The status of the concept of 'phoneme' in psycholinguistics.
ABSTRACT The notion of the phoneme counts as a break-through of modern theoretical linguistics in the early twentieth century. It paved the way for descriptions of distinctive features at different levels in linguistics. Although it has since then had a turbulent existence across altering theoretical positions, it remains a powerful concept of a fundamental unit in spoken language. At the same time, its conceptual status remains highly unclear. The present article aims to clarify the status of the concept of 'phoneme' in psycholinguistics, based on the scientific concepts of description, understanding and explanation. Theoretical linguistics has provided mainly descriptions. The ideas underlying this article are, first, that these descriptions may not be directly relevant to psycholinguistics and, second, that psycholinguistics in this sense is not a sub-discipline of theoretical linguistics. Rather, these two disciplines operate with different sets of features and with different orientations when it comes to the scientific concepts of description, understanding and explanation.
- SourceAvailable from: Per Henning Uppstad[Show abstract] [Hide abstract]
ABSTRACT: Phonology has been a central concept in the scientific study of dyslexia over the past decades. Despite its central position, however, it is a concept with no precise definition or status. The present article investigates the notion of 'phonology' in the tradition of cognitive psychology. An attempt is made to characterize the basic assumptions of the phonological approach to dyslexia and to evaluate these assumptions on the basis of commonly accepted standards of empirical science. First, the core assumptions of phonological awareness are outlined and discussed. Second, the position of Paula Tallal is presented and discussed in order to shed light on an attempt to stretch the cognitive-psychological notion of 'phonology' towards auditory and perceptual aspects. Both the core assumptions and Tallal's position are rejected as unfortunate, albeit for different reasons. Third, the outcome of this discussion is a search for what is referred to as a 'vulnerable theory' within this field. The present article claims that phonological descriptions must be based on observable linguistic behaviour, so that hypotheses can be falsified by data. Consequently, definitions of 'dyslexia' must be based on symptoms; causal aspects should not be included. In fact, we claim that causal aspects, such as 'phonological deficit', both exclude other causal hypotheses and lead to circular reasoning. If we are to use terms such as 'phonology' and 'phoneme' in dyslexia research, we must have more precise operationalizations of them.Dyslexia 09/2007; 13(3):154-74. · 1.12 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: The Symbolic Grounding Problem is viewed as a by-product of the classical cognitivist approach to studying the mind. In contrast, an epigenetic interpretation of connectionist approaches to studying the mind is shown to offer an account of symbolic skills as an emergent, developmental phenomenon. We describe a connectionist model of concept formation and vocabulary growth that auto-associates image representations and their associated labels. The image representations consist of clusters of random dot figures, generated by distorting prototypes. Any given label is associated with a cluster of random dot figures. The network model is tested on its ability to reproduce image representations given input labels alone (comprehension) and to identify labels given input images alone (production). The model implements several well-documented findings in the literature on early semantic development; the occurrence of over- and under-extension errors; a vocabulary spurt; a comprehension/production asymmetry; and a prototype effect. It is shown how these apparently disparate findings can be attributed to the operation of a single underlying mechanism rather than by invoking separate explanations for each phenomenon. The model represents a first step in the direction of providing a formal explanation of the emergence of symbolic behaviour in young children.Connection Science 01/1992; 4:293-312. · 0.71 Impact Factor
Article: On defining 'dyslexia'.Scandinavian Journal of Educational Research 01/1995; 39(2):139-156. · 0.27 Impact Factor
The Status of the Concept of ‘Phoneme’
Per Henning Uppstad · Finn Egil Tønnessen
at different levels in linguistics. Although it has since then had a turbulent existence across
altering theoretical positions, it remains a powerful concept of a fundamental unit in spoken
language. At the same time, its conceptual status remains highly unclear. The present article
aims to clarify the status of the concept of ‘phoneme’ in psycholinguistics, based on the
scientific concepts of description, understanding and explanation. Theoretical linguistics has
provided mainly descriptions. The ideas underlying this article are, first, that these descrip-
tions may not be directly relevant to psycholinguistics and, second, that psycholinguistics
in this sense is not a sub-discipline of theoretical linguistics. Rather, these two disciplines
operate with different sets of features and with different orientations when it comes to the
scientific concepts of description, understanding and explanation.
The notion of the phoneme counts as a break-through of modern theoretical lin-
Keywords linguistic sign · arbitrariness · connectionism · functional linguistics
in psycholinguistics. In this context, several claims are made and argued for. First, we claim
that identifying the conceptual status of the phoneme is an enterprise of demarcation relative
to the scientific concepts of explanation, understanding and description. Second, we claim
that there is confusion concerning the linguistic sign with regard to these three concepts.
Third, we claim that these two claims are strongly related, and that they must both be inves-
tigated in order to clarify the conceptual status of the phoneme. The notion of the phoneme
a stronger focus on explanation. In the present article, developmental arguments are brought
into play to shed light on the key issue of explanation.
We consider the concept of understanding to be primarily about the immediate expe-
rience of a phenomenon, involving empathy. Understanding is not static, but may deepen
according to the principle of the hermeneutic circle and may interact with the concepts of
explanation and description. The concept of description is primarily about the relationship
between language and the reality referred to by means of language. For this reason, we can
talk of the hypothetical status of descriptions. The description is not objective, owing to a
lack of precision in the language and to restrictions in our perceptual abilities. Finally, the
elaborate a procedure for distinguishing between explanation, description and understanding
as regards the phoneme.
In the following, we first describe and analyse aspects of the phoneme as it is conceived
of in the recent history of linguistics, and then go on to question established ideas about the
linguistic sign which are used in linguistic explanations. Next, we discuss the concept of the
phoneme from the perspective of the concepts of description and explanation. And finally,
it is described how the phoneme should be related to the domain of description, and why
greater value should be placed on non-segmental descriptions in psycholinguistics.
The Phoneme: A Controversial Notion
The Blurred Notion of ‘Phonology’
We object to the abandonment of the notion of the phoneme: even though the distinc-
et al. 2002, p. 3)
The Jakobsonian position quoted above is interesting in at least two respects. First, it claims
the relevance of the notion of the phoneme at a time when major branches of phonological
theory are moving away from it; and second, it uses the term ‘language structure’ instead
of ‘phonology’, the reason being the blurring of the latter notion: ‘And we avoid the words
of “phonology” and “phonological”—terms which Jakobson himself had helped to launch
(Jakobson et al. 2002, p. 4).
While the Jakobsonian position remains structuralist, these two respects pinpoint two
pertinent problems concerning branches of contemporary research on speech, writing and
dyslexia: (1) the status of the concept of ‘phoneme’; and (2) the flawed conception of ‘pho-
focus on the first one; the second problem is dealt with at length in Uppstad and Tønnessen
Instead of stating that the phoneme has its place in language structure, one should ask:
Where does it have its place? Even though the notion of ‘phoneme’ has had a turbulent exis-
tence in linguistics over the past century, it is still a widespread conceptualization of sound
structure in interdisciplinary research fields involving speech and writing, as well as in the
pedagogical domain of written language. Instead of judging whether the phoneme exists or
be evaluated primarily with respect to the purpose it serves, and not as an ontological entity.
The Concept of ‘Phoneme’ in Linguistic Theory
In structuralism, the concept of ‘phoneme’ and the notion of ‘distinctive features’ came to
constitute the foundation for structuralist description of language in general. It is, no doubt,
a very powerful conception of sound structure, even though Jakobson et al. (1952) presented
a model of acoustic features which was a less abstract conception of sound structure than
traditional, phoneme-based structuralist theory. In generative phonology, however, the pho-
neme has been degraded to a bundle of highly abstract features (Chomsky and Halle 1968),
which involves less focus on the phonemes as units in favour of their constituent features.
The development of auto-segmental-metric theory can be said to have fulfilled the degrada-
tion or discarding of the phoneme. Chronologically, these changes at the level of theories
have moved towards a discarding of the phoneme, towards a non-segmental phonology.
Nevertheless, the different positions referred to here are all highly present in contemporary
What we find strange is that in discussions of the linguistic sign, the phoneme always
crops up, and the other conceptions of sound structure are omitted. It is a well-known fact
that the notion of ‘phoneme’ is strongly associated with the structuralist understanding of
the arbitrary linguistic sign. Given the new assumptions regarding sound structure, should
we perhaps reflect more on the notion of the linguistic sign when discussing the conceptual
locus of the phoneme?
The Linguistic Sign: Aspects of Autonomous Linguistics in Functional Grammar
including psycholinguistics, adheres to this view of language structure. However, most func-
tional approaches to language deny autonomous linguistics, by claiming that structure is a
product of cognition and communication. In the field of functional grammar, the inseparable
relationship between expression and meaning is used as the main argument. Still, when it
comes to the sign, to the lexicon, we may ask whether aspects of autonomous linguistics
are not actually maintained in functional approaches to language. For instance, in his func-
tional grammar, Givón states: ‘The peripheral sensory-motor codes of human language are
the domain of phonetics, phonology and neurology’ (Givón 2001, p. 11). He also holds on
to the Aristotelian doctrine of arbitrariness for the linguistic sign but, like most functional-
ists, he rejects the arbitrariness position for grammar: ‘It is true, of course, that Aristotle’s
doctrine of the arbitrariness of the linguistic sign—thus the arbitrariness of cross-language
diversity—referred only to the coding of concepts (“words”) by sounds and letters. But lat-
ter-day structuralists unreflectively extended the arbitrariness doctrine to grammar.’ (Givón
2001, p. 4). Most functionalists concern themselves with grammar, especially syntax; and
cognitive linguists deal in particular with the lexico-semantic level. Therefore, functionalists
seem to be occupied with issues of ‘iconicity’ or ‘the inseparability of form and function’
for these parts of grammar, but not for the linguistic sign. In other words, for functionalists
the linguistic sign has come about:
Just as autonomous linguistics distinguishes between a speaker’s purely linguistic
knowledge, determined by the language faculty, and his non-linguistic knowledge,
derived from pragmatic competence and the conceptual system, so autonomous pho-
the abstract linguistic system which is claimed to underlie the physical data. (Taylor
1991, p. 28)
In structuralism and generativism, this autonomy is maintained by the strict divisions of
langue–parole and competence–performance, respectively. This can also be seen in Saus-
sure’s position, even though his focus is on the relationship between expression and meaning
as similar to that between the two sides of a sheet of paper: ‘[le son] n’est pour [la langue]
qu’une chose secondaire, une matière qu’elle met en œuvre’ (Saussure et al. 1969, p. 164).
regard to arbitrariness. If language structure is the product of communication and cognition,
why are the issues of phonology and the linguistic sign treated differently?
The Phoneme in Functional Linguistics
In functional linguistics, the idea of the prototype is applied to the phoneme. Unlike in the
structuralist approach, the basic principle of description here does not come from phonol-
ogy. Rather, the principle of prototypes first gained success in semantics and has then spread
to other parts of linguistic description (Taylor 1991). That is probably why semantics and
syntax have attracted the most attention in functional grammar. According to the principle
of prototypes, category membership is a matter of varying distance from a prototype, with
no clear boundaries between categories. The idea of prototypes stands in contradiction to the
generative claim of either ‘marked’ or ‘unmarked’ features (binary values) as well as to the
‘necessary’ and ‘sufficient’ conditions of structuralist and generative theory. Interestingly, as
mentioned above, in structuralism—and in generative theory—, description by means of the
principle of necessary and sufficient conditions first achieved success in phonology and was
only later applied to other parts of linguistic description. This gave phonology a fundamen-
tal position, and this is probably one reason why the conception of necessary and sufficient
spective of functional linguistics, however, features are no longer binary—instead, category
membership is a matter of degree. This shift of perspective represents a threat to structuralist
and generative theory because it does not meet those theories’ need for logical coherence;
what we have instead is a fuzzier logic (Flanagan 1991; Tønnessen 1995). What is more, this
kind of logic is probably better suited to behavioural studies in that the phoneme as defined
the phoneme as defined by necessary and sufficient conditions. While the derivations made
in phonology certainly become more fuzzy and complex, the explanatory power is strongly
Another interesting aspect of the notion of ‘phoneme’ in functional linguistics is the
very fact that the cradle of functionalist theory is semantics, not phonology. Because of the
semantic focus, the phonemic properties of sound may be hypothesized and explored, and
not remain deadlocked in their role as foundation of the theory. In our view, this is a decent
starting-point in searching for the status of the concept of the phoneme in language structure.
As this starting-point has open ends, it meets the standards of good scientific method and
may yield new assumptions about structure.
whether we deal with explanation in any case. We do not think so. What we are dealing with
is first and foremost matters of description, even though some conceptions of the phoneme
are more relevant to the domain of explanation.
In the next section, we wish to investigate the notion of the arbitrary linguistic sign, keep-
The question to be answered is: With this conception of the linguistic sign, do we ever enter
the domain of explanation, or have we simply adopted aspects of description and dressed
them up as causal relationships? As we see it, we cannot discuss the phoneme without also
discussing the linguistic sign. Maybe there has been too much talk of units, and to little talk
of the linguistic sign? And finally, in all this reasoning about the phoneme, we must ask: Are
we dealing with matters of explanation or description?
On Defining ‘Language’ in Psycholinguistics
The discipline of psycholinguistics arose from the discipline of linguistics, and the purpose
was to separate the study of psychological aspects of language from the study of language
structure. This is, in our view, an unfortunate separation of what is psychological and not,
and it contains the implication that these different aspects can be studied in their own right.
We believe that this distinction is artificial and that there should be a definition of ‘language’
which goes beyond this artificial distinction. In the following, we will define ‘language’ as
‘a set of codes with potential for meaning’. In doing so, we are not confined by a represen-
tational view of language. Instead, we define language by performance—as a set of codes
with potential for meaning. As regards the term ‘potential’, it should be underscored that this
does not mean that performance lacks meaning, nor that performance has a defined, original
meaning, and nor that there is a final, ideal meaning to be reached. Instead, it is underscored
that meaning is dynamic and indeterminate, and is realized in overt language behaviour (see
Rommetveit 1974). On this view, which is close to Quine’s position, there is no isolated
semantic representation which is associated with a phonological representation at a certain
point in the language development of children and adolescents, but rather a progression of
association from the time of birth, building performance as potential for meaning. And this
ings. In accordance with this conception of language, we will talk of ‘traces of meaning’ in
order to underscore that meaning is realized in multiple ways at every point in development.
The Linguistic Sign in Psycholinguistics
The Arbitrary Linguistic Sign Serves the Domains of Description and Understanding
grounds of lexical contrasts: minimal pairs. This position is problematic because it does not
include an explanation of how contrasts came about in the history of a particular language or
for the acquisition of a sound structure. This is unfortunate, and the fallacy of this reasoning
is made clear by the early Jakobson’s claim that the young child’s babbling is irrelevant to its
phonological development (Jakobson 1941). This is exactly where the problem arises: how
can one talk of phonemes as lexical contrasts and still consider the babbling as relevant?
Owing to the dominance of structuralist and generative theory (and cognitive psychology)
over the past decades, this problem has not been clearly stated, which is a consequence of
these theories’ incapability to explain development and change in psycholinguistics. It is
often claimed that language is symbolic and that the vocabulary spurt in children is a result
of symbolic processing (Clark 1993). This is probably true, but probably also far from being
the whole truth.
The rest of the truth may lie in the problem that can be identified in the relationship
between lexical (or symbolic) and sub-lexical (or sub-symbolic) processes. If phonological
structure and the linguistic sign are defined solely at a lexical (symbolic) level, we may
ask what has happened to phonological structure and the linguistic sign prior to this point
in development. As is claimed by Plunkett et al. ‘[t]here remains the fundamental problem
of providing an account of the “hook-up” between symbolic and sub-symbolic processes.’
(Plunkett et al. 1992, p. 294). They point out two different perspectives on this problem. The
first one is called ‘the symbol-grounding problem’ and is formulated as ‘How do syntac-
tically governed and semantically systematic symbol tokens receive a referential-semantic
interpretation in relation to sensory and perceptual information?’. The second, called ‘the
symbol-emergence problem’, is formulated as ‘How do syntactically governed and semanti-
cally systematic symbol types get into the cognitive system?’ (Plunkett et al. 1992, p. 294).
The symbol-grounding problem focuses on the symbolic, lexical level and seeks to ground
and focuses on development towards the lexical, symbolic level. Plunkett et al. further claim
that from the perspective of the symbol-emergence problem, there exists no symbol-ground-
ing problem, because it is solved as a by-product of the former problem. According to them,
there is a good reason why the symbol-emergence problem has gained little attention: ‘More
approach, they claim to present a framework combining both developmental and functional
issues. Interestingly, they relate the separation of these developmental and functional issues
to imaginations of a ‘language of thought’, ideas originating back to St Augustine where lan-
guage is merely a nomenclature of already-existing concepts (Gumperz and Levinson 1996).
These are persistent ideas in linguistics and psychology over the past century, and they
turn up even in theoretical frameworks where they are theoretically rejected, for instance
in functional linguistics. In our view, the separation of developmental and functional issues
concerning the linguistic sign represents a major challenge to functional linguistics, espe-
cially with regard to phonology. When focusing on the linguistic sign, we should ask: Does
our level of description serve our particular object of study, or has our description become
The strict division between a symbolic level and a sub-symbolic level must be regarded as
artificial, even though proponents of the symbol-grounding problem claim that it is natural.
The strict division into a lexical and a sub-lexical level leaves the way open for theories and
models based on representations. Child-language research has traditionally valued vocabu-
lary, in the way it can be identified by adults:
[...] decontextualization is itself a process with a relative rather than an absolute
end-point and the extent to which the meanings of individual lexical items can be
characterized as context independent is itself a function of the communicative context.
(Plunkett et al. 1992, p. 297)
tinuous, with the mechanism being qualitatively the same before and after the spurt: ‘As an
increasing number of words become decontextualized in the child’s vocabulary, new means
of representation emerge’ (Plunkett et al. 1992, p. 297).
Explanation Requires the Non-Arbitrariness of the Linguistic Sign
Saussure defines the linguistic sign as the association of a form (or signifier) with a meaning
(or signified) (Saussure et al. 1969). In this relationship, Saussure chooses not to focus on
the association itself, but instead on philosophical points which are indeed important and
which made structuralist theory a major factor in linguistics over a whole century. Still, we
ask whether some important points have been lost through the maintenance of the accepted
linguistic sign is arbitrary in two important ways. First, the association of a particular form
with a particular meaning is arbitrary (Taylor 1991). This is easily shown when a common
meaning is mapped onto different expressions in different languages. Second, the meaning
itself is arbitrary, because there is no obvious reason why the meanings given names in
various languages—and no other meanings—should have names; language is not simply a
nomenclature for already-existing concepts. The first point may today be considered as quite
obvious, and the second may be considered as not absolute with regard to claims made in
cognitive linguistics (Lakoff and Johnson 1999).
Concerning the notion of the phoneme in functional linguistics, we claim that holding
on to the arbitrariness doctrine on the grounds of these philosophical arguments is a side-
track. The realization that the concept of ‘horse’ is expressed by ‘cheval’ in French and by
‘Pferd’ in German is irrelevant to the question of how symbols enter the cognitive system
of the young child. As we have claimed, Saussure makes a choice of focus concerning the
linguistic sign, and the other obvious choice he could have made was to focus on the con-
vention, on the characteristics of the association. At the time, Saussure had justifications for
making his choice, but we think we can make a case for how the ideas of our time provide
us with good reasons for looking into this again. In a quotation from Taylor, the focus on
arbitrariness is underlined, while the obvious alternative candidate is put in brackets: ‘There
is no reason (other than convention) why the phonetic form [red] should be associated with
the meaning ‘red’ in English.’ (Taylor 1991, p. 6). If we focus instead on the convention,
and thus on the characteristics of the association, it is not obvious that the linguistic sign
is arbitrary. As regards the association, the linguistic sign is in fact non-arbitrary, because
it is not accidental that the association of a signifier with a signified is as it is, owing to
convention. Conventions may well have arisen in a more or less accidental manner, but an
established convention is non-accidental from the point of view of those subscribing to it.
We are left with the experiences we have with language. Association is the basis of conven-
tion—and therefore of symbolicity. We will look into aspects of this argument in the next
The Non-Arbitrariness of Association
If we focus on the symbol-emergence problem, we ask how the unity of word and experience
enters the cognitive system. And that is a very pertinent question. It amounts to selecting
association as the core of the linguistic sign, and to considering the linguistic sign as a con-
tinuum of different strengths and varieties of associations. In other words: there is no strict
distinction between a sub-symbolic and a symbolic level. In defining the linguistic sign,
Saussure presents it as the opposite of the symbol (it should be remarked that Saussure’s
notion of ‘symbol’ is different from the notion of ‘symbol’ used about the lexical level). He
the linguistic sign is arbitrary:
saussurian symbolthe linguistic signmetaphor
Fig. 1 The continuum of association in the non-representational view of language. The formerly distinct
levels of representation are here considered as located on a continuum of association strength and nature
Le symbole a pour caractère de n’être jamais tout à fait arbitraire; il n’est pas vide, il
y a un rudiment de lien naturel entre le signifiant et le signifié. (Saussure et al. 1969,
In most linguistic theories, the linguistic sign is defined by the Saussurian conception of
arbitrariness. In our view, this is too strong a claim of form and meaning as separate entities.
If we focus instead on association as the core of the linguistic sign, there is no important
distinction between the Saussurian symbol and the linguistic sign, because in both cases it is
a matter of form connected to traces of meaning (see the section “On Defining ‘Language’ in
and nature of the association, and they may therefore represent positions on a continuum.
The Linguistic Sign: A Continuum of Association
but probably also stronger. The Saussurian distinction between symbol and linguistic sign
can be justified only if one insists on keeping form and meaning separate. If the focus is on
developmental—and explanatory—aspects of the linguistic sign, this strict division cannot
This is so because structuralism, generativism and cognitive psychology are all unable to
explain historical or individual change and development. In this paradigm, the Saussurian
focus on arbitrariness in defining the linguistic sign provides a static, descriptive level of
representation on which theory has been built.
The inseparability of meaning and form is a central point in both connectionism and
functional linguistics; but still, when it comes to the linguistic sign, the assumptions of the
cognitive/structuralist paradigm are transposed to a large extent. The point of this article is
therefore to pinpoint the demand for a new conception of the linguistic sign in a suggested
combination of connectionism and functionalism, and further to show how an alternative
conception of the linguistic sign meets good standards of empirical science. These ambi-
tions can be achieved only with a new paradigm, simply because their basic assumptions are
incompatible with those of the old paradigm. Two of the positions (symbol and linguistic
sign) in Fig.1 are judgedbySaussure asqualitatively different andopposites.The alternative
view is to consider them as positions on a continuum of complexity and strength as regards
association. As we all know, associations are to some extent unique, and we can therefore
not exclude the possibility of unique associations. Where such associations belong on a con-
tinuum depends on what aspects of language we study. In diachronic studies of a language,
unique associations represent extreme and irrelevant positions, but for the young child or
for the adult second-language learner, unique associations are at the core of their language
development. In social communication, and through frequent interaction, associations are
made common to different degrees. And they are made common in different ways. This is
equally true of all contextual associations, which also vary in strength and variety. On this
alternative view, the linguistic sign is not opposed to the symbol; instead, the non-arbitrary
linguistic sign encompasses the whole range of the continuum, because this variety of asso-
ciation is reality for the language user—infant or adult. If competent adult speakers perform
exactly what Saussure called symbols, why should this be excluded from the level of rep-
resentation? Such questions make us look at association as the core of the linguistic sign,
and make us leave behind the representational view of the relationship between form and
From the alternative view described above, it is considered simplistic that the child, at a
certain point in time, searches for expressions that fit her/his experiences, and that this point
of development is what is called the vocabulary spurt. Rather, it is suggested that form and
functions of his son from nine months of age (Halliday 1975) can be interpreted according to
such a position. With reference to the continuum of the linguistic sign proposed above, it is
that associate with some traces of meaning. In this context, it is meaningless to talk of arbi-
trariness. It is also suggested that attention guides the mechanism of associating sound with
traces of meaning. This means that language development is seen as a matter of association
of cues of various sizes and kinds, where the characteristics of association are continuously
changing. It is not a matter of simply mapping a lexical meaning with an expression, but
rather a continuous alteration of cues that have potential for meaning. No wonder even the
concept of ‘word’ is difficult in the linguistic literature (Dixon and Aikhenvald 2002)! The
vocabulary spurt is then considered as a constructive experience relating to the continuum of
the linguistic sign. Having reached this position in development, the child’s exploitation of
is more, the mechanisms of association are still the same, but their outcome is different. We
use words, but we base our judgement of their function primarily on our associations—that
is, our experience of their intertwined form and function:
On this epigenetic view, the symbolic processing level in human cognition is not an
independent and autonomous system of symbols projecting onto the non-symbolic
level, but the emergent outcome of interactive re-organizations of categorization pro-
developmental mechanism. (Plunkett et al. 1992, p. 310)
Yes indeed, the linguistic sign is the association of form with meaning. Unfortunately, the
Saussurian double focus on ‘form’ and ‘meaning’ introduces two distinct levels of represen-
investigated. Still, it should be remarked that Saussure goes very far in unifying these levels
of representation without leaving the distinction of langue and parole:
On a vu [...], à propos du circuit de la parole, que les termes impliqués dans le signe
linguistique sont tous deux psychiques et sont unis dans notre cerveau par le lien de
l’association. Insistons sur ce point.
Le signe linguistique unit non une chose et un nom, mais un concept et une image
acoustique. Cette dernière n’est pas le son matériel, chose purement physique, mais
plus abstrait. (Saussure et al. 1969, p. 98)
As we see it, the representational view of language is bound to confuse sign and object:
‘Confusion of sign and object is original sin, coeval with the word’ (Quine 1977). Because
there cannot be identified any space for the semantic representation, it is tied to the referent.
Association and Language
When we consider association to be at the core of the linguistic sign, we avoid the repre-
sentational position and we may define language by performance—as a set of codes with
potential for meaning. Quine’s position is interesting in this respect, as he considers individ-
uation of objects to be a matter of progressed development, not the initial state of association
(Quine 1977). This position is very pertinent with regard to mainstream, ‘representational’
thinking. The idea that the child goes from signal to sign—or, in other words: enters the
symbolic stage—is valid according to a representational view, again owing to the concept
of the arbitrary linguistic sign. But, if association is at the core of the linguistic sign, and if
individuation of objects is an emergent property of association, there is a whole new frame-
work for discussing phonological development. Paradoxically, it is in this new framework
that the Saussurian slogan ‘dans la langue il n’existe que la différence’ becomes really pow-
erful. The core of the non-arbitrary linguistic sign—association—is nourished by the same
of ‘difference’ has become far more powerful as it is linked to association. Still, we must
also consider ‘difference’ differently. Saussure’s slogan is based on necessary and sufficient
conditions for category membership. With an understanding of categorization as a matter
of gradience and prototype, similarity to the prototype is central to categorization, not only
contrast with other categories.
The Phoneme: Descriptive or Explanatory Power?
If we focus on difference and similarity as basic factors for association, we no longer favour
lexical (representational) contrast as the basis for the phoneme. Structuralist theorists are
of course also aware of this aspect of difference, but here the ‘sub-lexical’ distinctions are
degraded and seen as secondary to lexical distinctions. Roman Jakobson and Linda Waugh
claim that the distinction between ‘distinctiveness proper’ and ‘sense determination’ is cru-
cial to understanding how sound functions in language (Jakobson et al. 2002). What is more,
without such a distinction the notion of ‘phonology’ is blurred.
From the position of the symbol-emergence problem, phonology must be defined differ-
ently than from the structuralist position. From this perspective, Lacerda and Lindblom’s
model contains interesting aspects of the phoneme as prototype (Lacerda and Lindblom
1997). Their approach is based on the following assumptions: (1) early language acquisition
is a matter of two processes (memory and perception); (2) variation is the learner’s friend;
(3) the system emerges as a consequence of rather non-linguistic processes; and (4) the child
has no linguistic knowledge at the time of birth:
According to the model, categories with prototypical properties emerge in a self-
organizing way as a consequence of the distributional patterns that the exemplars
cumulatively form in the perceptual representational space and in phonetic memory.
(Lacerda and Lindblom 1997, p. 14)
Lacerda and Lindblom’s model is a simulation of isolated vowels cut from infant-directed
speech. The simulations show how vowels perceived as belonging to the same category in
speech also activate areas that are not associated with that category, for instance how some
for each category. The simulations of vowel categories appear as neat X-ray pictures of the
phonemic prototypical properties:
[...] categories of speech sounds are fuzzy clouds of multi-dimensional (multi-modal)
sensory representations that are spontaneously structured by cross-correlation, due to
their simultaneity. (Lacerda and Lindblom 1997, p. 31)
However, this neat simulation of phonemic properties is placed in a quite strict framework
concerning the linguistic sign and the focus on the phoneme as unit. In our view, the question
Therefore, the main problem presented in the introduction of this article remains unsolved,
that is: the question of the conceptual status of the phoneme. To some extent, this is due to
Lacerda and Lindblom’s own theoretical framing of the simulations. Their position is cer-
tainly functionalist, but their adaptation of the conventional view of the linguistic sign and
their focus on language structure as primarily phonemic may not fully exploit the potential
of their findings. Interestingly, the strict phonemic structure is not part of the simulation—
it is added by conventional theory. The limitations of memory are causally linked to the
phonemic structure, but this point is only hypothetical and cannot be derived directly from
the simulation. Moreover, there are other potentials that could be exploited in Lacerda and
Lindblom’s simulation: the phonemic properties shown are less important parts of language
structures. The notion of phonemes as ‘fuzzy clouds’ fits with the insight of the phoneme as
prototype with exemplars at different distances from the core. This insight carries primarily
descriptive power. However, this description has a better potential for explanation than the
structuralist notion of ‘phoneme’, because it is less abstract and demands fewer supporting
hypotheses. The notion of phonemes as fuzzy clouds may explain why some words ‘pop
up’ in the mind more readily than others and it may explain some ‘slips of the tongue’ or
some errors, confusions and hesitations. It may also explain some developmental character-
istics: as the nucleus grows ‘harder’—in comparison to the loose cloudy environment—the
person may become less hesitant and more accurate when choosing words and expressions.
However, it still remains a crucial question how this insight can be operationalized differ-
ently from traditional segmental theory. We claim that this can be done only if one dares to
think differently about the linguistic sign. In our view, there is a tendency towards unfortu-
nate theoretical framing in Lacerda and Lindblom’s model, despite all its highly interesting
aspects. We would claim that their position represents a kind of functionalism that explains
the opposite—, and that they thereby represent a position closer to the position that Plunk-
ett et al. call the symbol-grounding problem. Their explication of the vocabulary spurt is a
way to combine the established view of a universal phonemic structure with the established
view of the arbitrary linguistic sign. In other words, the phoneme as hypothetical description
has been mixed up with matters of explanation. This is unfortunate, because the units of
description should not be taken a priori in the causal explanation, in order to avoid circular
The Status of the Concept of ‘Phoneme’: Description
but not separated. We also need a coherent theory, which means that the explanation must
to the domain of explanation. The problem arises when we use phonemic properties as part
of a causal relationship. Descriptions are hypothetical, but descriptions can be evaluated as
more or less adequate according to how well they fit the explanation and vice versa. With
regard to the reasoning about the arbitrariness of the linguistic sign, the phoneme as ‘fuzzy
cloud’ is considered to have a stronger descriptive power than the traditionally defined pho-
neme. While we are fully aware that most structures can usefully be described in various
ways, the point here is to differentiate between the domains of description and explanation.
Talking of explanation in phonology, we reject the assumption of the ‘symbolic’ level as a
level having more to do with language than earlier individual development. Development is
not a matter of going from signal to symbol—rather, it is a matter of language: of going from
language to language. Further, language is a set of overt codes with potential for meaning,
where these codes and their potential for meaning are established by differences and sim-
ilarities concerning associations, without any important division between what we believe
should be identified as form and meaning, respectively. Taking Quine’s point of view—that
the vocabulary spurt in the young child is primarily a matter of a rapid progression of associ-
ations, where the associations of differences and similarities concerning stimuli which have
gained the child’s attention reach a point where symbols emerge. Therefore, the view of a
as an important factor for the vocabulary spurt is considered to be a weak supporting hypoth-
esis, because this argument is valid only if we know for certain that language is primarily
stored as symbols instead of as a complex web of associations. And we do not know that.
From this perspective, the phoneme as fuzzy cloud and prototype is simply a hypothetical
and illustrative description of one effect of the basic mechanism in language, namely the
association of differences and similarities of material which has gained attention in interac-
tion between human beings. According to this position, the conceptual locus of the phoneme
can be hypothesized to be primarily descriptive. Thephonemic (and segmental) properties of
a certain language make a language structure, but not the structure. This is in line with what
auto-segmental-metric theory has shown us about structure in the world’s languages. From
this, it can be hypothesized that the phonemic effect is different in the world’s languages—
not just concerning different inventory, but also concerning the degree to which prototypical
effects are attached to segments. Tone languages are hypothesized to have a different proto-
typical effect concerning the segment than, for instance, agglutinative languages with little
language bias in linguistics (1982), which highlights the problem of description. In his view,
the notion of ‘phoneme’ is nourished by the alphabetical and segmental structure of (some)
What we call ‘language structures’ should therefore be considered as constructions, and
for the purpose of behavioural studies they should not be abstracted from overt language
behaviour. This is because the notion of ‘structure’ rests on the same ontological relativity as
all conceptual enterprise (Quine 1977) and is therefore restricted by the purpose of descrip-
tion and the domain of behaviour. From such a position, the different structural descriptions
represent a continuum of adequacy for behavioural studies. Adequacy is related to the pur-
pose of description. As regards the ‘phoneme’, the purpose of description in autonomous
linguistics is far from the purpose of description in behavioural studies. This issue has rarely
been highlighted in the discussion of the phoneme. What a phoneme ‘is’ depends on the pur-
about this concept should be avoided. While we consider both structuralist and generativist
descriptions as adequate descriptions of language structure, our claim is that they are distant
from the part of the continuum which is hypothesized to be adequate for behavioural studies.
In the present article, the status of the concept of ‘phoneme’ in psycholinguistics has been
characterized as vague; still, the phoneme defined as prototype and fuzzy cloud is important
as an example of a basic mechanism of language: association. What is more, the widespread
to be simplistic, because it shows only the tip of the iceberg. It covers only those aspects of
contrast that we are able to understand, for instance that there is a difference between /hæt/
and /kæt/. In order to be able to explain, we must search for descriptions that may serve our
purpose of explanation. And that involves leaving the domain of understanding. We thereby
go into phenomena that we may explain, but not understand. A parallel example can be given
from medicine, where the phenomenon of depression is (often) explained on the basis of the
neurotransmitter serotonin; however, we are not able to understand depression by means of
Focusing on association as the core of the non-arbitrary linguistic sign, we have made
an attempt to identify the boundaries of explanation, description and understanding when
dealing with the categorization of sound in empirical studies of behaviour. This implies a
larger focus on phonetics. Our claims of the phoneme as a concept with poor explanatory
power, does not imply a rejection of the categorical nature of language, but rather a matter
of relativizing the phonemic, segmental structure as the primary, universal sound structure
of the world’s languages.
Chomsky, N., & Halle, M. (1968). The sound pattern of English. London: Harper and Row.
Clark, E. V. (1993). The lexicon in acquisition. Cambridge: Cambridge University Press.
Clark, H. H., & Clark, E. V. (1977). Psychology and language: An introduction to psycholinguistics. New
York: Harcourt Brace Jovanovich.
Derwing, B. (1973). Transformational grammar as a theory of language acquisition (Vol. 10). Cambridge:
Cambridge University Press.
de Saussure, F., Bally, C., Sechehaye, A., & Riedlinger, A. (1969). Cours de linguistique générale 3rd
ed. Paris: Payot.
Dixon, R. M. W., & Aikhenvald, A. Y. (2002). Word: A cross-linguistic typology. Cambridge: Cambridge
Flanagan, O. (1991). The science of mind (2nd ed.). Cambridge, MA: MIT Press.
Givón, T. (2001). Syntax: An introduction (Rev. Ed.). Amsterdam: J. Benjamins.
Gumperz, J. J., & Levinson, S. C. (1996). Rethinking linguistic relativity. Cambridge: Cambridge University
Halliday, M. A. K. (1975). Learning how to mean: Explorations in the development of language.
London: Edward Arnold.
Jakobson, R. (1941). Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell.
Jakobson, R., Fant, C. G. M., & Halle, M. (1952). Preliminaries to speech analysis: The distinctive features
and their correlates. Cambridge, MA: MIT.
Jakobson, R., Taylor, M., & Waugh, L. R. (2002). The sound shape of language (3rd ed.). Berlin: Mouton
Lacerda, F., & Lindblom, B. (1997). Modelling the early stages of language acquisition. In Å. Olofsson
& S. Strömqvist (Eds.), Cost A8: Cross-linguistic studies of dyslexia and early language develop-
ment (pp. 14–34). Luxembourg: Office for Official Publications of the European Communities.
Lakoff, G., & Johnson, M. (1999). Philosophy in the flesh: The embodied mind and its challenge to
Western thought. New York: Basic Books.
Linell, P. (1982). The written language bias in linguistics. Linköping: Universitetet.
Newmeyer, F. J. (1986). The politics of linguistics. Chicago: The University of Chicago Press.
Plunkett, K., Sinha, C. G., Møller, M. F., & Strandsry, O. (1992). Symbol grounding or the emergence of
symbols? Vocabulary growth in children and a connectionist net. Connection Science, 4(3/4), 293–312.
Quine, W. V. (1977). Ontological relativity and other essays. New York: Columbia University Press.
Rommetveit, R. (1974). On message structure: A framework for the study of language and communica-
tion. London: Wiley.
Taylor, J. R. (1991). Linguistic categorization: Prototypes in linguistic theory. Oxford: Clarendon Press.
Tønnessen, F. E. (1995). On defining ‘dyslexia’. Scandinavian Journal of Educational Research, 39(2), 139–
Uppstad, P. H., & Tønnessen, F. E. (2007). The notion of ‘phonology’ in dyslexia research: cognitivism—and
Beyond. Dyslexia, 13, 154–174.