Review of The Internally-headed Relative Clause Construction
|Review's editor: Monica Macaulay
In his book, Enrico explores the Internally Headed Relative Clause (IHRC) Construction across several American Indian languages. While Enrico's own work has largely been focussed on Haida, he is obviously familiar with the work of linguists working in many other American Indian languages, referring to data from a large number of these languages, particularly (Masset) Haida, Lakhota, Navaho, and Cuzco Quechua.
Working with Combinatory Categorial Grammar (CCG), the author focuses in on the semantics of IHRCs and the grammatical relationship between the head and the two clauses. He concludes that there are two basic syntactic mechanisms for IHRCs which languages can draw from, 1) the head category (defined as np/(s\np), and 2) the adjective category (defined as (np\np)). The head category accounts for the existence of relative pronouns and that, for many languages, definiteness takes matrix scope over the relative clause. The adjective category, on the other hand, is manifested where there are a) restrictions on arguments that can be heads, b) null heads, c) incorporated heads, and d) where definiteness restrictions apply.
Chapter 1, “The Gradual Discovery of Internally Headed Relative Clauses,” is an overview of what IHRCs are, why they are significant, and of previous literature on IHRCs in American Indian languages. IHRCs, where a noun phrase (np) semantically belongs to the matrix clause but is syntactically within the modifying clause, are particularly interesting because ''their syntax seems starkly at odds with their semantics'' (p. 1). While many linguists worked on IHRCs through the 1970s, Enrico points out several ''mysteries'' related to IHRCs which still remain, some of which were mentioned by other authors but not incorporated into the analysis, and others only obliquely mentioned by previous researchers. These mysteries include the fact that we find long distance dependency between the head and the matrix clause without apparent movement, that IHRCs cannot occur with definite heads in some languages but not others, and that many languages can have null, and/or incorporated heads.
Chapter 2, “Combinatory Categorial Grammar,” is a general overview of CCG and some of the specific categories that Enrico employs in his analysis. He lays out the basic assumptions that he works from, important because Enrico's own assumptions and those underlying CCG differ greatly from other syntactic theories. He argues that CCG is particularly suited for this kind of analysis because it focuses entirely on the overt phonological form, and because it gives greater freedom in the manner of semantic combination. Enrico expands on automatic type raising (2.3.2), presuppositions and prodrop (2.3.3), and lexically prescribed binding (2.3.4), as these will play a significant role in his later analysis.
In Chapter 3, “The Analysis of IHRCs,” Enrico outlines the possible CCG Derivations of IHRCs and how those apply specifically to the unaddressed mysteries of IHRCs. First, he lays out the two possible derivations (3.1). Enrico argues that these are the only two ways to account for the fact that the head (which he argues to be a noun phrase (np) and not a noun (n)) is semantically an argument of each of two verbs in two distinct clauses, the matrix clause and the relative clause. The first possible derivation is the ''head category,'' in which the head np must take the relative clause residue as an argument, i.e. np/(s\np) in SOV languages. The second possible derivation is the ''adjective category,'' in which the relativizer is ''a deverbal modifier of the head, looking for the latter to satisfy its own subcategorization inherited from its verb base,'' i.e. (np\np(head)\$ in SOV languages (Enrico 2013:32).
In Section 3.1, Enrico gives several examples employing both possible derivations in a variety of constructions including the presence of postpositions, postpositional case markers, and alienable versus inalienable possessors. The two derivations seem fairly equivalent, Enrico points out, with the adjective category being slightly more favorable in handling case markers and relative clause stacking. There are differences, though, which he shows in Section 3.2. The adjective category is unable to combine directly with a definite head as its innermost argument must be indefinite. The head category derivation, on the other hand, is easily able to handle definite heads with matrix scope. Enrico shows that it may be possible to avoid the restrictions on the adjective category, eliminating the need for the head category, but this leads to other unwanted implications.
The adjective category is the preferable derivation in languages where definite heads and quantified heads are prohibited (3.2.3) and in accounting for negative polarity and irrealis heads with negative indefinite relative pronouns (3.2.4). The adjective category is necessary in null-headed IHRCs (3.4), IHRCs with incorporated heads (3.5), cases where relative pronouns are inside the IHRC (3.3.1), non-finite IHRCs with nontense verb morphology (3.2.4), and for dealing with the definiteness restriction in Lakhota, Navaho, and Yuman (3.3.2).
Chapter 4, “Constraints on the IHRC Construction,” attempts to account for cross-linguistic variation on the IHRC. Unlike Chapter 3, this chapter is much more speculative and its analyses are less conclusive, relying on some potentially controversial assumptions. Previous accounts have been unable to account for these variations in the structures of IHRCs, Enrico argues, because of a ''dependence of supposed constraints on movements as syntactic evidence.'' (Enrico 2013:77). Enrico instead focuses on semantics and parsing, arguing that together they lead to a gradient of acceptability rather than a clear judgment of acceptable or unacceptable. (This is, he acknowledges, very difficult to prove in dying languages.)
Enrico addresses the variation in the definiteness and quantifier restrictions in Lakhota (4.2), arguing that the requisite indefinite reading of Lakhota IHRCs is explained by the frequency of the adjective category over the head category and the impossibility of using it with definite heads (presumably the hearer will interpret the IHRC as a adjective category derivation without any counterevidence due to the lack of overt definite determiners).
Section 4.3, “Embedded Heads I: Heads in Complement and Adjunct Clauses,” looks at the long distance semantic relation between the internal head and the np comprising the whole IHRC, specifically in Haida. The analysis in Chapter 3 will handle embedded heads, but 'and', ''coordination of unlike categories,'' and reversing the order of composition, must be admitted in order to handle violations of the Element Constraint (EC) which disallows embedding heads in conjoined clauses (Enrico 2013:81). This is not totally unprecedented, Enrico points out, as this type of conjunction is found in other Haida conjunctions.
Section 4.4, “Embedded Heads II: Heads embedded in Relative Clauses,” is significantly longer than the other sections in Chapter 4, and the purpose of this section is very unclear. Enrico begins by pointing out that Lakhota and Mohave freely allow IHRCs embedded in lower relative clauses, while most languages (such as Quechua, Navaho, Haida, and Mohawk) reject them because of ''crossing coreference'' (p. 82). For instance, in the following example (Enrico's example (13), Section 4.4.1, p. 82; with simplified transcription), ''the relation between 'man' and 'gun picked up' is interrupted by 'dog bit' '' (p. 81).
*[[hastiin leechaa'i bishxash] e'e be'eldooh neidiita'] (n)e'e nahal'in
[[man dog 3.PERF.3.bite] REL gun 3.PERF.3.pick.up] REL bark
Intended (Enrico's translation): ‘The dog who the man who bitten by picked up the gun is barking.’
In Section 4.4.2, Enrico shows that the adjective category cannot be used to derive these examples, which are all ungrammatical. Therefore forward crossed composition is necessary for deriving such examples, as backward crossed composition is ruled out because of the need to preserve SOV word order.
These types of sentences, Enrico points out, are very limited in Lakhota and Mojave, with only two examples (one from each language) in the literature. These are not even good examples as ''the heads are in the wrong order'' and they may not actually mean what they are intended to (he gives several alternative interpretations). Thus, it may very well be the case that these languages do not actually allow IHRCs to be embedded in lower relative clauses.
Why are these sentences rejected, then, in most languages (Section 4.4.3)? Enrico argues that listeners, who parse left to right, hear the sentence and initially assume that the second verb must be the matrix verb and that the preceding relative clause is an argument of that second verb. In two-headed type relative clauses, the listener will hear yet another higher matrix clause verb and be unable to parse its meaning in relation to what they have already assumed is a single head rather than two relative clauses. Assuming, though, that this is simply a limitation in parsing, Enrico points out that ''enough exposure would [therefore] presumably enable a learner to develop the proper order-sensitive derivational algorithm'' (p. 86). Thus, we must either conclude that this type of construction is not actually possible in Mojave or Lakhota or take the very unsatisfying position, for this reader at least, that this is simply a case of difficulty in parsing which the hearer would learn to accommodate eventually.
In Section 4.4.4, Enrico shows that there are some cases in which Haida does allow embedded heads in infinitival relative clauses. This is why forward crossed composition must be available,
because this is the only mechanism which explains why Haida allows these constructions.
In Section 4.5, “Question Words in IHRCs: Presuppositional Islands in Haida, Lakhota, and Navaho,” Enrico explores why languages differ in whether they allow question words in relative clauses and, most significantly, why some languages apparently allow question words in some relatives but not in others. Again, this is related to the fact that in many American Indian languages, IHRCs are definite and are, therefore, presuppositional islands. Presuppositional islands restrict extraction because, by definition, the presupposition must be a true proposition for the discourse to proceed. Thus, those IHRCs which can be identified as definite cannot allow extraction of a question word.
Chapter 5, “Conclusion,” summarizes the two derivational categories briefly before transitioning to ''[the] reasons for choosing to work on semantics within a categorialgrammar framework'' (p. 1) as promised in Chapter 1. Only the first paragraph of the ''Conclusion'' is related to IHRCs or relative clauses at all, and there are only a few arguments for why a CCG analysis is preferable to a phrase structure analysis. Enrico argues that ''categorial grammar gives us freedom in the manner of semantic combination'' (p. 102), but he does not give any examples of how this freedom is useful or necessary for his analysis, or how phrase structure fails to yield the right semantic composition. The rest of this short section is focused on showing why the phrase structure hypothesis falls short, arguing that phrases do not exist and are not observable (p. 103).
Enrico is obviously well read in both relative clauses and IHRC constructions as well as the work of other linguists working on similar and related languages. The bibliography relating to other Native American languages is extensive and he does refer to a few papers written within other syntactic frameworks as well. He gives a nice summary of the types of constructions they show, including constructions or patterns which the authors themselves did not notice. This is supplemented by his own extensive fieldwork in Haida. He includes both example sentences and derivations from a wide variety of languages, which were for the most part very helpful, clear, and numerous.
This book is best intended for an audience already familiar with Combinatory Categorial Grammar (CCG) as well as a basic understanding of formal semantics and lambda abstraction. This is unfortunate as it will inevitably limit the number of people who will read Enrico's work. He does give a short summary of the basics of CCG in Chapter 2, but it is written less as an introduction to CCG than as a refresher for readers already familiar with it, and to orient them to his way of using the terminology. The book is obviously not intended as a textbook, or as a comprehensive review of CCG, but it is unfortunate that it is accessible to such a restricted audience. Still, the dedicated reader who wants to take advantage of Enrico's work may still find it worthwhile to familiarize themselves with the basic tenets of CCG through other resources as the research itself is very thorough and unique in the issues that it addresses.
Many readers will be disappointed to find that the book has neither a Table of Contents nor an Index. A Table of Contents, in particular, would have made the book easier to navigate, and would have required little effort on the part of the publisher to include.
There are other technical details which, while they do not detract from the significance of Enrico's work, also make it harder for the reader. The book is fairly short and the addition of a few more pages of exposition, while not strictly necessary, would have helped in making the book more accessible to the reader. Most significantly, there are several points where abbreviations are inconsistent (Principles and Parameters is both PP and P&P) or undefined (for example, page 31: ''Call this principle 'CR'''.) There are several points where Enrico uses the term ''ablaut'' and ''ablaut clauses'' with no clear explanation for what he means, but clearly it is not intended in the typical phonological sense. While it is very helpful to have such extensive examples, many did not include the metalanguage (English gloss). Sometimes this was because it was an alternative derivation of an example already glossed, but in other cases I was not able to find the meaning of the word anywhere in the text (for example, (6) on page 36). This did not significantly detract from Enrico's analysis, but it does distract from it.
Section 4, particularly the discussion in 4.4, is extremely difficult to read, without any clear goal or purpose, and the reader is left to draw their own conclusions about what this section is trying to show. The section begins by pointing out that some languages (Lakhota and Mojave) have no restriction against IHRC heads embedded in relative clauses, but Enrico then goes on to attempt to derive such sentences from ungrammatical examples. It is several pages before he notes that the Lakhota and Mojave examples which he does have are extremely questionable, yet he still concludes in Section 4.4.3 that with enough exposure these constructions would eventually be found acceptable even by speakers who reject them. In Section 4.4.4, he shows that there are cases where Haida does allow embedded heads in embedded infinitival relative clauses. This is apparently significant because this requires forward crossed composition something which he points out on page 84, in another section entirely.
Enrico, unfortunately, ends his book with an unjustified diatribe against other frameworks, particularly ''the phrase structure hypothesis.'' His criticism is particularly harsh and biased, with very few new arguments to back up his stance. He gives general arguments, but none are particularly new, and none have to do with IHRCs. This takes away from his careful presentation of previous research, his own underlying assumptions, and his detail in laying out conclusions throughout the majority of his text. This overly harsh critique detracts from the overall tone of the book, and he misses an opportunity to specifically address other accounts of IHRCs and why they fall short of his analysis in a more unbiased fashion.
| ABOUT THE REVIEWER:
Anne Beshears is currently a PhD student at Queen Mary University of London. She completed her MA in Linguistics from Jawaharlal Nehru University (JNU) in New Delhi. Her current research is focused on comparative and relative/correlatives in Hindi and Marwari, two Indo-Aryan languages spoken in North India. Her research interests include syntax, semantics, the syntax-semantics interface, syntactic variation, and semantic fieldwork.