LINGUIST List 14.2983

Sat Nov 1 2003

Review: WritingSys/Phonology:Neef, Neijt & Sproat(2002)

Editor for this issue: Naomi Ogasawara <>

What follows is a review or discussion note contributed to our Book Discussion Forum. We expect discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for review." Then contact Simin Karimi at


  1. Tobias Thelen, The Relation of Writing to Spoken Language

Message 1: The Relation of Writing to Spoken Language

Date: Thu, 30 Oct 2003 18:22:26 +0000
From: Tobias Thelen <>
Subject: The Relation of Writing to Spoken Language

Neef, Martin, Anneke Neijt and Richard Sproat, eds. (2002) The
Relation of Writing to Spoken Language, Niemeyer, Linguistische
Arbeiten 460.

Announced at

Tobias Thelen, University of Osnabrueck


Neef, Neijt and Sproat present a volume that consists of ten papers
which mostly grew out of talks given at the international workshop
Writing Language in Nijmegen on August 28-30 2000. In their
introduction they identify a set of general questions that are covered
in the papers from different perspectives. The main questions concerns
the relation of writing and spoken language, more specifically the
questions of how natural writing is, the notion of orthographic depth,
the relation between orthography as an object of research for
theoretical linguistics and the psycholinguistic investigation of its
use and the processes involved. In their introduction, the editors
raise the question of what kind of constraints are involved in reading
and writing and if non-local constraints are necessary to adequately
describe spelling and reading. The status of orthography and writing
is addressed in asking whether writing is derived from spoken language
or if it follows autonomous rules. Finally, they give attention to
the balance between reading and writing and state that reader- and
reading-oriented models of orthography might be more suitable than
writer- or writing-oriented ones.

The papers are organized in four sections. Each of them investigates
one phenomenon or theoretical question from different point of views
and different research backgrounds.


Section 1: Consistency

Anneke Neijt - The Interfaces of Writing and Grammar

Neijt addresses Sproat's consistency hypothesis (Sproat 2000) and
applies it to Dutch spelling. Dutch has a deep orthography, e.g. final
devoicing, it shows differences between native and non-native words,
and is best described by a two-step derivation: phoneme-to-grapheme
conversion of morphemes are followed by graphotactic rules. Neijt
tries to show that phoneme-to-grapheme rules are based on information
from different levels. She demonstrates that the Orthographically
Relevant Level (ORL) is different for native and non-native words. For
phoneme-to-grapheme rules she argues that some morphemes are spelled
according to an underlying phonemic form while some require a more
superficial level. This can be seen as contradicting the consistency
hypothesis. Graphotactic rules as presented by Nunn are not autonomous
but need orthographical and phonological information as shown for
diaeresis placement, syllabification degemination and stress
representation. Neijt concludes that Dutch orthography cannot easily
be described by the consistency hypothesis. Information from
different levels of language processing is involved.

Richard Sproat - The Consistency of the Orthographically Relevant
Layer in Dutch

Sproat directly replies to the aforementioned article and tries to
show that the examples given by Neijt do not violate the Consistency
Hypothesis. The devoicing examples given might stem from two different
processes, one applied before, the other applied after the ORL. Stress
information might be encoded by diacritics that don't show up in
surface orthography - Sproat argues that this solution is not worse
than looking at phonology and orthography simultaneously. The
different rules for different parts of orthography (Latinate
vs. native part) observed by Neijt are categorized differently by
Sproat: one is an obligatory morpho-phonological alternation, the
other one an optional phonetic alternation. In other cases,
etymological information is needed to decide which set of rules is to
be taken. This information must somehow be encoded in the
representation. For the need of information from different levels,
such as syntax or semantics, Sproat states that ''level'' means
''stage in derivational process'', not ''level in some
hierarchy''. So, all information accessible at a given stage of the
derivational process is present in the ORL, too. Sproat finally asks
why we should expect that consistency holds. He argues that
orthography should be as natural as possible to be usable. If
orthography was not, two things could happen to re-establish
consistency: spelling changes and spelling pronunciations.

Section 2: Cross-linguistic studies

Susanne R. Borgwaldt & Annette M.B. de Groot: Bevond Rime: Measuring
the complexity of Monosyllabic and Polysyllabic Words

The authors propose a new methodology for measuring bidirectional
consistency of spelling-sound correspondences in psycholinguistic
research. In contrast to other methods, the proposed one would be
suitable for both monosyllabic and polysyllabic words. They state that
the findings about the consistency of a spelling system vary on the
size of units (letter, rhymes, sliding windows, sublexical linguistic
units). Thus, a flexible methodology is necessary for several
application domains (speech synthesis, machine learning). Their new
method splits written and spoken words in units like onset, nucleus
and coda. Then, body (onset + nucleus) and rime (nucleus + coda) are
compared with respect to consistency of mapping. For polysyllabic
words, ambisyllabic segments and variable syllable boundaries may
impose problems, so coda and following onset are treated as one
unit. In the next step overlapping sublexical units (OSLUs) are
constructed as pairs, triples, or quadruples etc. of adjacent
sublexical units. Each of these patterns is compared to estimate
consistency. The result is a list of consistency scores for each OSLU
and a combined consistency score for the entire word. The authors show
that their method is more accurate than other methods as it is able to
disambiguate a lot of otherwise inconsistent spellings or
pronunciations. They claim that the proposed method is language
independent as well as easy to compute. Finally, the results of a
sample analysis are presented for Dutch and German.

Dorit Ravid & Steven Gillis: Teachers' Perception of Spelling Patterns
and Children's Spelling Errors: A Cross-linguistic perspective

Ravid and Gillis present the results of studies on teachers' knowledge
about morphological and morpho-phonological cues in spelling
homophonous graphemes in Hebrew and Dutch. They ask two questions: Is
children's knowledge and use of morphology matched by their teachers'
ability to explain spellings and what differences exist between
Belgian and Israeli teachers' ability to explain spellings. In a first
study children were given a spelling test with 4 sets of 8 homophonous
items, with different status of recoverability. For Hebrew the
predictions were matched: fewer errors for words with more cues, more
errors for words with fewer cues. For Dutch speaking children, the
predictions were not met as they scored relatively low on
morphologically motivated sets and relatively high on morphologically
unmotivated sets. The author present a possible explanation: Growing
up with a morphologically complex language like Hebrew makes it easier
to use morphological clues. Alternative explanation could concern the
teachers' knowledge about spelling rules. In a second study the items
from the children's test were randomized and presented to teachers in
a pairing task and a motivation of choice task. In the pairing task,
the expectations from the children's experiments were not
matched. Belgian teachers have been able to identify the pairs, Hebrew
teachers had greater difficulties. In the motivation task, Belgian
teachers scored better than Hebrew. For all Hebrew results, students
of teachers training colleges scored lower than university students,
for the Dutch results both groups had comparable results. The result
of this study is that the teachers' metalinguistic knowledge of
spelling patterns is a mirror image of children's performance. A
possible explanation could be that the acquisition of language
patterns and rules is implicit and natural. Metalinguistic explanation
has to be learned and is difficult. Teachers have to find simple
rules to teach for complex rules. This is easier for Dutch than for
Hebrew. A list of the test items used concludes the article.

Section 3: Diacritics and Punctuation

Vincent J. van Heuven: Effects of Diaeresis on Visual Word Recognition
in Dutch

Van Heuven presents a study on the effects of diaeresis in Dutch
spelling on the recognition of words. As Dutch words can contain long
sequences of vowel letters, the diaeresis is used to indicate
beginning of syllable. The guiding question for the study was how the
diaeresis affects the visual word recognition process, and whether the
proper use of diaeresis enables adult Dutch readers to recognize words
more efficiently. The experimental setting was to present words with
correct use of diaeresis and three types of diaeresis errors:
omission, addition and transposition. Subjects were asked to decide
whether a string presented is a Dutch word or not. The stimulus
material consisted of 240 letter strings, 120 of which were existing
Dutch words (40 filler, 80 crucial words), 120 orthographically and
phonologically legal Dutch nonsense strings. Three versions of the
crucial words have been used: the correct spelling, a spelling with
diaeresis error, and a spelling with a minimal spelling error. These
versions have been presented in an evenly mixed way for each of 120
students. The results of the experiment show a clear effect of error
type: the correct versions and the versions with diaeresis errors do
not differ whereas misspelling score significantly lower. The length
of the vowel sequence has a small effect, but has no interaction with
the error type, i.e. both diaeresis errors (omission and
transposition) have no effect. For the insertion of illegal diaeresis
the correct acceptance rate drops significantly. The addition of
diaeresis detracts from the recognizability of the word form but the
elimination of diaeresis does not create problems for the experienced
adult reader. The author concludes that it is possible that the
diaeresis, however, is of considerable help for inexperienced readers
and learners of Dutch orthography.

Jochen Geilfu�-Wolfgang: Optimal Hyphenation

The author presents a constraint based modelling of German hyphenation
in an optimality-theoretic framework. He gives structural
well-formedness constraints for orthographic syllables that to not
refer to phonological syllables. For his investigation he uses data
from the 2000 edition of Duden Rechtschreibung. The author shows that
eight properly ranked constraints yield correct results for all the
cases in question: - Ons: orthographic syllables begin with a
consonant grapheme - Nuc: orthographic syllables have a vowel grapheme
- *ComplexOns: orthographic syllables have at most one consonant
grapheme at their left edge - *ComplexNuc: orthographic syllables have
at most one vowel grapheme - AlignL: Left edges of stems and
morphological words coincide with left edges of orthographic syllables
- RecoverGrapheme: graphemes cannot be split - Margins: Every margin
of an orthographic syllable is a possible margin of an orthographic
word The author states that with this modelling orthographic syllables
are defined and related to the question of complex or noncomplex
graphemes, graphotactical constraints and morphological information,
but not phonological information. He concludes that word splitting in
German is based on orthographic syllables.

Ursula Bredel: The Dash in German

In contrast to the usual characterization of the dash (Gedankenstrich)
in German orthography as a polyfunctional mark, the author suggests a
uniform characterization. She motivates this approach by the
observation that both syntactical and intonational explanations of
punctuation omit the aspect of linking as a medium-specific
property. The dash has earlier been characterized as a polyfunctional
punctuation mark, that functions on a grammatical level, on a text
organizational level or on a pragmatic level. Attempts to
characterize the dash uniformly compromise the notions of
''interrupting'' and ''changing''. Bredel gives an overview of
development of the dash in German writing. Historically, it had two
different functions: marking omissions, marking end of lines. The
modern uses of dash have in common a certain change of focus. Bredel
shows this notion for parantheses, change of speaker/topic and
announcements. She concludes the dash prepares the reader to perceive
subsequent material under changed focus.

4. Sharpening in German

Christina Noack: Regularities in German Orthography: A Computer-Based
Comparison of Different Approaches to Sharpening

Noack uses a computational modelling of orthographic rules to
determine consistency of German orthography, namely the phenomenon of
sharpening (consonant letter doubling). Three different rule proposals
are investigated: a morpho-phonological approach proposed by Adelung
in 1788, a syllable-based approach proposed by Maas in 1997 and a
segment-based approach from the official rules in 1902. A word list,
taken from Adelung's ''Kleines W�rterbuch'' was taken as a corpus for
a simulation of the three rules. The rule-based data processing system
ortho 3.0 takes phonological, morphological and lexical information
about words as input and generates a spelling using a dynamic rule
apparatus. The results of the experiment show that the three rules
yield different kinds of mistakes. The morpho-phonological and the
syllable-based approaches proved superior to the segment based
approach. The author states that sharpening in German not simply marks
shortness but marks units which embrace several phonemes, namely

Martin Neef: The Reader's View: Sharpening in German

Neef presents a way to describe writing systems from a reader's
perspective. He describes reader-based orthography as output-oriented,
as it asks for constraints on written forms. The ''recoding
principle'' is central to this approach: The written form has to make
possible an unambigiguous recoding of the spoken form. Neef summarizes
writer-based analyses of sharpening and especially addresses the
problem of prosodically determined explicit forms which sometimes are
marginal and sometimes do not exist at all. A reader-based sharpening
constraint is proposed: ''If in a word less than two consonant letters
follow adjacently a simple vowel graph, this vowel graph must not
correspond to a centralized vowel''. Neef demonstrates that this
constraint covers large parts of German orthography and systematically
lists exceptions. German Orthography is highly inconsistent in the
marking of vowel quality if the vowel grapheme is followed by more
than one consonant letter. Neef calls this orthographic
underspecification. Examining the question whether stress influences
the marking of sharpening, Neef modifies his sharpening constraint to
allow readings as centralized vowels in unstressed vord-final
syllables. Neef concludes that ''writing systems aim at being
consistent not for the writer but for the reader.''

Thomas Lindauer: How Syllable Structure affects Spelling: A Case Study
in Swiss German Syllabification

Lindauer investigates the question how writers make use of their
implicit phonological knowledge. He describes the two main
explanations for sharpening (see Noack and Neef) and then focuses on
two phenomena in Swiss German: spelling of ss and a common misspelling
of unnecessary doubling of fricative consonants after long
vowels. Swiss German permits light stressed syllables in contrast to
standard German, but this vanishes when speakers speak Swiss Standard
German. There are true geminates in Swiss German and Swiss Standard
German which are best audible with fricatives. Because these
geminates also occur after long vowels, a syllable based rule for
sharpening leads to mistakes. According to Lindauer, a stem based rule
is much more suitable for teaching marking of sharpening to Swiss
German speaking learners. This leads to the conclusion that it is
appropriate to teach different spelling rules for different regions
that show different phonological phenomena.


The papers in this volume cover a wide range of current research
topics and offer various linguistic approaches to analyse writing
systems and their use. However, three limitations to this broad
coverage apply: (1) The workshop on which most of the papers have been
presented was especially rewarding because it brought together
researchers and research results from both theoretical linguistics and
psycholinguistics. Most of the psycholinguistic papers have been
published separately in a special issue of Journal of Written Language
and Literacy. So the volume is more focussed on theoretical
linguistics than the workshop was. On the other hand this circumstance
made it possible to present a more coherent set of papers that relate
to other in many points. (2) Most of the papers deal with data and
questions from German and Dutch, two relatively similar writing
systems. This leads to a good overview of current research on these
two writing systems and makes the volume especially interesting for
linguists interested in German and Dutch and delivers detailed
analyses of some core phenomena from these orthographies. As a
drawback, the papers and thus the entire volume lacks a bit of a
cross-linguistic perspective. Some interesting insights into
comparative research of very different writing systems are given by
the paper from Ravid and Gillis and Ravid, for example. (3) The
questions raised in the introduction are mostly not addressed directly
in the papers. From a more general point of view, the three papers
contributed by the editors are the most interesting. Most of the other
papers deal with relatively specific questions and make it sometimes
difficult to relate them to the guiding questions of the book. This,
of course, is inevitable given the nature of the volume as a
collection of workshop contributions.

All in all, Neef, Neijt and Sproat present an up-to-date collection of
papers that diverge in methodology, background and generality. The
book is especially interesting for linguists interested in current
discussions of German and Dutch orthography but also shows fields of
future research on the more general question of the relation of
writing and spoken language.


Sproat, Richard. 2000. A Computational Theory of Writing Systems.
Cambridge: Cambridge University Press.


Tobias Thelen is a Ph.D. student at the University of Osnabr�ck,
Germany. He works on computational models of German orthography and
the automatic analysis of children's spellings.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue