A Cognitive Semiotic Approach To Sound Symbolism

Sign Systems Studies 38(1/4), 2010
Cross-modal iconicity:
A cognitive semiotic approach to sound symbolism
Felix Ahlner, Jordan Zlatev
Centre for Languages and Literature, Lund University
Box 201, 221 00, Lund, Sweden
e-mail: felix.ahlner@gmail.com, jordan.zlatev@ling.lu.se
Abstract. It is being increasingly recognized that the Saussurean dictum of the

arbitrariness of the linguistic sign is in conflict with the pervasiveness of the
phenomenon commonly known as sound symbolism. After first presenting a
historical overview of the debate, however, we conclude that both positions have
been exaggerated, and that an adequate explanation of sound symbolism is still
lacking. How can there, for example, be (perceived) similarity between expressions
and contents across different sensory modalities? We offer an answer, based on
the Peircian notion of iconic ground, and G. Sonessons distinction between
primary and secondary iconicity. Furthermore, we describe an experimental study,
in a paradigm first pioneered by W. Khler, and recently popularized by V.
Ramachandran, in which we varied vowels and consonants in fictive word-forms,
and conclude that both types of sounds play a role in perceiving an iconic ground
between the word-forms and visual figures. The combination of historical
conceptual analysis, semiotic explication and psychological experimentation
presented in this article is characteristic of the emerging paradigm of cognitive
semiotics.
Cross-modal iconicity: A cognitive semiotic approach to sound symbolism 299
1. Introduction
Is the relationship between the expression and content poles of the

linguistic sign fundamentally arbitrary, as it is typically claimed,
following the famous dictum of the father of modern linguistics (see
Lyons 1968), Ferdinand de Saussure, or is there some kind of natural
connection between the two? This question dates back to antiquity,
but we would claim is still not satisfactorily resolved. Even the
most committed proponents of the arbitrariness dictum recognize the
existence of onomatopoetic words, but tend to regard these as
primitive curiosities, existing outside and alongside the language
system proper. The alternative viewpoint, commonly known as sound
symbolism (Hinton et al. 1994), has gained ground during the last few
decades. For example, a result that has been replicated a number of
times is that when both adults and children (without autism) are given
two fictive words like bouba and kiki and asked to decide which one
denotes a roundish and which a pointy figure, they agree up to 95%
that bouba suits best the roundish one (for example, Ramachandran,
Hubbard 2001). How are we to explain this, and is it possible to find a
dialectical synthesis that could help resolve the debate between
arbitrariness and sound symbolism? This is the main question that
we address in this article.
We proceed in three steps. First, we present a historical and
theoretical overview of the debate, involving theoretical positions in
linguistics and semiotics, and relate these to relevant empirical
findings. This review shows surprising conceptual shortcomings in the
discussion, including vague and inconsistent definitions of key terms
such as arbitrariness, sound symbolism and iconicity. Hence, in a
second step we endeavor to present a consistent semiotic analysis of
the (possible) relation of resemblance between linguistic sound
patterns and the objects that they denote, allowing us to make sense of
results such as those mentioned in the previous paragraph. Such
exercises of conceptual analysis and explication are necessary, but not
300 Felix Ahlner, Jordan Zlatev
sufficient to understand the phenomenon in detail. Therefore, in the

spirit of the emerging paradigm of cognitive semiotics (see Zlatev
2009), integrating methods and theories developed in the disciplines
of cognitive science with methods and theories developed in semiotics
and the humanities1, we turn in our third step to the description of a
psychological experiment aiming to decide whether the relevant
(cross-modal) similarity between linguistic expression and content is
found at the level of segments (vowels and consonants) or in the
combination of segments, and possibly in whole sound patterns.
While by no means claiming that this three-step exercise will
provide the resolution to the age-old debate, we will suggest that it
casts it in a partially new light, with the potential for a synthesis
between (aspects of) positions that have been usually regarded as
incommensurable: the conventionality and motivatedness of lexical
meaning. We will conclude by suggesting how our analysis, and in
particular the notion of cross-modal iconicity, can be productive for
future research.
2. Historical and theoretical background
2.1. Ancient Greece
As many big questions in Western thought, the debate between the

proponents of linguistic conventionalism vs. naturalism goes back to
ancient Greece. In fact, language was not the primary focus of the
ancients. The laws of society had for a long time been regarded as
immutable and natural, but if they could be regarded as determined by
convention, they could be questioned and debated, which was
fundamental for the idea of democracy (see Lyons 1968; De Cuypere
2008). Since language, in the form of public discourse, constituted the
1 www.cognitivesemiotics.com
means through which these discussions were carried out, it was

inevitable that the debate would also come to involve the proper
understanding of its essence.
Platos famous dialogue Cratylus, from around 400 BC, is the
principle reference for this early debate.2 The two main stances that
had developed up to that time were that language is either given by
nature (phsei) or by convention (thsei). In the dialogue, these are
represented by Cratylus and Hermogenes, respectively.3 The former
argues that a words sound-form is somehow naturally connected to
its meaning. Therefore, a particular sound-pattern can be deemed
more or less appropriate for expressing a particular meaning. The
latter argues that words are man-created and passed on through
generations, thus constituting the core of a conventional system of
communication: language. As a supporter of phsei, Cratylus presents
examples of how certain sounds are better suited for certain meanings,
since the movements of the tongue and mouth imitate or resemble
what the word means. For example, [r] is said to naturally express
rapidity and motion, and [o] roundness. The proposed sound-
meaning resemblances are then exemplified by appropriate words,
such as gngylon (round). Hermogenes then contests this by giving
examples of words containing the same sounds, but whose meaning is
quite different. In the end of the dialogue, the two debaters find
themselves in a stalemate; some sound forms seem indeed to re-
semble their meanings, while many others do not. They also agree
that in order to gain knowledge of reality [] one should study
reality as such, instead of taking recourse to the study of language.
(De Cuypere 2008: 14f)
Albeit far removed from the concerns of modern linguistics and
semiotics, Cratylus shows that the basic positions of the nature-
2 See Sedley (2003) for a recent comprehensive monograph on Cratylus.

3
Cratylus and Hermogenes were historical persons, but the opinions expressed
in the dialogue were ascribed to them by Plato.
convention debate were set early in Western thought. Several other

characteristics also seem to have lived on during the centuries. Firstly,
we can note the vague terms in which the phsei position tends to be
presented (natural connection, imitation). Secondly, the conventio-
nality of sound-meaning pairings tends to be construed as an anti-
thesis to there being any substantial resemblance (or other motivated
relation) involved; the step from here to arbitrariness is small. In fact,
conventionality and motivatedness are far from antithetical. We
return to this below, but let us here only mention the fact that
approximately half of the signs of American Sign Language, which is
undoubtedly a conventional semiotic system, are judged to be iconic,
that is, their shapes resemble their meanings (Woll, Kyle 2004).
Thirdly, the dialogues non-conclusive outcome has become a
recurrent theme; given the extremeness of the positions, combined
with vague formulations it is hard to see how it could have been
otherwise.
2.2. Saussure, structuralism and the arbitrariness

of the linguistic sign
As a key representative of both linguistic structuralism and semiotics,

Ferdinand de Saussure strived to move away from a nomenclatural
view of language, according to which words are attached to things and
ideas that exist independently of language and human thought.
Instead, the linguistic sign was claimed to involve not a link between
a thing and a name, but between a concept and a sound pattern
(Saussure 1959[1916]: 66).4 The linguistic sign thus constitutes one
4
As well-known, Saussures most famous publication Cours de linguistique
gnrale, derives from his students lecture notes based on Saussures lectures at
the University of Geneva 19061911. Current research shows that these may have
misrepresented his thinking in various respects (Bouissac 2010).
psychological entity whose two parts are intimately conjoined and

mutually dependent, each automatically evoking the other.
Saussure claimed that the primordial principle of the linguistic
sign is its arbitrariness. By this, he meant that a sound pattern and a
concept have no natural connection that motivates them to be
linked together in a linguistic sign. This is to be shown by observing
that sound-concept links differ between languages, for example, the
concept HORSE is linked to the sound patterns [hs] in English,
[val] in French, and [ku] in Czech. Neither sound pattern seems to
have any inherent equine qualities. Saussure concluded: Because the
sign is arbitrary, it follows no law other than that of tradition, and
because it is based on tradition, it is arbitrary (Saussure 1959[1916]:
74). One may wonder about the validity of the second sentence: does
cultural transmission (tradition) necessarily imply arbitrariness? The
conflation of the features conventionality and arbitrariness when it
comes to linguistic signs has been commonplace, both before and after
Saussure,5 but this hardly guarantees its correctness.
Another basic tenet of structuralism was that signs are delimited
and defined by their contrast with other signs in the specific language
system. Saussure (1959[1916]: 115f) mentions how the English word
sheep and the French mouton may have the same signification, but
whereas sheep contrasts with mutton in English, in French mouton can
denote both the living animal and its meat; hence sheep and mouton
have different meanings. Such language-specific structure, even more
prominent in grammar than in lexical semantics is indeed an
important characteristic of human languages. But is it also arbitrary?
Saussure, and many in his footsteps seemed to believe so but on
what ground? Finally, Saussure acknowledged the existence of
onomatopoetic words which seemed to contradict the principle of
5
Compare: Every existing form of human speech is a body of arbitrary and
conventional signs for thought, handed down by tradition from one generation to
another (Whitney 1867: 32).
arbitrariness, but downplayed their role by stating that such words

are never organic elements of a linguistic system. [] Moreover, they
are far fewer than is generally believed (Saussure 1959[1916]: 69f).
Over the course of the 20th century, counterarguments from
linguistic functionalists, based on a wealth of linguistic evidence, have
made many of these claims difficult to maintain. However, before
turning to these, it is important to remember that Saussures point of
view was in line with the linguistic consensus of the time. In 19th
century Europe, linguistics as a field was to a large extent occupied
with comparisons and reconstructions within the Indo-European
language family. An exaggerated belief in the methods of comparative
linguistics led linguists to believe that, through the application of
sound changes, human language could be traced back to its very origin.
In this pursuit, onomatopoeia was considered too crude to have been
able to give rise to the number and variety of related word roots existing
across the Indo-European languages. Max Mller called onomatopoeia
a chaotic anarchy, and in his popular Lectures on the Science of
Language he wrote that [the] onomatopic system would be most
detrimental to all scientific etymology and no amount of learning and
ingenuity displayed in its application could atone for the lawlessness
which is sanctioned by it (Mller 1864: 94). The historical changes
under study were the true natural growth of words, whereas
onomatopoetic words were [] artificial flowers, without a root. They
are sterile, and are unfit to express anything beyond the one object
which they imitate (Mller 1869: 361). Whitney drew the concise
conclusion that [there is no inner or] essential connection between idea
and word [] in any language upon earth (Whitney 1867: 32).
Not only did research focus almost exclusively on Indo-European
languages, many also considered these to be superior to other
languages. Inflecting languages such as Greek, German, Latin and
Sanskrit were thought to be the best tools for expressing human
thought, and they were thus placed on the highest stage of linguistic
development (see Losonsky 1999). As an example, Crowthers gram-
mar of Yoruba deemed the lack of inflection in the Yoruba verb a

negative character (Crowther 1852: 12). Joh. Bernhard Schlegel
made an important contribution in his description of the Kwa
language Ewe, but at the same time he noted that the language still
finds itself in an initial stage.6
These mutually reinforcing ideas led to what perhaps was the main
problem, namely that the linguists of that time worked with a skewed
database, in which the data came from only a fraction of the worlds
thousands of languages. The introduction to Saussures Course states
that one aim of linguistics should be to describe and trace the history
of all observable languages (Saussure 1959[1916]: 6), but the book
itself mentions only a handful non-Eurasian languages, none of which
merit any illustrative examples.
The 20th century saw a wealth of new branches of linguistics, and
most linguists have recognized the importance of a diverse database of
languages. However, since much of this development has taken place
within a Saussurean framework,7 the question of sound symbolism
[has been] reduced to the status of curious but irrelevant oddities in
modern linguistics in general (Hamano 1998: 211).
2.3. Cross-modal sound symbolism
In opposition to the dictum (sometimes approaching a dogma) of the

arbitrariness of the linguistic sign, the notion of sound symbolism was
used during most of the 20th century to highlight the existence of a
6
Schlegel 1857, quoted in http://ideophone.org/early-sources-on-african-
ideophones-schlegel/.
7 For example, in a recent textbook, Yule (2006: 10) boldly states It is generally
the case that there is no natural connection between a linguistic form and its
meaning. [] There are some words in language with sounds that seem to echo
the sounds of objects or activities and hence seem to have a less arbitrary con-
nection. English examples are cuckoo, CRASH, slurp, squelch or whirr. However,
these onomatopoeic words are relatively rare in human language.
diverse array of linguistic phenomena in which a sound unit such as a

phoneme, syllable, feature, or tone [can] go beyond its linguistic
function as a contrastive, non-meaning-bearing unit, to directly
express some kind of meaning (Nuckolls 1999: 228), or a direct
linkage between sound and meaning (Hinton et al. 1994: 1). This has,
however, been a minority position, and as Hamano (1998: 3) remarks
[discussion] of sound symbolism in natural languages often trigger
laymens curiosity but linguists skepticism.
The main argument in favor of sound symbolism is that resemb-
lances between sound patterns and meanings go beyond clear, but
admittedly peripheral, cases of onomatopoeia such as the English
words meow and bang. In this case there is indeed a degree of resemb-
lance between the words sound patterns and the sounds produced by
certain animals or inanimate objects. This is uncontroversial since
both expression and content involve the same sensory modality:
audition. However, the number of non-arbitrary signs in the languages
of the world would increase dramatically if similarities across moda-
lities can also be considered, or what is sometimes called synesthetic
sound symbolism (Hinton et al. 1994). Let us consider in order two
different kinds of phenomena that bare evidence for this: first existing
expressions in the lexica of a variety of languages, and then experi-
mental studies with fictive or unknown expressions, showing speakers
ability to be able to actively find such cross-modal correspondences.
2.3.1. Ideophones
A huge number of expressions that are claimed to display a non-

arbitrary relation between expression and meaning are attested in
languages on all continents, but are perhaps best known from West
Africa and East Asia. Different terms are used in different scholarly
traditions, but the most common ones are ideophones in Africanist
literature (Childs 1994), expressives for Austroasiatic languages (Diff-
loth 1976; Svantesson 1983), and mimetics for Japanese and Korean
(Hamano 1998; Sohn 1999; Ivanova 2006). Such terminological

profusion is confusing, since the phenomenon referred to appears to
be the same, as illustrated below. To avoid confusion, we will con-
sistently use the term ideophones in the remainder of this article.
As examples of ideophones, consider the Japanese expressions
koro-koro small object rolling continously and goro-goro large object
rolling continuosly (Kita 2008). We may observe several typical
features: (a) reduplication, complete or partial, is common though not
necessary characteristic, (b) there are often, but again not always,
contrasts in the sound-forms corresponding to differences in meaning,
(c) to the extent that there is a resemblance between sound-patterns
and meanings, this is at a rather abstract level: it is not the sound of
objects rolling that is being imitated, and the contrast between [k]
and [] can hardly be said to correspond to an auditory feature related
to size.
In fact, a class of Japanese ideophones (mimetics) sometimes called
psychomimes (Baba 2003) are even said to express mental states, and
it is far from clear how this can be regarded as a relation of resemb-
lance or imitation between sound patterns and meanings (see
Ikegami, Zlatev 2007). (1) shows one such example, borrowed from
Kita (2008: 31).
(1) Taro-wa sutasuta-to haya-aruki-o si-ta

Taro-TOP IDPH-COMP haste-walk-ACC do-PAST
Taro walked hurriedly
To consider a different language, (2) and (3) used by Dingemanse,

show ideophones in the Niger-Congo language Siwu spoken in Ghana.
The examples are taken from spontaneous discourse.8
8
Dingemanse, Mark. How to do things with ideophones: observations on the
use of vivid sensory language in Siwu. Paper presented at the SOAS Research
Seminar, June 3 2009, London.
(2) --fi mnymny

3SG-FUT-shine IDPH.sparklingly
It will shine sparklingly
(3) -to -tu kk it tskwtskw

3SG-PROG 3SG-cut fowl head IDPH.sawing
He is cutting off the fowls head (in a sawing or cutting manner)
Again, we notice features (a) and (c) listed above (Dingemanse does
not mention any contrasts in these particular examples), and we could
perhaps add two more: ideophones are rather hard to translate, and do
not lend themselves easily to paraphrases with other expressions.
Dingemanse, for example, describes Siwu ideophones as typically
evoking a sensory event as a whole rather than describing just one
aspect of it. For example, in (2) above, mnymny is not simply an
intensifying word such as very, and it can only be used with verbs
such as fi (sparkle, shine). Finally, when speakers of Siwu are asked
about their perceived function of such ideophones, they typically reply
that without them, speech is bland, and that they make stories more
interesting. Similar accounts of ideophones are also given by Japanese
speakers.
Several dictionaries of Niger-Congo and Bantu languages contain
thousands of entries for ideophones (Childs 1994). Diffloth (1976)
reports similar numbers from Austroasiatic languages. There are
dictionaries of Japanese especially devoted to ideophones, and they are
also found in many ordinary dictionaries (Ivanova 2006). One
preliminary conclusion is that ideophones are indeed not far fewer
than is generally believed, as Saussure claimed. Also the pure fact that
there could be dictionaries of such items, shows, firstly, that despite
what was suggested above, their meaning can be approximately
rendered through other linguistic expressions, and they are not as
sterile, [] unfit to express anything beyond the one object which
they imitate as Mller (1869: 361) claimed. Secondly and more

importantly, their possible presence in a dictionary shows that they are
truly conventional, that is, part of the common knowledge, shared by
the speakers of the respective languages (see Itkonen 2003, 2008;
Zlatev 2007). Since by all accounts they are not at the same time
arbitrary, ideophones clearly show the mistake of conflating conven-
tionality and arbitrariness.
At the same time, their non-arbitrariness is rather difficult to pin-
point. As shown above, the statement that there is a synesthetic
relation between expression and content across modalities is not
altogether clear. Which modalities? Is this a relation of similarity
(resemblance) and if so, along which dimensions can such cross-
modal similarity be established? Furthermore, since ideophones are as
was pointed out conventional, it is not obvious that any potential
similarity plays any cognitive role in their synchronic use. To address
these issues we need to turn to psychological evidence.
2.3.2. Evidence for on-line mapping between

sound patterns and meanings
Since the 1920s, numerous experiments have been performed to show

that when people are asked to match certain kinds of meanings with
fictive or unknown word forms, the outcome is significantly higher
than chance. Sapir (1929) investigated what he called phonetic
symbolism among 500 English speakers. He presented participants
with fictive word pairs differing only in their vowel for example, mil
and mal and two contrasting meanings to the words in each pair,
for example, small table and big table. The results showed that in
80% of the cases the participants assigned the word containing [a] to
the larger object, and [i] to the smaller one.
The same year, Khler (1929) introduced the fictive words takete
and baluma (later changed to maluma), which he predicted would be
matched with a pointy, angular figure, and a smooth, wavy figure,

respectively. His own experiments, and later those of others confirmed
Khlers hypothesis with results showing that speakers of different
languages do tend to match words and such figures far more often
than chance predicts (Nuckolls 1999; Westbury 2005; Hunter-Smith
2007). The phenomenon has been called shape symbolism.
The actual word pairs used in these experiments have varied, but in
general, one word contains voiceless stops [p, t, k] and closed un-
rounded vowels [i, e], and the other contains voiced stops [b, d, ],
sonorants [m, n, l, ] and open rounded vowels [u, o, ]. Firth (1957)
mentions kikeriki and oombooloo, and Ramachandran and Hubbard
(2001) introduced bouba and kiki, which have become the latest
standard. In their influential study, 95% of the participants matched
bouba with the round, wavy figure, and kiki with the pointy, angular
figure. The explanation of this phenomenon has, however, again been
given in rather vague terms: [] the representation of certain lip and
tongue movements in motor brain maps may be mapped in non-
arbitrary ways onto certain sound inflections and phonemic represen-
tations in auditory regions (Ramachandran, Hubbard 2001: 10). This
description is in line with Ramachandrans theory of the phenomenon
of synaesthesia, the involuntary linking of certain senses due to innate
cross-wiring. But as Ikegami and Zlatev (2007) observe, such shape
symbolism is qualitatively different from true synaesthesia in not
being involuntary: seeing certain figures is not accompanied auto-
matically with certain sound perceptions, or vice versa. Furthermore, a
condition for the success of such cross-modal mappings is that figures
and sound patterns can be found to differ along the same qualitative
dimension. In this particular case, this could be said to be the
dimension ROUNDNESS (or conversely POINTEDNESS). In the
studies performed by Sapir, it was SIZE.
In a related type of experiment, participants have been given two
contrasting words (smallbig, roundflat, etc.) in a familiar language
and two words in an unfamiliar language, and asked to attempt to
match these. Studies of this kind have been conducted since the 1930s,
involving several unrelated languages (Hunter-Smith 2007: 23f).
Brown et al. (1955) asked English speakers to match word pairs in
English with Chinese, Czech and Hindi. For all three unfamiliar lan-
guages, over half of the word pairs were matched correctly signifi-
cantly better than chance, sometimes by over 90% of the participants.
Two conditions for the success of such experiments could be
observed. First, participants are to match a pair of familiar words, with
a pair of unfamiliar ones. Second, the familiar words should contrast
along a given dimension (as noted above for shape symbolism) and
thus form antonym pairs, for example, smallbig, roundflat, bright
dark. For example, when Maltzman et al. (1956) performed a study in
which English speakers were asked to match antonyms in Croatian
and Japanese (both of which were unknown to the speakers), they
failed to perform this at levels higher than chance. When they were
also given the antonyms in English, they performed significantly better
than chance. In another study, Brackbill and Little (1957) did not use
antonyms, but rather a list of highly frequent words in Chinese,
Japanese and Hebrew. English speakers were asked to match given
word pairs across languages, but failed.
In the next section, we offer an explanation of these findings, but
let us first take stock.
2.4. From sound symbolism to cross-modal iconicity
As shown in this overview, the age-long debate on whether linguistic

signs are fundamentally conventional or natural has been troubled
with conceptual unclarity and unnecessary polarization, both of which
increased when conventionality was equated with arbitrariness at the
end of the 19th century, most famously by Saussure. On the one hand,
it is impossible to deny that linguistic signs are semiotic conventions,
or even norms, since they are commonly known and there are public
criteria for their correct application, as shown by the possibility of

describing them in dictionaries (Itkonen 2003, 2008). However, this
does not exclude that their origin may be due, at least in part, to non-
arbitrary criteria such as partial similarity between expression and
content, and that this similarity is maintained, at least for certain
(classes of) expressions and more so for some languages than others.
On the other side, proponents of sound symbolism, from Cratylus
onwards, seem to have overstated their claims. Even if some form or
degree of resemblance (between sound pattern and meaning) can be
established for some expressions, and more so in some languages than
others, such resemblance would not be sufficient for understanding
what the sound patterns mean, if the pairings were not also conven-
tional. Even the most sound-symbolic linguistic expressions are not
like realistic drawings (pictures), were the resemblance relation is
indeed sufficient, as we will explicate in the next section.
What seems to have been lacking is a consistent conceptual frame-
work in which to analyze the non-arbitrariness of (some) linguistic
signs. To this testifies the loose definitions of the phenomenon: it is
not sufficiently explained in which ways sound patterns are supposed
to imitate, directly express or resemble meanings, especially if this
is to take place across sensory modalities.
In the next section, we propose to provide the outlines of such a
conceptual framework, borrowing seminal concepts from the other
founding father of semiotics, C. S. Peirce, (the first being Saussure)
interpreted and further developed by G. Sonesson (1989, 2007, 2009b).
In such a perspective, the term sound symbolism becomes less
adequate, since for Peirce a symbol is a sign that expressly lacks any
motivated relation (ground) between expression (representamen)
and content (object). Since the professed relation between expression
and content observed in ideophones and cross-modal mappings
(reviewed in Section 2.3) is that of similarity (resemblance), a more
appropriate term for the phenomenon appears to be cross-modal
iconicity.
3. A semiotic analysis of cross-modal iconicity
3.1. Signs and their ground
Semiotics is usually characterized as the study of signs,9 but this is

hardly adequate since semiotics studies meanings of all kinds from
those inherent in direct perception and sensorimotor experience, to
cultural phenomena such as music and dance, to visually perceived
representations like pictures and gestures, to language, and its variety
of meanings, on the levels of prosody, lexis, grammar and discourse. If
all of these are characterized as signs, the concept becomes extensi-
vely broad (see Sonesson 1989, 2007). Hence, we find it better to
regard semiotics as the systematic study of meaning (Fuller 1997: 30;
compare Sonesson 2009a; Zlatev 2009). Within this study, the concept
of sign nevertheless retains a crucial place, and the question of how the
capacity to use signs emerged in evolution, and develops in ontogeny
is of vital importance (Donald 1991, 2001; Mandler 1996; Deacon
1997; Tomasello 1999, 2008; DeLoache 2004; Sonesson 2007, 2009a;
Zlatev 2007, 2008, 2009).
Peirce defined his central concept many times, but the following is
perhaps the most famous one:
[a] sign, or representamen, is something which stands to somebody for

something in some respect or capacity. It addresses somebody, that is, creates in
the mind of that person an equivalent sign, or perhaps a more developed sign.
The sign stands for something, its object. It stands for that object, not in all
respects, but in reference to a sort of idea which I have sometimes called the
ground of the representamen. (CP 2.229)
The sign (as a whole) can be said to involve at least three entities
interacting in the process of semiosis (see Fig. 1): a representamen
which stands for an object, to somebody (the interpreter). The
9
For example, Chandler, Semiotics for Beginners, http://www.aber.ac.uk/media/
Documents/S4B/
representamen gives rise to an effect in the mind of that person,

namely an interpretant (De Cuypere 2008: 30ff; Parker 1998: 144ff). It
is crucial that the representamen and the object are related only in
some respect or capacity and not in all respects; otherwise they
would be identical and inseparable. This basis of relating the represen-
tamen to the object is a fourth entity in semiosis, the ground.
Figure 1. A schematic illustration of semiosis (N.B. in order to simplify the

following figures, the interpretant will not be drawn).
Based on the nature of the ground that is, in what way a repre-
sentamen stands for an object Peirce divided signs into three so-
called ideal types, the first of which is of central importance for our
analysis. In an iconic sign, the ground is that of similarity, or more
precisely when the representamen and object are found to share
certain similar qualities independently of each other. The typical
example is a picture and its visual similarity to that which it depicts.
The second type is an indexical sign, where the ground is not based on
similarity, but on contiguity in time and space. The third type is the
symbolic sign. This differs from the earlier two by being based only on
convention, for example, using $ as a sign for US dollar. It is im-
portant to bear in mind that this taxonomy presents ideal types, and
that real-world signs usually contain properties from more than one of
the three, that is, more than one type of ground.
3.1.1. Primary and secondary iconicity
Since actual signs involve combinations of grounds, Sonesson (1997)

introduced an important distinction between iconic signs. In primary
iconic signs, the iconic ground is sufficient for establishing the sign,
and any possible conventional ground plays a secondary role. The
obvious example of this is again a drawing: while stylistic conventions
of drawing differ enormously, knowledge of them is not required in
order to recognize that a particular drawing depicts, say, a human face.
On the other hand, if the sign is established by other means by
convention, by pointing to a referent while uttering its name (its
representamen) or by simply telling the interpreter what something
means and any possible iconic ground is found by the interpreter
only later (if at all), this would be a case of a secondary iconic sign.
Primary iconicity: the perception of an iconic ground obtaining between

two things is one of the reasons for positing the existence of a sign
function joining two things together as expression and content. Secondary
iconicity: the knowledge about the existence of a sign function between two
things is one of the reasons for the perception of an iconic ground between
these same things. (Sonesson 1997: 741, our emphases)
Sonesson illustrates primary iconicity with pictures, and secondary

iconicity with so-called droodles: simple drawings whose ground be-
comes evident only once it is pointed out (Fig. 2). But the distinction is
perhaps even more relevant in discussing iconicity in language.
Figure 2. A primary iconic sign: a picture of an elephant, and a secondary iconic

sign (a droodle) showing (for example) a person playing trombone in a wardrobe,
or perhaps a paper clip placed under an envelope.
3.1.2. Iconicity in language
Jakobson (1971[1965]) first introduced the Peircean concept of

iconicity into linguistics, and over the last 30 years, it has become a
popular term in the cognitive-functional school of linguistics, whose
representatives have applied it in analyses of syntax and morphology
(Haiman 1980; Croft 2003; Itkonen 2004, 2005). At the same time,
there has been persistent disagreement on how to apply the notion.
For example, Croft (2003) claims that the principle one expression
one meaning, that is, the tendency to limit polysemy (one expres-
sion many meanings) and synonymy (several expression one
meaning) is a case of linguistic iconicity. Itkonen (2004) justifiably
objects that this is an overextension of the concept, since expressions
and meanings do not exist independently of one another (which, as
pointed out above, is a requirement for iconicity), but constitute the
two sides of the linguistic sign; else we would be back to the
nomenclature view of language criticized by Saussure.
Already Jakobson suggested that plural endings illustrate an iconi-
city of quantity: longer expressions correspond to a larger quantity.
On the other hand, Haspelmath (2008) objects that the phenomenon
is better explained through frequency of use: more commonly oc-

curring types of word (the ones in singular) have shorter forms since
this is simply more economical. There are problems to this analysis,
but if we accept it, we could use Sonessons distinction between
primary and secondary iconic signs to explicate it: to the extent that
singular/plural forms of nouns can be viewed as iconic to lower/higher
quantity, such an iconic sign is only of the secondary type; it is
established by convention, and the similarity between expression and
meaning (found by linguists) is epiphenomenal: it plays no role either
in synchronic interpretation or in diachronic origin of the plural
morphemes. Sonesson himself has made suggestions along these lines:
As for iconicity in language and in music, it most of the time seems to
be secondary (Sonesson 2009b: 51). De Cuypere (2008) argues for a
similar position.
However, if this were altogether true, it would effectively devalue
iconicity as an explanatory concept in linguistics. It would, for
example, make the iconicity of ideophones analogous to that of
droodles: possibly vicarious similarities that can be perceived only
after the signs have been established by convention. However, there is
something deeply unsatisfactory to this latter claim. On the one hand,
the sheer number of ideophones in at least some languages seems to
speak against it. Even more, it would make the on-line cross-modal
mappings that have been experimentally attested inexplicable. There-
fore, Ikegami and Zlatev proposed that there is indeed more than
secondary iconicity in language:
We would like to suggest that the distinction between primary and secondary
iconicity is more of a cline, defined by the degree to which the sign function
(i.e. knowing what an expression represents) is necessary for perceiving the
similarity involved in iconicity. From this perspective, the bouba/kiki pheno-
menon is somewhat intermediary in the cline. (Ikegami, Zlatev 2007: 270)
As it stands, however, this formulation is also unsatisfactory: it is not

really clear how phenomena such as shape symbolism can be
regarded as lying between primary and secondary iconicity, since the

two are defined in a complementary way. In the following, we provide
a possible resolution to this dilemma.
3.2. Cross-modal iconicity as a combination

of primary and secondary iconicity
Let us again consider the two types of experiments described in

Section 2.3.2, those involving the matching between fictive words and
figures, and those involving the matching between words in a familiar
language and an unfamiliar one and attempt to provide a semiotic
explication of their findings in terms of the notions discussed in the
previous sub-section.
3.2.1. Matching fictive words and figures
In the experiments carried out in the tradition inaugurated by Khler

and made famous by Ramachandran and colleagues more recently,
there are always two representamina (the fictive words such as bouba
and kiki) and two perceptual objects exhibiting some kind of contrast
(round vs. pointy figures, small vs. big objects, etc.). The interpreter
will succeed in performing the matching to the extent that he or she
can discern a similarity, that is, an iconic ground, between the
representamina and objects in order to create two signs. Schematically,
this can be presented as a sequence of three steps (see Fig. 3). Step 1
shows the initial conditions in the experiment with two separate sets
of representamina (R1, R2) and two objects (O1, O2). Step 2 shows
that two parallel grounds are discerned (dashed lines), constituting a
composite analogous ground: R1 is to O1 as R2 is to O2.10 On this

basis in Step 3, the linking of representamina and objects is established,
creating two (presumed) signs.
Figure 3. Matching fictive words to figures by finding a composite analogous

ground.
Is this to be regarded a matter of primary or secondary iconcity? We

propose that it is not one or the other, and not something in
between (as suggested by Ikegami and Zlatev (2007)), but rather a
combination of the two, as follows:
First, the interpreter is told that a combination of sign relations
exists between the pairs of representamina R1 and R2 and perceptual
objects O1 and O2. This is a precondition for discerning the com-
posite analogous ground in Step 2. That is, knowledge of the existence
of sign relationships, is one of the reasons for the perception of an
iconic ground, that is, conforming to the definition of secondary
iconicity, given earlier.
Once this analogous ground is perceived, however, it serves as the
basis for positing specific sign relations between R1O1 and R2O2 in
Step 3. That is, quoting again Sonessons definition of primary
10
One R-O ground would be, strictly speaking, sufficient for performing the
analogy if there are only two (presumptive) signs involved, as an anonymous
reviewer points out.
iconicity (see above): the perception of an iconic ground obtaining

between two things is one of the reasons for positing the existence of a
sign function joining two things together as expression and content.
Observe that the precondition for the success of the second step is
that the contrastive relationship between the two expressions
(representamina) on the one side is found to correspond to an
analogous contrastive relationship between the two objects. This kind
of second-order iconicity is characteristic of analogy (Itkonen 2005),
that is, similarity of relations rather than objects. It also presupposes
the characteristic feature of sign-to-sign contrast emphasized by
structuralism, though not the tenet of the arbitrariness of linguistic
sign. Therefore, we may conclude that the primary iconicity involved
in the process is doubly conditioned by, on the one hand, prior
knowledge that there are (yet unspecified) sign relations (secondary
iconicity), and the principle of linguistic contrast, on the other hand.
We will return to this in what follows.
3.2.2. Matching familiar antonyms with ones

in an unfamiliar language
In the case of matching two antonymic words R1 and R2 in a familiar

language (for example, large and small) to corresponding words R3
and R4 in an unfamiliar language, we have no perceptual objects, but
since the first two words are familiar, the interpreter will have recourse
to their meanings via the sign relationship and can imagine, for
example, a LARGE and SMALL object. This is shown as Step 1 in
Figure 4. In Step 2, due to secondary iconicity, an analogous ground is
discerned between representamina and objects. In Step 3, a similar
analogous ground is found, R3O1 and R4O2, and on this basis a
sign relationship is posited to exist between these (that is, primary
iconicity) in Step 4. Thereby R1 is matched to R3, and R2 to R4.
Figure 4. Matching familiar antonyms with those in an unknown language, by

finding a higher-order analogy between two composite analogous grounds.
The rate of success in this type of experiments was typically lower than
those in which fictive words were matched to figures,11 which can be
explained by the following two factors. First, the matching is per-
formed across sets of composite grounds, that is, a third-order relation.
Given this complexity, it is nevertheless remarkable that such experi-
ments meet success at all! Second, all word-forms (representamina)
are actual existing words which may or may not (for example, due to
11 Some antonyms give very high degrees of correct matching, while others
show results identical to chance. In some cases, different languages appear to have
reversed grounds, leading to results significantly worse than chance (Hunter-
Smith 2007).
historical sound change) have a degree of iconic ground characteristic

of fictive forms, explicitly chosen like bouba and kiki or mil and mal to
exhibit (cross-modal) similarity with corresponding figures. On the
other hand, this testifies that iconic grounds are in many cases
preserved in languages, and many signs are indeed not arbitrary. This
also suggests that cross-modal iconicity in language, far from being
epiphenomenal, can under appropriate conditions have cognitive con-
sequences, for example, in foreign language learning.
Again, however, one of the appropriate conditions is prior know-
ledge that what we have in the two languages are synonymous lin-
guistic signs: remember that experiments were unsuccessful when two
sets of representamina in unfamiliar languages were to be matched.
Therefore secondary iconicity (in Step 1) was a precondition for the
operation of primary iconicity (in Step 4). Also as pointed out in 2.3.2,
participants were able to solve the task only when O1 and O2 were
contrasted on the same quality dimension (SIZE, ROUNDNESS etc),
but not when expressions were picked on factors such as frequency.
Finally, the principle of inter-sign contrast was assumed here as well,
as part of the design of the experiment, and the instructions given to
participants.
3.3. Summary and further questions
In this section, we showed how moving beyond the vague formu-

lations of sound symbolism and adopting a semiotic framework
distinguishing between primary and secondary iconicity allowed us to
make sense of experimental results, and to confirm the cognitive and
linguistic reality of a degree of primary iconicity (in combination with
secondary iconicity and inter-sign contrast along uniform dimen-
sions). Furthermore, discerning the (composite) iconic grounds neces-
sary for performing tasks such as those that were here explicated
presupposes the ability to find similarities across the modalities of
hearing (and speaking) and at least the modality of vision.
Many questions, however, remain. How is such cross-modal map-

ping achieved? Do speakers perceive the form of their vocal apparatus,
haptically and propioceptively, for example as round in forming the
sound [o] when mapping to a roundish figure, or is the matching
performed on a deeper, more abstract or even amodal level? Is the
matching performed on the basis of segments (such as phonemes), or
even distinctive features, such as voiced/unvoiced, or is it rather a
matter of transitions between segments and whole sound patterns? If it
is a matter of segments, do vowels or consonants play a more deter-
mining role for establishing what was here called the iconic ground?
The conceptual explication offered here cannot help us to answer
these questions, and we need to turn to so-called empirical methods.
Since we are not aware of studies which have been designed explicitly
to address questions such as those given above, we performed a case
study of our own, which we describe in the following section.
4. A case study of cross-modal iconicity
4.1. Goals and method
We performed an experiment in the tradition of Khler (1929) and

Ramachandran and Hubbard (2001), described in Section 2.3.2, asking
Swedish participants to match fictive words and visual figures. In
contrast to earlier studies, we decided to utilize a larger range of word
pairs, and a more systematic comparison of different types of sounds:
vowels, consonants and combinations of these. The purpose of this
was to attempt to discern the relative importance of vowels and con-
sonants, as well as the patterns of these taken as a whole, for perceiving
an iconic (cross-modal) ground that would allow the participants to
correctly match the fictive words and figures, as analyzed in Section 3.
First, we selected maximal phonetic contrasts between consonants
and between vowels, based on the synaesthetic properties that have
been ascribed to such sounds in previous studies, and on the basis of

our own intuitions.
Consonants:
voiceless obstruents (hard, sharp, pointy)
vs.
voiced sonorants (soft, smooth, heavy):
[p, t, k, t] vs. [m, l, n, ]
Vowels:
front close unrounded (sharp, small)
vs.
back open (round, large):
[i] vs. [u]
Second, we defined four pair-types of fictive words forming simple

reduplicative patterns of the form CVCV, using these phonetic
contrasts in different ways. These constituted four different conditions.
Conditions (a) and (b) contrasted individual sounds only, while (c)
and (d) contrasted combinations of these, as follows:
a) Two words with different vowels, but the same sonorant

consonant, for example, lili vs. lulu.
b) Two words with different consonants, but with the same vowel
[i], for example, kiki vs. nini.
c) Incongruent combination: a word with a hard consonant and
round vowel was contrasted with a word with a soft consonant
and sharp vowel, for example, tutu vs. lili.
d) Congruent combination: a word with a hard consonant and a
sharp vowel was contrasted with a word with soft consonant
and round vowel, for example, titi vs. lulu.
The actual pair-types used in the experiment are given in Table 1 in

Section 4.3. The task for the participants was to match the words in
each pair of figures such as those shown in Figure 5, which were
similar to those used by Ramachandran and Hubbard (2001), that is, a
sharp pointy star-like figure, and a roundish amoeba-like figure.
Figure 5. The contrastive figures used in the experiment: the star and the
amoeba. All figures were of these types, but with slight variation, so that no parti-
cipant was shown two identical figures.
4.2. Hypotheses
We formulated a set of three conditional hypotheses, on the basis of

the contrasts between the four types of word pairs (conditions).
H1. If the iconic ground between sound patterns and figures is
above all discerned on the basis of contrast between soft and hard
consonants, then conditions (b), (c) and (d) should give rise to a high
rate of matching of the words with these consonants to the amoeba
and star figures, respectively, while condition (a) should lead to a
lower rate, or to no effect higher than chance.
H2. If the iconic ground is above all due to contrasts between
round and sharp vowels, then a higher rate of correct matching
should be found in conditions (a), (c) and (d), and a lower rate or no
effect at all in condition (b).
H3. If the iconic ground is due to the combination of consonants
and vowels, (and possibly the transitions between them), then the
congruent condition (d) should lead to the highest rate of correct
matching, followed by the cases in which there was contrast in only

vowels (a) and consonants (b), and least of all in the incongruent
condition (c).
Our expectation was that H3 would find the strongest support.
4.3. Participants and procedure
In the experiment, 20 participants (9 female, median age: 25) were

recruited among students at Lund University, and compensated with a
small sum, corresponding to a lunch ticket. The task was administered
through the software E-prime. The 20 participants were divided into
four groups of 5 (numbered 1 to 4 in Table 1). Each participant in the
4 groups was given a specific version of the software, which presented
the participant with one instance of each of the 4 word pairs
conforming to the conditions (a), (b), (c) and (d), without hearing the
same fictive word twice, as shown in Table 1.
Table 1. The spoken stimuli used in the experiment, ordered by word pair type (A
to D) and participant group (1 to 4).
A. Vowel B. Consonant C. Incongruent D. Congruent
C2V1 vs. C2V2 C1V1 vs. C2V1 C1V2 vs. C2V1 C1V1 vs. C2V2
1 lili lulu titi mimi pupu ii kiki nunu
2 mimi mumu pipi ii kuku nini titi lulu
3 ii uu kiki nini tutu lili titi mumu
4 nini nunu titi lili tutu mimi pipi uu
When the target figures (see Figure 5 above) were presented on the
screen, randomized for left-right order, a pre-recorded voice was
heard saying (in Swedish) a sentence conforming to the schema: Lets

say that one of these figures is called <Word X> and the other <Word
Y>, which one would you call <Word X/Word Y>? The fictive words
for the variables <Word X> and <Word Y> were taken automatically
from a list consisting of the word pairs given in Table 1, in a carefully
arranged way. The order in which the elements of the word pair were
played was randomized, so that the harder sound pattern would
sometime come first, and sometime second. Correspondingly, the
question of which sound pattern was to be explicitly matched with
which figure (and which left to be matched to the other by logical
inference), half of the time concerned <Word X>, and half of the time
<Word Y>. Participants had the possibility to repeat the spoken
stimuli as many times as they wanted, by pressing a certain key. In our
debriefing questions, however, none of the participants reported that
they had done this.
The order in which the conditions (a), (b), (c) and (d) were pre-
sented was also randomized, and the figures were slightly varied in
each of the four conditions, but always having a star-like figure and
an amoeba-like figure. Their relative left-right position was also
randomized, to fully eliminate any possible bias that could arise, for
example, from the left-to-right direction of writing in Swedish. The
response was to be given by pressing one of two keys on the keyboard.
Interspersed between these key tasks were 4 filler tasks. One
consisted of the participant hearing a short sound clip in a foreign
language, and then choosing from three alternatives which language it
was. The other tasks were logical questions, similar to those used in
IQ tests.
The purpose of avoiding presenting a given sound pattern (fictive
word) twice was that once the participant had matched a particular
sound to a figure, this was likely to affect the decision on a subsequent
trial, in which the fictive word would have participated in a different
contrast. For example, if a participants first word pair were lili vs. lulu,
and lili was matched with the star-like figure, and then the second
pair were lili vs. kiki, it is likely that the participant would have been
disinclined to give the name lili to the amoeba figure, despite the fact
that kiki may sound sharper to them in this context.
Each participant performed the experiment in a closed room,
sitting alone at the computer. On completion, participants were asked
a few debriefing questions in a short interview carried out by the first
author, in order to attempt to understand something about their
reasoning when performing the matching task. These interviews were
recorded, and key portions were subsequently transcribed.
4.4. Results
In order to evaluate the three hypotheses, the quantitative data was

analyzed by pooling the replies from all 20 participants for each of the
four conditions. The results are shown in Table 2.
Table 2. Results from the word-to-figure matching tasks, divided by conditions.
Condition star figure amoeba figure
A. Vowel contrast [i] 90% [i] 10%

[u] 10% [u] 90%
B. Consonant [p, t, k, t] 80% [p, t, k, t] 20%

contrast [m, l, n, ] 20% [m, l, n, ] 80%
C. Incongruent [u]+[p, t, k, t] 65% [u]+[p, t, k, t] 35%

combination of [i]+[m, l, n, ] 35% [i]+[m, l, n, ] 65%
V and C
D. Congruent [i]+[p, t, k, t] 90% [i]+[p, t, k, t] 10%

combination of [u]+[m, l, n, ] 10% [u]+[m, l, n, ] 90%
V and C
Statistical analysis (binominal distribution) of the results showed that

the results for conditions (a), (b) and (d) corresponded highly
significantly to the expected (correct) matching (obstruents and [i]
matched to the star figure, sonorants and [u] to the amoeba figure)
(p < 0.01). In the case of (c) there seemed to be a certain bias for
matching obstruents to star and sonorants to amoeba, despite the
fact that the vowel contrast should have pulled in the opposite
direction, as seen by the results in condition (a), but the bias for (c)
was markedly lower (p = 0.13), and not significantly higher than
chance.
Concerning H1, stating that the iconic ground would be perceived
above all on the basis of the consonant contrast, the results from
condition (b) and (d) were supportive, but those in (a) and (c)
contradictive. Hypothesis H2 was most clearly contradicted by the
results in both conditions (b) and (c), despite the fact that (a) showed
that even a contrast in the vowels alone was sufficient for a very high
rate of expected matching (90%). The strongest evidence that it was
not either vowels or consonants that serves as the ground for matching
to the figures was (c) the only condition in which the rate of
expected matching fell below significance level.
The results did not correspond entirely to H3, since the rate of
expected matching in condition (d) was identical to that in (a) and
nearly so in (b). But the finding that vowels and consonants play a role
on their own, and that they significantly lowered the matching rate in
our incongruent combination (c), allows us to conclude that H3 was
the hypothesis most clearly supported among the three. Therefore, we
may conclude that the results showed that both vowels and consonants
independently, and in combination, contribute to establishing the
iconic ground in cross-modal iconicity of the type investigated in the
study.
4.5. Discussion
While our relatively low number of participants seemed to give rise to

a ceiling effect of sorts leveling out the possible differences between
conditions (a), (b) and (d) it is not implausible to conclude that the
iconic ground was due to the whole sound pattern gestalt, rather than
to individual segments. Consider one example of word pairs from the
congruent condition (d). The sound patterns [kiki] and [mumu]
contain sounds whose articulations differ greatly: [k] requires non-
vibrating vocal chords, and a blocking of the vocal tract, followed by a
release of air. The vowel [i], on the other hand, requires a steady flow
of air from the lungs, passing through vibrating vocal chords. These
differences result in an abrupt transition between sounds in the sound
pattern [kiki]. This can be seen in the first spectrogram in Figure 6.
Inversely, the sounds [m] and [u] are very similar to each other,
leading to a less abrupt transition. Both [m] and [u] are pronounced
with vibrating vocal chords, and [m] involves a closing of the lips,
interfering minimally with the tongues position for [u]. In [mumu],
the lips are rounded and repeatedly pressed against each other, adding
to a haptic-proprioceptive softness of the sound pattern.
A broad qualitative analysis of the debriefing interviews showed
that participants had some awareness of the cross-modal similarities
between figures and sound patterns, but (unsurprisingly) did not
formulate this with sufficient meta-linguistic knowledge to allow us to
judge whether they focused on segments (vowels and consonants) or
on the whole patterns. The majority of the replies were such as those
given in (4)(7) (translated from Swedish), in which the participants
mention sounds more generally, while only (8) mentions conso-
nants and (9) is rather indeterminate.
Figure 6. Spectrograms of the fictive words [kiki] and [mumu], as used in the
experiment. The white part in the middle of [kiki] is silence (that is, a blocking of
the vocal tract before a release of air). These spectrograms show 05,000 Hz and
are 400 ms in duration.
(4) I associated certain sounds with certain shapes.

(5) One of the words sounded like something soft, and the other
as something hard.
(6) The figures looked like the sounds in their names.
(7) I felt how the words felt and sounded.
(8) Hard consonants are sudden sounds.
(9) Hard sounds are shorter. Drawn out sounds are softer.
Interestingly, few of the participants spontaneously mentioned

vowels, despite the fact that in Swedish schools, for reasons of
spelling rules, the vowels [i], [e], [y], [], and [] are usually called soft
vowels (mjuka vokaler), whereas [a], [u], [], and [o] are called hard
vowels (hrda vokaler). We were concerned that this fact could affect
the participants performance, and in the debriefing we explicitly

asked them whether they had thought about these terms during the
experiment, but none claimed to have done so.
When asked to exemplify what they meant by sharp and soft
sounds, participants replied with both consonants and vowels, in an
expected manner. Voiceless obstruents were given as examples of
hard and pointy sounds, whereas the voiced sonorants were
described with adjectives such as soft and round. Although the
vowels [i] and [u] were the only ones used in the experiment, parti-
cipants sometimes offered other vowels to express the difference
between sharp and soft sounds. These were vowels that usually
followed a back/front division, so that for example, [o] and [] were
considered soft, round vowels, and [e] and [y] were considered
pointy, hard vowels.
Concerning the still more difficult question how the participants
managed to establish the similarity between figures and sound patters,
the descriptions of the participants were not detailed enough to be able
to judge, though the reply given in (7) suggests that haptic (or pro-
prioceptive) perception of the vocal tract during actual pronunciation
was important. Indeed, some of the participants reported to have
pronounced the words to themselves before making their choices. For
the time being, we cannot offer more to the how question than the
somewhat speculative account offered by Ikegami and Zlatev,
describing the study performed by Ramachandran and Hubbard
(2001) using the expressions bouba and kiki:
If we start with the shapes, the cross-modal mapping between vision and touch
would allow them to be perceived as soft and sharp [respectively],
motivating the use of these quasi-synaesthetic metaphors as a natural way to
describe these figures. From the side of the expressions the production of the
velar stop /k/, even more so combined with the front, unrounded vowel /i/
involves obstructions and narrowings in the vocal tract, which can similarly be
perceived as sharp and edgy. On the other hand, the shape of the vocal tract
and the lips in the production of /u/ in bouba are quite literally roundish and
the passage of air is soft. The mappings between the senses Vision-Touch-
Proprioception-Sound in internal meaning space thus provides for a correspon-
dence between the shapes and the labels that would be impossible otherwise. A
robot, or a Martian with a very different kind of body [] would not be able to
perceive the iconicity involved. (Ikegami, Zlatev 2007: 225)
5. Overall summary and further research
We began this article by asking whether the cognitive semiotic

approach combining methods and theoretical concepts from the
humanities and cognitive sciences could help provide a dialectical
synthesis that could help resolve the debate between arbitrariness and
sound symbolism. The three steps that we took to address this
question are rather typical for this approach. First, to take a historical
perspective, looking at how the problem has been framed over the
centuries, as well as more recently, and what kinds of arguments and
evidence have been considered relevant. Second, to attempt to re-
frame the problem, in (preferably) clearer concepts, and thirdly, to
make this reframing amenable to empirical investigations, in order to
enter a cycle of cross-fertilization with more pronouncedly philo-
sophical perspectives, such as those of semiotics and phenomenology
(see Gallagher, Zahavi 2008; Sonesson 2007, 2009c; Zlatev 2009).
In the process, we found reasons to propose not a Grand Dia-
lectical Synthesis, but several more minor ones, on all of these three
levels. Let us here summarize our findings, relating, where appropriate,
to questions deserving future research.
5.1. The historical perspective
As pointed out when summarizing Section 3, both extreme sides in the

age-long (and continuing) debate have been in error: those claiming
that language is essentially arbitrary (at least on the level of simple
signs), and those claiming something similar to what the early Witt-
genstein famously stated: that language is a picture of reality. The
error of the first group, who we may (non-arbitrarily) call arbitrarians
was due in part to the conflation of the concepts of convention and
arbitrariness, and in part due to a skewed database of linguistic
evidence, disregarding non-European languages, and of course, the
signed languages of the deaf, which were only recognised (in some
quarters still grudgingly) as true languages some 40 years ago. The
error of the second group, the sound symbolists, was again due in part
to lack of adequate concepts: just remember the profusion of terms
used to describe the phenomenon: sound symbolism, shape
symbolism, phonetic symbolism, mimetics, expressives, ideopho-
nes, psychomimes, synaesthemes etc. in which sounds imitate,
directly express, or resemble meanings. Many sound symbolists also
show clear traces of a binary logic (that can still be observed in some
cognitive-functional quarters nowadays), contrary to that of the
arbitrarians: since language is not arbitrary, it must be essentially
motivated, with resemblance, or iconicity as a chief factor.
There have of course been many exceptions to such binary
opposition between the two extremes (for example Givn 2002), some
of these quite old, as can be seen in the following quote from Jespersen
(1964[1922]: 397):
Yes, of course it would be absurd to maintain that all words at all times in all
languages had a signification corresponding exactly to their sounds, each sound
having a definite meaning once for all. But is there really much more logic in the
opposite extreme, which denies any kind of sound symbolism and sees in our
words only a collection of wholly accidental and irrational associations of sound
and meaning?
But such voices have somehow been dwarfed over the ages, it seems.
With the widening of the linguistic database during the 20th century,
Saussures statement that sound-symbolic words are never organic
elements of a linguistic system, became increasingly hard to maintain.
Hinton et al. (1994) contains several descriptions of sound symbolic
systems in languages from all over the world, including less exotic
languages including English (Rhodes 1994; Oswalt 1994). But we sense
a new possible error: when sound symbolism can no longer be denied,
there are tendencies to present it as an ordinary linguistic parameter,
like word-order, or an areal feature, like the presence of noun classi-
fiers or click consonants. For example, Svantesson (1983) classifies
sound symbolic words in the Austroasiatic language Kammu as a
distinct word class, with its own morphological, syntactic and into-
national features. But why should symbolic words constitute a unique
word class in languages like Kammu and Japanese, but not in English
or Spanish? To regard it simply as a typological/areal feature is hardly
explanatory.
To some extent, there is indeed evidence that (clearly) sound
symbolic expressions are treated differently from those that are not, by
their speakers and by processes of language change. Dingemanse
(2009: 840f), for example, describes Siwu ideophones as displaying
deviant phonotactic patterns. Hamano (1998: 86) points out that
Japanese sound symbolic words have retained the word-initial sound
[p], which, through sound change, has otherwise become [h] in
Japanese. This can be seen as the result of a perceived connection
between the sound [p] and its sound symbolic connotations of an
abrupt and explosive movement. Even Jespersen (1964[1922]: 406)
mentioned that English cuckoo [kku] has not undergone the regular
sounds changes that would have resulted in [kku].
Kita (1997, 2001) has attempted to provide a theoretical account of
such a dual code system. On the basis of considerable empirical basis,
he argues that the meaning of Japanese ideophones (mimetics)
consists of an affect-imagistic dimension, where language has direct
contact with sensory motor and affective information (Kita 1997:
320), while non-mimetic expressions such as quantifiers, logical
operators, and the members of ordinary noun classes have their
meanings in an analytic dimension (compare Ikegami, Zlatev 2007).
There is a degree of appeal to such accounts, but again without

additional explanations, it is mysterious why certain languages would
rely heavily on the affect-imagistic dimension compared to others.
Furthermore, a strict separation between sound-symbolic and non-
sound symbolic expressions would seem to be contradicted by results
such as those reviewed in Section 2.3.2, where the meaning of ordi-
nary antonyms in unknown languages was in many cases guessed
correctly (when presented with corresponding familiar antonyms, and
asked to match the two pairs).
What is much more plausible, we propose, is an evolutionary expla-
nation. Iconicity is a key factor in the emergence of new expressions
as can be seen in signed languages (where new signs emerge from
gestures), or on the basis of iconicity not between sound and meaning,
but between the meanings of different expressions, as (computer) mouse.
At the same time, with conventionalization, the role of iconicity
diminishes as again seen in signed languages, where arbitrary signs
are deemed to be at least as many as iconic signs (Woll, Kyle 2004).
Centuries of writing and language standardization in European lan-
guages would have further contributed to this, and language change
provoked by language contact perhaps even more. Many of the
languages heavy on ideophones lack one or both of these characteristics:
writing/standardization and extensive language contact. From this
perspective, there is a substratum of sound-symbolism in all languages,
and the degree to which this is to be found may be due to factors such as
those here suggested.12 This is clearly a hypothesis that would deserve
further investigation.
12 Drawing from Ecuadorian Quechua, Nuckolls suggests another possible

factor, namely that ideophones thrive within animistic cultures (Nuckolls 2004:
132f). Her proposal is largely that people in animistic cultures use ideophones to
relate to the surrounding world much more profoundly than in Western non-
animistic cultures: I argue for a view of ideophones as a type of cultural
discourse through which speakers align themselves with nonhuman life forms
and forces of nature. This alignment is suggested by the special performative
5.2. The cognitive semiotic explication
By utilizing a cognitive semiotic framework, and in particular Sones-

sons (1997) notions of primary and secondary iconic signs, we believe
to have clarified the phenomenon of sound symbolism even more. But
here, as well, we were lead to a sort of dialectics. On the one hand, we
agree with Sonesson that out of context, sound symbolism would
appear to be a matter of secondary iconicity since few could under-
stand the meaning of even the most transparent ideophone on first
mention, without the help of context to unlock its key. But at least in
some cases in particular when provided with two contrastive
expressions and two contrastive meanings this secondary iconicity
becomes the ground for understanding previously unknown signs,
that is, becomes transformed, so to speak, into primary iconicity.
This analysis helped us make sense of the successful results of
studies of matching fictive words and visual figures, and of matching
known and unknown antonyms, and also to interpret the reasons
behind some of the unsuccessful ones. Somewhat surprisingly, the
analysis vindicated one of the key notions of structuralism, the
contrastive relation between linguistic signs. Just as meanings and
sounds can be defined by what they are not, the key condition for
unlocking the iconcity of fictive and real words (transforming it from
secondary to primary, as formulated in the previous paragraph) seems
to be to allow them to stand in a contrastive relationship, along a
single semantic dimension, as shown in Figure 7.
properties of ideophones, which collapse the distinction between a speech event

and a narrated event, thus compelling a speaker to become an action, event, or
process, in order to communicate about it (Nuckolls 2004: 131, our emphases).
Figure 7. The dimension of SMOOTHNESS, with the two extreme ends smooth
and sharp, constituting the basis for matching visual and phonetic contrasts.
This could explain why there is less primary iconicity, even with the
help of context, in (spoken) language than what one could assume
from the hypothesis laid out at the end of the previous subsection. Far
from all words in a language can be placed on one or another
dimension, as in Figure 7. Adjectives, or other expressions of quality,
are most susceptible to this, whereas nouns like horse obviously much
less so. Hence, the prediction would be that even if such a non-
gradable expression were initially strongly sound symbolic (iconic),
this iconicity would not be preserved with conventionalization and
historical change.
Indeed, Hinton and colleagues list a number of features that are
likely to be expressed by sound symbolic words, namely salient
characteristics of objects and activities, such as movement, size, shape,
color, and texture (Hinton et al. 1994: 10). Note that all of these are
qualities that are more or less gradient, that is, can be placed on
dimensions such as SPEED, SIZE, ROUNDNESS, SMOOTHNESS etc.
Dingemanse points out that sound symbolic words typically involve
characteristics that are discernable by the six senses (sight, touch,
hearing, interoception, taste, smell), as well as emotion, and as we
would expect from our emphasis on cross-modality, combinations of
these.13
13
Dingemanse, Mark What do we really know about ideophones? Paper pre-
sented at the 6th World Congress of African Linguistics, August 21, 2009, Kln.
Typological studies such as these, however, have their limitations,

since the processes in which the iconicity was established was some-
time in the (indefinite) past, and we have no way of knowing whether
it is alive enough in the minds of speakers, or only a historical relic.
This led us, as many others since Sapir (1929) and Khler (1929) to
experimentation.
5.3. Experiment
We found it lacking, especially from a linguistic perspective, that

previous experiments with fictive word matching to figures simply
picked two expressions that seems to their authors to be soft vs. hard
such as bouba and kiki. Hence, we chose our sound patterns carefully,
on the basis of systematic contrasts between vowels and consonants.
Thus, we could formulate the question, utilizing the semiotic termino-
logy introduced in Section 3: is it contrasts in vowels or in consonants
that mostly lead to the perception of an iconic ground between sound
patterns and figures? The dialectic answer was: both, and (most likely)
their combination. Still, it was rather surprising that when consonants
and vowels were combined incongruently from a sound-symbolic
perspective, there was a certain bias in the direction of the consonants.
Perhaps this is not surprising, since it is the obstruent consonants that
introduce the most salient contrast to all other segments, both the
vowels and the sonorant consonants (see Figure 6). If we had instead
used a less salient contrast for the consonants (for example, voiceless
vs. voiceless obstruents) as the basis for a possible ground for
matching to sharper vs. softer figures, the bias in favor of conso-
nants would probably have been diminished. Such studies are an
obvious further extension of our research.
Other extensions include the use of other dimensions such as SIZE
(close vowels to small figures), and more challengingly, COLOR
(different vowels are known to give rise to different colors in some
forms of synaesthesia, and the French poet, representative of the

school of symbolism, Rimbaud (1986) claimed a particular mapping:
I invented the colors of vowels: A black, E white, I red, O blue, U
green. But as we pointed out, synaesthesia is both involuntary, and
known to give rise to arbitrary associations due to cross-wiring
between the senses, which can be more or less idiosyncratic. COLOR
is not a single dimension but at least two dimensions: HUE and
LIGHTNESS. Whereas we would predict intersubjective agreement in
a task involving matching sounds to the LIGHTNESS dimension
(open vowels dark, close vowels light) it is much less obvious
what the ground would be for perceiving similarities between sounds
and dimension of HUE (the colors of the rainbow).
This discussion actualizes the point we made at end of Section 4: it
is still not completely clear how the ground for cross-modal iconicity
is perceived, that is, how we are able to establish similarities across
different modalities. Figure 7 gives the impression that there is an
abstract amodal dimension, in this case of SMOOTHNESS, between
the senses of vision and sound, but this is doubtful, to say the least. We
would prefer an account of step-by-step matching between senses, as
suggested in the quotation from Ikegami and Zlatev (2007), but as
admitted, this is still rather speculative. Cognitive semiotics, aided by
neuroscience, would hopefully make progress in this direction with
time. It is known that persons with lesions in the angular gyrus,
known to be involved in cross-modal integration, have difficulties with
performing the fictive words-to-figures matching task (Ramachandran,
Hubbard 2001), and more recently it has also been strongly suggested
that this also applies for people with autism (Oberman, Ra-
machandran, in press).
Lastly, we should mention the factor of writing. In designing the
experiment, we took two steps to minimize any potential role of
written language: the sound forms were presented orally, and their
order in the pairs randomized, as well as that of the figures, to avoid a
possible bias from the left-to-right direction of writing. But we can
nevertheless not exclude that (Swedish) writing may have played at

least some role in performing the task, 14 since spelling is not
completely arbitrary either. The letters in the Latin alphabet for
voiceless obstruents are indeed quite pointy, for example, <k> and
<t>. Similarly for the letter for the front close vowel <i>. On the other
hand, back open vowels tend to be round, <o> and <u>. Of course,
the mappings are not completely systematic, since English also uses
the arguably smoother <c> and <q> for the sound [k], which most
Romance languages use as their standard. Still, we have no guarantee
that such secondary iconicity could not be unlocked in some contexts.
Indeed, Westbury found an interaction between visual form and
phonology (Westbury 2005: 16) in two experiments using Latin
characters. It would be interesting to compare our results, performed
with Swedish speakers, with results from languages using a script that
does not show similar differences. In Burmese writing, for example,
nearly all letters contain a semicircle, see Table 3.
Table 3. Table showing approximate representation of the sound sequences

[bumo] and [kite] in different scripts.
Script [bumo] [kite]
Burmese
Korean
Gujarati
Georgian
Hiragana
14 For example by visualizing the words spelled when hearing them. In one of
the debriefing questions we asked whether the participants had done so. None
reported to have done so, but mysterious are the ways of the brain.
Another worthwhile comparison would be with children who have not

yet learned the (Swedish) writing system. In a study of this sort,
Maurer et al. (2006), found evidence for successful sound-figure
matching even in pre-literate (English-speaking) toddlers, which is
also what our cross-modal account would predict. But from what age
the phenomenon appears still remains to be investigated.
To conclude, we hope to have shown that our cognitive-semiotic
approach to sound-symbolism, and the notion of cross-modal
iconicity can contribute to going beyond the debate with arbitrariness,
and to help focus the discussion not so much on the existence of the
phenomenon, but on its nature. In doing so, we are bound to move
towards a more balanced and thus more adequate account of human
language, shaped through the dialectics of convention and iconicity.15
References
Baba, Junko 2003. Pragmatic function of Japanese mimetics in the spoken dis-
course of varying emotive intensity. Journal of Pragmatics 35: 18611889.
Bouissac, Paul 2010. Saussure: A Guide for the Perplexed. London: Continuum.
Brackbill, Yvonne; Little, Kenneth B. 1957. Factors determining the guessing of mea-
nings of foreign words. Journal of Abnormal and Social Psychology 54: 312318.
Brown, Roger W.; Black, Abraham H.; Horowitz, Arnold E. 1955. Phonetic
symbolism in natural languages. Journal of Abnormal and Social Psychology
50: 388393.
15 We want to thank the research program at Centre for Cognitive Semiotics

(CCS), Lund University, for providing the necessary intellectual environment for
conducting this research, as well as for economic support for carrying out the
empirical study. We wish also to express our gratitude to the Humanities Lab at
Lund University for supplying equipment for conducting the experiment. Special
thanks go to Joost van de Weijer for help with the statistical analysis. Finally, we
gratefully acknowledge Gran Sonesson, Lars-ke Henningsson, Timo Maran,
and an anonymous reviewer for critical feedback on an earlier version of this
article. We have followed their advice in many, but not all, respects and should
therefore be held responsible for any remaining shortcomings.
Childs, George Tucker 1994. African ideophones. In: Hinton et al. 1994: 178204.
CP = Peirce, Charles S. 19311958. Collected Papers of Charles Sanders Peirce.
Cambridge: Harvard University Press. [Hartshorne, Charles; Weiss, Paul
(eds.), 19311935; vols. 78. Burks, A. W. (ed.) 1958; In-text references are to
CP, followed by volume and paragraph numbers]
Croft, William 2003. Typology and Universals. Cambridge: Cambridge University
Press.
Crowther, Samuel Ajayi 1852. A Vocabulary of the Yoruba language, Together with
Introductory Remarks by the Rev. O. E. Vidal, M. A. London: Seeleys.
De Cuypere, Ludovic 2008. Limiting the Iconic: From the Metatheoretical
Foundations to the Creative Possibilities of Iconicity of Language. Amsterdam:
John Benjamins.
Deacon, Terrance 1997. The Symbolic Species: The Co-evolution of Language and
the Brain. New York: Norton.
DeLoache, Judy S. 2004. Becoming symbol-minded. Trends in Cognitive Sciences
8: 6670.
Diffloth, Grard 1976. Expressives in Semai. Austroasiatic Studies, Oceanic
Linguistics Special Publication 13: 249264.
Dingemanse, Mark 2009. Ideophones in unexpected places. In: Austin, Peter K.;
Bond, Oliver; Charette, Monik; Nathan, David; Sells, Peter (eds.), Proceedings
of Conference on Language Documentation and Linguistic Theory 2. London:
SOAS, 8397.
Donald, Merlin 1991. Origins of the Modern Mind: Three Stages in the Evolution of
Culture and Cognition. Cambridge: Harvard University Press.
2001. A Mind so Rare: The Evolution of Human Consciousness. New York:
Norton.
Firth, John Rupert 1957. Papers in Linguistics 19341951. London: Oxford Uni-
versity Press.
Fuller, Gillian 1997. Reading the readers: The role of semiotics in visitor research.
Visitor Behavior 12(34): 3031.
Gallagher, Shaun; Zahavi, Dan. 2008. The Phenomenological Mind: An Intro-
duction to Philosophy of Mind and Cognitive Science. London: Routledge.
Givn, Tom 2002. Bio-Linguistics. The Santa Barbara Lectures. Amsterdam:
Benjamins.
Haiman, John 1980. The iconicity of grammar: isomorphism and motivation.
Language 56: 515540.
Hamano, Shoko 1998. The Sound-Symbolic System of Japanese. Standord: CSLI
Publications.
Haspelmath, Martin 2008. Frequency vs. iconicity in explaining grammatical

asymmetries. Cognitive Linguistics 19: 133.
Hinton, Leanne; Nichols, Johanna; Ohala, John J. (eds.) 1994. Sound Symbolism.
New York: Cambridge University Press.
Hunter-Smith, Sarah 2007. Understanding without Babblefish: Reviewing the
Evidence for Universal Sound Symbolism in Natural Languages. Unpublished
thesis from Swarthmore College.
Ikegami, Takashi; Zlatev, Jordan 2007. From non-representational cognition to
language. In: Ziemke, Tom; Zlatev, Jordan; Frank, Roslyn M. (eds.), Body,
Language and Mind, vol 1: Embodiment. Berlin: Mouton, 241283.
Itkonen, Esa 2003. What is Language? A Study in the Philosophy of Linguistics.
Turku: Turku University Press.
2004. Typological explanation and iconicity. Logos and Language 5: 2133.
2005. Analogy as Strucure and Process. Amsterdam: Benjamins.
2008. The central role of normativity for language and linguistics. In: Zlatev,
Jordan; Racine, Timothy P.; Sinha, Chris; Itkonen, Esa (eds.), The Shared
Mind: Perspectives on Intersubjectivity. Amsterdam, Philadelphia: Benjamins,
279305.
Ivanova, Gergana 2006. Sound symbolic approach to Japanese mimetic words.
Toronto Working Papers in Linguistics 26: 103114.
Jakobson, Roman 1971[1965]. Quest for the essence of language. Selected Writings
II. The Hague: Mouton, 345359.
Jespersen, Otto 1964[1922]. Language, its Nature, Development and Origin.
London: George Allen & Unwin Ltd.
Kita, Sotaro 1997. Two-dimensional semantic analysis of Japanese mimetics.
Linguistics 35: 379415.
2001. Semantic schism and interpretive integration in Japanese sentences with
a mimetic: a reply to Tsujimura. Linguistics 39: 419436.
2008. World-view of protolanguage speakers as inferred from semantics of
sound symbolic words: A case of Japanese mimetics. In: Masataka, Nobuo
(ed.), The Origins of Language, Unraveling Evolutionary Forces. Tokyo:
Springer, 2538.
Khler, Wolfgang 1929. Gestalt Psychology. New York: Liveright.
Losonsky, Michael (ed.) 1999. Wilhelm von Humboldt: On Language: On the
Diversity of Human Language Construction and its Influence on the Mental
Development of the Human Species. New York: Cambridge University Press.
Lyons, John 1968. Introduction to Theoretical Linguistics. London: Cambridge
University Press.
Maltzman, Irving; Morrisett, L; Brooks L. O. 1956. An investigation of phonetic

symbolism. Journal of Abnormal and Social Psychology 53: 249251.
Mandler, Katherine 1996. Language in Cognitive Development. The Emergence of
the Mediated Mind. Cambridge: Cambridge University Press.
Maurer, Daphne; Pathman, Thanujeni; Mondloch, Catherine J. 2006. The shape of
boubas: Soundshape correspondences in toddlers and adults. Developmental
Science 9(3): 316322.
Mller, Max 1864. Lectures on The Science of Language. London: Longman, Green,
Longman, Roberts & Green.
1869. Lectures on The Science of Language, vol. 2. New York: Charles Scribner
and Company.
Nuckolls, Janis B. 1999. The case for sound symbolism. Annual Review of Anthro-
pology 28: 255282.
2004. To be or to be not ideophonically impoverished. In: Proceedings of the
Eleventh Annual Symposium about Language and Society Austin, Texas
Linguistic Forum 47: 131142.
Oberman, Lindsay; Ramachandran, Vilayanur S. (in press). Preliminary evidence
for deficits in multisensory integration in autism spectrum disorders: The
mirror neuron hypothesis. In: Pineda, Jaime A. (ed.), Role of Mirroring in
Social Cognition. Totowa, New Jersey: Humana Press.
Oswalt, Robert L. 1994. Inanimate imitatives in English. In: Hinton et al. (eds.)
1994: 293306.
Parker, Kelly A. 1998. The Continuity of Peirces Thought. Nashville: Vanderbilt
University Press.
Ramachandran, Vilayanur S.; Hubbard, E. M. 2001. Synaesthesia a window into
perception, thought and language. Journal of Consciousness Studies 8: 334.
Rhodes, Richard 1994. Aural images. In: Hinton et al. (eds.) 1994: 276292.
Rimbaud, Arthur 1986. Collected Poems. Penguin Classics. [Oliver Bernard, ed.]
Sapir, Edward 1929. A study in phonetic symbolism. Journal of Experimental
Psychology 12: 225239.
Saussure, Ferdinand de 1959[1916]. Course in General Linguistics. New York: The
Philosophical Library.
Schlegel, Joh. Bernhard 1857. Schlssel der Ewesprache, dargeboten in den
Grammatischen Grundzgen des Anlodialekts. Stuttgart.
Sedley, David 2003. Platos Cratylus. Cambridge: Cambridge University Press.
Sohn, Ho-Min 1999. The Korean Language. Cambridge University Press.
Sonesson, Gran 1989. Pictorial Concepts. Lund: Lund University Press.
1997. The ecological foundations of iconicity. In: Rauch, Irmengard; Carr,

Gerald F. (eds.), Semiotics Around the World: Synthesis in Diversity.
Proceedings of the Fifth International Congress of the IASS, Berkeley, June 12
18, 1994. Berlin & New York: Mouton de Gruyter, 739742.
2007. From the meaning of embodiment to the embodiment of meaning: A
study in phenomenological semiotics. In: Ziemke, Tom; Zlatev, Jordan; Frank,
Roslyn M. (eds.), Body, Language and Mind, vol. 1: Embodiment. Berlin:
Mouton, 85127.
2009a. New considerations on the proper study of man and, marginally,
some other animals. Cognitive Semiotics 4: 133168.
2009b. Prolegomena to a general theory of iconicity. Considerations on
language, gesture, and pictures. In: Willems, Klaas; De Cuypere, Ludovic
(eds.), Naturalness and Iconicity in Language. Amsterdam: John Benjamins,
4772.
2009c. The view from Husserls lectern. Cybernetics and Human Knowing, 16
(34): 107148.
Svantesson, Jan-Olof 1983. Kammu Phonology and Morphology. Travaux de
lInstitut de Linguistique de Lund 18. Lund: CWK Gleerup.
Tomasello, Michael 1999. The Cultural Origins of Human Cognition. Cambridge:
Harvard University Press.
2008. Origins of Human Communication. Cambridge: MIT Press.
Westbury, Chris 2005. Implicit sound symbolism in lexical access: evidence from
an interference task. Brain and Language 93: 1019.
Whitney, William D. 1867. Language, and the Study of Language: Twelve Lectures
on the Principles of Linguistic Science. New York: Charles Scribners Sons.
Woll, Bencie; Kyle, James 2004. Sign language. Encyclopedia of Language and
Linguistics. Oxford: Elsevier.
Yule, George 2006. The Study of Language. Cambridge: Cambridge University
Press.
Zlatev, Jordan 2007. Embodiment, language and mimesis. In: Ziemke, Tom; Zlatev,
Jordan; Frank, Roslyn M. (eds.), Body, Language and Mind, vol 1: Embodiment.
Berlin: Mouton, 297337.
2008. From proto-mimesis to language: Evidence from primatology and social
neuroscience. Journal of Physiology, Paris 102: 137152.
2009. The semiotic hierarchy: Life, consciousness, signs, language. Cognitive
Semiotics 4: 169200.
:
-
, -
-
.
, -

. , ,
-
?
.
. -
, .
. . -
-
, -
-
. -
,
, ,
.
Modaalsuste-vaheline ikoonilisus:
kognitiivsemiootiline lhenemine hlikusmboolikale
ha enam on mistetud, et Saussurei seisukoht keelemrgi arbitraarsest

loomusest on vastuolus hlikusmboolika nime all tuntud laialt levi-
nud fenomeniga. Prast levaate esitamist selleteemalisest ajaloolisest
debatist nitame, et mlemad positsioonid on liialdatud ja rahuldav
seletus hlikusmboolika nhtusele puudub jtkuvalt. Kuidas niteks on
vimalik tajuda sarnasust vljenduste ja sisu vahel erinevates sensoorsetes
modaalsustes? Oma vastuses sellele ksimusele tugineme Peircei mistele
ikoonilisuse alus ja G. Sonessoni eristusele primaarse ja sekundaarse
ikoonilisuse vahel. Lisaks sellele esitame oma empiirilist uurimust para-
digmas, mille esimeseks teerajajaks oli W. Khler ja mida hiljuti on

populariseerinud V. Ramachandran. Selles uurimuses varieerisime vo-
kaale ja konsonante vljameldud snavormides ning judsime jreldu-
sele, et nii konsonandid kui vokaalid mngivad olulist rolli snavormide
ja visuaalsete kujundite vahelise ikoonilisuse aluse tajumisel. Ajaloolise
kontseptuaalse analsi, semiootiliste seletuste ja pshholoogiliste eks-
perimentide kombinatsioon, nagu see on esitatud kesolevas artiklis, on
aluseks kujunevale kognitiivse semiootika paradigmale.

A Cognitive Semiotic Approach To Sound Symbolism

Uploaded by

Copyright:

Available Formats

A Cognitive Semiotic Approach To Sound Symbolism

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Cognitive Semiotic Approach To Sound Symbolism

Uploaded by

Copyright:

Available Formats

Sign Systems Studies 38(1/4), 2010

Abstract. It is being increasingly recognized that the Saussurean dictum of the

Is the relationship between the expression and content poles of the

sufficient to understand the phenomenon in detail. Therefore, in the

2. Historical and theoretical background

2.1. Ancient Greece

As many big questions in Western thought, the debate between the

means through which these discussions were carried out, it was

2 See Sedley (2003) for a recent comprehensive monograph on Cratylus.

convention debate were set early in Western thought. Several other

2.2. Saussure, structuralism and the arbitrariness

As a key representative of both linguistic structuralism and semiotics,

psychological entity whose two parts are intimately conjoined and

arbitrariness, but downplayed their role by stating that such words

mar of Yoruba deemed the lack of inflection in the Yoruba verb a

2.3. Cross-modal sound symbolism

In opposition to the dictum (sometimes approaching a dogma) of the

diverse array of linguistic phenomena in which a sound unit such as a

A huge number of expressions that are claimed to display a non-

(Hamano 1998; Sohn 1999; Ivanova 2006). Such terminological

(1) Taro-wa sutasuta-to haya-aruki-o si-ta

To consider a different language, (2) and (3) used by Dingemanse,

(2) --fi mnymny

(3) -to -tu kk it tskwtskw

they imitate as Mller (1869: 361) claimed. Secondly and more

2.3.2. Evidence for on-line mapping between

Since the 1920s, numerous experiments have been performed to show

matched with a pointy, angular figure, and a smooth, wavy figure,

2.4. From sound symbolism to cross-modal iconicity

As shown in this overview, the age-long debate on whether linguistic

criteria for their correct application, as shown by the possibility of

3. A semiotic analysis of cross-modal iconicity

3.1. Signs and their ground

Semiotics is usually characterized as the study of signs,9 but this is

[a] sign, or representamen, is something which stands to somebody for

representamen gives rise to an effect in the mind of that person,

Figure 1. A schematic illustration of semiosis (N.B. in order to simplify the

3.1.1. Primary and secondary iconicity

Since actual signs involve combinations of grounds, Sonesson (1997)

Primary iconicity: the perception of an iconic ground obtaining between

Sonesson illustrates primary iconicity with pictures, and secondary

Figure 2. A primary iconic sign: a picture of an elephant, and a secondary iconic

3.1.2. Iconicity in language

Jakobson (1971[1965]) first introduced the Peircean concept of

is better explained through frequency of use: more commonly oc-

As it stands, however, this formulation is also unsatisfactory: it is not

regarded as lying between primary and secondary iconicity, since the

3.2. Cross-modal iconicity as a combination

Let us again consider the two types of experiments described in

3.2.1. Matching fictive words and figures

In the experiments carried out in the tradition inaugurated by Khler

composite analogous ground: R1 is to O1 as R2 is to O2.10 On this

Figure 3. Matching fictive words to figures by finding a composite analogous

Is this to be regarded a matter of primary or secondary iconcity? We

iconicity (see above): the perception of an iconic ground obtaining

3.2.2. Matching familiar antonyms with ones

In the case of matching two antonymic words R1 and R2 in a familiar

Figure 4. Matching familiar antonyms with those in an unknown language, by