Nerbonne, John, Klaus Netter, and Carl Pollard (eds), 1994:

German Head-Driven Phrase Structure Grammar

(CSLI Lecture Notes, Number 46)
Stanford, California: Centre for the Study of Language and Information. Pp. xi + 404.
ISBN 1-881526-30-5 (hardback) £40.00 (US$49.95); 1-881526-29-1 (paperback) £16.95 (US$21.95).

reviewed by

Jonathan West

Department of German Studies, University of Newcastle upon Tyne, NE1 7RU
[Download this Review (15K, Rich Text Format)]
Copyright Notice:
First published in Web Journal of Modern Language Linguistics in association with the publishers (to be announced). © 1997 Jonathan West.
The moral rights of the author(s) to be identified as author(s) of this work are asserted in accordance with §§.77 and 78 of the Copyright, Designs and Patents Act 1988. This work may be reproduced without the consent of the author, in part or in whole in any manner and in any medium subject only to the two following conditions:
(a) no charge shall be made for the copy containing the work or the excerpt,
(b) a copy of this notice shall precede the work or the excerpt.

Head-Driven Phrase Structure Grammar (HPSG) is one of a number of competing grammatical theories students have to come to terms with these days. This book was written because "the time seemed ripe for making the case anew that this formally explicit and computationally useful framework also serves the purpose for which it was originally intended: the formulation of empirical hypotheses about natural languages (and in the limit, about universal grammar)" (p. 1). It is the editors' wish "to see the value of HPSG as a grammar theory explored more deeply" (p.1), presumably by germanists as well as general linguists. In that case, it is hard to understand why there is no general introduction to the theory, especially as this approach has been applied hitherto largely to English. The editors admit that "[s]ome readers may be disappointed not to find an extensive introduction to the background theory of Head-Driven Phrase Structure Grammar in this preface (and some prospective readers have already indicated that they would appreciate one)" and refer instead to Pollard and Sag (1994) -- 440 pages!! --, which is intended to replace Pollard and Sag (1987), but it helps to have read the former as well. Doubtless an "extensive introduction" would have overloaded the present volume, but a short introduction would have improved its readability immensely, as the wise referees no doubt pointed out. At the very least a glossary with examples is needed for I suspect that even generative grammarians will find this hard going. Perhaps the truth is that the volume is really intended for the cognoscenti.

A review such as this cannot hope to make good that deficit, even if the reviewer understood all the details. However, it will help readers to know that HPSG is a formal system of signs: it proposes that all signs at minimum possess the two attributes PHON and SYNSEM. PHON is the basis for phonetic and phonological interpretation; SYNSEM is "a complex of linguistic information" [...] "more or less analagous to the information that is distributed between the levels of D-structure and LF in current transformational models" (Pollard and Sag 1987:15). The best analogy I can think of is to say that HPSG is not algorithmic in the sense that many computer programs are, which use a formal procedure on a data set and produce an output; instead, it is rather like Prolog, in which the program and the data set are the same. A Prolog program solves problems concerned with objects and the relationships between them (Clocksin and Mellish 1987: 1); HPSG is concerned with objects of different sorts and the relationships between them, captured by the attributes of the objects, rule schemata and universal principles (Pollard and Sag 1987:17, 395-403). We are therefore dealing with feature structures / lexical entries rather than labelled trees and rewrite rules, although I suspect that these (like the analagous computer programs) are in essence notational variants.

The Introduction (pp. 1-10) includes summaries of the issues tackled in each contribution to the volume and thus preempts a reviewer's job. In the first of these, Erhard Hinrichs (who supplies the data -- "the examples [...] reflect the grammaticality judgments of the alphabetically first author", they aver in a construction of which Sir Humphrey Appleby would have been proud) and Tsuneko Nakazawa address the problem of "Linearizing Finite AUXs in German Complex VPs" (pp. 11-37). They focus on the so-called auxiliary flip construction, whereby the auxiliary precedes the non-finite verbs in subordinate clauses as in [1] rather than following it as in [2]

[1] Ich wußte, daß er das Examen hat bestehen können.
[2] Ich wußte, daß er das Examen bestanden hat.

This construction is accounted for in terms of "feature values that are specified in lexical entries of triggering verbs" (p. 13), and it is indeed triggered by a small finite set of modals and other verbs, which appear to be the set of infinitive verbs without fahren, gehen, kommen and spüren (there is no discussion of this).

Roberst Kasper's contribution "Adjuncts in the Mittelfeld" (pp. 39-69) is concerned with problems of ordering. The fact that in a sentence such as [3] allows permutation of the Mittelfeld constituents {gestern}, {ihrem Mann} and {diese Geschichte} "poses a problem for the phrase structure schemata that are typically assumed in HPSG, because these schemata require all nonsubject complements to be combined with the verb at once into a phrase" (p. 39).

[3] Sie hat gestern ihrem Mann diese Geschichte erzählt.

Kasper's solution is a "new analysis of adverbial modification in terms of flat phrase structures in which adjuncts appear as sisters of complements" (p. 66). Kasper may be rediscovering the wheel here, as dependency grammar has long recognized that complements and adjuncts are dominated by the same node (Engel 1982:119). Use of flat phrase structures "can be ragarded as an attempt to partially decouple the principles of semantic composition from phrase structure configurations" (p. 66f.), but this again is a restatement of what has long been accepted, namely that, in contradistinction to English, constituent order in German does not reflect syntactic relationships: the appearance of an element in the Vorfeld, or departures from the normal (i.e. statistically most common) order of constituents in the Mittelfeld are determined by semantic factors, such as explicit connection to the previous text, emphasis and the like (Engel 1988:334-344).

Tibor Kiss investigates (in rather dodgy English, the editors may care to note) "Obligatory Coherence: The Structure of German Modal Verb Constructions" (pp. 71-107). He correctly states that modal verbs differ from other "control verbs" (these are not precisely defined) by prohibiting extraposition and pied-piping and follows Bech (1957) in calling these constructions coherent "if the governed verb is nonfinite and both the governor and the governee form a unit of a certain kind" (p. 73). Kiss wishes to show "how Bech's descriptive distinction can be made explicit in terms of phrase-structural representations, given the rules and principles of HPSG. [The] main thesis [is] that optional coherence may result in a verbal cluster, but in the case of obligatory coherence the formation of a verbal cluster is required" (italics are Kiss's). I sense a circular argument here. Besides, Bech's descriptive distinction is already explicit, as it describes sub-classes of verbs using distributional criteria.

John Nerbonne's contribution deals with "Partial Verb Phrases and Spurious Ambiguities" (p. 109-150) and, mindful of the relationship between the ability to occupy the Vorfeld and status as a sentence constituent, seeks to account for pairs such as [1a] and [1b]:

[1a] Das Buch lesen wird er schon können.
[1b] Lesen können wird er das Buch schon.

The disjuncture between the Mittelfeld structure implied by these examples, and others such as [2]

[2] Einen Hund füttern, der Hunger hat, wird jeder wohl können.

leads Nerbonne to deny "that PVP fronting is to be explained as the displacement (or alternative expression) of a potential Mittelfeld constituent" (p. 114). For the solution, do not lose the erratum slip, which resolves "the framework of ?" as "the framework of Pollard and Sag 1994".

Mike Reape's contribution, "Domain Union and Word Order Variation in German" (pp. 151-197) (not "in Germanic", as in the introduction), questions the common assumption that the terminal yield of a surface syntax tree has any necessary direct implications for word order, which are accounted for by word order domains. These are ordered sequences of constituents, which can be combined, as long as the relative sequence of elements remains the same. His example (p. 157) is the sub-clause

[3] daß es3 ihm2 jemand1 zu lesen3 versprochen2 hat1

where "each NP is separated from its head by other constituents", the subscript figures showing the dependency relationships such that "es is the direct object of [...] zu lesen" and so on. This should yield six possible orders for the sub-clause, and it seems to do just that:

[3.1] daß es ihm jemand zu lesen versprochen hat
[3.2] daß es jemand ihm zu lesen versprochen hat
[3.3] daß ihm es jemand zu lesen versprochen hat
[3.4] daß ihm jemand es zu lesen versprochen hat
[3.5] daß jemand ihm es zu lesen versprochen hat
[3.6] daß jemand es ihm zu lesen versprochen hat

Reape then relates this phenomenon to extraposition, subjectless constructions, pronominalization, scrambling, the scope of adjuncts and negation and the intraposition of VPs in the Mittelfeld. I have problems with the basic analysis and therefore with the approach as a whole. The main clause realization of this construction would be [3a], and in the present tense [3b]

[3a] Jemand hat ihm versprochen, es zu lesen.
[3b] Jemand verspricht ihm, es zu lesen.

where the infinitive clause commutes with an accusative neuter pronoun das / es, resulting in [3c]

[3c] Das verspricht ihm jemand. / Jemand verspricht es ihm.

How does an analysis which sees {jemand} as depending on {hat} account for the fact that {verspricht} commutes with {hat versprochen}? Are all the variants [3.1-6] equally likely?

Andreas Kathol addresses the problem of "Passives without Lexical Rules" (pp. 237-272) and this may usefully be taken with Carl Pollard's "Toward a Unified Account of Passive in German" (pp. 273-296). In the case of Kathol, I cannot understand why he ignores the evidence of German. He notes of German that "passive in that language is case sensitive. Unlike in English, where notions such as 'primary object' determine which argument can undergo passivization, the crucial factor in German is that the argument has to bear accusative case. Thus trying to promote to subject genitive objects [...] or dative objects [...] yields ungrammatical results. One of the examples he gives is [4], which produces ungrammatical results [4a] when the genitive is "promoted", but is passivized correctly in [4b], a construction which he studiously ignores.

[4] Ich gedenke des Toten.
[4a] *Der Tote wird gedacht.
[4b] Des Toten wird gedacht.

That this is no slip of the pen is demonstrated by similar evasion -- or avoidance -- of the correct passive forms of [5] and [6]:

[5] Ich helfe dem Jungen.
[5a] *Der Junge wird geholfen.
[5c] Dem Jungen wird geholfen.

[6] Ich schenke dem Jungen ein Buch.
[6a] *Der Junge wird ein Buch geschenkt.
[6b] Dem Jungen wird ein Buch geschenkt.

Pollard does not fall prey to this error, but his correct data (p. 274) seems to implicate him, the editor, in the sin of the contributor. Either way, it does not inspire confidence.

The other contributions are as follows. In "Argument Structure and Case Assignment in German" (pp. 199-236), Wolfgang Heinz and Johannes Matiasek "develop a theory that derives the occurrence of the case forms -- especially structural case forms -- from general principles" (p. 201). Klaus Netter progresses "Toward a Theory of Functional heads: German Nominal Phrases" (pp. 297-340) and this is followed by Dale Gerdemann's "Complement Inheritance as Subcategorization Inheritance" (pp. 341-363). The final contribution, by Brigitte Krenn and Gregor Erbach, addresses the problem of "Idioms and Support Verb Constructions" (pp. 365-396).

I mention these authors and their work for completeness and information only, as the point alluded to in my introductory paragraph is amply supported by the papers described in slightly greater detail. Most address interesting German grammatical phenomena, but, in order to make the case for HPSG, some comparison with other descriptive paradigms (e.g. HPSG vs. Categorial Grammar; HPSG vs. Lexical-Functional Grammar) would have been necessary. Contrastive treatments of a range of key phenomena, showing where the latter fail and where HPSG might provide a solution, would surely have been the way to approach this question. As it is, one is left with the impression of a collection of essays treating various phenomena -- sometimes discussing the same phenomenon and producing contradictory solutions -- which could have been dealt with just as well within the framework of another theory. My impression is that the suitability of HPSG as a vehicle for the description of German is at best unproven and that only a methodologically more rigorous investigation with stronger editorial control can decide its usefulness one way or the other.


Clocksin, W. F. and C. S. Mellish, 1987. Programming in Prolog. Berlin, Heidelberg, etc.: Springer-Verlag. (Third, Revised and Extended Edition. First published 1981.)

Engel, Ulrich, 1982. Syntax der deuschen Gegenwartssprache. Berlin: Schmidt. (Grundlagen der Germanistik, 22. Second edition, first published 1987.)

-- 1988. Deutsche Grammatik. Heidelberg: Groos.

Pollard, Carl and Ivan A. Sag, 1994. Head-Driven Phrase Structure Grammar. Stanford, California: Centre for the Study of Language and Information. Chicago and London: University of Chicago Press. (Studies in Contemporary Linguistics)

Return to The Issue 2 Contents Page |