26.2028, Review: General Ling; Ling Theories; Neuroling; Psycholing: Preston, Newmeyer (2014)

The LINGUIST List via LINGUIST linguist at listserv.linguistlist.org
Thu Apr 16 21:02:21 UTC 2015

LINGUIST List: Vol-26-2028. Thu Apr 16 2015. ISSN: 1069 - 4875.

Subject: 26.2028, Review: General Ling; Ling Theories; Neuroling; Psycholing: Preston, Newmeyer (2014)

Moderators: linguist at linguistlist.org (Damir Cavar, Malgorzata E. Cavar)
Reviews: reviews at linguistlist.org (Anthony Aristar, Helen Aristar-Dry, Sara Couture)
Homepage: http://linguistlist.org

*************    LINGUIST List 2015 Fund Drive    *************
Please support the LL editors and operation with a donation at:


Editor for this issue: Sara  Couture <sara at linguistlist.org>

Date: Thu, 16 Apr 2015 16:58:40
From: Geoffrey Sampson [sampson at cantab.net]
Subject: Measuring Grammatical Complexity

Discuss this message:

Book announced at http://linguistlist.org/issues/25/25-4329.html

EDITOR: Frederick J. Newmeyer
EDITOR: Laurel B Preston
TITLE: Measuring Grammatical Complexity
PUBLISHER: Oxford University Press
YEAR: 2014

REVIEWER: Geoffrey Sampson, University of South Africa

Review's Editor: Helen Aristar-Dry


For many decades it was an axiom of linguistics that all human languages are
equally complex; the claim was put forward as uncontroversial in many
influential textbooks.  Sometimes the point was expanded by saying that
languages might differ in the complexity of particular structural subsystems,
but greater complexity in one system would be balanced by less in another, for
instance Charles Hockett (1958: 180–1) said that Fox has a more complex
morphology than English but a correspondingly simpler syntax.  This idea has
19th-century roots in both Britain and America.

In the 21st century the axiom has been challenged.  The earliest widely-read
counterarguments were probably those contained in Miestamo et al. (2008) and
Sampson et al. (2009).  Surprisingly, to some (at least) of us who were
responsible for the challenges, we seemed to be pushing at an open door.  As
the editors of the volume under review say (p. 7), ''linguists of all stripes
are increasingly willing to entertain the idea that one language might indeed
be simpler or more complex than another''.  Many linguists now take that
position for granted, though quite recently it was heresy.

The present book is the outcome of a workshop held at Seattle in March 2012
under the title ''Formal linguistics and the measurement of grammatical
complexity''.  After an Introduction by the co-editors it contains thirteen
chapters in which various contributors examine different aspects of this

One enlightening way to characterize the book as a whole might be to compare
it with the 2008 and 2009 volumes cited above, all three being collections
stemming from conferences (in the 2008 and 2009 cases held in Helsinki and in
Leipzig respectively).  Although quite comparable in many respects, the new
book contrasts with the previous two both geographically and methodologically.
 Almost all early dissent from the equal-complexity axiom arose among European
linguists, and although North American academics often attend meetings in
Europe, few participated in the Helsinki or Leipzig workshops.  Even in the
present volume, derived from a meeting on the US west coast, a majority of
contributors are European-based, but in this book there is also a respectable
representation of American and Canadian scholars.  And that correlates with a
difference in the methodological flavour of many chapters.  The Helsinki and
Leipzig volumes were fairly thoroughly empirical, with few appeals to abstruse
theoretical concepts.  Many contributors to the present book, on the other
hand, are chiefly concerned to bring the issue of language complexity into
relationship with generative-linguistics theorizing, including the concept of
Universal Grammar and the principles of the Minimalist Programme.

This is not to suggest that European linguists are all empiricists and
American ones all generativist.  As a statistical tendency there may be some
truth in that, but it is certainly not an absolute rule, and within this book
for instance Chapter 7, by Andreas Trotzke of Germany and Jan-Wouter Zwart of
the Netherlands, focuses on Minimalism, while Chapter 10, by the American
David Ross, is quite empirical and un-theory-laden.  But it is natural that a
conference organized in the USA will have made an effort to attract European
scholars whose interests chime with American majority concerns, and the result
is a book whose overall flavour is rather different from the 2008 and 2009

To outline each chapter separately would make for an over-long review, but a
simple list of chapter-titles (after the co-editors' Introduction) will give
an impression of the range of topics covered:

2  Major contributions from formal linguistics to the complexity debate, by
John Hawkins

3  Sign languages, creoles, and the development of predication, by David Gil

4  What you can say without syntax: a hierarchy of grammatical complexity, by
Ray Jackendoff and Eva Wittenberg

5  Degrees of complexity in syntax: a view from evolution, by Ljiljana

6  Complexity in comparative syntax: the view from modern parametric theory,
by Theresa Biberauer, Anders Holmberg, Ian Roberts, and Michelle Sheehan

7  The complexity of narrow syntax: Minimalism, representational economy, and
simplest Merge, by Andreas Trotzke and Jan-Wouter Zwart

8  Constructions, complexity, and word order variation, by Peter Culicover

9  Complexity trade-offs: a case study, by Kaius Sinnemäki

10  The importance of exhaustive description in measuring linguistic
complexity: the case of English _try and_ pseudocoordination, by Daniel Ross

11  Cross-linguistic comparison of complexity measures in phonological
systems, by Steven Moran and Damián Blasi

12  The measurement of semantic complexity: how to get by if your language
lacks generalized quantifiers, by Lisa Matthewson

13  Computational complexity in the brain, by Cristiano Chesi and Andrea Moro

14  Looking for a ''Gold Standard'' to measure language complexity: what
psycholinguistics and neurolinguistics can (and cannot) offer to formal
linguistics, by Lise Menn and Cecily Jill Duffield


This book is very rich in ideas, many of which cry out to be explored further
and evaluated positively or negatively by a reviewer.  Within the bounds of a
reasonable review it is impossible to do this comprehensively, and I shall
limit myself to taking up a few points which strike me as particularly
significant -- another reviewer might well choose different issues to discuss.

One recurring problem is that the empirically-minded contributors and those
who believe in Universal Grammar seem to argue past one another rather than
engaging seriously.  For instance, David Gil argues, as he has frequently done
elsewhere, that languages which fail to encode logical distinctions which are
basic to English and other European languages must often be seen as ''vague''
rather than ''ambiguous'' with respect to those distinctions.  If a young
American child produces the utterance ''Mommy sock'', this could be
interpreted into adult English either as ''Mommy's sock'' or as ''Mommy is
putting a sock on me'', i.e. as a case of attribution or of predication; Gil
suggests that we may be wrong to suppose that the distinction is meaningful in
the child's mind, and whether or not that is so he argues, alluding to
extensive evidence, that there are languages spoken by adults in which such a
contrast is meaningless.  Gil believes that linguists often fail to notice
this because of mental blinkers imposed on us by the dominance of
European-derived cultures and languages in the modern world.  Ray Jackendoff
and Eva Wittenberg's chapter is intended as an answer to Gil, but they begin
by writing (p. 66) ''We assume that ... people across the planet ... think the
same thoughts, no matter what kind of grammatical system they use''.  Surely,
one cannot refute an empirically-based claim by simply ''assuming'' its
converse?  Gil is saying that members of different cultures _don't_ think all
the same thoughts, and to my mind he is obviously right.  It is true that, in
the 21st century, every language used as the official language of a recognized
State is capable of rendering the intricate logical structures of, say, the
United Nations Charter, but that is because in the world as it has evolved, in
order to be taken seriously as a nation any human group has been required to
change its language to make it capable of expressing the same logical
distinctions as are found in the Latin-influenced languages of Europe. 
Languages spoken far from centres of power, or before the period of European
cultural dominance, often lack or lacked logical distinctions that are crucial
to European languages, and contain(ed) other distinctions which European
languages do not make; and the language structures are the best guide we have
to the nature of the speakers' thought-patterns.  (For fuller discussion see
Sampson et al. 2009: 15–18, Sampson and Babarczy 2014.)

Incidentally, even as a believer in Universal Grammar, Lisa Matthewson
demonstrates, using evidence from a native language spoken in British Columbia
and Washington State, that languages are _not_ ''equally capable of expressing
any idea'' and that UG ''does not plausibly contain a list of meanings which
all languages must be able to express'' (p. 263).  I do not understand how
Jackendoff and Wittenberg can hope to refute arguments like Gil's and
Matthewson's via what the former describe in their opening line as a ''thought

A related problem is that the generative linguists' chapters often presuppose
familiarity with technical terminology which is not part of a general
linguistic education, in the way that terms like ''phoneme'' or ''NP'' are. 
Ljiljana Progovac discusses the construction of ''TP'' and ''vP'' elements (p.
85).  Theresa Biberauer et al. (p. 107) ''take Merge to recursively combine
two syntactic objects alpha and beta ... [which] may be drawn from the
Lexicon/Numeration (External Merge), or ... from within alpha or beta
(Internal Merge)''.  Lisa Matthewson asserts (p. 244) that an NP of the form
_every N_ ''can neither be analyzed as being of type e, nor as being of type
<e,t> ... [but] must be of type <<e,t>,t>''.  I have been studying linguistics
fairly intensively for fifty years, but I am afraid that these references mean
nothing to me.  In view of what I have seen elsewhere of the genre of
discourse in which they are at home, I wonder whether some of them mean
anything much at all; they are not explained in this book.  Biberauer et al.
write that their remark quoted above, together with more in the same vein,
amounts to ''a fairly mainstream set of technical assumptions'', but there is
evidently more than one mainstream.

A different kind of problem concerns the concept of ''complexity'' itself. 
Contributors differ in how far (if at all) they believe it is possible to
measure the complexity of a human language.  At one extreme, Theresa Biberauer
et al. are prepared to quote specific figures for the complexity of individual
languages, offering the following table (p. 125):

 Japanese 1.6
 Mohawk 1.8
 Mandarin 2
 Basque 2
 English 3
Others discuss properties which make for greater or less complexity in a
language, without claiming to be able to put a figure on overall complexity. 
David Ross makes the healthy suggestion that, for a computation of overall
language complexity, quite out-of-the-way pieces of grammar (his example is
the English _try and Verb_ construction) would be just as relevant as central
syntactic properties such as discussed by Biberauer et al. (e.g. consistency
with respect to Greenbergian word-order universals):  all are part of what a
speaker has to learn, even if some constructions are only infrequently
deployed in speech.  (Ross makes this point partly in order to object (p. 206)
to John McWhorter's claim that pidgins and creoles are simpler than ''old''
languages, though it is not clear that Ross has actual counterevidence to
McWhorter's claim.)  Some (e.g. John Hawkins, p. 29) urge that it is
meaningless to talk of one language being more complex than another because
there is no metric which could make the overall complexity of different
languages commensurable.  (However, I have argued elsewhere (Sampson 2014),
via an analogy with legal systems, that unquantifiability would not be a bar
to recognizing greater or less complexity among cultural constructs such as
languages.)  Many of these various points of view are pursued without a
consideration of what we mean in general by the word ''complex'' -- and that
is reasonable enough, since after all it is a standard word of everyday
English.  But some contributors do think it is important to pin down what we
mean by calling a system ''complex'', and they mainly do so by referring to
the ideas of the physicist Murray Gell-Mann (who apparently claimed explicitly
that his definition of complexity applied to languages among other things).

Gell-Mann defines the complexity of an entity in terms of the minimum length
of a description of it.  That puzzles me, because description length seems
wholly dependent on the language used for describing.  Consider two languages,
one of which makes nouns plural by reduplication, i.e. saying the singular
form twice, while the other suffixes to the singular form a phoneme sequence
derived from that form by deleting each even-numbered phoneme and then
inserting the vowel /i/ wherever two consonants are adjacent, and the
consonant /b/ wherever two vowels fall together.  Intuitively we would surely
agree that the second pluralization rule is more complex than the first.  Yet,
if we extended English by adopting a term ''thaumatize'', meaning ''repeat
after deleting each even-numbered phoneme, then inserting ... [etc.]'', the
''complex'' pluralization rule could be described very concisely:  ''the
singular is thaumatized''.  Gell-Mann's approach seems to require agreement on
a fixed description-language (which might include the word ''repeat'' but not
''thaumatize'').  But identifying what that description-language should
comprise might be more difficult than glossing the ordinary English word

(Incidentally, Daniel Ross says (p. 204) that his complexity concept is
specifically the one that Gell-Mann calls ''effective complexity'', but I can
find no relationship between Ross's exposition and Gell-Mann's ''effective
complexity'', at least as that is defined in Gell-Mann (1995).)

The difficulties of defining complexity perhaps make a suitable note on which
to bring this review to a close.  There are many other points in this book
which would be well worth discussion, but rather than exhausting readers'
patience let me instead warmly recommend that they read and ponder the book
for themselves.

The book is well-produced, with few printing errors, though the prime sign, as
in ''A'-movement'', is repeatedly (e.g. p. 115) shown as an apostrophe; the
Old English letter thorn is more than once given as a capital where it should
be lower case (p. 152); and the German verb 'wollen' is misprinted on p. 163
as 'wöllen'.  I noticed one point which appears to be a serious factual error:
 on p. 294 Lise Menn and Cecily Jill Duffield state that the ''edh'' sound of
e.g. ''this'', ''then'', is the most frequently occurring English consonant,
but every source I have consulted puts that phoneme quite a way down the
frequency list.  I cannot recognize Trotzke and Zwart's definition of
''context-free phrase-structure grammar'' (p. 142) as equivalent to standard
definitions of that term.  When, in a footnote on p. 40, David Gil says that
nominative and accusative case markings are identical for many Russian verbs,
I think ''verbs'' must be a careless slip for ''nouns''.


Gell-Mann, M.  1995.  What is complexity?  Complexity 1.16–19.

Hockett, C.F.  1958.  A Course in Modern Linguistics.  New York:  Macmillan.

Miestamo, M., K. Sinnemäki, and F. Karlsson, eds.  2008.  Language Complexity:
Typology, Contact, Change.  Amsterdam:  John Benjamins.

Sampson, G.R.  2014.  Complexity in language and in law.  Poznan Studies in
Contemporary Linguistics 50.169–77; www.grsampson.net/ACil.htm.

Sampson, G.R. and A. Babarczy.  2014.  Grammar Without Grammaticality: Growth
and Limits of Grammatical Precision.  Berlin:  de Gruyter.

Sampson, G.R., D. Gil, and P. Trudgill, eds.  2009.  Language Complexity as an
Evolving Variable.  Oxford:  Oxford University Press.


Geoffrey Sampson MA, PhD (Cambridge), FBCS, Professor Emeritus, is a Research
Fellow in linguistics at the University of South Africa, having retired from
the Sussex University School of Informatics in 2009. His books and articles
have contributed to most areas of linguistics, and also include works on
statistics, computer science, political thought, and ancient Chinese poetry.
His most recent book is a new edition of 'Writing Systems' (Equinox, 2015).
Homepage: <www.grsampson.net>.

LINGUIST List: Vol-26-2028	
Visit LL's Multitree project for over 1000 trees dynamically generated
from scholarly hypotheses about language relationships:

More information about the LINGUIST mailing list