Book Review: 

German in Head-Driven Phrase Structure Grammar
John Nerbonne, Klaus Netter and Carl Pollard, editors
Stanford, California: CSLI
1994


The book is a collection of eleven papers that were presented in the
workshop "German Grammar and HPSG" on August 1991. It contains
analyses of various linguistic problems in German in the framework of
Head-Driven Phrase Structure Grammar (HPSG) written by the most
prominent researches in this area. As such, the book gives an
interesting and thorough view on the state of the art in the HPSG
approach to German Grammar.

The good Introduction (pp. 1 - 10) provides general information on the
book and the papers it contains. Apart from a brief description of all
the contributions, the editors provide a general framework to the
variety of the papers in this volume, relating them and classifying
them by their subjects.

As noted by the editors, a prior acquaintance with HPSG theory, as
given in Pollard and Sag 1987 (PS1) and Pollard and Sag 1994 (PS2), is
mandatory: virtually all of the papers are incomprehensible otherwise.
However, most contributors give the necessary details needed for
understanding their solutions, so that readers that are familiar with
HPSG should face little trouble, if any at all, when reading the book.

The papers are classified into four categories: Chapters 1 to 5 are
concerned with issues such as word order, verb clusters and verbal
complexes, all of which are crucial in German and much less prominent
in English. Many of these works tackle the problem of defining
constituents, and it seems the even that criteria for determining what
constitutes a constituent are as yet unclear.

Chapters 6 to 8 address the related phenomena of case assignment and
passivization. All three chapters share many assumptions and
cross-refer to each other. Nevertheless, the mechanisms that they
suggest differ, and an interesting picture of the treatment of those
phenomena is obtained.

Chapter 9 concentrates on the notion of 'heads' and attempts to define
functional heads in the HPSG framework, drawing on data from nominal
phrases in German.  Finally, chapters 10 and 11 discuss the structure
of the lexicon.  Chapter 10 deals with complement inheritance in
morphology, in an approach that contrasts the one taken by Categorial
Grammar. Chapter 11 discusses idioms and the right way to represent
them.

In what follows I briefly review specific chapters of the book.  My
proficiency in German would not allow me to comment on the linguistic
data presented by the authors. Therefore, I always assume the data to
be correct and complete. I must note that the book is perfectly
accessible to readers with little or no knowledge of the German
language.

In the first paper, "Linearizing AUXs in German Verbal Complexes",
Erhard Hinrichs and Tsuneko Nakazawa propose a treatment of the
auxiliary flip construction that further elaborates their previous
works. The linguistic data are presented clearly and extensively and
explained well, and the deficiencies of previous analyses are
discussed. The authors suggest an elegant solution that requires few
modifications to the current theory. The solution supports different
idiolects by slightly modifying lexical entries of the verbs that
trigger the phenomenon. This approach goes well with HPSG's main
trend, namely that information stems from the lexicon and rules are as
general as possible.

In "Adjuncts in the Mittelfeld" Robert Kasper addresses the problems
that the distribution of adjuncts in German poses for HPSG phrase
structure schemata. The problem is explained well, and three different
approaches for solutions are discussed and compared. By introducing a
small number of amendments to HPSG's original schema, the author
generalizes the theory and neatly provides for the problems he
presents. It is worth noting that Kasper deals mainly with semantic
aspects of the sentence structure. A great number of rules, schemata
and lexical entries are depicted, contributing to the reader's feeling
that the solution is feasible.

"Partial Verb Phrases and Spurious Ambiguities" by John Nerbonne
presents the problem of spurious ambiguity: 
"...under treatments in which fronted constituents must always
correspond to potential Mittelfeld constituents, the Mittelfeld must
support a great variety of otherwise unmotivated constituent
structures."
Nerbonne suggests an alternative analysis that doesn't assume a complex
Mittelfeld structure nor uses traces. The paper promotes an analysis
in which partial verb-phrases (PVPs) are considered constituents only
in fronted positions. This rather unusual approach is justified by the
author on the basis of several criteria; however, as Nerbonne himself
observes, it replaces structural ambiguity with lexical ambiguity.
Nevertheless, the paper is thought provoking and formally detailed.

The problems of constituency also play a major role in Mike Reape's
"Domain Union and Word Order Variation in German". Without presenting
much data, but by referring the reader to a paper by Klaus Netter,
Reape suggests rather radically that word order must not be determined
by the yield of the syntax tree, taken from left to right. Rather, he
introduces the notion of "word order domain" and the operation of
"domain union" to define a compositional mapping from syntactic
structures to phonology. As can be seen from the many examples in the
paper, this approach supports hard-to-tackle data such as cross-serial
subordinate clauses, extraposition, subjectless constructions,
scrambling and more. While it seems to me that such an approach leads
to over-generation, it certainly deserves further study.

In "Argument Structure and Case Assignment in German", Wolfgang Heinz
and Johannes Matiasek aim at developing a theory that will derive the
different cases of German by general principles. A very brief
discussion of other approaches to case assignment precedes a
characterization of a sort hierarchy for the values of the CASE
feature and a simple Case Principle, whose application is demonstrated
in detail. The authors take care of German past participle,
auxiliaries and passives, and they also show the interaction of case
assignment with control phenomena. However, it is clear that the
treatment of case assignment must be interleaved in a more general
theory of argument structure, and it is still far from clear how this
could be done.

Passive constructions are the topic of "Passive without Lexical
Rules", by Andreas Kathol, and "Toward a Unified Account of Passive in
German", by Carl Pollard. Kathol challenges the approach by which
lexical rules are the best means for handling passivization. First, he
presents many arguments for treating dative passive as a genuine
passive construction. Then he shows the difficulties that such an
assumption poses for lexical rules. Finally, he uses approaches of
handling auxiliaries for passives, considering passive auxiliaries as
some kind of raising verbs. The solution is extended to impersonal
passive and its consequences are discussed in detail, including a
comparison with Pollard's paper. Then the possibility for double
application of passive is discussed. The second part of the paper
presents a solution for adjectival passives using a sort hierarchy
rather than lexical rules. The paper gains extra significance due to
the lack of efficient approaches for implementing lexical rules in
HPSG.

Pollard builds upon Kathol's data and borrows some of his techniques,
but takes them to a different direction. His aim is to unify the
treatment of personal and impersonal passives, and he does so by using
the ERG feature whose value, he suggests, should be either an empty
list or a singleton list. Considering both papers together, it seems
that there is still much to be done in this area, and Pollard lists
many open questions.

Klaus Netter, in "Towards a Theory of Functional Heads", attempts to
define functional heads in the HPSG framework, based upon data from
German nominal phrases. The key question is: What constitutes the head
of a noun phrase: a noun or a determiner? As determinerless
constructions and the declension phenomenon show, no single 'pure'
approach is satisfying, and the author suggests a mixed solution,
requiring neither empty determiners and epsilon-rules nor disjunctive
subcategorization lists. Netter defines 'lexical category' versus
'functional' one, but there is no clear distinction between the two:
each category carries some substantive (MAJOR) features as well as
some functional (MINOR) ones. The functional categories select the
substantive ones, but functional heads parasitically acquire the MAJOR
features from their complements. Diverse examples are given from the
domain of nominal categories, and the solution is then refined to
support adjectives as adjuncts to nominal heads and to allow for the
description of declension classes. Finally, HPSG's attitude to
specifiers is presented and the advantages of the author's solutions
are discussed. The paper is clear and well explained, apart from the
linguistic data, that are given with no translation. It is puzzling to
see whether the solution Netter suggests still holds for other
languages: for example, Hebrew, Arabic and modern Greek have
constructions of the form Det-Noun-Det-Adj which might require
different approaches.

In "Complement Inheritance as Subcategorization Inheritance" Dale
Gerdemann challenges the Categorial Grammar approach of treating
complement inheritance as a case of functional composition. He
describes the analysis suggested by Moortgart and lists its advantages
and drawbacks. Function composition is independently motivated by many
linguistic phenomena, but the author lists several problems with its
application to complement inheritance. As an alternative he suggests
employing inheritance in the subcategorization list, much in the same
way as auxiliary verbs are treated by Hinrichs and Nakazawa. The paper
shows how the problems discussed above are solved and presents
examples from English and Dutch to convince that the solution is
suitable to various languages. Many problematic cases are also
discussed, and it seems that a complete theory of complement
inheritance is still needed.

The last paper, "Idioms and Support Verb Constructions" by Brigitte
Kern and Gregor Erbach, has to do with the complex problems that
idiomatic expressions pose for the linguist. They distinguish between
idioms, support-verb constructions and compositional collocations
(which are easy to handle and are not discuss further). After
presenting a wide variety of examples for idioms and support verb
constructions, enough to convince one that these expressions are
rather flexible, syntactically, although their semantics is not
compositional, the authors get to the main question: Are idiomatic
expressions lexical signs or phrasal ones? For fixed idioms the answer
is, naturally, the former. However, the paper shows that this is the
better treatment of complex idioms, too, including support verb
constructions. A brief review of a possible HPSG approach to idioms is
presented, with minor deviations from the standard theory. Finally,
the authors discuss the argument structure and the semantics of
support verbs.

Some general comments have to be made regarding the book as a whole.
As this is a collection of works, it lacks uniformity in many aspects.
For example, some of the papers refer to HPSG as described in PS1,
while others refer to the later version (PS2). This is most bothering
in questions such as weather or not the subject is a part of the
SUBCAT list. Feature structures are graphically depicted differently
among the papers. Each paper has its own bibliography list: As many of
the entries are common to more than one paper, it would have made more
sense to compile one common bibliography list for all the papers. This
would have also eliminated the embarrassing mistakes in the references
to PS2, which appears in three or four different variants throughout
the book.

The computational framework assumed by different authors is also
different: some assume various definite clause relations as inherent
to the formalism, some define sort hierarchies that are not bounded
complete (see Carpenter 1992), some use features to define sorts:
many of these extensions are not supported by current implementations
of HPSG.

While a general introduction to HPSG might be redundant, the book
could benefit from some introduction regarding the presumptions that
most authors share about the HPSG treatment of German (such as the
approaches of Hinrichs and Nakazawa). This could eliminate the need to
rephrase such assumptions in many of the papers.

In summary, this volume is important for every one that is interested
either with Germanic linguistics or in HPSG, but it will prove most
useful for the (rather large) community of researchers that are
interested in both. It would be very interesting to see if all the
solutions that are presented in this volume could be unified into one
coherent grammar, that would lead to a wide-coverage grammar for the
German language. It is also desirable that the theoretical solutions
brought by the contributors will be implemented and tested on some
computational platform such as ALE, CUF, Troll or TFS. While I know
that such efforts have been done, for example for the first paper,
none of the authors shows any results from a working system. Such an
implementation will not only help in convincing that the particular
analysis is coherent, but also contribute to the acceptability of HPSG
as one of the best implementable linguistic theories.

References:

Carpenter, B. (1992). The Logic of Typed Feature Structures.
Cambridge: Cambridge University Press.
Pollard, C. & Sag, I. (1987). Information-Based Syntax and Semantics.
Stanford: CSLI.
Pollard, C. & Sag, I. (1994). Head-Driven Phrase Structure Grammar.
Stanford: CSLI.

Reviewed by:

Shuly Wintner
Computer Science Department
Technion, Israel Institute of Technology