Back to Book Reviews

Back to Cercles



Symmetry Breaking in Syntax


Hubert Haider


Cambridge Studies in Linguistics, 136

Cambridge: University Press, 2013.

Hardcover. xviii + 264 p. ISBN 978-1107017757. £65.00


Reviewed by Laure Gardelle

École Normale Supérieure de Lyon


This book is the result of twenty years’ research by the author on the verb-object (VO) / object-verb (OV) divide across languages. While most existing analyses derive the OV order from VO via transformations which Haider describes as a mere ‘patch-up strategy’ [x], one major claim made here is that one type is not more basic than the other: languages are universally asymmetric and their phrase structure is right-branching.

Syntactic structures are said to be symmetric ‘if for a class of (sub-)trees [A B], there is a corresponding class of (sub-)trees that differ only with respect to the order of their immediate sub-constituents, i.e. [B A], within a given language, or cross-linguistically’ [1]. Language structures are asymmetric because ‘the inverse order of a well-formed sequence of phrases in a complex phrase structure is in general not a well-formed sequence’, whether in a given language or cross-linguistically [ibid.] This claim finds its roots in research published in the mid-1980s, first Barss & Lasnik (1986), then Larson (1988) and subsequent works. Before that, sequences such as double-object constructions (e.g. give the reader a hint vs. dem Leser einen Hinweis geben) were considered to have mirror constructions, with a head-initial structure in English (and complements found after the head) as opposed to a head-final one in German (with the complements before the head). The English construction was said to be left-branching, as opposed to right-branching for German, i.e. the two languages were said to have symmetric structures. But a number of works found that left-branching analyses were not adequate for head-initial double-object constructions. For instance, Barss & Lasnik (1986) argued that both head-initial and head-final projections were right-branching.

The author’s own theory, which relies on that fundamental idea, is based more specifically on two concepts: the BBC (Basic Branching Constraint, initially termed ‘Basic Branching Conjecture’) and the PDI (Principle of Directional Identification, a directionality parameter for the identification relation between head and dependents). Once combined, they allow for both VO and OV, and even for a third type of constituent order, called ‘Type III’ – which has not been fully investigated yet, due to the traditional focus on the binary opposition between VO and OV. Note that to Haider, directionality is a basic feature of the grammar of languages, whereas Generative Grammar usually considers linearization to be an epiphenomenon [65].

The BBC is defined as follows:


‘BBC.=def The structural build-up (merger) of phrases (and their functional extensions) is universally right-branching.’ [3]


The BBC is associated with an identification requirement between the head and its dependents, defined as follows: ‘merged positions in the projection of the head of a phrase need to be properly identified by the head under canonical directionality’ [4]. Canonical directionality is what produces head-final or head-initial surface structures; it is dependent on the PDI:


‘Principle of Directional Identification (PDI): a merged phrase P must be properly identified. Def.: A merged phrase P is properly identified by the head of the host phrase h° if

(i) P is in the canonical directionality domain of h°, and

(ii) P and (an extension of) h° minimally, mutually c-command each other.

(Extension of h°= def. projection of h° [-maximal])’ [4]


As a corollary, OV languages (those with head-final structures) are ‘the straightforward instantiations of the BBC in combination with the PDI’: the branching direction is naturally to the right. VO structures (head-initial organisation) are more complex and require shell structures. However, their major advantage over OV structures is that the head is given early, which is preferred for bottom-up processing. In other words, neither OV nor VO structures are optimal (the former give the head later, the latter have more complex shell structures), which explains why each type is found in a comparable proportion of languages. The author argues for the existence of a third type, Type III, characterised by ‘the option of underspecified canonical directionality’ [5]. It is found in early stages of Indo-European languages or in languages such as Yiddish or Slavic.

The book is divided into nine chapters, most of them drawing on previous research by Haider. Chapter 1 presents the theory and wonders why asymmetry is a fundamental property of syntactic structures. This question is explored further in chapter 2, which argues that this asymmetry is the result of a Darwinian-type natural selection process, viz. adaptation by cognitive selection. The following six chapters propose evidence for the existence of the BBC and PDI in various domains: applied phrase structuring [chapter 3], language typology [chapter 4], the Germanic VO / OV split [chapter 5], adverbial positions in OV and VO languages [chapter 6], resultative predicates [chapter 7] and nominal structures [chapter 8]. Finally, chapter 9 compares the author’s model with a competing one, LCA (Linear Correspondence Axiom, Kayne 1994), which also relies on right-branching but makes use of massive phrase movement. The chapters can be read separately, without the reader having to look up other parts of the book, which accounts for the ‘considerable overlap’ between them [xii]. The book itself can be complemented by Haider (2010), The Syntax of German, which investigates other aspects of right-branching structures. It focuses on German but also includes comparisons with English.

Here are some of the main ideas developed by the author. First, the reason why natural languages should only have right-branching structures is thought to be the cognitive evolution of their grammars. Evolution is defined as ‘adaptation by selection out of a pool of variants’ [7] during the acquisition process – a child does not copy the grammar of the language, but reconstructs it from the received input. The adaptive advantage of right-branching structures is that they allow for faster language processing, and processing time is a limited resource. In other words, ‘right-branching structures guarantee optimal grammar-parser congruence’ [20], the BBC is a ‘parser-friendly asymmetry principle’ [27], and this explains why it is part of Universal Grammar, ‘the grammar formation capacity provided by our brains’ [23].

The BBC model is shown to provide simpler analyses than those that postulate symmetric structures. For instance, about double object constructions, Larson (1988)’s claim that in English, a left-branching VP is transformed into a right-branching one through two movements (head movement and raising of the direct object) and subsequent ‘structure pruning’ [50] is not only arbitrary, but it involves unnecessary technical devices. The BBC model is much simpler: all it requires is head movement. Another example is the Minimalist concept of ‘little-v’, which in Haider’s approach is simply one of the V-positions in the shell structure of complex head-initial VPs – similarly, in NPs, Haider finds no need for ‘little-n’. The BBC constraint (right-branching) also predicts the absence of certain structures in some of the world’s languages. For instance, it explains why there could not be a so-called ‘little-v’ in head-final (OV) languages, or why languages with mandatory expletive subjects have to be VO languages (without pro- or topic-drop) – for a fuller list, see Haider [67 ff.] Another advantage of the BBC model is that it accounts for the position of adverbials, for instance of negation. In head-final VPs (OV languages), the canonical position of the verb is at the bottom of the phrase, so that any position within the VP c-commands it. Negation can therefore be placed within the VP. Conversely, in head-initial VPs (VO languages), the canonical position for the verb is the advanced position in the shell structure, so that any item that must c-command the canonical position of the verb must precede the VP.

The author also investigates the VO / OV split in Germanic languages, a singularity for languages in a contiguous area and therefore an enigmatic phenomenon: West Germanic languages (e.g. Afrikaans, Dutch, Frisian, German) have an OV structure, although in surface sentences this property is partially hidden by the pan-Germanic V2 property. To Haider, it is very unlikely, for a number of reasons, that the parent language had either an SVO or an SOV order and then underwent a split (towards SOV or SVO respectively). He claims rather that the proto-language belonged to Type III, i.e. had an underspecified or flexible directionality property for verbal heads. There was then a syntactic change towards a specified directionality. The author rests his case with the example of Yiddish (a Type III language), Latin and Romance languages, and finally English. For the latter, he reviews existing theories, which argue either for co-existing VO / OV patterns before specialisation (competing-grammars model, open-parameter model) or for a reduction to SVO in order to minimise syntactic movement (excessive-movement model). He concludes that the idea of Old English as a Type III language is in fact better and simpler, and goes on to hypothesise why flexible directionality shifted to a VO organisation in English. He relates this shift to three independent factors: variation in the distribution of verbs inherent to Type III, verb movement by V-second, and variation of verb order in the verb cluster in languages that move verbs.

Further evidence of the BBC constraint is found in resultative structures, which to the author prove that head-final organisation (OV languages) is not derived from head-initial (VO) projections. In OV languages the position of result predicates is invariably pre-verbal and V-adjacent, whereas in VO languages they can be found either adjacent to the verb or in a more distant position. This conclusion of Haider’s is based on a rejection of the analysis of result predicates as small clauses. For instance, they drank the teapot empty should not be analysed as drink + the small clause the teapot empty as its complement, contrary to traditional analyses, because the verb drink never takes a proposition as its complement – and a small clause is a proposition. For the same reason he also rejects Kratzer (2005)’s analysis of the structure as similar to a serial verb construction. Instead, he argues for a complex-predicate analysis, with the teapot as the object of drink empty.

The author then extends his analyses to noun phrases. In Germanic OV languages the canonical directionality of heads is not uniform: N° and P° are phrase-initial whereas V° and A° are phrase-final. But the BBC constraint also applies. For instance, it accounts for constraints on N+N compounding: it is productive only with head-final compounds, as in Germanic languages, whereas it is highly constrained for head-initial compounding (e.g. in the Romance languages). Similarly, there cannot be head-initial recursive compounding. Another instance is recursive cluster nominalisation from infinitives, which is not available for head-initial structures (VO).

Finally the author compares his approach to competing models. First of all, universal right branching is not unanimously accepted today. For instance, to Liao (2011) and Citko (2011), the core system of syntax is symmetric. To Liao, symmetry breaking is only an interface effect, while to Citko, merge, move and labelling (that is, syntactic categorisation) are basically symmetric, i.e. ambi-directional. Haider, however, finds it difficult to see the advantage of symmetry, which is not even a preferred organisation in biology. He attributes this bias to the authors’ focus on VO languages, more specifically on small portions of the grammar of those languages. Secondly, the BBC approach is not the only existing model among linguists who favour universal right branching. The author devotes a whole chapter to a comparison with LAC (Linear Correspondence Axiom, Kayne 1994), which differs mainly by the role given to asymmetry and linearisation. In the BBC model, OV and VO differ simply by the positioning of the heads and the presence of shell structures for VO. LCA, on the other hand, relies heavily on phrasal movement: VO is viewed as axiomatic, and OV is derived from VO structures by ‘massive movement of non-head constituents’ [211]. This goes against the intuition that grammars should not evolve in such a way as to hamper the processing routines. Besides, the claim that, for instance, German (OV) is derivationally more complex than Icelandic (VO) appears less convincing than the BBC approach, which views them as simply implementing opposite directionality values for identification. Also, the BBC offers predictions that LCA cannot make. For instance, it explains why there cannot be quirky subjects in German – for a fuller list, see Haider [221 ff.] The author concludes that LCA works as a theory, but that a simpler account such as the BBC is superior.

All in all, the broad range of linguistic phenomena to which the theory is applied and the predictive power of the BBC and PDI in a number of domains make Haider’s model of syntactic structure both consistent and convincing. It could be regretted that a handful of conclusions rely on assumptions which themselves do not seem to be ascertained. For instance, the author states that directionality has to be regarded as a basic feature of grammar, otherwise ‘it is at odds with the presumed organization of grammar as an algorithm that allows fast mappings between linearizations and structures’ [66]. Such a conception of grammar is not itself the mainstream approach, as Haider notes [212], and it is based on a series of hypotheses. Similarly, the idea that the cognitive evolution of grammars is primarily driven by the need for fast language processing is enticing, but does not seem to rely on any existing theory of language acquisition or evolution – at least no framework or references are given. This makes some of the assertions less convincing, for instance: ‘The production side, on the other hand, is under no time pressure, but the perception side is. As a speaker, I may consume as much time as I need for the production of an expression’ [8]. It could be objected that in interaction, production is also limited in time and that fast reaction (production) might be as important as fast processing. An assertion such as ‘Languages do not change instantaneously, but grammars may’ [113] also appears rather peremptory. Such statements, however, are rare and only concern minor elements in the book.

The theory will also benefit from a confrontation with a higher number of languages – which could not possibly have been carried out here given how broad and rich the study already is –, especially in view of the fact that the proposed model goes against several mainstream approaches. For instance, the author argues that in English, the reason why the verb and its nominal complement must not be separated by intervening items such as an adverbial (*analyse carefully problems) is that the verb would then no longer minimally c-command the object. This constraint is described as ‘a PDI effect for head-initial phrases’ [142]. In French, however, another VO language, such word order would not be problematic (analyser avec soin des problèmes). The reader is therefore left to wonder whether there might be language-specific constraints within the framework of the BBC and PDI, or whether the model might be too broad. This, however, will be left for subsequent research to assess. In the meantime, the linguistic phenomena described for Germanic languages and several others totally fit the proposed model.

As can be seen from those remarks, the book, which is very accessible and stimulating, should be recommended to any linguist interested in syntactic structures, Germanic languages and/or language evolution.




Barss, Andrew & Howard Lasnik (1986), ‘A note on anaphora and double-objects’. Linguistic Inquiry 17: 347-354.

Citko, Barbara (2011), Symmetry in Syntax : Merge, Move and Label (Cambridge Studies in Linguistics). Cambridge: University Press.

Haider, Hubert (2010), The Syntax of German. Cambridge: University Press.

Kayne, Richard (1994), The Antisymmetry of Syntax. Cambridge, MA: MIT Press.

Kratzer, Angelika (2005), ‘Building resultatives’. In Claudia Maienborn & Angelika Wöllstein (eds.), Event Arguments : Foundations and Applications. Tübingen: Niemeyer : 177-212.

Larson, Richard (1988), ‘On the double-object construction’. Linguistic Inquiry 19 : 335-391.

Liao, Wei-wen (2011), The Symmetry of Syntactic Relations, unpublished PhD dissertation, University of Southern California.


Cercles © 2013

All rights are reserved and no reproduction from this site for whatever purpose is permitted without the permission of the copyright owner.

Please contact us before using any material on this website.