ChapterPDF Available




Non-concatenative morphological phenomena appear on the face of it to require a powerful morphological component, capable of more than straightforward addition of affixes consisting of segmental material. We propose that the full range of non-concatenative phenomena may be completely accounted for in piece-based terms using analytical tools that are independently necessary. These phenomena include mutation, infixation, ablaut or melodic overwrite, subtraction, metathesis, reduplication, and templates. First, lexical entries for affixes may be underspecified, lacking information about segmental or featural content, which is then filled in by the phonology. Second, the way in which the content of affixes associates to the word may be prespecified in crucial respects, and phonology may be faithful to prespecified structure. In either case, non-concatenative effects are exclusively down to the phonology, not the morphology, which is purely additive. © Editorial matter and organization Jochen Trommer 2012. © The Chapters their several authors 2012. All rights reserved.
Laura J. Downing & Barbara Stiebels (ZAS Berlin)
1 Introduction
One of the defining principles of human language is that the relation between form and meaning is
fundamentally arbitrary (Saussure 1983: 100). Even in related languages, words with the same
meaning can be different enough in form to provide a barrier to understanding: e.g., head (English),
Kopf (German), huvud (Swedish). Yet, it is also clear that some classes of lexical items, some
morphological processes and some aspects of morphological and syntactic structure have an iconic
(non-arbitrary) motivation. The modern study of such iconic aspects of language is usually traced
back to Peirce (1932), whose theory of the sign distinguishes between symbols and icons. Symbols
exhibit an arbitrary relation between form and meaning, while iconic signs are characterized by
their homology with the entity signified. A very concrete form of iconic relation is illustrated by the
onomatopoeic forms in (1a), which show that words for a rooster’s crow are recognizably similar in
unrelated languages due to the intended similarity between the words and sound they are imitating.
Iconicity, in Peirce’s work, also includes more abstract relations between meaning and form, like
“diagrams”: linguistic structures that mirror in some way relations among concepts or elements of
discourse. One type of diagrammatic iconicity is illustrated in (1b), where the order of coordinated
clauses in a sentence almost universally is understood as reflecting the order of the events related.
Another instance of diagrammatic iconicity, in (1c), represents the interesting case where the
number of repetitions of the stem (reduplication vs. triplication) mirrors somehow the increasing
number of shuddering events:
(1) Iconicity in form, structure and relation
a. Rooster’s crow (Childs 1994: 189)
English cock-a-doodle-doo
Japanese kokekoko
Hebrew kukuRiku
Kisi kukuluukuu
b. Iconic order of coordinated clauses
Veni, vidi, vici
‘I came, I saw, I conquered.’ (Julius Caesar)
c. Mokilese (Harrison 1973)
roar ‘give a shudder’
roar~roar ‘be shuddering’
roar~roar~roar ‘continue to shudder’
It is diagrammatic iconicity which is considered most relevant to linguistic analysis of language
structure. As work like Newmeyer (1992) and Haspelmath (2008) notes, in the past 25 years it has,
in fact, become more and more popular in the functionalist literature, especially, to propose that
many aspects of language structure are motivated by iconicity.1 This chapter obviously does not
* We would like to thank the editor, two anonymous reviewers, Dieter Wunderlich and Wendy Sandler for thoughtful comments
on an earlier version, which led to numerous changes in the content and presentation of this chapter. The usual disclaimers
1 Note that this literature does not reflect a consistent use of the notion of iconicity. As Haspelmath (2008) comments, some
authors equate iconicity with non-arbitrariness. Such a broad view of iconicity qualifies all structures that could receive a
provide the scope to review all of these recent arguments for or against iconicity. Since this chapter
is included in a book on (morphological) exponence, we will focus on morpho-phonological
phenomena where iconicity is regularly mentioned as a motivating factor.2 Our choice of topics is
also naturally guided by our interest in particular languages and theoretical issues. The aim of the
chapter is to provide a balanced critical review of the use of iconicity to explain recurring cross-
linguistic patterns in the form and structure of words, providing both the evidence that supports
analyzing certain phenomena in terms of iconicity as well as counterexamples to overly strong
universalist claims. In spite of the fact that iconicity is often associated with functional approaches
to language, as we shall see, iconicity principles are also often implicitly or explicitly invoked in
formal approaches to morphological structure.
The order of presentation is, roughly, from the less abstract types of iconicity found in sound (or
shape) symbolism, through more abstract types of diagrammatic iconicity.3 We begin, in section 2,
with a discussion of those phenomena that come closest to the understanding of iconicity as a
relation of homology (e.g. sound/shape symbolism). We then turn to different types of
diagrammatic iconicity. In section 3, we discuss the iconicity of complexity, that is, the
generalization that more complex forms realize more complex meanings. In section 4, we introduce
the concept of isomorphism, whose role for iconicity has been highlighted by Haiman (1985). We
will show that isomorphism (one-to-one correspondence in form-meaning pairs) is a central idea in
analyses of paradigm structure. Isomorphism is also relevant for understanding the role of iconicity
in motivating unmarked linear ordering and cohesion of elements, as we will show in sections 5 and
6. In connection with the iconicity of linearization and cohesion, we will also focus on the fact that
one must distinguish an absolute and a relative notion of iconicity: the former relates to structures
that are iconic by themselves, the latter to instances of paradigmatic contrast. Here, the mapping
between competing forms to distinct, though closely related meanings is iconic. In the final section,
we conclude and point out desiderata for future research.
2 Homology
As noted in the introduction, the onomatopoeic vocabulary found in the lexicon of all languages
represents a striking example of iconic homology. The notion of homology is defined in terms of
reference. The iconic nature of onomatopoeias is based on a relation between the sign and its
referent, as the form of the words clearly is intended to imitate the sounds the words refer to. The
much-cited case of the quote in (1b) Veni, vidi, vici in which the sequence of verbs iconically
reflects the temporal order of the denoted events, is likewise based on a referential relation.
It is obvious that the linguistic means to encode homology to some entity in the external world
is restricted. The linguistic sign can encode homology only in terms of particular segmental or
suprasegmental material or in linear terms (order of elements or degree of cohesion between
elements). Indeed, onomatopoeia is considered to comprise a small and unproductive corner of
human language. Other forms of iconic sound-meaning homology (or sound-hand shape homology,
functional motivation as iconic. In contrast to this loose understanding, Ramat (1995) proposes that iconicity is a relation of
similarity and does not, for instance, imply properties such as transparency. Transparent structures, hence, would not be iconic
per se in this view.
2 According to Haiman (1985a) morphological structures tend to be less iconic than syntactic structures because morphologically
complex forms may undergo lexicalization and reanalysis and even erosion. According to Dressler (2005) there is a higher
amount of iconicity in extra-grammatic morphology (e.g. zigzag). However, hypocoristics, being extra-grammatic as well, often
involve truncation (e.g. Australian English sandie ‘sandwich’, tabbie ‘tablet’, connie ‘conductor’), which is not transparently
3 One aspect we will not deal with in this paper is Iconicity of categorization (e.g. Hopper & Thompson 1985, Wierzbicka 1995),
according to which systematic patterns of conceptualization lead to analogous categorization in morphosyntax.
in Sign Languages) have been argued to play a more important role in structuring the lexicon in
many languages, however. In sections 2.1 and 2.2, below, we present a critical overview of two
types of iconic homology that surveys like Hinton et al. (1994) and Nuckolls (1999) have found to
be common enough in the world’s languages that they qualify as universal tendencies: size
symbolism and the frequency code, and ideophones (or expressives).4 In section 2.3, we briefly
discuss the issue of iconicity in sign languages.
2.1 Sound/shape symbolism
The strongest claims for universality in sound symbolism are found in the literature on size (or
magnitude) symbolism: e.g., Nuckolls (1999), Ohala (1994), Ultan (1978). The iconic (i.e., non-
arbitrary) form-meaning relationship conveyed by size symbolism is that, in languages with size (or
distance) related morphology, high tones, vowels with high second formants (notably /i/ and other
front vowels) and high frequency consonants (palatals) are associated with small size or small
distance. In contrast, low tones, vowels with low second formants (notably /u/ and other back
vowels), and low frequency consonants are associated with large size or large distance. Two main
types of evidence are cited to support the claim that this relationship represents a potential language
universal. First, typological studies (notably, Ultan 1978) surveying languages for size-related
morphology (or at least what he calls “patterned occurrence” of size symbolism) demonstrate, first,
that size/distance symbolism is widespread, occurring in about a third of the languages in Ultan’s
survey. They also demonstrate a strong tendency for small size (or distance) to be indicated with
high frequency sounds and large size (or distance) to be indicated with low frequency sounds; other
phonological properties are seldom recruited for size symbolism. The sort of data presented to
support this conclusion is illustrated in (2). For example, in (2ai), we see that in Shona, the final
vowel of proximal demonstratives copies the first vowel, while their distal correlates contain a low
frequency vowel, /o/. The English examples in (2aii) illustrate the widespread pattern of ending
diminutive nicknames (italicized) with the high front vowel, [i]. In (2b), we find an example from
Ewe showing High tone correlating with small size (cf. the tone in the bolded words). And in (2c)
are examples from Japanese mimetic vocabulary, showing a correlation between consonant
palatalization and small size (or semantically related qualities, like instability, immaturity and
cheapness, etc.):
(2) Size-sound symbolism
a. Distance symbolism in Shona (Ultan 1978: 531)
i. Proximal demonstratives: uyu, aya, iyi; Distal correlates: uyo, ayo, iyo
ii. English nicknames: John / Johnny, Thomas / Tommy, Julia / Julie, Susan / Susie
b. Size symbolism in Ewe (Ultan 1978: 532); accents indicate tone
àtìgo lè gòlìì ‘the barrel is cylindrical’
vs. kpéví lè gólí ‘the little stone is rounded’
c. Japanese mimetics (Hamano 1994: 148-149)
i. suru-suru ‘something passes smoothly’
syuru-syuru ‘something goes through a narrow space and makes a noise’
ii. noro-noro ‘slow movement’
nyoro-nyoro ‘a snake’s wriggly and curving movement’
4 Although reduplication has been analyzed as a kind of sound-symbolism (repetition of form is said to correlate with repetition of
meaning), we discuss it in section 3, below, which deals with Iconicity of complexity, because we think that the aspect of
homology (similarity between referent and linguistic expression) is less prominent in reduplication than the way it mirrors
semantic complexity via morphological complexity.
iii. kata-kata ‘something solid and square hits a hard surface and makes a
homogeneous sound’
katya-katya ‘hard objects such as keys hit each other and make a variety of
As these examples and the collection of articles in Hinton et al. (1994) show, the claim that size
symbolism is cross-linguistically widespread depends on a metaphorical interpretation of size as
including not only the concept of physical smallness but also brightness or lightness, quickness,
proximity, and attitudes such as affection, intimacy or disdain (Nuckolls 1999: 230).
The other source of evidence for the universality of the correlation between high frequency-
small size and low frequency-large size is psychological experiments. Sapir (1929) reports what
appears to be the first of these experiments, which presented subjects with invented word pairs,
identical except for the vowel: e.g., [mil] / [mal]. Sapir found that when subjects were presented
with paired words, the word with /a/ was 80% more likely to be judged as appropriate for naming a
larger object than the word with /i/. Sapir’s judgment experiment inspired many follow up
experiments, thoughtfully surveyed in Nuckolls (1999) and Hunter-Smith (2007). For example,
Imai et al.’s (2008) study of the acquisition of Japanese mimetic verbs like those in (2c), above,
shows not only that adult speakers of both Japanese and English judge novel mimetic forms to be a
better fit to the action if they follow size-symbolic patterns. It also shows that Japanese 3-4 year
olds are significantly better at generalizing the meaning of mimetic verbs to novel forms than they
are at generalizing the meaning of non-mimetic verbs to novel forms. Other experiments have
presented subjects with size-related words from a variety of languages unknown to the subjects (and
unrelated to the subjects’ native language), rather than invented words. While the interpretation of
these experiments remains controversial, it must be acknowledged, as both Nuckolls (1999) and
Hunter-Smith (2007) note, that no satisfactory alternative explanation to iconicity has been
provided for why subjects with different language backgrounds have above-chance success rates in
these experiments at guessing the meaning of unfamiliar size-related words, when the words match
the universal size-symbolic patterns: high / front vowels denote small size and low / back vowels
denote large size.
Because size symbolism is geographically widespread and influences subjects’ assignment of
meaning to unknown words in psychological experiments, work since Sapir (1929) has proposed
that its non-arbitrariness is iconic and has a biological basis. One proposal is that /i/ is associated
with smallness and /a/ with bigness, due to the relative space between the articulators when
pronouncing high / front vs. low / back vowels. More recently, Ohala (1994) has argued that both
the attested range of phonetic patterns and the range of concepts and attitudes associated with size
symbolism are best explained by the Frequency Code Hypothesis. According to this hypothesis,
high frequency sounds high vowels, high tone, palatal consonants are associated not only with
small size but also with attitudes like affection, disdain, submissiveness because smaller human
beings (and other animals) naturally have voices (or make other sounds) that have a higher
frequency than those of larger animals. If small size is associated with non-adult and/or non-
dominant status, higher frequency sounds would iconically be associated with words denoting these
semantic fields. Ohala (1994) argues that this hypothesis has the further advantage of linking size
symbolism to cross-linguistically common functions of high pitch vs. low pitch patterns in
intonation. Questions and other expressions of uncertainty are often accompanied by high or rising
pitch, while assertions are accompanied by low or falling pitch (Bolinger 1978; Gussenhoven
2004). This pattern follows iconically from the Frequency Code Hypothesis, if one assumes that
asking questions is associated with non-dominance while assertion is associated with a dominant
The disadvantage of the Frequency Code Hypothesis and other work making a strong claim
for the universality of the iconic size symbolism and intonation patterns it accounts for – is that it is
too strong. If these patterns are biologically based then we might expect them to be more
consistently incorporated into natural languages. However, as noted above, Ultan’s (1978)
typological study shows that only roughly a third of the languages sampled have size-related
symbolism in their morphology. More problematic is the fact that one finds numerous exceptions
and counterexamples to the hypothesis. Citing Ultan’s (1978) typological study again, one can note
that while 80-90% of the languages with size-symbolic patterns equate high frequency sounds with
small size – a strikingly high proportion, to be sure – some 10-20% of the languages in the sample
therefore do not. As Diffloth (1994) notes, a universal “tendency” is not the same as a universal.
And a universal should not be reversed in any language, yet this is what Diffloth shows occurs in
the pattern of “expressive” formation (i.e., ideophones) in Bahnar. As shown in the data below, in
this language high frequency vowels (i.e., high and mid tense vowels) are associated with bigness
and low frequency vowels (mid lax vowels) are associated with smallness:
(3) Bahnar expressives (Diffloth 1994: 110)
a. /blooŋ-blooŋ/ ‘of numerous reflections caused by rays of light on a large object’
vs. /blɔɔŋ-blɔɔŋ/ ‘id., small object’
b. /bliil-ɲip/ ‘of a large scintillating fire…’
vs. /blɛɛl-ɲɛp/ ‘id., small fire’
c. /jul-kәjul/ ‘of a large creature, trotting about’
/jɔl-kәjɔl/ ‘id., small creature’
The role of the frequency code in explaining crosslinguistic tendencies in intonation patterns faces a
similar problem, namely, the “universal” correlation between rising pitch and questions is reversed
in a number of languages. For example, as Rialland (2007) demonstrates, a majority of the 78
genetically diverse languages in the Sudanic belt included in a survey of question intonation do not
use high or rising tones to mark questions but rather use low tones or some other “lax” prosody.
In sum, the correlation between high frequency sounds and small size (and related concepts) is
common enough that it seems to be iconic in the sense of being non-arbitrary. However, what
remains controversial is the source of this iconic relation.
2.2 Ideophones
The lexicon of most languages contains ideophones: types of sound-symbolic words which have in
common that they intend to “give a vivid representation of an idea in sound” (Doke 1935: 118) and
typically function to dramatize a narration (Kilian-Hatz 2001: 156). Some of these classes of words
are clearly iconic. For example, as noted in the introduction, onomatopoeic words aim to mimic in
sound what they denote:
(4) English onomatopoeic words
cock-a-doodle-doo, meow, boo-hoo, clap, boom, buzz, thwack
Some onomatopoeic words also make iconic use of size-symbolism like clink vs. clunk – and/or
iconic reduplication, like ding-dong or clickety-clack, in English. Other classes of words with
sound-symbolic patterning are not so straightforwardly iconic. For example, English phonaesthetic
word sets show a regular association between some word-initial phonemic sequence like word-
initial gl – and a related set of meanings – in this case, having to do with brightness or light:
(5) English gl phonaesthemes
glossy, glitter, glimmer, glow, glare, gleam, glance, glint, glisten
The form-meaning association found in phonaesthemes like these is non-arbitrary, as work like
Bolinger (1949) points out, in the sense that it defines a systematic pattern (albeit, with exceptions,
like glove or glum),5 yet one cannot say that the sound mimics the meaning. Similarly, the bolded
words in the Zulu sentences below are recognized by Zulu speakers as having non-arbitrary formal
properties which mark them as belonging to the class of ideophones, yet the form-meaning
associations are not straightforwardly iconic:
(6) Zulu ideophones (Voeltz & Kilian-Hatz 2001: 1)
a. Tula uthi tu!
‘Keep perfectly silent!’
b. Kubomvu tubhu.
‘It is bright red.’
c. Wathi uyahamba wathi twa obukwini.
‘As he went he sank deep into the bog.’
In English and other European languages, ideophones have received relatively little serious
linguistic attention, as they are considered a peripheral, even childish or whimsical, part of the
lexicon, common in oral narrative, children’s literature and comic books but mostly banned from
the formal standard written language (Childs 2001; Nuckolls 1999; Oswalt 1994; Rhodes 1994).
However, as the articles (including a lengthy bibliography) in Voeltz & Kilian-Hatz (2001) make
clear, ideophones are well represented in the languages of the world, and in many languages they
make up a productive and sizable proportion of the lexicon. Childs (1994: 179) cites the following
figures for selected African languages: 8,000-9,000 ideophones in Gbaya; 2,600-3,000 in Zulu;
25% of the lexicon of Nupe; 7-11% of the lexicons of Kisi and Yag Dii. Watson (2001: 400)
estimates that ideophones make up a significant proportion (10-30%) of the words in many
Southeast Asian languages. A recurring theme in the cross-linguistic literature on ideophones is that
they are defined as a class by having marked phonological, morphosyntactic and/or semantic
properties (Nuckolls 1999). Klamer (2001; 2002) argues that their relative markedness lends them
diagrammatic iconicity: one finds a match between marked form and marked status in other areas of
the grammar. To assess this claim, let us review some of the marked properties which have been
attributed to ideophones.
Klamer (2001: 167; 2002) defines ideophones as phonologically marked in the sense of not
following constraints that are active for other parts of the vocabulary. They often use segments or
prosody outside of the usual phonemic inventory and violate phonotactic constraints of the
language. Klamer (2001: 171; 2002) cites Kambera to illustrate: in this language a set of marked
vowels à, è, ò, ù occurs only in ideophones. (See (7), below.) Numerous other examples are
listed in Childs’ (1994) detailed survey. In Tumbuka tone is contrastive in ideophones but not in the
rest of the lexicon. In Southern Bantu languages like Zulu, ideophones commonly are monosyllabic
(see (6a/c), above), in violation of a general bisyllabic minimality constraint holding for other
words. In many languages, at least some ideophones can be lengthened indefinitely or repeated
more than once to iconically indicate duration or repetition. This is not possible in words with
grammatically contrastive length – where one finds an absolute contrast in the duration of long and
short vowels – or morphological reduplication – which normally only allows a single repetition.
5 According to Bergen’s (2004) study of English phonaesthemes, in an online version of Webster’s 7th collegiate dictionary, fully
39% of the word types that begin with ‘gl’ have definitions related to light or vision. Bergen’s priming study shows that this
form meaning correlation is frequent enough that ‘gl’ has morphemic status in English, a result which Bergen argues is
problematic for compositional theories of morphology while supporting connectionist models like that of Bybee (1995a, b). It is
beyond the scope of this chapter to take up the relevance of phonaesthemes and other sound-symbolic vocabulary for this
debate. The interested reader should consult Bergen (2004) for thoughtful discussion.
Finally, in many languages, ideophones are pronounced at a distinctively higher or lower register
than other words and are set off by a pause.
This prosodic setting-off of ideophones is often matched by morphosyntactic isolation or
markedness. Indeed, a recurring cross-linguistic syntactic property is that ideophones are
syntactically “aloof”, set apart from other items in the sentence. When they are integrated into the
syntactic structure, they often remain distinct in being obligatorily preceded by a dummy verb with
a meaning related to “do” or “say.” This can be illustrated with onomatopoeic forms in English:
e.g., The dog went, “bow-wow. Childs (1994: 187) gives many more examples like this, as do
several articles in the Voeltz & Killian-Hatz (2001) volume.
In Kambera, the phonological markedness of ideophones is matched by a morphological
property: verbs derived from ideophonic roots are the only words formed with a circumfix, /ka-
…-k/ (Klamer (2001: 171; 2002). Note, too, the marked vowels (indicated with a grave accent, as
noted above), which are distinctive for ideophonic roots:
(7) Kambera ideophones
Ideophonic root Verbal form Gloss
a. ngùru ka-ngùru-k ‘murmur’’
b. hèri ka-hèri-k ‘tear something’
c. tòru ka-tòru-k ‘rattle’
d. pàka ka-pàka-k ‘smack’
e. wàdi ka-wàdi-k ‘blink’
f. rèri ka-rèri-k ‘ablaze’ (fire) / ‘shine’ (ring)
Ideophones are also often morphosyntactically marked in having no morphology (Childs 1994;
Kilian-Hatz 2001; Nuckolls 2001). In Tumbuka and Southern Bantu languages (like Zulu), for
example, ideophones have no class agreement or other affixes, unlike all other word classes. This
property can also make it hard to assign them to a lexical category in many languages.
Klamer (2001: 167; 2002) proposes that these formally marked properties of ideophones are
typically mirrored by semantic markedness, and this allows us to identify diagrammatic iconicity in
the form-meaning relationship in ideophones. Semantically unmarked items, Klamer proposes, have
a general interpretation, while more marked items are more specific and concrete. Indeed, a general
property of ideophones is that they often convey very specific imagery, as in the rain ideophones
found in Kisi (Childs 1994: 189); note the iconic potential for indefinite lengthening in the first two
words to indicate duration of the sound:
(8) Kisi rain ideophones
a. wa-a-a-… ‘sound of rice being sown or gentle rain’
b. bia-a-a-… ‘sound of soft rain’
c. bakala-bakala ‘sound of rain falling in single droplets’
The use of size-symbolism, onomatopoeia and reduplication in forming some ideophones also gives
them a specific or concrete semantics due to the iconic nature of these word types (Watson 2001).
These word classes, then, have both directly iconic properties – their form mimics their meaning
as well as diagrammatic iconic properties – specific, non-arbitrary semantics. Rather paradoxically,
Klamer (2001, 2002) also proposes that ideophones are semantically or functionally marked
because they are often non-referential and optional, simply underscoring the meaning of an adjacent
referential word or adding liveliness or directness to the narration.
Both types of unmarked semantic properties of ideophones can be illustrated with examples
from Pastaza Quechua. Nuckolls (2001: 274) notes that the ideophone polang, with a meaning
related to emerging upwards from water, can appear with several verbs of motion, refining the
meaning of the verb, while another, more onomatopoeic ideophone with a similar meaning – bhuxx
– can only be used when describing the movement of a particular freshwater dolphin:
(9) Pastaza Quechua ideophones (underlined) of emerging from water
‘“(At first) it wasn’t there and then later polang it would emerge,” saying this, he would
tell (us).’
bhuxx bhuxx
‘Climbing upriver, those dolphins went, emerging bhuxx bhuxx.
These semantic and functional properties, of course, hold generally for adjuncts and modifiers:
adverbs, for example, also are optional, non-referential and add specificity and, potentially,
liveliness to the meaning of the referential words they modify. While the dramatic or poetic
function of ideophones often makes them easy enough to identify, it remains problematic, as Childs
(1994) points out, how to translate this observation into a precise definition that distinguishes
ideophones semantically from other word types and characterizes what is marked about their
Whether or not ideophones in any particular language illustrate diagrammatic iconicity depends,
then: on whether or not they show marked formal properties; on whether they show marked
semantic or functional properties (much harder to define); and, crucially, on whether there is a
correlation between markedness of form and markedness of meaning in that language. Klamer
(2001, 2002) argues that Kambera satisfies all of these criteria. As we saw in (7), above, Kambera
ideophonic verbs have marked formal properties: they uniquely contain marked vowels and are
derived by infixation. They also can be considered semantically marked as they express very
specific sense impressions: “sound, touch, taste, smell, feeling, emotion and sigh (including
lexicalizations of movements of the body and of body parts)” (Klamer 2001: 169). Ideophones in
this language meet the definition of a word class iconically correlating marked form with marked
Less perfect correlations are found in languages where ideophones either do not consistently
have a marked form or do not consistently have a marked meaning. Both types of ideophones are
easy enough to find. Phonaesthemes like the English set in (5) can be considered ideophones, due to
their use of sound symbolism: we find a systematic non-arbitrary correlation between form and
meaning. They also meet the criterion of semantic markedness: the words express specific sense
impressions. However, these words are not formally marked: gl is a well-formed syllable-initial
consonant cluster, found in many non-ideophonic English words (glove, glum, glue, etc.). They also
are morphosyntactically unmarked, easily categorized as nouns or verbs with the usual potential for
affixation and occurring in the appropriate position in a sentence.
Nuckolls (2001) demonstrates the inverse markedness mismatch for the ideophone tak in
Pastaza Quechua. This ideophone is phonologically marked: regular words of the language seldom
end in consonants; only ideophones regularly do. However, it is not always semantically marked.
As Nuckolls shows, while all the diverse uses of the ideophone can be unified as involving some
form of contact, only in some cases, like (10a), is the contact concrete and specific, as we expect
from an ideophone. In others, it has a more abstract, aspectual interpretation linked to punctuality or
completiveness of the action. This is illustrated in the sentence in (10), which describes the
successive coilings of a snake’s body (dziriri) until the final positioning (tak) of its tail in mid-air:
(10) Pastaza Quechua (Nuckolls 2001: 281)
a. tak as the sound of contact
‘When it is touched tak from above, it clings.’
b. tak as ‘contactless contact’ with completive or punctual meaning
dziriri dziriri dziriri dziriri dziriri tak
‘Then here (it coiled itself) dziriri [x 5] and (placed) its tail tak above.’
In short, while the dramatic and expressive function of ideophones leads one to expect that they
should show iconicity, clearly-defined iconic properties – whether direct or diagrammatic – are
often difficult to identify.6 As work like Bolinger (1949) and Bergen (2004) argue, though, the
existence of a statistically significant set of phonaesthemes (e.g., English glint, glimmer, gleam,
etc.) or size-sound symbolic vocabulary in a language lends them psychological reality as non-
compositional morphemes, with the ability to motivate language change or to be the source of
neologisms. For example, in 16th century English, sacke underwent an irregular change in
pronunciation to sag, most likely due to ‘phonaesthemic attraction’ to words like drag, flag, lag
which share the meaning of ‘slow, tiring, tedious motion’ (Bergen 2004: 307). Developments like
this would reinforce the non-arbitrariness of these sound-meaning correspondences in the language,
lending them the iconic property of isomorphism, discussed in section 4, below.
2.3 Role of iconicity in sign language
The role of iconic homology in motivating the form of words in sign languages is a long-standing
issue in the field, but with the opposite emphasis than in spoken language. For spoken language, it
is taken for granted that for most words the form is arbitrarily linked to the meaning. Iconicity must
be explained and argued for. For sign language, in contrast, an important research goal of modern
linguistic research has been to demonstrate the arbitrary and abstract properties of signs and sign
language morphosyntax, in order to counter the misleading earlier characterization, found for
example in the brief discussion in Bloomfield’s (1933: 144) highly influential work, that sign
language is merely composed of iconic gestures. This section reflects this different emphasis and
begins by sketching the important arguments showing sign language words and word-formation
processes have many of the same arbitrary properties as spoken language. Then we present some
aspects of the lexicon and morphology of sign language which show iconic properties. In some
cases, these iconic properties parallel ones discussed above for spoken languages. In other cases,
they reflect linguistic resources unique to sign language.
Modern sign language linguistics is generally considered to begin with Stokoe’s (1960) ground-
breaking work demonstrating that sign language words and morphemes are not, in fact, iconic
gestures akin to pantomime. Instead, signs are composed of a fixed, language-specific repertoire of
hand configurations, places of articulation and movements. Like the phonemes of spoken language,
each of these components is meaningless on its own. Only well-formed combinations of these
components define meaningful morphemes. Klima & Bellugi (1979) report on a number of
experiments and studies confirming that the signs of sign language are componential, not
6 As one of the reviewers points out, Ghomeshi et al.’s (2004) study of intonation in different English reduplication patterns
shows that intonation can also iconically express emotive connotations of reduplication patterns. One might expect that, in some
languages, particular ideophones could have a distinctive intonation pattern which might provide a different impression of how
iconic they are and also might provide another parameter for evaluating their phonological markedness. Unfortunately, this kind
of information is not readily available and will have to remain a topic for future research.
transparently iconic symbols, quite distinct from pantomime. To test how transparently iconic sign
language is, Klima & Bellugi (1979) carried out a series of perception experiments which required
naïve non-signers to assign a meaning to lexical signs (in ASL – American Sign Language)
referring to both concrete and abstract nouns. Not a single subject was able to guess the meaning of
81 of the 90 signs when the subjects were not provided with a choice of possible meanings. Even
when subjects were asked to choose a meaning from a given list of possibilities, they did no better
than chance at guessing the meaning. These results clearly show that ASL words are non-iconic
enough that any sign language is as opaque to those who do not know it as any spoken language is.
Klima & Bellugi (1979) also show that there is a great deal of variation among sign languages,
which would be unexpected if sign languages were all just pantomimes involving iconic gestures.
The sign for tree, for example, is iconically based in several different sign languages, yet different
aspects of tree have been symbolized in each sign language. This demonstrates the arbitrariness
underlying even rather iconic signs: the community of signers must choose what aspects of a
referent to represent and then codify a representation within the constraints imposed by the
parameters defining possible signs in a particular sign language. Because signs are symbols, not
pantomimes, it is also possible for the same sign to have quite different meanings in different sign
languages. Klima & Bellugi (1979: 152) provide several examples of this in their detailed
comparison of Chinese Sign Language (CSL) with ASL: the CSL sign for ‘father’, for example, is
identical to the ASL sign for ‘secret’; the CSL sign for ‘explain’ is identical to the ASL sign for
Other evidence for non-iconicity in sign language is that the diachronic development of signs
typically shows increase in arbitrariness in order to satisfy well-formedness constraints on sign
shape and sign language ‘phonotactics’ (Frishberg 1975). For example, the sign for ‘home’ in ASL
was originally a compound composed of two rather transparently iconic signs, ‘eat’ (made by
bringing one closed hand towards the mouth) and ‘sleep’ (made by leaning the head towards both
hands placed together to form a pillow). In the current sign for ‘home,’ these original components
have been so drastically reduced, to make the sign fit ASL word well-formedness constraints, that it
is considered one of the most opaque signs of ASL. In another test for the iconicity of signs, Klima
& Bellugi (1979: 30) found that not a single subject guessed that the gestures involved in the sign
for ‘home’ are related to eating and sleeping.
Synchronic evidence internal to sign language for the non-iconicity and componentiality of
signs comes from the fact that mistakes by deaf children acquiring sign languages are often counter-
iconic but logical given the regularity and componentiality of sign language grammar. (For
example, using an incorrect, articulatorily simpler hand configuration with the correct place of
articulation and movement.) Also, slips of the hand by adult native signers show that iconicity plays
no role in motivating mistakes, which instead usually involve substituting one parameter of the sign
(like place of articulation or hand configuration) for another. Klima & Bellugi (1979) discuss the
significance of such mistakes for the linguistic analysis of ASL structure in some detail.
However, like spoken language, sign language has areas of the lexicon and grammar which are
iconic. Similar to spoken language, reduplication is a productive morphological process with a
range of meanings consistent with the Iconic Principle of Reduplication (13), discussed in the next
section. ASL reduplicative constructions express concepts like duration or iteration or distribution
or intensity (Klima & Bellugi 1979; Sandler & Lillo-Martin 2006: 232; Wilbur 2005). Many sign
languages have what is called a classifier construction which has several properties that are
reminiscent of the ideophones of spoken language. In many signs only one hand – the dominant
hand – forms the hand configuration which is contrastive for that particular sign. The non-dominant
hand plays a passive role and assumes an unmarked hand configuration. However, it is also possible
for the non-dominant hand to modify the meaning of the sign by assuming one of the classifier hand
configurations: for example, an extended single pointer finger for a thin, straight shape; all four
fingers extended for a wide, straight shape; two fingers pointing down for a human; the entire hand
pointing upward for a tree. Classifier morphemes can also involve modifications of the contrastive
movement associated with a sign or of its place of articulation. Like ideophones, they are most
common in – even essential to – lively narrative discourse. Like spoken language ideophones,
classifiers have a marked form: they often violate well-formedness constraints on hand
configuration and movement for two-handed signs. And like ideophones in spoken languages, they
are semantically marked in Klamer’s (2001) sense of serving to make the meaning of their referents
more specific and concrete. For example, they can be used to indicate the difference between a car
skidding or wandering uphill. Or they can make a construction’s meaning more specific in a way
that seems analogous to class agreement: the classifier construction depicting a falling event is
made with the human legs classifier configuration if a human falls but with a different configuration
if a telephone pole falls (Sandler & Lillo-Martin 2006: 78-83). As we can see, the classifiers are
iconic (once one knows their meaning), in the sense that one can see that the classifiers represent
directly some visual aspect of what they denote. This makes them seem more like onomatopoeia in
spoken language, except that the visual medium of sign language allows for more aspects of real
world referents to be mimicked in language form.
Sandler & Lillo-Martin (2006: 501-503) show that the visual medium of sign language also
allows for some forms of iconicity particular to sign language. In the Sign Language of the
Netherlands (SLN), unselected (non-contrastive) finger shape is often used iconically (van der
Kooij 2002). For example, the unselected fingers of the SLN sign APPLY-LIPSTICK can be signed
with unselected fingers extended if the lipstick has been applied in a thick layer, but with closed
unselected fingers if the lipstick has been applied more modestly. In SLN and Israeli Sign Language
iconicity blocks a phonological process called Weak Drop, which involves omitting the non-
dominant hand in some symmetrical two-handed signs, if the non-dominant hand expresses too
much iconic semantic information. Finally, in SLN and ASL – perhaps other sign languages – when
marked or anomalous hand configurations are sporadically found outside of the classifier system,
they are always clearly motivated by iconicity.
In short, while the lexicon and morphosyntactic structure in sign language, as in spoken
language, is predominantly arbitrary, iconicity of homology does seem to play a stronger role in
these languages, due to the inherently greater potential signs have compared to sound for mimicking
their referents.
3 Iconicity of complexity
Iconicity of complexity is the most relevant notion of iconicity for morphologically complex
structures because it gives a natural account for the fact that all languages that allow for complex
morphological structures have concatenative morphology (among possible other means of
exponence). For this reason, we discuss in some detail its role in diverse areas of prosodic and non-
prosodic morphology. Iconicity of complexity reflects the generalization that more complex
meanings are expressed by more complex forms.7 This can be schematized as follows:
7 It is worth noting that some cognitive studies have questioned the implicit view of the direction of the isomorphism of
form/structure and referent/meaning. Generally, the relationship is seen as a mapping from meaning/reference to form/structure.
However, as Slobin has shown in various papers (e.g. Slobin 2005 and references therein), this relation is not unidirectional.
There are data that suggest that language may shape perception (Slobin’s “thinking for speaking”), more specifically that “the
signified appears to be dependent on the signifier” (Slobin 2005:308).
(11) a. Iconic complexity b. Non-iconic complexity
M1 (σ) F1 /X/ M1 (σ) F1 /X/
M2 (σ α) F2 /X+z/ M2 (σ α) F2 /X+z/
If two forms F1 and F2 differ in terms of extra (supra-)segmental material z and two semantic
representations M1 and M2 differ in terms of an extra meaning component α, then M1 should be
assigned to F1 and M2 to F2 as in (11a); the inverse assignment (M1 to F2 and M2 to F2) as in
(11b) would be non-iconic; it would imply that the meaning component α is associated with a
truncation operation that eliminates the segmental material z from the base. An illustrative case is
found in the encoding of comparative/superlative:
hypothetical language
In English – and many other languages – the comparative and superlative are morphologically more
complex than the positive; the superlative is encoded by a marker that contains more segmental
material than the comparative. Hungarian even shows a derivational relation between the
superlative and the comparative. The English and the Hungarian pattern are expected if the
comparative and superlative involve semantic enrichment of the positive form – as generally
assumed in the literature (see, for instance, Bierwisch 1989b, Kennedy 2001).8 Quite unexpected is
the pattern indicated for the hypothetical language in (12): here, the positive exhibits the most
complex form, while the comparative and superlative are derived via truncation. No such pattern
has been reported so far.
It is obvious that concatenative morphology seems to encode Iconicity of complexity best in that
greater formal complexity manifests itself in additional segmental material. Apophonic encodings
such as ablaut are harder to evaluate in terms of formal complexity.9 In our view the overwhelming
cross-linguistic preference for structures that follow Iconicity of complexity is not very remarkable
at all, given the differences in expressive power between a system of morphological marking that
implements “more form – more meaning” in a more or less agglutinative/concatenative morphology
and one that does not. For example, no inventory of truncating operations could ever have the
potential to distinguish the set of meanings that may be distinguished by separate overt morphemes.
There are generally only a few ways to distinguish various meanings by means of distinct
truncations and still keep the morphological base (usually the root/stem) recoverable. If languages
display operations such as truncation, infixation or reduplication, their number is usually small,
being restricted to a small set of semantic operations. The extensive study on subtractive
morphology by Weeda (1992) shows that simple truncation (i.e., the resulting form is specified in
prosodic terms) only has a limited function (usually deriving hypocoristics, vocatives or
abbreviations), whereas subtractive truncation (i.e. the part of the base that is retained in truncation
8 Note that the data on comparatives and superlatives have been used to point out that one must distinguish between iconicity that
is given in terms of reference (the cases of homology discussed above) and one that is given in terms of meaning (see Nöth
2008). If the increasing formal complexity of positive – comparative – superlative (e.g. largelargerlargest) were interpreted
as an increase in the referential domain (sometimes dubbed Iconicity of quantity), the gradation of the corresponding antonyms
(e.g. smallsmallersmallest) would be non-iconic because the comparative of these adjectives refers to a decreased degree of
the respective property. See Haspelmath (2008) for additional critiques of the explanatory value of Iconicity of quantity.
9 Usually, the various means of exponence are ranked according to their degree of iconicity. Dressler (2005), for instance, points
out that agglutinative morphology is more iconic than stem alternation (e.g. sing/sang), which in turn is more iconic than
conversion (i.e. zero-marked category shift). Truncation/subtraction is regarded to be non-iconic.
is specified in prosodic terms; see section 3.2, below) may have paradigmatic function and occur in
inflection or word formation (e.g. plural in Koasati, past perfective in Tohono O’odham). However,
so far no language has been reported with a high number of truncating morphemes.
In the following sections, we discuss specific domains of prosodic and non-prosodic
morphology where it can be argued that Iconicity of complexity plays an explanatory role. In
sections 3.1 and 3.2, we show that the output shapes characteristic of reduplication and other
morphological constructions subject to prosodic minimality reveal a strong correlation between
more meaning/function and more form. In section 3.3, we discuss the role of markedness
hierarchies in a formal account of differential object marking, while section 3.4 shows that iconicity
needs to be relativised with respect to specific lexical items. Section 3.5 discusses the implicit role
of iconicity of complexity in selected current formal frameworks.
3.1 Reduplication
Reduplication is defined as the systematic repetition of all or part of the phonological material of a
base for derivational or inflectional purposes. While it is uncommon in European languages, it is an
extremely widespread and productive morphological process, found in all language families,
including Indo-European (Rubino 2005). Even though, in principle, reduplication, like other
morphological processes, could express any meaning, Moravcsik’s (1978) typological survey joins
many others in showing that one finds a strong tendency for reduplicative morphology to have an
iconic meaning. Kouwenberg & La Charité (2005: 534) state this generalization especially
(13) The Iconic Principle of Reduplication:
More of the same form stands for more of the same meaning.
This principle encodes two types of iconicity. First, a reduplicative morpheme, like other
concatenated morphemes, illustrates iconicity of complexity. Adding the reduplicative morpheme to
a base makes its output more phonologically and more semantically complex. Reduplication also
illustrates iconic homology: repetitive form mirrors repetitive meaning. Moravcsik’s (1978: 317)
survey of reduplication shows that reduplication is very commonly used to indicate increased
quantity of participants or of events or to indicate an increased amount of emphasis. Kouwenberg &
La Charité (2005) provide numerous examples of reduplicative forms which illustrate reduplicative
iconicity of homology from Caribbean Creole languages, and the other articles in Hurch (2005), as
well as Moravcsik’s (1978) survey, confirm that the Iconic Principle formalizes a recurring
crosslinguistic tendency in the meaning of reduplicated forms. A small, representative sample of
reduplication patterns illustrating these meanings is given in (14). The Pangasinan data in (14a)
illustrates the use of reduplication to form the plural of nouns; the Sundanese data in (14b)
illustrates the use of reduplication to indicate repeated or continuous action; and the Sundanese data
in (14c) illustrates the use of reduplication to indicate increased emphasis; the reduplicated form is
(14) a. Pangasinan nouns (Rubino 2005: 11)
toó ‘man’ to-tóo ‘people’
amigo ‘friend’ a-mi-migo ‘friends’
báley ‘town’ bal-báley ‘towns’
plato ‘plate’ pa-pláto ‘plates’
b. Sundanese verbs (Moravcsik 1978: 319)
guyon ‘to jest’ gu-guyon ‘to jest repeatedly’
godәg ‘to shake the head’ go-godәg ‘to keep on shaking the head’
c. Sundanese intensive reduplication (Moravcsik 1978: 321)
hayaŋ ‘want’ hayaŋ-hayaŋ ‘want very much’
rame ‘jolly’ rame-rame ‘be very jolly’
As we saw above in discussing size symbolism, iconic homology must be interpreted in a
metaphorical way, to cover a range of concepts related to increased quantity, including not only
plurality or repetition but also distributed, reciprocal, continuous or habitual action, along with
augmentatives and intensives.
Paradoxically, reduplication also often has what appears to be the opposite range of meanings
from those expected from the Iconic Principle (13). Even in the same languages, one finds
reduplication used to indicate diminution, attenuation or scattered repetition (as opposed to
intensive repetition). Kouwenberg & LaCharité (2005) provide examples of this use of
reduplication from Caribbean Creole languages. Other articles in Hurch (2005) and Moravcsik’s
(1978: 322-324) survey provide similar examples from a range of other languages. A representative
sample of reduplication patterns illustrating these meanings is given below:
(15) Diminutive or attenuated or scattered reduplicative meanings
a. Ndyuka (Kouwenberg & LaCharité 2005: 540)
guun ‘to be green’ guun-guun ‘to be greenish’
lepi ‘to be ripe’ lepi-lepi ‘to be not quite ripe’
booko ‘to break’ booko-booko ‘to be somewhat broken, dilapidated’
b. Nez Perce (Moravcsik 1978: 322)
té:mul ‘hail’ temul-té:mul ‘sleet’
xóyamac ‘child’ xoyamac-xóyamac ‘small child’
c. Zulu verb stem reduplication (Downing 1999: 15)
-bambéla ‘catch for’ -bambe-bambéla ‘catch for now and again’
-bóna ‘see’ -boná-bona ‘see here and there’
-fundísa ‘teach’ -fundi-fundísa ‘teach now and again’
Kouwenberg & LaCharité (2005) propose that these apparently opposite meanings also satisfy the
Iconic Principle in (13). Repetition is conceptually related to scattering – as well as concentration
of a property or activity, paving the way for meanings related to attenuation or diminution to
While the Iconic Principle does seem to be well supported, expressing a genuine crosslinguistic
tendency, one can easily find reduplication patterns which do not straightforwardly satisfy the
Principle, either because the same form has not been repeated or because the meaning does not
express either repetition or scattering. To begin with problems involving non-repetition of
reduplicative form, reduplication is a morpho-phonological process, and as such the form is subject
to prosodic constraints. For example, as shown in Moravcsik’s (1978) survey – and subsequent
surveys – reduplication involves, in principle, a single copy of the base form. One does not find a
base repeated several times in order to iconically represent that an action has been repeated several
times. Further, the reduplicative form is often not exactly the same as the base form. This can be
seen in the data in (14) and (15), which contain both total reduplication patterns and reduplications
which copy only an initial one- or two-syllable string. The partial reduplications convey the same
range of meanings as total reduplications, even though the reduplicated form is not repeating a
meaningful string. Infixing reduplication – rather common with vowel-initial bases, as shown by the
Pangasinan example a-mi-migo in (14a), above – represents an even more striking deviation from
iconic repetition of form and meaning. As shown by the Xhosa data in (16), the repeated form
(underlined) in an infixing pattern interrupts the base form and, like other types of partial
reduplication, does not repeat a meaningful base string:
(16) Infixing reduplication in Xhosa (Cassimjee 1998)
a. C-initial stems
-phátha -phathá-phatha ‘touch/ X here and there’
-sebénza -sebe-sebénza ‘work/ X here and there’
b. V-initial stems – infixing
-álátha -á-lathá-latha ‘to point at/ X here and there’
-óphúla -ó-phulá-phula ‘to break/ X now and again’
Reduplication does not always satisfy the Iconicity Principle, then, because the repeated form does
not have any meaning. As a result, it cannot lead to ‘more of the same meaning’.
The meanings conveyed by reduplication also are not always straightforwardly linked to notions
of repetition or scattering. For example, Kulikov’s (2005) study of reduplication in the Vedic verb
notes that the use of reduplication to convey perfective aspect is not so readily amenable to an
iconic interpretation as the intensive or frequentative or even the present continuous tense-aspects,
which also are formed using reduplication. Indeed, Rubino (2005) lists a host of grammatical
functions conveyed by reduplication in different languages e.g., inchoative, singular absolutive,
reflexive, causative which have no obvious iconic meaning. Finally, as work like Inkelas & Zoll
(2005) and Orie (1997) makes clear, apparent reduplication can sometimes be linked to no meaning,
as it fulfils purely prosodic purposes. For example, in Chechen, chained clauses are marked by a
second position clitic which immediately precedes the inflected main verb. The clitic must be
preceded by another element in the clause. If no other element is present, the main verb is
reduplicated to fill the obligatory pre-clitic position. (The clitic is underlined and the reduplicated
form and base are bolded in the example below):
(17) Chechen second position clitic (Inkelas & Zoll 2005: 9)
‘Ahmad stayed (for a while) and left.’
The reduplication is clearly a dummy form, filling an obligatory syntactic ‘slot’. It does not repeat
the meaning of the base or have a meaning of its own. Orie (1997: 58) argues that the apparent
reduplication found in some Yoruba deverbal nouns has a similar explanation. (Note that the
deverbal noun has a habitual or continuous meaning consistent with the Iconic Principle):
(18) Yoruba deverbal nouns (Orie 1997: 58)
Base verb Deverbal noun
a. ‘dance’ -jó ‘dancing’ cf. ì-jó ‘dancing’
b. ‘dye’ -kú ‘dying’
c. só’ ‘fart’ -só ‘farting’
d. ‘split’ -là ‘splitting’
Vowel-initial syllables in Yoruba can only be Mid-toned or Low-toned; High-toned syllables must
be consonant-initial. The copy of the base-verb initial consonant in the deverbal nouns is, Orie
(1997) proposes, therefore not motivated by a morphological process of reduplication. Rather, it has
a prosodic motivation: it satisfies the requirement that High-toned syllables begin with a consonant.
In sum, reduplication represents a morphological construction where we can identify two types
of iconicity. First, reduplication often involves homology: the correlation between repeated form
and increased quantity (and related concepts) is common enough that reduplication often satisfies
the Iconic Principle. Reduplication by definition also involves iconicity of complexity: doubling all
or part of a base morpheme yields a form which is both phonologically and semantically more
complex than the base.
3.2 Prosodic minimality and morphological complexity
Concepts related to the iconicity of complexity play an important role in recent theories of prosodic
morphology which aim to derive morpheme minimality and maximality requirements from more
general phonological asymmetries characterizing roots compared to affixes and their combination
into derived stems. Morphemes can be classified as roots or affixes based on asymmetries in their
semantic or functional complexity. Roots provide the basic meaning of a word, while affixes
modify the basic meaning or specify the grammatical function of the root. Roots, then, can be
considered more semantically complex than affixes because they are the semantic heads of words
(Dresher & van der Hulst 1998), while affixes are dependents.
This asymmetry in the semantic complexity of roots and affixes is reflected, in many languages,
in asymmetries in phonological complexity. As Downing (to appear) shows, in Khoisan languages
and in Bantu languages, the full range of phonemic contrasts is only realized in the root (or, even
the initial CV of the root); only a reduced phoneme inventory is found in affixes (and non-initial
root syllables). For example, in Ju|’hoansi, a Khoisan language, Miller-Ockhuizen (2001) reports
that the 89 contrastive consonants are only found root-initially; drastically fewer consonants are
found in other positions. Marked vocal phonation is also only found in the root-initial syllable in
this language. As noted in work like Beckman (1997), such phonological asymmetries in roots
compared to affixes could have a psycholinguistic motivation, as they would aid listeners in
identifying the central meaning-bearing part of a word.
The root-affix asymmetry is reflected not only in the featural make-up of these two morpheme
types but also in their respective canonical size. Even in languages where roots and affixes can both
be monosyllabic, roots are often required to consist minimally of a heavy syllable (CVC or CVV),
while affixes can consist of a light (CV) syllable. As work like Niepokuj (1991) argues, this
asymmetry is found in the diachronic grammaticalization of original free words to affixes. The
English suffix /-ly/, for example, is historically derived from the free word, like. Its current affix-
like status is reflected in its reduction in size. Work like Urbanczyk (1996) and Downing (2006)
shows that this asymmetry is fairly widespread in the languages of the world. For example, in most
Salishan languages as well as in Bantu languages, roots are minimally CVC while affixes are
typically shorter (CV or VC). Another widespread generalization about minimal word size is that
derived words (stems) are frequently required to contain minimally two syllables (compared to one,
for underived words). This generalization holds for Semitic languages, Bantu languages and
Turkish, to name just a few (Ussishkin 2000, 2005; Downing 2006). For example, Inkelas & Orgun
(1995) demonstrate that derived words are subject to a disyllabicity condition for many speakers of
Istanbul Turkish, while monomorphemic words are minimally monosyllabic (and, with few
exceptions, bimoraic) for all speakers, as shown in (19):
(19) Turkish minimality (Inkelas & Orgun 1995: 769-774)
a. Disyllabicity condition for derived words
* ye-n
‘eat-PASS’ (=’be eaten!’)
cf. ye-di
* de-n
‘say-PASS’ (=’be said!’)
cf. de-mek
b. Monosyllabic underived words
Work like Downing (2006), Dresher & van der Hulst (1998), Itô (1990) and Ussishkin (2000, 2005)
propose that there is an iconic motivation for why derived stems, roots and affixes are subject to
different minimality constraints. The greater semantic or morphological complexity of these three
morpheme types is mirrored in their (minimal) prosodic complexity. More specifically,
semantically complex roots are minimally required to be prosodically branching while affixes are
not. Morphologically complex derived stems are minimally required to branch into two syllables,
while morphologically simplex roots can be branching monosyllables.
Evidence for this iconic approach to minimality comes not only from recurring static
generalizations about the size of possible words (i.e., roots or stems) in diverse languages.
Minimality requirements also motivate insertions (or block deletions) to allow words to achieve (or
maintain) a minimal size. For example, nasal fusion a process which accompanies nasal
prefixation in a number of Indonesian languages – is blocked before monosyllabic bases in
Javanese. As Uhrbach (1987) shows, the output of nasal affixation – a stem-forming process – (and
fusion) must be minimally disyllabic, even though monomorphemic roots can be monosyllabic:
(20) Nasal fusion in Javanese (from Uhrbach 1987: 233, fig (11))
a. Polysyllabic bases
‘shave someone’
‘return something’
‘to write’
b. Monosyllabic bases
ŋәcet (*ɲet)
‘(to) print’
ŋәbom (*mbom)
‘(to) bomb’
Minimality requirements also play a central role in constraining the size of prosodic morphological
constructions like reduplications and truncations. Recent work like McCarthy (2000), Downing
(2006) and Urbanczyk (1996) proposes that minimal size requirements on prosodic morphemes
should fall out from general minimality constraints on stems, roots and affixes. If one has
independent evidence for categorizing a reduplicative morpheme or a truncation as a particular
morpheme type derived stem, root or affix – then its minimal size falls out from that
categorization. For example, Itô (1990) provides a detailed study of loanword truncations in
Japanese. One striking result of the study is that free truncations (that is, ones that can be words) are
minimally disyllabic. Examples are given in (21); the portion of the full word omitted in the
truncation is in parentheses:
(21) Japanese disyllabic loanword truncations (Itô 1990: 219)
As Itô (1990: 218) points out, it is unexpected for truncations to be subject to a disyllabic
minimality condition, as many common lexical items in the native Japanese vocabulary are
monomoraic: e.g., su ‘vinegar’, na ‘name’, no ‘field’, te ‘hand’. To account for the minimality
condition on free truncations, Itô (1990: 227) proposes there is a disyllabic minimal word
requirement, holding only of derived words. That is, the disyllabicity requirement falls out from the
fact that truncations are derived stems, which iconically are required to minimally branch into two
To provide an example from reduplication, Urbanczyk (1996) shows that Lushootseed, a
Salishan language spoken along the Northwest Coast of North America, has a distributive
reduplicative morpheme with the canonical form, CVC:
(22) Lushootseed distributive reduplication (Urbanczyk 2000, fig (24))
‘fly here and there’
‘all tangled up’
‘seeking a woman to marry’
‘many white folks’
As Urbanczyk (1996) argues, the CVC shape of the distributive reduplicative morpheme is best
accounted for if it is categorized as a root, as the canonical form of root morphemes in Lushootseed
is CVC: according to Urbanczyk, 68% of all roots in Lushootseed are CVC. Further, these
reduplicative forms attract stress, which usually targets root-initial position. In Lushootseed, then,
the minimal CVC form of roots and root-like reduplicative morphemes follows from the iconic
requirement that roots, as semantically complex morphemes, should be prosodically branching.
While this approach to minimality, which motivates prosodic branching as iconically reflecting
semantic complexity and/or morphological complexity, accounts for a strong recurring tendency, as
Downing (2006) notes, it is not always easy to provide independent evidence that a particular
reduplicative morpheme is best categorized as a stem, root or affix. Since this approach crucially
relies on the matching morpheme type to prosodic minimality, it does not straightforwardly account
for cases where evidence for morpheme type is lacking.
3.3 Differential object/subject marking
One phenomenon that has been taken as an instance of Iconicity of complexity and that also has
received some attention not only in the functional but also in the formal linguistic literature is that
of linking splits that fall under so-called differential object or differential subject marking,
respectively. (See, for instance, Silverstein 1976, Bossong 1985, Comrie 1989.) The observation is
that NPs or DPs may receive overt case marking in subject or object position depending on their
sortal and referential properties. NPs/DPs that denote highly referential or sortally high-ranked (e.g.
human or animate) entities tend to receive overt case marking in object position (accusative), but
tend to remain caseless in subject position. Conversely, NPs/DPs that denote unspecific or
indefinite or sortally low-ranked entities tend to receive overt case marking in subject position
(ergative), but remain caseless in object position. This has been described as “markedness reversal”.
The role-specific disposition for certain referential and sortal properties results from general cross-
linguistic lexicalization patterns. Situations are lexicalized from the perspective of the
causing/agentive entity. Languages have basic verbs such as hit, but no basic verbs meaning ‘be
hit’. Therefore, animate/human and highly individuated referents tend to occur in subject position.
The following scale, which summarizes findings from Silverstein (1976), describes the overall
(23) NP/DP scale
1 > 2 > 3/DEM > proper nouns > human > animate > inanimate nouns
agent patient
accusative ergative
This scale with a few exceptions (see e.g. Filimonova 2005) can be understood as an
implicational scale. If a category of a certain point of this scale is marked by accusative, then all
other categories higher on the scale will also receive accusative marking in the respective language.
Likewise in the other direction: if a certain category instantiates ergative, then all other categories
lower on the scale will do so as well.
Generally, the more expected distribution of an NP/DP remains unmarked, whereas the less
expected distribution is overtly marked. The observed case marking pattern can receive a double
interpretation as an instance of iconicity or as an instance of economy. Aissen (2003) claims that
iconicity is involved: the more marked a direct object qua object is, the more likely it is to be
overtly case-marked. Economy, then, would be the conflicting tendency to avoid case marking.
Note that moving along the scale could only mean semantic enrichment in one direction: moving
down the scale is usually seen as a negative specification of the features [definite], [human],
[animate]. Therefore, a different understanding of Iconicity of complexity is involved in differential
subject/object marking. It is the marked context that receives a more complex marking. The specific
implementation of differential subject and object marking varies cross-linguistically (see data in
Dixon 1994 and formal accounts by Aissen 1999, 2003, Kiparsky 2001, Stiebels 2000, 2002).
Aissen assumes the preference scales in (24), which in turn are derived by Harmonic Alignment
(Prince & Smolensky 1993) from the relational scale in (25a) and the definiteness/specificity scale
in (25b). These scales are empirically well-founded and play an important role in various
(25) a. subject (SU) > object (OJ)
b. Pronoun (PRO) > proper name (PN) > definite (DEF) > specific (SPEC) > nonspecific
The preference scales in (24) can be reinterpreted as a ranking of markedness constraints:
(26) a. *SU/NSPEC » *SU/SPEC » *SU/DEF » *SU/PN » *SU/PRO
Via local conjunction of *C, which penalizes caseless NPs/DPs, with the rankings in (26), Aissen
derives the following constraint hierarchies, which are meant to account for the category-specific
preferences for overt case marking:10
(27) a. *SU/NSPEC & *C » *SU/SPEC & *C » *SU/DEF & *C » *SU/PN & *C » *SU/PRO &
b. *OJ/PRO & *C » *OJ/PN & *C » *OJ/DEF & *C » *OJ/SPEC & *C » *OJ/NSPEC &
Aissen (2003) identifies the local conjunctions in (27) as iconicity constraints (favouring
morphological marking for marked configurations). The ranking in (27a) expresses that caseless
nonspecific subjects are worse than caseless specific subjects, which in turn are worse than caseless
definite subjects, etc. For objects, the reverse pattern holds: i.e. caseless pronominal objects are
worse than caseless proper names, etc. Deriving the markedness hierarchies by means of Harmonic
Alignment allows Aissen to avoid arbitrary constraint rankings, which are in principle possible in
optimality-theoretic approaches. The prediction is that no constraint ranking should emerge in
which, for instance, *OJ/DEF is ranked higher than *OJ/PRO.
Note that Aissen’s analysis implements the assumption of an underlyingly reduced case
inventory. Overt case (e.g. accusative or ergative) is enforced by category-specific local
conjunctions. In contrast, Stiebels (2000, 2002) assumes an underlyingly rich case inventory. In her
account, case is blocked by contextualized markedness constraints that exclude case for a category
that shows a strong disposition for a certain argument role (i.e. subject or object). Despite these
conceptual differences (formal markedness in the marked context vs. economy in the unmarked
context), both these approaches, along with Kiparsky (2001), are able to derive the occurring case
distributions.11 Differential object marking has also been analyzed within Bidirectional Optimality
Theory (Zeevat & Jäger 2002) and in an evolutionary game-theoretic approach (Jäger 2007).
Haspelmath (2008:14) argues that the technical machinery used by Aissen is not necessary
because frequency distribution already explains the fact that the rarer elements tend to be overtly
coded. However, this critique misses a crucial point of Aissen’s account and other formal analyses,
namely: these accounts try not only to describe a strong tendency in the world’s languages but also
to derive the specific case marking of a given language. There is no evidence so far that the
language-specific case patterns correlate with language-specific frequency distributions of the
respective categories, i.e. that languages with a fully generalized accusative would show an equal
distribution of animate/definite and inanimate /indefinite NPs in object position.
3.4 The lexical nature of markedness
Iconicity of complexity must be relativised not only with respect to morpheme type (affix vs. root
vs. derived form) but also with respect to lexical item.12 Quite a number of categories display
lexeme-specific markedness values, e.g. number marking (e.g. Wierzbicka 1985, Acquaviva 2008),
aspect marking (e.g. Krämer & Wunderlich 1999 for Yucatec Maya), causative/anti-causative
10 Confusingly, local conjunction is evaluated as a logical disjunction, penalizing candidates that fail both halves of the
conjunction. (See Downing 1998, Stiebels 2002 for more discussion of how constraint conjunctions are evaluated.) For this
reason, a local conjunction defines the ‘worst of the worst’ (i.e., the most marked).
11 However, the approaches differ in their predictions for the kind of case pattern that a language with a 3-way split should exhibit:
ACC/NOM NOM ERG/NOM (Aissen/Kiparsky) vs. ACC/NOM ERG/ACC/NOM ERG/NOM (Stiebels). The former pattern is not
attested, whereas the latter is.
12 See also the discussion in Haspelmath (2008), who, however, discusses some of the same phenomena under the rubric of relative
vs. absolute frequency of the respective markers.
marking (Haspelmath 1993), and marking of possessor extension/reduction (see section 6.1).13 As
Haspelmath (1993) shows with data from a number of languages, verbs that denote changes that
need not be externally caused may show an unmarked inchoative form (e.g. ‘freeze’, ‘dry’, ‘sink’,
‘melt’), while the causative variant is overtly derived. Predicates that often require an external cause
to instigate the change show an unmarked causative (e.g. ‘split‘, ‘break’, ‘close’, ‘open’, ‘gather’),
while the inchoative/decausative variant is overtly derived (e.g. German sich öffnen ‘open (itr.,
The lexeme-specific preference for a certain value of a grammatical category is exploited in so-
called “polarity” marking (see Wunderlich, this volume). This notion goes back to Meinhof (1912)
and refers to cases in which the relation between two signifiers/exponents A and B and two
signified elements X and Y is such that A represents X in some contexts and Y in the others,
whereas B is the signifier for the complementary contexts. A phenomenon that illustrates this
property nicely is so-called ”inverse number marking”. Here a polyfunctional number marker turns
a “basic” singular into a plural and a “basic” plural into a singular. Anttila & Bodomo (2009) and
Grimm (2009) have analyzed inverse number marking in Dagare (see Wunderlich, this volume, for
examples); Steins (2000) and Harbour (2007) have analyzed the more complex system of Kiowa
described in Watkins (1984).
Languages may also exhibit a polarity system in the area of voice/diathesis morphology.
Wunderlich (this volume) cites the causativizing/detransitivizing marker -e in Japanese (e.g. tat-u
‘stand’ vs. tat-e-ru ‘raise’, or-u ‘break’ (tr.) vs. or-e-ru ‘break’ (intr.)). In Olutec, the marker yak-
encodes both passive and causative. It functions as causative with intransitive verbs (see (28a)) and
as passive with transitive verbs (see (28b)). Hence, it shifts the valency feature [transitive] to the
opposite value, similar to the Japanese case.
(28) Olutec (Zavala 2006:290)
‘that is when I killed that little parrot’
‘but where is land going to be given away?’
In semantic terms, the shift of the [transitive] feature is not easily modelled. The causative is
usually represented as the addition of a causative relation, which introduces a new highest argument
(e.g. (29a)), whereas passive binds the highest argument (e.g. (29b)).
(29) a. λx λs DIE(x)(s) λx λu λs’ s CAUSE(u, DIE(x)(s))(s’)
b. λy λx λs OFFER(x,y)(s) λy λs x OFFER(x,y)(s)
It is far from obvious how this could be expressed by means of an underspecified semantic entry.
The Japanese data are easier to account for. Generally, this kind of polyfunctionality is cross-
linguistically rare.
3.5 Modelling iconicity of complexity
With Iconicity of complexity taken for granted as an unmarked setting for the relation between
phonological and morphological structure, alternative theoretical analyses may thus be evaluated
according to the degree to which they model this principle. As noted above, analyses of
comparatives, for instance, could ignore that the comparative form is more complex than the
13 Note that “anti-causative” and “decausative” can be used interchangeably.
positive or that there is generally one marker for deriving comparatives of positive and negative
antonyms (big – bigger, small smaller). That a certain analysis gains support by considering the
formal marking, is a way to integrate aspects of iconicity into theoretical linguistics. This would
mean, for instance, that one takes the fact that some languages overtly derive the superlative from
the comparative (e.g. Hungarian or Spanish) as evidence for a semantic analysis of the superlative
that enriches or further specifies the representation of the comparative.
A formal constraint that in some sense implements effects of Iconicity of complexity is the
monotonicity requirement: i.e. no information should be deleted/overridden by morphosyntactic
operations that create complex structures (e.g. Sag et al. 1986, Wunderlich 1996, Bresnan 2001 and
previous work in LFG). Koontz-Garboden (2008, 2009) formulates the following Monotonicity
(30) Monotonicity Hypothesis (MH)
Word formation operations do not remove operators from lexical semantic representations
(= conceptual structures).
This monotonicity assumption has consequences for lexical semantics: a morphologically complex
form should include the semantic representation of its base in its own representation. The major
challenge to the monotonicity assumption discussed in the literature is decausative verbs, which
have already been mentioned above (see Koontz-Garboden 2008, 2009, Haspelmath 2008). These
verbs represent inchoative verbs derived from causative verbs, illustrated by the following examples
taken from Koontz-Garboden:
(31) a. Pima
hain ‘break (tr.)’ 'e-hain ‘break (itr.)’
b. Quechua
paska ‘open (tr.)’ paska-ri ‘open (itr.)’
According to Koontz-Garboden (2008), these data suggest the following derivation, in which the
causative meaning component is deleted by decausativization markers:
(32) λx λy λs λe [v [CAUSE(v,e) EFFECTOR(v,y) BECOME(e,s) THEME(s,x) φ(s)]]
λx λs λe [BECOME(e,s) THEME(s,x) φ(s)]
Various authors have assumed an operation along these lines (e.g. Reinhart 2002, Härtl 2003,
Kaufmann 2007). In contrast, Koontz-Garboden assumes that the causative meaning component is
retained in the decausative verbs, referring to certain data in Italian (modifier da se ‘by itself’) and
Spanish (negation of decausative verbs).14 Thus, no violation of monotonicity would occur.
Another influence of Iconicity of complexity can be seen in accounts that discuss the choice for
features and feature values. In Minimalist Morphology, for instance, features and feature values are
chosen such that marked forms are represented by features which bear a plus value in this context
(Wunderlich & Fabri 1995, Wunderlich 1996). Thus, a feature [plural] is justified, with [+plural]
referring to forms such as cat-s, children, ox-en, as well as [past], with [+past] referring to laugh-ed
or went, because [-plural] and [-past] do not receive a special marking. In such a framework,
positive values are taken as informative, whereas negative values represent information inferred as
defaults. In this view, number and tense would normally not be specified in terms of features
14 Note that Koontz-Garboden’s tests crucially rely on a contrast of the decausative verbs with their causative counterparts (e.g. it
is not the case that x broke y but y broke by itself). The admissibility of such contrastive readings, which are not possible with
statives or underived inchoatives, does not imply that the causative component found in the transitive counterpart is retained.
That underived inchoatives behave differently from decausative verbs may be due to the fact that the former often do not refer to
changes that may be externally caused.
[present] or [singular], unless strong morphosyntactic evidence suggests the assumption of such a
4 Isomorphism
Though not a central focus of many linguists dealing with iconicity, isomorphism has been treated
as one aspect of iconicity. Bolinger (1977) characterizes isomorphism as “the natural condition of
language […] to preserve one form for one meaning, and one meaning for one form” (ibid., x).
Haiman (1985a) splits the concept of iconicity into isomorphism and motivation. Haiman defines
isomorphism as Bolinger does, as a one-to-one correspondence in form-meaning pairs, whereas
motivation concerns iconicity in complex expressions. It is obvious that languages implement
isomorphy to a greater or a lesser degree, but so far, no language has been proven to be
strongly/completely isomorphic in all respects.
There are various potential challenges to the strict one-to-one correspondence between form and
meaning. For example, the following common morphological phenomena appear to violate
isomorphism: polysemy, synonymy, homonymy, meaningless morphs (e.g. thematic vowels) and
zero forms (e.g. third person singular forms in verbal paradigms of many languages). In evaluating
the various potential violations of isomorphism, Haiman (1985a) points out that homonymy is
incontestable, whereas the existence of total synonyms (i.e. complete interchangeability of forms in
all contexts and registers) can be questioned. Both homonymy and synonymy can have a strong
impact on word formation and the morphosyntax of a given language. A language with a high
degree of homonymy may try to reduce the number of homonyms by extending the forms with
elements that specify that homonym semantically, leading to increased formal complexity. Haiman
(1985a) cites data from Mandarin Chinese, but also English examples, such as funny-haha and
funny-queer. Potential synonymy has a very systematic effect in that either one form is blocked
(German Dieb ‘robber’ blocks *Stehl-er, the agent nominal derived from stehlen ‘steal’; see van
Oostendorp, this volume, for more discussion of blocking effects), or that the two competing
expressions are differentiated in semantic terms or in their socio-linguistic value (i.e., register).
Some of the iconicity effects result from these contrasts. (See section 6.2, below.)
In the domain of inflectional paradigms, isomorphism is a crucial assumption in explaining both
paradigm uniformity (or analogy) and syncretisms. Paradigm uniformity (that is, the isomorphic
constraint that all realizations of the same morpheme should have the same form) is appealed to in
analyses of sound systems to explain surface exceptions to a regular sound change or a regular
synchronic phonological process.15 For example, in North American English, the ‘t’ in capitalistic
is pronounced with a flap, while the ‘t’ in militaristic is aspirated. This difference is unexpected, as
the two words have, on the surface, the same number of syllables and the same stress pattern.
Paradigm uniformity provides an explanation for this difference, as Davis (2005) shows. The flap of
capitalistic is preserved from the base form, capital; similarly, the aspiration of militaristic is
preserved from military. That is, the pronunciation of ‘t’ is uniform for the paradigm of words
derivationally related to each base. (See Albright & Fuss, this volume, for further discussion of
issues related to paradigmatic effects.)
Syncretisms (i.e., overlaps in the form of morphemes with distinct functions), in contrast, are a
type of homonymy and therefore are potential exceptions to isomorphism. Their impact on a
paradigm, however, depends on the nature of the syncretism. Systematic syncretism (to be derived
15 See the introductions to Lahiri (2003) and Downing et al. (2005) for overviews of this topic and several recent papers motivating
the role of analogy and paradigm uniformity in both the diachronic and synchronic sound systems of a variety of languages.
by underspecified entries) is a less crucial violation of isomorphism than unsystematic syncretism,
which generally involves either disjunctive specifications of the item or two or more distinct entries,
both of which indicate homonymy.16 Müller’s (2004) analysis of syncretism in Icelandic postulates
the following principle (which might be more aptly named, the ‘anti-Syncretism Principle’):
(33) Syncretism Principle (Müller 2004: 236)
Identity of form implies identity of function (with a certain domain, and unless there is
evidence to the contrary).
As a consequence, syncretisms are by default regarded as analyzable in terms of underspecification
or similar devices. According to Müller, (33) is an instance of a more general iconicity requirement;
we can see that it restates the definition of isomorphism given above:
(34) Iconicity Principle (Müller 2004: 238)
Similarity of form implies similarity of function (within a certain domain and unless there is
evidence to the contrary).
It follows from (33) and (34) that the number of homonymous forms should be reduced in the
analysis of a given inflectional system.17
Whereas homonymy is of an accidental nature and does not show cross-linguistically recurrent
and systematic patterns, polysemy generally does (see, for instance, Bierwisch 1983 and the papers
in Pustejovsky & Boguraev 1996, Ravin & Leacock 2000). There are varying degrees of
systematicity in polysemy. Besides the systematic cases of “constructional polysemy” (e.g. enjoy a
book = ‘enjoy reading the book’) and “sense extension” (e.g. grinding processes such as the derived
mass readings of lamb or chicken) discussed in work like Copestake & Briscoe (1996)), there are
also patterns that occur in quite a number of languages, but not with the same generality and
frequency as the above-mentioned phenomena. Haiman (1985a) discusses two such cases: the
cross-linguistically frequent similarity of conditional protases and polar questions, and the frequent
similarity of conditional protases and topic marking. Another common case comes from the area of
voice/diathesis morphology (see Wunderlich 1993a, Dixon & Aikhenvald 2000 for a general
overview). Whereas many languages differentiate between various argument-extending operations
(e.g. causative and applicative) and various argument-reducing operations (e.g. passive, antipassive,
reflexivization), some languages display polyfunctional markers, covering several distinct
operations of argument extension or argument reduction. An example of this is a generalized
detransitivizer in languages in which the marker functions as both passive and antipassive
depending on the verb and the syntactic context (e.g. Cooreman’s (1994) analysis of Warrungu and
Diyari). A generalized detransitivizer may also subsume passives, reflexives, and anti-causatives, as
the following examples from Amharic illustrate. The potential readings depend on the lexical
semantics of the verbal base or on the sortal properties of the argument’s referents.18
16 The question of whether a specific instance of syncretism is systematic in Müller’s sense or not is often dependent on the use
and interpretation of the respective morphosyntactic features and their values.
17 Müller (2004) also proposes that Icelandic noun declension exhibits an iconic mapping such that more specific or more limited
functions are realized by markers that contain less sonorous segments. The phonological side of this iconic mapping relation is
less convincing: first, all of the markers discussed by Müller are unmarked in the sense that they are among the most common
sounds (/i, u, a, s/) of the world’s languages. Second, the sonority hierarchy is not a general phonological markedness hierarchy.
Taking epenthesis as indicator for low markedness one has to acknowledge that the high vowels /i, u/ are very common
epenthetic segments; and /s/ is the least marked continuant consonant.
18 Middles (e.g. in Ancient Greek and Fula) also usually cover a set of distinct (mostly detransitivizing) usages; see Wunderlich
(2001) and Kaufmann (2004, 2007) for analyses and underspecified representations of middles.
(35) Amharic (Amberber 2000:325)
‘Aster cut the rope’
‘the rope was cut by Aster’
i) ‘the bottle was broken (by the boy)’
ii) ‘the bottle broke’
‘Aster washed herself’
A polyfunctional argument extension marker may, for instance, subsume causative and applicative,
as illustrated in the following for Kinyarwanda -IIsh, which encodes both an instrumental
applicative, as in (36a), and a causative, as in (36b). Tuggy (1988) reports a similar polyfunctional
causative/applicative marker for Tetelcingo Nahuatl.
(36) Kinyarwanda (Kimenyi 2004:1535)
‘the boy is cutting a tree with an ax’
‘the man is making the boy cut the tree’
The challenge of polyfunctional markers lies in their representation: can a single representation
cover all readings? Are multiple entries necessary? The latter implies that one is actually dealing
with homonymy. The Kinyarwanda marker -IIsh may be specified as REL(u,s), denoting a general
relation of an argument u to the situation s denoted by the base verb. This relation may be added as
first conjunct ([REL(u,s) & VERB(x,y)(s)]) or second conjunct ([VERB(x,y)(s) & REL(u,s)]), which
then, by further elaboration at the conceptual level, receives a reading as causative or applicative.
Other polyfunctional markers may pose greater analytical challenges, however.
Meaningless morphs usually have prosodic functions or are relics from once meaningful
morphs. Their existence can usually not be explained in purely morphosyntactic terms. The
postulation of zero forms (‘Ø’) is generally theory-dependent. There are approaches like Minimalist
Morphology that try to avoid zero morphemes more or less completely (Wunderlich & Fabri 1995,
Wunderlich 1996), and others that would postulate a zero morpheme for each non-overt category.
Non-overt categories do not constitute a homogeneous class: a zero form in a paradigm usually
receives its meaning due to paradigmatic contrast with other forms. Therefore, one form can remain
covert due to economy. (See Trommer, this volume, for detailed discussion of issues related to zero
exponence.) A different case is represented by conversion, i.e. covert category shift. Depending on
the category shift, different degrees of semantic enrichment are involved. Regular event
nominalization does not include semantic enrichment (only a category shift; see, for instance,
Bierwisch 1989a), whereas denominal verbs are enriched by an event argument and meaning
components such as CAUSE or BECOME (see Kiparsky 1997). Cross-linguistically, such covert shifts
are not unusual: note the high number of denominal verbs in English or the categorially unspecified
roots in Polynesian languages such as Tongan (see Broschart 1997).19
Besides understanding isomorphism as a requirement that the form and meaning of lexical items
stand in one-to-one correspondence, one could also think of isomorphism as a one-to-one
correspondence between various levels of linguistic representation. As Hyman (2003) suggests, in
an “ideal language”, morphological complexes would converge in terms of semantics
(compositionality), syntax (affix order mirroring syntactic derivations), morphological layering
(sequencing correlating with order of affixation) and phonological cyclicity. Deviations from
isomorphism between two or more levels are usually analyzed as bracketing paradoxes. Egg (2006)
explicitly declares structures with bracketing paradoxes to be non-iconic. In the general discussion
of bracketing paradoxes, its potentially non-iconic character is seldom discussed. The question of
whether a structure displays a bracketing paradox is in many cases theory-dependent. In
frameworks such as Autolexical Syntax (Sadock 1991), the levels of representation are seen as
parallel and autonomous, as a way to avoid the problems that bracketing paradoxes may pose for
linguistic representation. In general, much more research is needed on the potential nature of
bracketing paradoxes and their role in linguistic change (elimination or creation of paradoxes) and
their role in language processing and acquisition. If no recurrent patterns and no strong
psycholinguistic effects can be found, isomorphism between levels of representation might be
considered more relevant for evaluating the conceptual elegance of linguistic frameworks (e.g. the
restrictiveness of the approaches) than for the organization of grammars.
As we will show in the following two sections, both sub-concepts of isomorphism are relevant
in analyses of the iconic nature of linear ordering and structural cohesion. Theoretical accounts of
linear ordering rely on isomorphism between syntactic or semantic representations and PF or
surface representations: that is, isomorphism between levels of representation. This is discussed in
section 5. Isomorphism is also established when a division of labour emerges such that the more
cohesive form is iconically aligned to the tighter semantic relation, whereas the less cohesive form
is aligned to a less tight semantic relation This is discussed in section 6.
5 Iconicity of linearization phenomena
There is a debate both in descriptive and in theoretical linguistics about whether morpheme orders
are arbitrary or follow underlying syntactic and/or semantic principles/constraints. Many analysts
propose that, in the unmarked case, morpheme order is not arbitrary. Instead, it shows iconic
isomorphism. As we show in this section, the debate about the role of isomorphism in defining
unmarked morpheme order has both an empirical/descriptive and a theoretical aspect. In descriptive
terms, the question arises as to which morphemes or morpheme combinations tend to show
unexpected sequencing and which other grammatical constraints could be responsible for this. In
theoretical terms, three questions emerge: a) which grammatical constraints in grammar lead to
iconic morpheme orders, b) which other factors besides the proposed grammatical constraints on
morpheme ordering contribute to iconic outcomes of morpheme sequences, and c) should
theoretical/formal accounts of morphological sequencing be modelled more in view of arbitrary
morpheme orders in a subset of the world’s languages or in a subset of a language’s morphology or
in view of the much more frequent canonical cases.
19 Zero forms are usually postulated for categories that quite regularly receive overt marking in natural languages. The
isomorphism debate does not really focus on the lack of overt marking of semantic operations that are necessary in semantic
composition (e.g. type shifts).
5.1 Extra-grammatical constraints on morpheme ordering
We begin by looking at principles of morpheme ordering in compounds, which in many cases do
not exhibit the sort of scopal or functor-argument relation between the respective stems/roots which
we show in the next section plays a role in ordering other types of morphemes. With respect to
verbal (V-V) compounds or verbal chaining (e.g. in Papuan languages, see Foley 1986), event-
related ordering constraints seem to play a crucial role. The order of the verb stems in the various
types of V-V compounds is determined by temporal sequence. If the subevents denoted by the verbs
involved are not contemporaneous but sequential, the verbs are ordered according to the temporal
sequence of the subevents denoted by the verbs. (Note the similarity to the veni, vidi, vici example
cited in the Introduction.) This is true both for head-initial as well as head-final verbal compounds.
For instance, in Japanese (e.g. Matsumoto 1996, Gamerschlag 2005), verbal compounds are head-
final. There are two types of verbal compounds with a sequential event structure: cause compounds
(e.g. suberi-otiru = slip-fall ‘slip’, ni-tumaru = boil-become.packed ‘preserve’) and means
compounds (e.g. osi-taosu = push-knock.over ‘knock over’, kami-tubusu = bite-crush ‘chew,
crunch’). Sequences of effect+cause (e.g. *tumari-niru ‘become.packed-boil’) or
consequence+instigation (e.g. *nugi-suteru ‘’) are ruled out. Gamerschlag
assumes this follows from a corresponding iconicity constraint, which is ranked higher than other
linearization constraints. The role of temporal sequencing constraints has also been pointed out for
serial verb constructions (see e.g. Tai 1985 for Chinese).
With respect to N-N compounds, sequencing constraints that derive from a given structure in the
external world, as in the case of V-V compounds, do not seem to exist. Dvandva compounds/co-
compounds would be relevant candidates for the study of iconic ordering constraints because no N-
stem inherently is better qualified as head noun in this construction. Wälchli (2005) has dealt with
the potential iconic nature of co-compounds, but he has mainly focused on the fact that co-
compounds tend to refer to semantically natural and tight coordinations, as shown in the Georgian
examples in (37), below:
(37) Georgian co-compounds (Wälchli 2005:11f.)
a. da-dzma
b. ts’ol-k’mar.i
c. mt’a-bar.i
d. mšvild-isar.i
Insights into sequencing come from studies on binomials or freezes, which, however, have a
different syntactic and prosodic structure from co-compounds (Müller 1997). Research on
binomials (freezes) like Cooper & Ross (1975), Birdsong (1995), Landsberg (1995) and Müller
(1997) – has revealed that culturally and cognitively determined salience relations, along with
prosodic constraints, determine the order of the elements. Generally, the more salient element
precedes the element perceived as less salient. Common salience rankings refer to qualities such as
animate vs. inanimate (man-machine), human vs. animate (man or beast), adult vs. adolescent
(father and son), male vs. female (brother and sister), positive vs. negative (pro and con),
proximate vs. distal (here and there), to name the most important factors.20 The key concept is some
20 The Georgian examples in (37a/b) do not follow the frequently observed ordering male before female entities.
kind of egocentricity in evaluating salience, with entities closer to “ego” being more salient and
realized first. With kinship terms, it may reflect former/current hierarchical relations between
family members in that society, with the noun referring to the higher-ranked person realized first.
Nevertheless, in order to prove that the assumed salience relations do in fact play a role,
experiments have to be carried out that verify the salience factor in the context of nonverbal tasks.
Without such evidence, the assumption of a salience-conditioned ordering of N-stems in co-
compounds is no more than a reification.
5.2 Grammatical constraints on morpheme orders
A variety of frameworks propose that morpheme order has an iconic basis, which reflects
independent scopal or relevance relations holding among the morphemes. We briefly review three
leading proposals here.
Bybee (1985a/b) and Givón (1985) both use the notion of “relevance” to account for overall
tendencies in verbal morpheme orders. Morphemes with higher relevance for the root are realized
closer to it or more often undergo fusion with the verb stem. Givón states this principle as follows:
(38) “The more relevant the operator is to the operand, and the more specific and exclusive it is
to the operand, the closer to the operand it will be placed.” [Givón 1985:208]
Bybee has undertaken an extensive cross-linguistic study on morpheme ordering. Her approach is
meant to account for the observation that derivational affixes strongly tend to appear closer to the
stem than inflectional affixes. Bybee argues that these affixes show a higher relevance for the stem
because they have a stronger semantic effect, changing, for instance, a stem’s lexical meaning, its
Aktionsart or the number of its arguments (diathesis/valence markers).21
Baker (1985, 1988) derives morpheme order from syntactic principles. In this theory, morpheme
order is determined mainly by syntactic derivations in accordance with the Mirror Principle:
(39) Mirror Principle (Baker 1985:375)
Morphological derivations must directly reflect syntactic derivations (and vice versa).
Generally, Baker assumes head movement of the respective morphemes, which are base-generated
as distinct syntactic heads. The possible output is constrained by restrictions on head movement
(cyclicity) and other well-formedness constraints (e.g. the Stray affix filter or the case filter). For
instance, the combination of causative and applicative in the order V-APPL-CAUSE is excluded
according to Baker (1988) due to violations of the case filter.
Approaches that follow Baker in a syntactic derivation of morpheme orders are legion (e.g. in
frameworks such as the Principles and Parameters approach or Minimalist Syntax or Distributed
21 Bybee also explains preferred morpheme orders within inflectional categories of the verb (relative distance to verbal stem:
aspect < tense < mood) by her notion of relevance. Wunderlich (1993b), interpreting Bybee’s findings more explicitly,
formalizes the observed morpheme order tendencies in terms of a hierarchy of functional categories, which mirrors the
sequencing of inflectional affixes:
(i) Hierarchy of functional categories (Wunderlich 1993b:63)
According to Wunderlich, (i) merges two hierarchies: a TMA-system (aspect (ASP), tense (TNS), MOOD) and a participant-
structure system (voice/diathesis (DIA), gender (GEN), number (NUM), person (PER), CASE). The TMA-system relates to the
referential arguments of verbs, whereas GEN, NUM and PER relate to the referential argument of nouns. As Wunderlich points out,
the hierarchy predicts not only possible affix orders but also possible fusions of categories: adjacent categories may fuse into
one morpheme. Furthermore, if a certain category is lexically specified (e.g. via root inflection or suppletion), lower-ranked
categories can only receive a default value. The observed tendencies in morpheme orders are not without exceptions, though.
(See, for instance, Trommer 2003 for a study of agreement morphology and tense/aspect.)
Morphology). We refer the reader to Cinque (1999) and Julien (2002) as exemplary cases. His
approach is not uncontroversial, however. It runs counter to the lexicalist hypothesis (e.g. Chomsky
1970, Di Sciullo & Williams 1987, Bresnan & Mchombo 1995) in assuming that word structures
are accessible to syntax. One also finds numerous counterexamples to the orderings motivated by
Baker’s syntactic principles. Though rare, the V-APPL-CAUSE order, which, as noted above, is
excluded by Baker, is still attested (Stiebels 2003). Likewise, one finds instances of structures
combining antipassive and applicative, which are ruled out by Baker as a violation of the cyclicity
constraint. We return to the combination of causative and applicative below.
Muysken (1986) interprets the Mirror Principle in terms of semantic scope. If an affix A has
scope over affix B, it must be external with respect to B, as illustrated in the following:
(40) a. Affix order: V-AFF1-AFF2-... vs. V-AFF2-AFF1-...
b. Semantic scope: AFF2(AFF1(V)) vs. AFF1(AFF2(V))
However, not all morpheme combinations invoke scopal restrictions. Rice (2000) distinguishes
three types of morpheme combination. Firstly, two morphemes A and B may lack any kind of
scopal relation. Therefore, no morpheme order concerning A and B is preferred. Both morpheme
orders may be possible, or a language may arbitrarily choose one option. Secondly, each of the two
morphemes may take the other one into its scope. Therefore, both morpheme orders are relevant
because they differ in their scopal interpretations. Thirdly, the scope relation is fixed such that only
morpheme A may take morpheme B into its scope. As a result, only the order where A is the outer
morpheme is possible. According to Rice, the first two cases are instances of local variability: there
may be language-internal or cross-linguistic variation regarding the actual morpheme orders. In
contrast, the third case is predicted to show global uniformity: i.e., all languages should display the
relevant morpheme order.
The first case of lack of scopal relations discussed by Rice could be a test case for additional
iconicity constraints that are independent of scopal restrictions. This point was taken up in section
5.1, above. The second case is illustrated by the following example from Bolivian Quechua, in
which the combination of the assistive and the hortative transparently exhibits two morpheme
orders with respective interpretational differences. The assistive (ASS) adds an assister argument to
the base verb, which is realized as subject. The hortative intensifier (HORT) expresses that the action
denoted by the verb is executed with a certain amount of energy.
(41) Quechua: Assistive/hortative (van de Kerke 1996:198)
‘she helped me wash the clothes energetically’
‘she helped me energetically wash the clothes’
The morpheme order in (41a) has the expected interpretation that the assisting action is executed
energetically, whereas the order in (41b) denotes the situation of energetic washing.
Deviations from the expected morpheme orders in multi-scopal settings occur in two guises. In
one type of case, language-specific constraints may restrict the combination of the morphemes to
one possible order, which receives a surface-true, i.e. compositional interpretation. Bolivian
Quechua, for instance, requires the repetitive affix -kipa ‘again’ (REP) to be internal to the causative
affix (CAUS), ruling out the inverse order. The interpretation is compositionally fixed to express the
repetition of the situation expressed by the base verb.
(42) Quechua: Causative/repetitive (van de Kerke 1996:176)
‘my mother made me rewash the clothes’
#‘again my mother made me wash the clothes’
However, one can also find cases in which a given morpheme order has both the compositional and
the non-compositional interpretation. The latter violates the Mirror Principle. Stiebels (2003)
classifies these morpheme orders as opaque. Whereas the restricted morpheme order in (42) implies
a complete gap for a certain morpheme combination, opaque affix orders lack a distinct PF for one
of the two readings. The combination of the hortative and the causative in Quechua provides an
example of opaque affix order: the surface order HORT-CAUSE has the additional non-compositional
interpretation that the causing event is executed energetically.
(43) Quechua: Hortative/causative (van de Kerke 1996:177)
a. ‘you should make Maria wash the clothes with energy’
b. ‘you must energetically make Maria wash the clothes’
Stiebels schematizes the possible settings of multi-scopal morpheme combinations as follows:
(44) Schema of attested affix orders in multi-scopal contexts
a. transparent b. restricted
c. opaque1 d. opaque2
As (44c/d) show, there are two potential cases of opacity: (44c) is polysemous with a transparent
and a scope-violating interpretation, whereas (44d) only has a reading that contradicts its morpheme
sequencing. This would be the worst outcome of morpheme sequencing for the Mirror Principle.
Given the fact that various authors have acknowledged the existence of morpheme orders that
violate Baker’s (1985, 1988) syntactic Mirror Principle (e.g. van de Kerke 1996, Hyman &
Mchombo 1992, Hyman 2003, Stiebels 2003), the question arises as to what the nature and the
causes of these deviations from what remains a very useful markedness constraint might be. One
can speculate that diachronic processes might play an important role. Nevertheless, other factors
can also contribute to opaque outcomes in morpheme orders. We would like to briefly discuss one
recurrent pattern in the combination of diathesis markers. Whereas the combination of diathesis
markers that introduce a new subject shows clear transparent structures, opaque patterns may occur
with the combination of two or more applicative markers or the combination of the causative and
applicative. (See Stiebels 2003 for instances of (44c).) Languages often show a V-CAUS-APPL order
even if the applied object is part of the caused subevent (e.g. an instrument of the action denoted by
the base verb), not the causing subevent. The Chichewa sequence lir-its-ir (‘cry-CAUSE-APPL’)
occurs for an instrumental applicative in which the instrument is used in the causing event as well
as for a benefactive applicative in which the beneficiary is related to the crying event (Hyman &
Mchombo 1992). Quechua provides another example:
(45) Quechua: causative/applicative (van de Kerke 1996:192)
a. ‘in my place my mother made Ana make a sweater’
b. ‘my mother made Ana make a sweater in my place’
c. ‘my mother made Ana make me a sweater’
Reading (45a) is compositional, whereas the other two readings, in which the beneficiary is related
to the subevent denoted by the base verb, are not. The semantic differences between the two orders
is schematized in a simplified manner in (46); APP denotes the relation integrated by the applicative:
(46) Combination of causative and applicative
λz λy λx λu λs' s [CAUSE(u,V(x,y)(s))(s') & APP(s',z)]
λz λy λx λu λs' s CAUSE(u, [V(x,y)(s) & APP(s,z)])(s')
Note that the non-occurring order V-APPL-CAUS would prevent a reading in which the applied
argument is related to the causing subevent. That this order is very rare cross-linguistically may
have its explanation in the already observed strong tendency that lexical items that refer to events
are linearized according to the chronological sequence of the denoted events. The causation is
always the initiating event, therefore it precedes the applicative independent of whether the applied
argument is part of the causing subevent or the caused subevent. The applied argument does not
affect the temporal order, therefore it does not seem to play a crucial role. The combination of
causative and applicative thus represents a case in which additional iconic ordering constraints seem
to outrank scope-related preferences.
5.3 Non-iconicity as a starting point – templatic approaches to morpheme
In contrast to the iconic approaches to morpheme order discussed in the preceding sections, work
like Inkelas (1993), Stump (1993), and Nordlinger (2008) argue that morphological sequencing is
more or less arbitrary and is accounted for by templates which simply stipulate the language-
specific ordering of morphemes by position classes. Thus, transparent/compositional and non-
transparent/non-compositional sequences are treated equally.
In our view, templatic approaches like these generalise the worst case in morpheme sequences
to all other instances of morphological ordering. The rather unrestricted generative power of these
formal devices is extended to cases that can be shown to be compositional and transparent. We
illustrate this point with Watters’ (2002) description of the nominal morphology of the Tibeto-
Burman language, Kham. Watters assumes the position class template given in (47), which contains
one prefix position and six suffix positions. The stem-adjacent suffix position is reserved for
number (plural), whereas the most distant suffix position is reserved for the nominaliser. The
positions in between are filled by various case markers. Six case markers compete for suffix
position 3. Watters tries to account for the flexibility of the locative and the adessive by means of
the optional suffix position 5.
(47) Kham: Position classes (Watters 2002:70)
The example in (48) instantiates the position classes indicated in (49). However, apart from the
unusual combination of ablative and locative case, their order is rather transparent, with no affix
being in a position that would contradict its scopal properties.
(48) Kham (Watters 2002:70)
‘with those from his house’
As (49) shows, Watters has to assume two levels of affixation because affixes of position 2 and 3
occur external to those of position 5 and 6. Such recursion is quite unexpected from the perspective
of a templatic approach because it suggests semantic transparency of the morphological complex.
The templates themselves do not implement recursion. Successive semantic enrichment via
affixation is illustrated in (50): possessor extension in (50a), possessor saturation/agreement in
(50b), the addition of a neighborhood region via the ablative in (50c), the addition of a locative
relation in (50d), the specification of the locatum argument as some kind of person/agent via the
nominalizer in (50e), the addition of the plural marker in (50f), and finally the addition of the
comitative relation in (50g).
(50) /zihm/; λu HOUSE(u)
a. /zihm/; λv λu [HOUSE(u) & POSS(v,u)]
b. /u-zihm/; λu v<-1,-2,-pl> [HOUSE(u) & POSS(v,u)]
c. /u-zihm-ni/; λu v<-1,-2,-pl> ~PROX[HOUSE(u) & POSS(v,u)]
d. /u-zihm-ni-ka/; λx u v<-1,-2,-pl> LOC(x, ~PROX[HOUSE(u) & POSS(v,u)])
e. /u-zihm-ni-ka-o/; λx u v<-1,-2,-pl> [PERSON(x) & LOC(x, ~PROX[HOUSE(u) &
f. /u-zihm-ni-ka-o-ra/; λx<+pl> u v<-1,-2,-pl> [PERSON(y) & LOC(x, ~PROX[HOUSE(u) &
g. /u-zihm-ni-ka-o-ra-sǝ/; λy x<+pl> u v<-1,-2,-pl> BE_ACCOMPANIED(y, [PERSON(x) &
LOC(x, ~PROX[HOUSE(u) & POSS(v,u)])])
As (50) suggests, there is no need for a template. The derivation applies in a completely
compositional manner. Therefore, the morphological structure of Kham nouns does not illustrate
arbitrariness in morpheme order.
22 ~PROX[x] denotes regions complementary to the proximity region of x. This is very simplified but suffices for purposes of
representation. Local cases usually conflate a LOC-relation and some specification of a neighborhood region.
Apart from being rather unrestricted, templatic approaches suffer from another deficiency: they
do not make any interesting cross-linguistic predictions for language acquisition. Arbitrary and non-
arbitrary orders should be equally easily acquired. Under the templatic view, children would just
have to learn the position classes of the respective language and the alignment of the morphological
markers to these classes. Children’s errors should then relate to wrong alignments. There is,
however, evidence that children may reanalyze arbitrary orders and replace structures that are less
iconic by more iconic ones. Slobin (1985), for instance, reports that children tend to replace certain
affixal scope markers (here negation; e.g. the Turkish example (51a)) by free forms (e.g. (51b)) or
realize them in a position that is more in line with their scopal properties:
(51) Turkish: replacement of affixal negation (Slobin 1985)
a. koy-ma-dı-m
On the basis of such data, templatic accounts should only provide an analysis of last resort for
clearly arbitrary morpheme sequences, and indeed, there is a body of work attempting to recast
templatic accounts in more iconic terms. For instance, Hyman (2003) assumes both a MIRROR
constraint (like Baker’s Mirror Principle, discussed just above) and morpheme-specific TEMPLATE
constraints. Depending on the respective ranking of these two types of constraints, transparent or
opaque affix orders emerge. This kind of OT approach does not explain why both arbitrary and non-
arbitrary morpheme ordering principles might co-occur, however. Trommer (2003), whose
approach is couched in an optimality-theoretic version of Distributed Morphology (Halle &
Marantz 1993, Harley & Noyer 1999), derives morpheme orders from the interplay of underlying
syntactic structures (determining the ordering of tense and aspect markers) and post-syntactic
morphological operations (e.g. alignment constraints for person and number). In such an approach,
the post-syntactic morphological component is usually responsible for non-transparent outcomes.
As we can see, while it can often be demonstrated that, in the default case, morpheme ordering
follows iconic principles, apparently arbitrary orderings appear to exist and remain a challenge.
6 Iconicity phenomena of cohesion
Various contrastive patterns of more cohesive vs. less cohesive structures have received attention in
the functional literature on iconicity. Some phenomena have been treated under the label of
Iconicity of conceptual distance, others under the label of Iconicity of independence (Haiman
1985a). The interesting point we would like to focus on is the relative nature of iconicity for these
phenomena: usually, the more cohesive structure shows a tighter semantic relation between the
elements involved than the less cohesive structure. Here, iconicity cannot be defined in absolute
terms, as for instance with homology.
6.1 Iconicity of conceptual distance
According to Haiman (1985a), the linguistic distance between expressions corresponds to their
conceptual distance. Linguistic distance is measured in terms of structural distance (number of
intervening prosodic/morphological/syntactic boundaries). Whereas linguistic distance can be more
or less easily evaluated in terms of intervening boundary nodes, the notion of conceptual distance is
harder to state more precisely. Unfortunately, the accounts we know of that use this notion do not
provide a more precise characterization.23 An oft-cited example (see, e.g., Newmeyer 1992) for
Iconicity of conceptual distance is the distinction of lexical (kill) vs. periphrastic causatives (cause
to die). The lexical causative, showing lexical fusion, denotes direct causation, whereas the
periphrastic causative, separated from the verb by a word boundary, denotes indirect causation,
which may involve intermediate subevents.
Iconicity of conceptual distance has also been proposed to account for the distinction in the
marking of alienable vs. inalienable possession found in many languages. Inalienable possession
usually reflects a tighter semantic relation between the possessor and the possessed noun than
alienable possession. Inalienable possession typically includes body-part expressions (e.g. arm, see
(52a)) and kinships terms (e.g. brother, see (52b)), which are relational already in their base entry.
Most sortal nouns are underlyingly non-relational and may thus be represented as shown in (52c).
These nouns must undergo possessor extension, as shown in (52d), in order to become relational
and take a possessor argument.
(52) a. arm λv λu [ARM(u) & PART_OF(u,v)]
b. brother λv λu BROTHER(u,v)
c. canoe λu CANOE(u)
d. poss(canoe) λv λu [CANOE(u) & POSS(v,u)]
Corresponding to the semantic distinction, inalienable possession is marked by morphosyntactic
means that exhibit a tighter morphosyntactic relation between the possessor and the possessed
nouns.24 Given the various formal differences between inalienable and alienable possession,
languages with bound and free possessive markers will usually use the bound markers to indicate
inalienable possession, whereas the free forms mark alienable possessors:
(53) Mekeo (Haiman 1985a: 131)
‘my younger brother’
eʔu ngaanga
‘my canoe’
‘your younger brother’
emu ngaanga
‘your canoe’
The difference between inalienable and alienable possession may not only involve bound vs. free
forms, but also involve more segmental material or additional morphology in the case of alienable
possession. This, then, would be an instance of Iconicity of complexity, which shows that the
various notions of iconicity are not fully disjoint in their extension. Since alienable possession
requires possessor extension, this operation may be indicated by an overt marker. (See examples in
Haiman 1983, 1985a.) The addition of further markers thus follows automatically.
As a mirror image, however, non-relational uses of relational nouns require binding of the
possessor marking, which may likewise be overtly coded. Mayan languages are representative in
this respect. “Absoluble” relational nouns (Lehmann 2003) show an overt marking of possessor
reduction (e.g. Yucatec tàatah ‘father (relational)’ tatah-tsil ‘father (non-relational)’), which can
be conceived of as some kind of “anti-possessive” (Stiebels 2006). According to Lehmann, the
languages’ lexical inventories may include “inabsoluble” and “impossessible” nouns, which cannot
23 The list of phenomena that have been regarded as cases to be explained by Iconicity of conceptual distance also includes
morpheme orders (Haiman 1983, 1985a). We follow Haspelmath (2008) in splitting Iconicity of conceptual distance into
cohesion and sequencing phenomena; the latter have been dealt with in the previous section.
24 The Oceanic language Daakaka (Kilu von Prince p.c.) exhibits various morphosyntactic alternatives for encoding possession. As
in many other Oceanic languages, body part expressions exhibit a bound possessive pronoun whereas inalienable nouns usually
take an unbound possessive pronoun. However, there is a certain class of body part expressions (referring to internal organs such
as ribs or bones) that not only takes a free possessive pronoun, but, in addition, uses a linker that does not occur with inalienable
nouns. These nouns thus instantiate the least tight structure.
be turned into non-relational or relational nouns, respectively. This example illustrates once again
that iconicity must be relativised with respect to particular lexical items.
6.2 Iconicity of independence
Under the notion of Iconicity of independence, Haiman (1983) relates the linguistic separateness of
expressions to the conceptual independence of the entities represented. Separate words or clauses
tend to denote separate entities or events, respectively. As in the case of Iconicity of conceptual
distance, a paradigmatic contrast between a more synthetic and a more analytic construction is to be
accounted for. The division of labour that emerges from this contrast is a meaning difference
between the two competing structures, such that the more synthetic construction correlates with a
lower degree of referential independence for the item to be morphologically integrated. The
phenomena discussed in the literature include, for instance, noun incorporation, the interpretation of
bound vs. free reflexives and the selection pattern of clause-embedding predicates. Each
phenomenon raises different questions concerning its iconic character.25
Noun incorporation (Mithun 1984) has been regarded as an instance of Iconicity of
independence because in many noun-incorporating languages the incorporated noun displays less
referential independence than a non-integrated noun. Whereas a non-integrated noun can be
definite/specific, as in the Ponapean example (54a), the incorporated noun can only be understood
non-specifically, as in (54b). The resulting V-N complex often denotes an event type, not an event
(54) Ponapean (Mithun 1984: 850)
‘I took all that medicine.’
‘I completed my medicine-taking.’
However, there is type of noun incorporation (type III in Mithun’s classification) in which noun
incorporation serves for manipulation of discourse structure. Here, the incorporated noun represents
a known entity that is backgrounded by means of noun incorporation. The following example from
Huahtla Nahuatl illustrates a typical discourse segment in which the entity to be put in the
background is realized as an incorporated noun in the answer (55b) to the question (55a).
(55) Huahtla Nahuatl (Mithun 1984: 861)
‘where is the knife? I want it now’
‘He cut the bread with it (the knife).’
The incorporated noun is clearly referentially accessible, albeit less discourse-salient. Sadock
(1980) had earlier pointed out that in Greenlandic, incorporated nouns may function as antecedents
25 In theoretical linguistics, the asymmetry between integrated vs. non-integrated forms has been under debate in discussing the
lexicalist hypothesis (see references above), which in its strongest form states that words are inaccessible to syntactic operations.
One aspect of this inaccessibility is the fact that words are regarded as anaphoric islands (see Postal 1969).
to pronouns, which he took as evidence for the syntactic nature of noun incorporation in
Greenlandic. The following example, taken from van Geenhoven (1998), illustrates this:
(56) Greenlandic (van Geenhoven 1998)
‘Aani has a dogi. Iti is called Miki.’
According to Farkas & de Swart (2003), noun-incorporating languages can be classified in terms of
the discourse transparency of incorporated nouns, i.e. the potential of incorporated nouns to
function as antecedents. Languages such as Greenlandic show discourse-transparent noun
incorporation, whereas many other languages do not allow discourse transparency. However, as
pointed out by Farkas & de Swart, noun incorporation is invariably characterized by one general
property: the incorporated noun must scope with the incorporating predicate and cannot have wide
scope relative to any operator or quantifier that scopes over the incorporating predicate.
Since noun incorporation is not generally characterized by lack of referential independence, one
may ask whether scopal inertness should be taken as a factor that iconically distinguishes bound
from unbound forms, replacing the concept of Iconicity of independence. In other morphological
domains, however, scopal inertness does not exhibit the same predictive force in terms of bound vs.
unbound forms. There are languages that realize sentential negation or sentence mood markers
exclusively by bound morphemes. However, systematic scopal differences compared to languages
with morphologically unbound forms have not been reported, as far as we know. Whether scopal
differences emerge in languages that exhibit both bound and unbound scopal markers with the same
function is a question for future research.
Unlike noun incorporation, which does not show a division of labour between bound/integrated
forms and unbound/non-integrated forms – at least in discourse-transparent languages – other
markers display effects that can be traced back to this division of labour. One illustrative case
would be the different potential of German negators to license replacive negation with sondern
‘but’ (Jacobs 1982). Free negators license replacive negation (nicht schön, sondern … ‘not beautiful
but’), bound negators do not (*un-schön, sondern lit. ‘unbeautiful but’). This distribution is
expected and iconic in so far as the prefix un- is embedded and thus not accessible as licensor of
sondern. (It also avoids synonymy.) The inverse pattern should not occur.
A further instance of Iconicity of independence discussed by Haiman (1983, 1985a) is the
systematic difference between bound and free reflexives. According to Haiman, free reflexives tend
to encode entities that are conceived of as distinct from the antecedent, whereas this is not the case
with reflexive affixes. An illustrative example would be the following minimal pair from
(57) Hungarian: free vs. bound reflexive (Haiman 1985a: 143)
‘I washed myself’
b. Meg-mos-akod-t-am
‘I washed’
In Russian, the affixal reflexive has turned into a kind of detransitivization marker (see (58b/d),
below), whereas the free reflexive has remained a true reflexive.
(58) Russian: free vs. bound reflexive (Haiman 1985a: 145f.)
a. bit’ sebja ‘hit oneself’
b. bit’-sja ‘bump into (accidentally)’
c. utomit’ sebja ‘exhaust oneself (on purpose)’
d. utomit’-sja ‘grow weary (spontaneously)’.
However, languages that display only one type of reflexive do not exhibit this interpretative
contrast, which shows that iconicity can only be claimed when paradigmatic contrast exists. In
Classical Nahuatl, for instance, the reflexive prefix functions both as a true reflexive (e.g. (59a))
and as a kind of intransitivizer for some verbs (e.g. (59b)).
(59) Classical Nahuatl: bound reflexive (Launey 2004:1437)
a. ni-no-pāka
‘I wash myself
b. m-iʔtoa on
‘that is what is said’
In German, it is free reflexives that serve both functions (sich waschen ‘wash oneself’ vs. sich
vergrößern ‘become bigger’).
The third case we would like to consider is the morphosyntactic realization of the complements
of clause-embedding predicates. Languages and their clause-embedding predicates vary as to which
kinds of sentential complements they select. While there are no cross-linguistically uniform
patterns, there are certain tendencies that are usually explained by relating a certain hierarchy of
clause-embedding predicates to a certain hierarchy of complementation structures. What is claimed
to be iconic is the mapping between these two hierarchies. The general idea goes back to Givón’s
“binding hierarchy” (1980, 1985), further developed in Givón (1990/2001). He proposes the
following principle:
(60) Referential cohesion and event integration (Givón 1990:527)
The more the two events coded in the main and complement clauses share their referents,
the more likely they are to be semantically integrated as a single event; and the less likely is
the complement clause to be coded as an independent finite clause.
Givón’s syntactic scale of complementation structures is given in (61), and his scale of predicate
classes in (62). The tightest morphosyntactic connection between an embedding predicate and its
complement is that of predicate raising (lexicalization as one verb, e.g. as a verbal compound or a
structure derived by verb incorporation). The least integrated structure would be that of a direct
(61) Syntactic scale of clausal arguments (Givón 1990:519)
Predicate raising > bare stem > infinitive > for-to > Subjunctive > indirect quote > direct
The scale is meant to reflect the structural similarity of each clausal argument type to root clauses,
which depends on the realization of the embedded subject, the inflectional potential of the
embedded clause’s head (tense, mood, aspect), the degree of morphosyntactic fusion of matrix and
embedded predicate, the presence of complementisers and further indicators of embedding. The
predicate scale in (62) evaluates the degree of event integration such that the highest predicate class
shows the highest degree of event integration. Concomitant factors are the influence exerted by the
matrix agent on the embedded agent, the success of manipulation (if present), and the commitment
and emotional involvement of the matrix subject with respect to the outcome encoded in the
complement clause.
(62) Scale from manipulation to cognition (Givón 1990:530)
successful intended causation > attempted manipulation > preference/aversion > epistemic
anxiety > epistemic certainty/uncertainty > utterance
The general preference, then, is for elements high on the predicate scale to occur in structures that
are high on the syntactic scale in (61). Since it is not possible to predict the language-specific
selection pattern of a given predicate, Givón and other typologists (e.g. van Valin 1993, Cristofaro
2003) usually interpret (62), or a slightly modified scale, as an implicational scale: the specific
selection pattern of a predicate at a certain point of the scale implies that all predicates equal or
higher on the scale will select a complement structure that is as high or higher on the syntactic
scale, i.e. an equally or even more tightly integrated complement.
Since our chapter is concerned with morphological exponence, only the tightest form of
integration (compounding, verb incorporation) is interesting here. It is a well-known fact that many
languages display morphological causatives. Likewise, preference/desiderative predicates such as
‘want’ may figure as incorporating verbs or some kind of lexical affixes. The scale in (62) would
gain support, if, for each predicate class, there is a language that has this predicate class as a cut-off
point for some type of complement structure (here, incorporation or compounding). Note that one
can find polysynthetic languages such as Yaqui (Guerrero 2006) and West Greenlandic, which
exhibit incorporating predicates that are at the bottom of the predicate scale, namely epistemic (e.g.
‘believe’) or utterance predicates. Consider the following examples from Yaqui:
(63) Yaqui (Guerrero 2006:170/186)
‘I believe Pedro to be buying a horse.’
‘Ramón says that the coyote is dying.’
The prediction for Yaqui, then, is that, with the possible exception of more specific utterance
predicates, all other clause-embedding predicates could possibly surface as incorporating verbs.
Generally, one cannot conclude from the descriptions available on clausal complementation in the
world’s languages whether all clause-embedding predicates behave in the predicted way. This
would require a thorough analysis of the whole inventory of a language’s clause-embedding
predicates and their array of selected complements. Therefore, further research is needed in order to
verify the validity of these implicational scales.26
Once again, iconicity is taken as a relative concept, defining the unmarked structural relation.
Lexical predicates are evaluated in terms of their “semantic fit” for a certain complement type.
Their relative potential to take clausal arguments with a rich variation in TMA-specifications and
embedded-subject referents is aligned to the structural richness of the respective complement type.
A “stronger” predicate (with more semantic variation in the complement clause) will also occur
with “stronger” complements (with more syntactic structure).
26 Stiebels (2007) points out that there may be problems with the proposed scales of predicate classes. For instance, in Hungarian
manipulative predicates select subjunctive complements, whereas perception predicates may select infinitival complements.
Cristofaro (2003), for instance, ranks manipulative predicates higher than perception predicates.
7 Conclusions and open questions
Let us conclude by reviewing the roles we have shown that iconicity plays in formal theories of
morphology. As noted in the introduction, the role of iconicity in explaining language structure has
been a major issue in functional linguistics, with the framework of Natural Morphology/Syntax
being one of its strong proponents: see Mayerthaler 1981, Dressler et al. 1987, Wurzel 1989. In
formal approaches, iconicity is seldom put explicitly on the table. However, we have demonstrated
in this chapter that iconicity may often function as an implicit guiding principle in the specific
formulation of formal analyses. Iconicity of complexity underlies most markedness hierarchies, in
all fields of linguistics, as Newmeyer (1992) points out. Paradigm uniformity is motivated by the
iconic principle of a one-to-one relationship between form and meaning. Baker’s (1985, 1998)
Mirror Principle relies on an isomorphism between syntax and surface order of morpheme
sequences. The same is true for alternative approaches that refer to principles of semantic scope. In
the context of Optimality Theory, authors have explicitly used iconicity constraints that interact
with more structure-related formal constraints. Since constraint interaction provides a mechanism
for formalizing that iconicity principles are tendencies, with principled exceptions defined by
higher-ranking constraints, Optimality Theory has brought the welcome development of
incorporating more formal explicitness in analyses that postulate that iconicity principles play a
crucial role.
Numerous open questions about the role of iconicity in language and linguistic analysis remain.
One obvious question concerns the psychological reality of iconicity, i.e. its role in language
processing and language acquisition. While Haiman (2008) is sceptical about the cognitive benefit
of iconicity (see the quote in (64)), Givón (1985:189) states the meta-principle in (65).
(64) “Iconicity seems at least at present to offer no proven cognitive benefits. [...] Now it seems
that neither of the traditional motivations for linguistic form (economy of effort for the
benefit of the speaker vs. clarity for the benefit of the hearer) can account for it [iconicity].”
[Haiman 2008: 45]
(65) The iconicity meta-principle (Givón 1985:189)
All other things being equal, a coded experience is easier to store, retrieve and communicate
if the code is maximally isomorphic to the experience.
The studies we have cited throughout the paper (Slobin 1985, Bergen 2004, Imai et al. 2008) show
that Givón is right in that iconic structures are easier to acquire and retrieve. However, there is still
a need for more systematic research on the acquisition and processing of iconic vs. less iconic
Another open question concerns the proposal found in work like Haspelmath (2008), for
instance, that the insights gained from the assumption that Iconicity of complexity, Iconicity of
quantity and Iconicity of cohesion can be derived from frequency facts. In this view, shared by
connectionist approaches like Bybee (1995a, b), the distribution of marked vs. unmarked values
corresponds to infrequent vs. frequent occurrence. Besides the fact that the claims of usage-based
accounts quite often lag far behind their foundation in thorough empirical studies (not all of these
claims are supported with empirical findings on frequencies), frequency distributions only signal
correlations. They do not provide an explanation because the frequency distributions themselves
need to be explained. As Haiman (2008) points out, productive patterns (whose tokens are often
very infrequent, at least in the beginning) show iconicity effects independent of frequency. From
the perspective of formal theoretical linguistics, explicit accounts of specific grammatical structures
do not appear to be likely to be replaced by observations about frequency-correlated tendencies
because the latter will not suffice to provide a precise and predictive account of these phenomena.
More typological research on the frequency of iconic structures compared to non-iconic structures
is certainly necessary to resolve this debate.
We end with desiderata for future research. First, as we have noted at numerous points in the
chapter, more typological studies of morphological structures said to illustrate iconic principles are
needed to test whether these principles represent cross-linguistically robust tendencies. Further, in
order to demonstrate that mappings are iconic, one must represent both the form and the meaning in
a way that allows for testable comparisons. Analyses appealing to iconicity instead often rely on an
intuitive understanding of the items’ meanings. More formal rigor in the representations of form
and meaning in studies appealing to iconicity, tested more systematically on a wider variety of
languages, is clearly needed.
A accusative-like agreement
ABL ablative
ACC accusative
ADE adessive
ADV adverbializer
AG agentive
ALL allative
ANIM animate
APPL applicative
ASP aspect
ASS assistive
CAUS causative
CH.CLT chaining clitic
CIS cislative
CL noun class agreement
COM comitative
COMPL completive aspect
DEF definite
DEICT deictic particle
DEL delative
DEP dependent verb form
DET determiner
DIM diminutive
DS different subject
E ergative-like agreement
ELA elative
F feminine gender
GEN.AGR gender agreement
HORT hortative
IN/ON interior/exterior region case
IND indicative
INF infinitive
INSTR instrumental
IRR irrealis
ITR intransitive
LAT lative
LOC locative
M masculine gender
N nominative-like agreement
NEG negation
NOM nominative
NOML nominalizer
NSPEC non-specific reference
OJ object
ORIENT orientation
P possessor agreement
PASS passive
PAST past tense
PERF perfect
PTCP participle
PL plural
PN proper name
PRES present tense
PRO pronoun
RED reduplicand
REFL reflexive
REP repetitive
SG singular
SPEC specific reference
SS same subject
SU subject
TOP topic marker
WIT.PAST witnessed past
Acquaviva, Paolo. 2008. Lexical Plurals: A Morphosemantic Approach. Oxford: Oxford University
Aissen, Judith. 1999. Markedness and subject choice in Optimality Theory. Natural Language and
Linguistic Theory 17, 673-711.
Aissen, Judith. 2003. Differential object marking: Iconicity vs. economy. Natural Language and
Linguistic Theory 21, 435-483.
Amberber, Mengistu. 2000. Valency-changing and valency-encoding devices in Amharic. In: R. M.
W. Dixon & Alexandra Y. Aikhenvald (eds.), 312-332.
Antilla, Arto & Adams Bodomo. 2009. Prosodic morphology in Dagaare. In: Masangu Matondo et
al. (eds.), Selected Proceedings of the 38th Annual Conference on African Linguistics.
Somerville, MA: Cascadilla Proceedings Project, 56-68.
Baker, Mark. 1985. The Mirror Principle and morphosyntactic explanation. Linguistic Inquiry 16:
Baker, Mark. 1988. Incorporation: a theory of grammatical function changing. Chicago: The
University of Chicago Press.
Beckman, Jill N. 1997. Positional faithfulness, positional neutralisation and Shona vowel harmony.
Phonology 14, 1-46.
Bergen, Benjamin K. 2004. The psychological reality of phonaesthemes. Language 80, 290-311.
Bierwisch, Manfred. 1983. Semantische und konzeptuelle Repräsentation lexikalischer Einheiten.
In: Rudolf Růžička & Wolfgang Motsch (eds.), Untersuchungen zur Semantik. Studia
grammatica 22, 61-99. Berlin: Akademie Verlag.
Bierwisch, Manfred. 1989a. Event nominalization: Proposals and problems. In Wofgang Motsch
(ed.), Wortstruktur und Satzstruktur (Linguistische Studien 194), 1-73. Berlin: Akademie der
Wissenschaften der DDR; Zentralinstitut für Sprachwissenschaft.
Bierwisch, Manfred. 1989b. The Semantics of Gradation. In Manfred Bierwisch & Ewald Lang
(eds.), Dimensional Adjectives, 71-261. Berlin: Springer.
Birdsong, David. 1995. Iconicity, markedness, and processing constraints in frozen locutions. In:
Marge E. Landsberg (ed.), 31-45.
Bloomfield, Leonard. 1933. Language. New York: Holt.
Bolinger, Dwight L. 1949. The sign is not arbitrary. Boletín del Instituto Caro y Cuervo 5, 52-62.
Bolinger, Dwight. 1977. The form of language. London: Longmans.
Bolinger, Dwight. 1978. Intonation across languages. In Joseph Greenberg. (ed.), 471-524.
Bossong, Georg. 1985. Empirische Universalienforschung: Differentielle Objektmarkierung in den
neuiranischen Sprachen. Tübingen: Narr.
Bresnan, Joan. 2001. Lexical-functional syntax. Oxford: Blackwell.
Bresnan, Joan & Sam A. Mchombo. 1995. The Lexical Integrity Principle: Evidence from Bantu.
Natural Language and Linguistic Theory 13, 181-254.
Broschart, Jürgen. 1997. Why Tongan does it differently: Categorial distinctions in a language
without nouns and verbs. Linguistic Typology 1, 123-165.
Bybee, Joan L. 1985a. Morphology: A study of the relation between meaning and form.
Amsterdam: Benjamins.
Bybee, Joan L. 1985b. Diagrammatic iconicity in stem-inflection relations. In: John Haiman (ed.),
Bybee, Joan L. 1995a. Regular morphology and the lexicon. Language and Cognitive Processes 10,
Bybee, Joan L. 1995b. Diachronic and typological properties of morphology and their implications
for representation. In Laurie Beth Feldman (ed.), Morphological Aspects of Language
Processing. Hilldale, NJ: Lawrence Erlbaum Associates, 225-246.
Cassimjee, Farida. 1998. Isixhosa Tonology: An Optimal Domains Theory Analysis. Munich:
Lincom Europa.
Childs, G. Tucker. 1994. African ideophones. In Leanne Hinton et al. (eds.), 178-204.
Childs, G. Tucker. 2001. Research on ideophones, whither hence?: The need for a social theory of
ideophones. In: Voeltz & Kilian-Hatz (eds.), 63-73.
Chomsky, Noam. 1970. Remarks on Nominalization. In: Roderick A. Jacobs & Peter S. Rosenbaum
(eds.), Readings in English Transformational Grammar. Waltham, Mass.: Blaisdell, 184–221.
Cinque, Guglielmo. 1999. Adverbs and functional heads: a cross-linguistic perspective. Oxford:
Oxford University Press.
Comrie, Bernard. 1989. Language universals and linguistic typology. 2nd ed. Oxford: Blackwell.
Cooper, William E. & John Robert Ross. 1975. World order. In: R. E. Grossman, L. San & T. J.
Vance (eds.), Papers from the parasession on functionalism. Chicago Linguistic Society, 63-
Copestake, Ann & Ted Briscoe. 1996. Semi-productive polysemy and sense extension. In James
Pustejovsky & Branimir Boguraev (eds.), 15-67.
Cooreman, Ann. 1994. A functional typology of antipassives. In: Barbara Fox & Paul J. Hopper
(eds.), Voice: Form and function. Amsterdam: Benjamins, 49-88.
Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press.
Davis, Stuart. 2005. The Paradigm Uniformity effect reconsidered. In Downing et al. (eds.), 107-
Diffloth, Gérard. 1994. i: big, a: small. In Leanne Hinton et al. (eds.), 107-114.
Di Sciullo, Anna-Maria & Edwin Williams 1987. On the definition of word. Cambridge, MA: The
MIT Press.
Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press.
Dixon, R. M. W. & Alexandra Y. Aikhenvald. 2000. Introduction. In: R. M. W. Dixon & Alexandra
Y. Aikhenvald (eds.), Changing valency: case studies in transitivity. Cambridge: Cambridge
University Press, 1-29.
Downing, Laura J. 1998. On the prosodic misalignment of onsetless syllables. Natural Language
and Linguistic Theory 16, 1-52.
Downing, Laura J. 1999. Morphological constraints on Bantu reduplication. Linguistic Analysis 29,
Downing, Laura J. 2006. Canonical Forms in Prosodic Morphology. Oxford: Oxford University
Downing, Laura J. to appear. Accent in African languages. In R.W.N. Goedemans, H.G. van der
Hulst & E. A. van Zanten, eds. Word Prosodic Systems in the Languages of the World. Berlin:
Mouton de Gruyter.
Downing, Laura J., T. Alan Hall & Renate Raffelsiefen. 2005. Introduction. In Downing et al.
(eds.), 1-16.
Downing, Laura J., T. Alan Hall & Renate Raffelsiefen (eds.). 2005. Paradigms in Phonological
Theory. Oxford: Oxford University Press.
Dresher, B. Elan & Harry van der Hulst. 1998. Head-dependent asymmetries in phonology:
complexity and visibility. Phonology 15, 317-352.
Dressler, Wolfgang U. 2005. Word-formation in Natural Morphology. In: Pavol Štekauer &
Rochelle Lieber (eds.), Handbook of Word-formation. Dordrecht: Springer, 267-284.
Dressler, Wolfgang U., Willi Mayerthaler, Oswald Panagl & Wolfgang U. Wurzel. 1987. Leitmotifs
in Natural Morphology. Amsterdam: Benjamins.
Egg, Markus. 2006. Anti-Ikonizität an der Syntax-Semantik-Schnittstelle. Zeitschrift für
Sprachwissenschaft 25, 1-38.
Farkas, Donka F. & Henriëtte de Swart. 2003. The semantics of incorporation: From argument
structure to discourse transparency. Stanford: CSLI Publications.
Filimonova, Elena. 2005. The noun phrase hierarchy and relational marking: Problems and
counterevidence. Linguistic Typology 9, 77–113.
Foley, William A. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge University
Frishberg, Nancy. 1975. Arbitrariness and iconicity: historical change in American Sign Language.
Language 51, 696-719.
Gamerschlag, Thomas. 2005. Komposition und Argumentstruktur komplexer Verben: Eine
lexikalische Analyse von Verb-Verb-Komposita und Serialverbkonstruktionen. Berlin: Akademie
Geenhoven, Veerle van. 1998. Semantic incorporation and indefinite descriptions. Stanford: CSLI
Givón, Talmy. 1980. The Binding Hierarchy and the typology of complements. Studies in language
4, 333-377.
Givón, Talmy. 1985. Iconicity, isomorphism, and non-arbitrary coding in syntax. In: John Haiman
(ed.), 187-219.
Givón, Talmy. 1990/2001. Syntax: An introduction. Vol. II. Amsterdam: Benjamins.
Ghomeshi, Jila, Ray Jackendoff, Nicole Rosen, & Kevin Russell. 2004. Contrastive focus
reduplication in English (the salad-salad paper). Natural Language & Linguistic Theory 22,
Greenberg, Joseph H. (ed.) 1978. Universals of Human Language, vol. 2: Phonology. Stanford,
Calif.: Stanford University Press.
Grimm, Scott. 2009. Number marking and individuation: A view from Dagaare. Paper presented at
ZAS Berlin, March 26th, 2009.
Gussenhoven, Carlos. 2004. The Phonology of Tone and Intonation. Cambridge: Cambridge
University Press.
Guerrero, Lilián (2006). The structure and function on Yaqui complementation. München: Lincom
Haiman, John. 1983. Iconic and economic motivation. Language 59, 781-819.
Haiman, John. 1985a. Natural syntax: Iconicity and erosion. Cambridge: Cambridge University
Haiman, John (ed.). 1985b. Iconicity in syntax. Amsterdam: Benjamins.
Haiman, John. 2003. Explaining infixation. In: John Moore & Maria Polinsky (eds.), The nature of
explanation in linguistic theory. Stanford: CSLI Publications.
Haiman, John. 2008. In defence of iconicity. Cognitive Linguistics 19, 35-48.
Hamano, Shoko. 1994. Palatalization in Japanese sound symbolism. In Leanne Hinton et al. (eds.),
Halle, Morris & Alec Marantz. 1993. Distributed Morphology and pieces of inflection. In: Kenneth
Hale and Samuel J. Keyser (eds.), The view from building 20: Essays in linguistics in honor of
Sylvain Bromberger, 111-176. Cambridge, MA: MIT Press.
Harbour, Daniel. 2007. Morphosemantic Number: from Kiowa Noun Classes to UG Number
Features. Dordrecht: Springer.
Harley, Heidi & Rolf Noyer. 1999. Distributed Morphology. Glot International 4, 3-9.
Harrison, Sheldon P. 1973. Reduplication in Micronesian languages. Oceanic Linguistics 12, 407-
Härtl, Holden. 2003. Conceptual and gramatical characteristics of argument alternations: The case
of decausative verbs. Linguistics 41, 883-916.
Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In:
Bernard Comrie & Maria Polinsky (eds.), Causatives and transitivity. Amsterdam: Benjamins,
Haspelmath, Martin. 2008. Frequency vs. iconicity in explaining grammatical asymmetries.
Cognitive Linguistics 19, 1-33.
Hinton, Leanne, Johanna Nichols & John J. Ohala (eds.). 1994a. Sound Symbolism. Cambridge:
Cambridge University Press.
Hinton, Leanne, Johanna Nichols & John J. Ohala. 1994b. Introduction: Sound symbolic processes.
In Leanne Hinton et al. (eds.), 1-12.
Hopper, Paul. J. & Sandra A. Thompson. 1985. The iconicity of the universal categories ‘noun’ and
‘verbs’. In: John Haiman (ed.), 151-183.
Hunter-Smith, Sarah. 2007. Understanding without Babblefish: Reviewing the Evidence for
Universal Sound Symbolism in Natural Languages. Swarthmore College Senior Thesis.
Hurch, Bernhard (ed.). 2005. Studies on Reduplication. Berlin: Mouton de Gruyter.
Hyman, Larry M. 2003. Suffix ordering in Bantu: a morphocentric approach. In: Geert Booij &
Jaap van Marle (eds.), Yearbook of Morphology 2002, 245-281.
Hyman, Larry M. & Sam A. Mchombo. 1992. Morphosyntactic constraints in the Chichewa verb
stem. Berkeley Linguistics Society 18, 350-364.
Imai, Mutsumi, Sotaro Kita, Miho Nagumo & Niroyuki Okada. 2008. Sound symbolism facilitates
early verb learning. Cognition 109, 54-65.
Inkelas, Sharon. 1993. Nimboran position class morphology. Natural Language and Linguistic
Theory 11, 559-624.
Inkelas, Sharon & Orhan Orgun. 1995. Level ordering and economy in the lexical phonology of
Turkish. Language 71, 763-93.
Inkelas, Sharon & Cheryl Zoll. 2005. Reduplication. Cambridge: Cambridge University Press.
Itô, Junko. 1990. Prosodic minimality in Japanese. Proceedings of CLS 26, 2: Parasession on the
Syllable in Phonetics and Phonology, 213-239.
Jacobs, Joachim. 1982. Syntax und Semantik der Negation im Deutschen. München: Fink.
Jäger, Gerhard. 2007. Evolutionary Game Theory and typology: A case study. Language 83, 74-
Julien, Marit. 2002. Syntactic heads and word formation. Oxford: Oxford University Press.
Kaufmann, Ingrid. 2004. Medium und Reflexiv: eine Studie zur Verbsemantik. Tübingen: Niemeyer.
Kaufmann, Ingrid. 2007. Middle voice. Lingua 117, 1677-1714.
Kennedy, Christopher. 2001. Polar opposition and the ontology of ‘degrees’. Linguistics and
Philosophy 24, 33-70
Kerke, Simon van de. 1996. Affix order and interpretation in Bolivian Quechua. Ph.D. Dissertation,
University of Amsterdam.
Kilian-Hatz, Christa. 2001. Universality and diversity: Ideophones from Baka and Kxoe. In F. K.
Erhard Voeltz & Christa Kilian-Hatz (eds.), 155-163.
Kimenyi, Alexandre. 2004. Kinyarwanda. In: Geert Booij, Christian Lehmann, Joachim Mugdan &
Stavros Skopeteas (eds.), Morphologie. Ein internationales Handbuch zur Flexion und
Wortbildung. Vol. 2. Berlin: Walter de Gruyter, 1527-1545.
Kiparsky, Paul. 1997. Remarks on Denominal Verbs. In Alex Alsina, Joan Bresnan & Peter Sells
(eds.), Argument Structure. Stanford: CSLI, 473-499.
Kiparsky, Paul. 2001. Structural case in Finnish. Lingua 111, 315-376.
Klamer, Marian. 2001. Expressives and iconicity in the lexicon. In F.K. Erhard Voeltz & Christa
Kilian-Hatz (eds.), 165-181.
Klamer, Marian. 2002. Semantically motivated lexical patterns: A study of Dutch and Kambera
expressives. Language 78, 258-286.
Klima, Edward & Ursula Bellugi. 1979. The Signs of Language. Cambridge, Mass.: Harvard
University Press.
Kooij, Els van der. 2002. Phonological categories in Sign Language of the Netherlands: the role of
phonetic implementation and iconicity. Ph.D. dissertation, Leiden University.
Koontz-Garboden, Andrew. 2008. Monotonicity at the lexical semantic-morphosyntax interface. In
Emily Elfner & Martin Walkow (eds.), Proceedings of the 37th annual meeting of the Northeast
Linguistic Society. Amherst, MA: GLSA.
Koontz-Garboden, Andrew. 2009. Anticausativzation. Natural Language and Linguistic Theory 27,
Kouwenberg, Silvia & Darlene LaCharité. 2005. Less is more: Evidence from diminutive
reduplication. In Bernhard Hurch (ed.), 533-545.
Krämer, Martin & Dieter Wunderlich.1999. Transitivity alternations in Yucatec, and the correlation
between aspect and argument roles. Linguistics 37, 431-479.
Kulikov, Leonid. 2005. Reduplication in the Vedic verb: Indo-European inheritance, analogy and
iconicity. In Hurch (ed.), 431-454.
Landsberg, Marge E. (ed.). 1995a. Syntactic iconicity and linguistic freezes: The human dimension.
Berlin: Mouton de Gruyter.
Landberg, Marge E. 1995b. Semantic constraints on phonologically independent freezes. In: Marge
E. Landsberg (ed.), 65-78.
Launey, Michel. 2004. Nahuatl (Uto-Aztecan). In: Geert Booij, Christian Lehmann, Joachim
Mugdan & Stavros Skopeteas (eds.), Morphologie. Ein internationales Handbuch zur Flexion
und Wortbildung. Vol. 2. Berlin: Walter de Gruyter, 1433-1453.
Lahiri, Aditi. 2003. Introduction. Aditi Lahiri (ed.), 1-14.
Lahiri, Aditi (ed.). 2003. Analogy, Levelling, Markedness. Berlin: Mouton de Gruyter.
Lehmann, Christian. 2003. Possession in Yucatec Maya. 2nd revised edition. Arbeitspapiere des
Seminars für Sprachwissenschaft der Universität Erfurt 10. University of Erfurt.
Matsumoto, Yo. 1996. Complex predicates in Japanese: a syntactic and semantic study of the
notion ‘word’. Stanford: CSLI Publications.
Mayerthaler, Willi. 1981. Morphologische Natürlichkeit. Wiesbaden: Athenaion.
McCarthy, John J. 2000. The prosody of phase in Rotuman. Natural Language and Linguistic
Theory 18, 147-197.
Meinhof, Carl. 1912. Die Sprache der Hamiten. Hamburg: Friederichsen.
Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60, 847-893.
Moravcsik, Edith. A. 1978. Reduplicative constructions. In Joseph H. Greenberg (ed.), Universals
of Human Language, vol. 3, Word Structure. Stanford, Calif.: Stanford University Press, 297-
Moravcsik, Edith A. 2000. Infixation. In: Geert Booij, Christian Lehmann & Joachim Mugdan
(eds.), Morphologie: Ein internationales Handbuch zur Flexion und Wortbildung. Vol. I. Berlin:
Walter de Gruyter, 545-552.
Muysken, Pieter. 1986. Approaches to affix order. Linguistics 24, 629-643.
Müller, Gereon. 1997. Beschränkungen für Binominalbildung im Deutschen: Ein Beitrag zur
Interaktion von Phraseologie und Grammatik. Zeitschrift für Sprachwissenschaft 16, 5-51.
Müller, Gereon. 2004. Syncretism and iconicity in Icelandic noun declensions: a Distributed
Morphology approach. In Yearbook of Morphology 2004, 229-271.
Newmeyer, Frederick J. 1992. Iconicity and generative grammar. Language 68, 756-796.
Niepokuj, Mary K. 1991. The historical development of reduplication, with special reference to
Indo-European. Ph.D. dissertation, University of California-Berkeley.
Nöth, Winfried. 2008. Semiotic foundations of natural linguistics and diagrammatic iconicity. In:
Klaas Willems & Ludovic De Cuypere (eds.), Naturalness and iconicity in language.
Amsterdam: Benjamins, 73-100.
Nordlinger, Rachel. 2008. Verbal morphology in Murrinh-Patha: evidence for templates. Ms.,
University of Melbourne. To appear in Morphology.
Nuckolls, Janis B. 1999. The case for sound symbolism. American Review of Anthropology 28,
Nuckolls, Janis B. 2001. Ideophones in Pastaza Quechua. In F.K. Erhard Voeltz & Christa Kilian-
Hatz (eds.), 271-285.
Ohala, John J. 1994. The frequency code underlies the sound-symbolic use of voice pitch. In
Leanne Hinton et al. (eds.), 325-347.
Orie, Olanike Ola. 1997. Benue-Congo Prosodic Phonology and Morphology in Optimality Theory.
Munich: Lincom Europa.
Oswalt, Robert L. 1994. Inanimate imitatives in English. In Leanne Hinton et al. (eds.), 293-306.
Peirce, Charles S. 1932. The icon, index and symbol. ms. reprinted in Collected papers of Charles
Sanders Peirce. ed. by Charles Hartshorne & Paul Weiss, vol. 2: Elements of logic. Cambridge,
Mass.: Harvard University Press, 156-173.
Postal, Paul. 1969. Anaphoric islands. In: Robert I. Binnick, Alice Davison, Georgia M. Green &
Jerry L. Morgan (eds.), Papers from the 5th regional meeting of the Chicago Linguistic Society,
Prince, Alan & Paul Smolensky. 1993. Optimality Theory: constraint interaction in generative
grammar. Ms., Rutgers University, New Brunswick and University of Colorado, Boulder.
[published 2004 by Malden, MA: Blackwell.]
Pustejovsky, James & Branimir Boguraev (eds.). 1996. Lexical semantics: The problem of
polysemy. Oxford University Press.
Ramat, Anna Giacalone. 1995. Iconicity in grammaticalization processes. In: Raffaele Simone (ed.),
Iconicity in language. Amsterdam: Benjamins, 119-139.
Ravin, Yael & Claudia Leacock (eds.). Polysemy: Theoretical and Computational Approaches.
Oxford: Oxford University Press.
Reinhart, Tanja. 2002. The theta system: An overview. Theoretical Linguistics 28, 229-290.
Rhodes, Richard. 1994. Aural images. In Leanne Hinton et al. (eds.), 276-292.
Rialland, Annie. 2007. Question prosody: an African perspective. In Tomas Riad & Carlos
Gussenhoven (eds.), Tone and Tunes, vol. 1: Typological Studies in Word and Sentence
Prosody. Berlin: Mouton de Gruyter, 35-62.
Rice, Keren. 2000. Morpheme order and semantic scope. Cambridge: Cambridge University Press.
Rubino, Carl. 2005. Reduplication: Form, function and distribution. In Bernhard Hurch (ed.), 11-29.
Sadock, Jerrold M. 1980. Noun incorporation in Greenlandic: a case of syntactic word formation.
Language 56, 300-319.
Sadock, Jerrold M. 1991. Autolexical syntax: a theory of parallel grammatical representations.
Chicago: The University of Chicago Press.
Sag, Ivan A., Ronald Kaplan, Lauri Karttunen, Martin Kay, Carl Pollard, Stuart Shieber & Annie
Zaenen. 1986. Unification and Grammatical Theory. In Proceedings of the Fifth Annual
Meeting of the West Coast Conference on Formal Linguistics, 238-254. Stanford: CSLI
Sandler, Wendy & Diane Lillo-Martin. 2006. Sign Language and Linguistic Universals.
Cambridge: Cambridge University Press.
Sapir, Edward. 1929. A study in phonetic symbolism. Journal of Experimental Psychology 12, 225-
Saussure, Ferdinand de. 1983. Cours de linguistique générale. Édition critique préparée par Tullio
de Mauro. Paris: Payot.
Silverstein, Michael. 1976. Hierarchy of features and ergativity. In: R. M. W. Dixon (ed.),
Grammatical categories in Australian languages. Australian Institute of Aboriginal Studies No.
22. New Jersey: Humanities Press Inc, 112–171.
Slobin, Dan I. 1985. The child as a linguistic icon-maker. In: John Haiman (ed.), 221-248.
Slobin, Dan I. 2005. Linguistic representations of motions events: What is signifier and what is
signified? In: Constantino Maeder, Olga Fischer & William J. Herlofsky (eds.), Outside-in –
inside-out: Iconicity in language and literature 4. Amsterdam: Benjamins, 307-322.
Steins, Carsten. 2000. How to account for non-concatenative phenomena in a morpheme-based
theory. In: Barbara Stiebels & Dieter Wunderlich (eds.), 105-122.
Stiebels, Barbara. 2000. Linker inventories, linking splits and lexical economy. In: Barbara Stiebels
& Dieter Wunderlich (eds.), 211-245.
Stiebels, Barbara. 2002. Typologie des Argumentlinkings: Ökonomie und Expressivität. Berlin:
Akademie Verlag.
Stiebels, Barbara. 2003. Transparent, restricted and opaque affix orders. In: Uwe Junghanns & Luka
Szucsich (eds.), Syntactic Structures and Morphological Information. Berlin: Mouton de
Gruyter, 283–315.
Stiebels, Barbara. 2006. From rags to riches: nominal linking in contrast to verbal linking. In: Dieter
Wunderlich (ed.), Advances in the theory of the lexicon. Berlin: Mouton de Gruyter, 167-234.
Stiebels, Barbara. 2007. Towards a typology of complement control. In: Barbara Stiebels (ed.),
Studies on complement control. ZAS Papers in Linguistics 47, 1-80.
Stiebels, Barbara & Dieter Wunderlich (eds.). 2000. Lexicon in focus. Berlin: Akademie Verlag.
Stokoe, W. C., Jr. 1960. Sign language structure. Studies in Linguistics Occasional Papers 8.
Buffalo: University of Buffalo Press.
Stump, Gregory T. 1993. Position classes and morphological theory. In Geert Booij & Jaap van
Marle (eds.), Yearbook of Morphology 1992. Dordrecht: Kluwer, 129-180.
Tai, James H.-Y. 1985. Temporal sequence and Chinese word order. In: John Haiman (ed.), 49-72.
Trommer, Jochen. 2003. The interaction of morphology and syntax in affix order. In Geert Booij &
Jaap van Marle (eds.), Yearbook of Morphology 2002. Dordrecht: Kluwer, 283-324.
Tuggy, David. 1988. Náhuatl causative/applicatives in Cognitive Grammar. In: Brygida Rudzka-
Ostyn (ed.), Topics in Cognitive Linguistics. Amsterdam: Benjamins, 587-618.
Uhrbach, Amy. 1987. A formal analysis of reduplication and its interaction with phonological and
morphological processes. Ph.D. dissertation, University of Texas.
Ultan, Russell. 1978. Size-sound symbolism. In: Joseph Greenberg (ed.), 525-568.
Urbanczyk, Suzanne. 1996. Patterns of Reduplication in Lushootsheed. Ph.D. dissertation,
University of Massachusetts-Amherst.
Ussishkin, Adam P. 2000. The Emergence of Fixed Prosody. Ph.D. dissertation, University of
California-Santa Cruz.
Ussishkin, Adam P. 2005. A fixed prosodic theory of nonconcatenative templatic morphology.
NLLT 23, 169-218.
Van Valin, Robert D. Jr. 1993. A Synopsis of Role and Reference Grammar. In: Robert D. Van
Valin (ed.), Advances in Role and Reference Grammar. Amsterdam: Benjamins, 1–164.
Voeltz, F. K. Erhard & Christa Kilian-Hatz. 2001. Introduction. In F. K. Erhard Voeltz & Christa
Kilian-Hatz (eds.), 1-8.
Voeltz, F. K. Erhard & Christa Kilian-Hatz (eds.). 2001. Ideophones. Typological Studies in
Language 44. Amsterdam: John Benjamins.
Wälchli, Bernhard. 2005. Co-compounds and natural coordination. Oxford: Oxford University
Watkins, Laurel J. 1984. A grammar of Kiowa . Lincoln, NB: University of Nebraska Press.
Watson, Richard L. 2001. A comparison of some Southeast Asian ideophones with some African
ideophones. In F. K. Erhard Voeltz & Christa Kilian-Hatz (eds.), 385-405.
Watters, David E. 2002. A grammar of Kham. Cambridge: Cambridge University Press.
Weeda, Donald Stanton. 1992. Word truncation in Prosodic Morphology. Unpublished Ph.D. thesis,
University of Texas at Austin.
Wierzbicka, Anna. 1985. “oats” and “wheat”: the fallacy of arbitrariness. In: John Haiman (ed.),
Wierzbicka, Anna. 1995. Adjectives vs. verbs: the iconicity of part-of-speech membership. In:
Marge E. Landsberg (ed.), 223-245.
Wilbur, Ronnie B. 2005. A reanalysis of reduplication in American Sign Language. In Bernhard
Hurch (ed.), 595-623.
Wunderlich, Dieter. 1993a. Diathesen. In: Joachim Jacobs, Arnim von Stechow, Wolfgang
Sternefeld & Theo Vennemann (eds.), Syntax. Ein Internationales Handbuch zeitgenössischer
Forschung. Vol. I. Berlin: Mouton de Gruyter, 730–747.
Wunderlich, Dieter. 1993b. Funktionale Kategorien im Lexikon. In: Frank Beckmann & Gerhard
Heyer (eds.), Theorie und Praxis des Lexikons. Berlin: Walter de Gruyter, 54-73.
Wunderlich, Dieter. 1996. Minimalist morphology: the role of paradigms. In: Geert Booij & Jaap
van Marle (eds.), Yearbook of Morphology 1995. Dordrecht: Kluwer, 93-114.
Wunderlich, Dieter. 2001. Prelexical syntax and the Voice hypothesis. In: Caroline Féry &
Wolfgang Sternefeld (eds.), Audiatur Vox Sapientiae. A Festschrift for Arnim von Stechow.
Berlin: Akademie Verlag, 487-513.
Wunderlich, Dieter. this volume. Polarity and constraints on paradigmatic distinctness.
Wunderlich, Dieter & Ray Fabri. 1995. Minimalist Morphology: An approach to inflection.
Zeitschrift für Sprachwissenschaft 14, 236-294.
Wurzel, Wolfgang Ullrich. 1989. Inflectional morphology and naturalness. Dordrecht: Kluwer.
Zavala, Roberto. 2006. Serial verbs in Olutec (Mixean). In: Alexandra Y. Aikhenvald & R. M. W.
Dixon (eds.), Serial verb constructions: a cross-linguistic typology. Oxford: Oxford University
Press, 273-300.
Zeevat, Henk & Gerhard Jäger. 2002. A reinterpretation of syntactic alignment. In: Dick de Jongh,
Henk Zeevat & Marie Nilsenová (eds.), Proceedings of the 3rd and 4th International
Symposium on Language, Logic and Computation, ILLC, Amsterdam.
Since (Zipf, George Kingsley. 1935. The psychobiology of language: An introduction to dynamic philology . Cambridge, MA: MIT Press; Zipf, George Kingsley. 1949. Human behavior and the principle of least effort. Journal of Consulting Psychology 13(3)), it has been known that more frequent lexical items tend to be shorter than less frequent ones, and this association between the length of an expression and its frequency has been applied to various grammatical patterns (syntactic, morphological, and phonological) and related to predictability or expectedness in the typological literature. However, the exact interactions of frequency and expectedness, their effect on shortening, and the mechanisms involved, are still not well understood. This paper proposes the Form-Expectedness Correspondence Hypothesis ( fech ), taking into account not only the frequency of expressions but their overall structure and distribution, and explores the fech in the domain of nominal inflection from a quantitative perspective.
When communicating speakers map meaning onto form. It would thus seem obvious for languages to show a one-to-one correspondence between meaning and form, but this is often not the case. This perfect mapping, i.e. transparency, is indeed continuously violated in natural languages, giving rise to zero-to-one, one-to-many, and many-to-one opaque correspondences between meaning and form. However, transparency is a mutating feature, which can be influenced by language contact. In this scenario languages tend to evolve and lose some of their opaque features, becoming more transparent. This study investigates transparency in a very specific contact situation, namely that of a creole, Haitian Creole, and its sub- and superstrate languages, Fongbe and French, within the Functional Discourse Grammar framework. We predict Haitian Creole to be more transparent than French and Fongbe and investigate twenty opacity features, divided into four categories, namely Redundancy (one-to-many), Fusion (many-to-one), Discontinuity (one meaning is split in two or more forms,) and Form-based Form (forms with no semantic counterpart: zero-to-one). The results indeed prove our prediction to be borne out: Haitian Creole only presents five opacity features out of twenty, while French presents nineteen and Fongbe nine. Furthermore, the opacity features of Haitian Creole are also present in the other two languages.
Constructional approaches to morphology and syntax are based on the idea that the Saussurean sign is not only a powerful device for modeling the relationship between the form and meaning of morphemes, but, if appropriately adapted, it can be usefully extended to any kind of morphological and syntactic structure. Such approaches have been shown to be able to effectively account for a wide range of morphosyntactic phenomena, but an underexplored area is how different kinds of signifiers become associated with both lexical and constructional meanings. This article considers this issue by exploring the range of variation found in the shapes of signifiers in morphological constructions. A particular focus will be signifiers that deviate from a canonical linear ideal and the role of templates in constraining the realization of signifiers. The kinds of meanings that specific kinds of signifiers can be associated with in signs will also be briefly considered. The primary goal of this article is to establish the study of possible signifier shapes as an important issue for Construction Morphology. It will also be argued that constructional approaches are especially well suited for analyzing generalizations holding among the signifiers in a given language.
This paper presents a study on the form-meaning relationship within a subpart of the lexicon, namely lexical reduplication. It compares the semantic classification of lexical reduplication in diverse languages, focusing on three quantitative studies on the distributions of semantic categories in three typologically and genetically unrelated languages, Bikol, Tibetan and Arabic. On this basis, it is argued that the common view that lexical reduplications are a “semantically arbitrary class” and as such irrelevant for studies on reduplication is not justified. It is in contrast claimed that the mapping of certain meanings on this specific phonological pattern is highly iconic, and that these word forms can be classified as a specific subgroup of expressives. This group even shows remarkable parallels with the prototypical meanings of morphological reduplication. An explanation for this – certainly controversially discussed – observation could be provided by the cognitive semiotic concept of “cross-modal iconicity” (Ahlner & Zlatev 2010).
This article advances a first systematic Functional Discourse Grammar (FDG) treatment of reduplication. Building on cross-linguistic arguments for reduplication's iconic motivation and non-concatenative derivational nature, principled advantages of FDG's functional-typological orientation over formal reduplicative models are programmatically demonstrated: Reduplication is differentiated from repetition in FDG's architecture, the basics for implementing reduplicative iconicity into the model are outlined, and several formalizations of the process based on existing FDG work on morphological derivation are suggested. Phonological characteristics of reduplication are mentioned briefly and mostly left for future FDG research.