helps us motivate the notion of PRO. We argue that a nominal may not receive two theta-roles and that therefore control-constructions require PRO. Saying that an argument can receive two theta roles begs the question why a sentence like John kills cannot mean John kills himself. Martin wonders whether one couldn’t equally well ‘say that nominals may bear two semantic roles only if one of them is assigned by a non-finite verb?’ One could of course, but from a semantic perspective it would really be questionable to state that non-finiteness would affect the meaning of a verb in this sense. Given the structure of the theory built up so far, a zero analysis is a more likely hypothesis. That doesn’t mean that we’re married to PRO. The PRO-hypothesis may very well be wrong, but assuming something like PRO is at this stage a more natural step than the alternative that Martin sketches.
The crucial idea we wanted to show, as we did in chapter 1, is that particular theoretical choices may lead to the postulation of zero elements, yielding again testable predictions. If these turn out to be incorrect, the theory should be modified or replaced by an alternative theory. But that does not entail that in showing how (linguistic) theories are built up, one should neglect the existence of zero elements; what needs to be shown is why linguists postulate them, and how their existence can be empirically and theoretically motivated and evaluated.
Finally, Martin criticizes our claims about case theory, primarily because we assume there to be a functional head, dubbed Fin by us, and more standardly known as I, T or AgrS in the generative literature, that is responsible for assigning or licensing nominative case when it if finite. However, we do not just posit that. We argue that what is needed linguistically is a restriction on the distribution of nominals that does not readily follow from semantic considerations and that such a theory requires a notion like finiteness for the licensing of nominatives. We subsequently address the question where finiteness is located and argue on the basis of distributional facts that we can identify a head, Fin, that is to be distinguished from the verbal head, even if the verb itself carries the finiteness morphology.
That works well to explain the distribution of nominatives, but raises questions about the nature of morphology. Why does one pronounce Mary s often sleep as Mary often sleeps? This is a typical example of where a particular theoretical innovation triggers new questions, and where the reader should be provided with a proper answer. Even though Martin says that ‘K&Z are not deterred by this and simply refer the reader to a later chapter’, we explicitly discuss this issue in chapter 8 and provide evidence for the separation of the Fin and V-heads. For one, we show that do-support can now be better understood. And, yes, we do so by assuming that English not is a head, a purely didactic choice for us. Even though the syntactic literature often takes not to be a head, Hedde has argued since 2004 that not is actually phrasal, but this would just unnecessarily complicate the explanation provided here. So again, didactics wins.
All issues above show that one has to make choices when writing textbooks, especially for young and untrained first year undergraduates in language and linguistics programs, the intended audience. One cannot immediately discuss all issues that are currently discussed in the syntactic literature, one cannot discuss all technical notions that scholars apply, and one cannot discuss the way particular phenomena can be attested in all languages investigated. One needs to make choices and in these, and the other chapters, our choice was to show how particular questions concerning data sets in one particular language lead to theoretical assumptions, whose consequences should be investigated as a next step. Our experiences so far suggest that this works successfully and many of our students have decided to study linguistics, even though their initial intention was not to do so. Naturally, that does not entail that such successes could not be yielded under other approaches. Presumably, they could, perhaps even better. But in order to be successful, any alternative approach would also be forced to make choices and therefore make itself vulnerable to the same kinds of criticisms as we have been.
Frank, S. L., Bod, R., & Christiansen, M. H. (2012). How hierarchical is language use? Proceedings of the Royal Society B: Biological Sciences, 279(1747), 4522–4531. http://doi.org/10.1098/rspb.2012.1741