Language and Evolution: Homepage Robin Allott


Sound symbolism is not a very satisfactory term but it is a familiar one to cover a phenomenon which has been noted and studied over very many years, the apparent appropriateness of the sound-structures of many individual words for their meanings. A better description for this might be 'natural expressiveness'. What is unsatisfactory about the term 'sound symbolism' is its use of the radically confused concept 'symbol'. This is one of the words used extensively and with confidence in many different disciplines, but with no clear idea of what a 'symbol' is. Often there is total contradiction between the use of 'symbol' by different authors. So, Peterfalvi (1970), author of perhaps the best book in French on sound symbolism (commended warmly by Roman Jakobson 1987), says that the term 'symbol' in its modern sense always includes the idea of a natural, and not conventional, analogical correspondence between the concrete form and the object which it symbolises. Bolinger(1963), a leading American linguist, takes the radically opposite view: meanings in which form imitates nature are called iconic: meanings that are arbitrary and conventional are called symbolic. A clash of symbols indeed!

More serious than the right name for the phenomenon of 'sound symbolism' is the cast-iron orthodoxy formulated by leading authorities in linguistics that the phenomenon does not exist at all. Saussure (1915) is notorious for the slogan 'l'arbitraire du signe', by which he denied onomatopoeia (and all other natural expressiveness of words) other than as marginal, and treated even apparent onomatopoeic words as no more than conventionalised forms. This dogma was restated and reinforced by Charles Hockett(1958, 1963) in his influential identification of what he described as 'design features' of language: The relation between a meaningful element in language and its denotation is independent of any physical and geometrical resemblance between the two. The semantic relation is arbitrary rather than iconic.. Other influential linguists concurred; Firth(1964) warned students to beware of sound symbolism, saying that the sounds of words in themselves paint nothing.

This certainty on the part of Saussure and his followers is all the more surprising and apparently perverse in the face of the exceptionally long history of evidence to the contrary, presented by equally perceptive and equally authoritative writers, and in the face of what has in this century become the large body of scientifically planned experiment establishing the reality of sound symbolism. The debate started with Plato (in the dialogue Cratylus):"Everything has a right name of its own, which comes by nature. A name is not whatever people call a thing by agreement, just a piece of their own voice applied to the thing, but there is a kind of inherent correctness (orthoteta tina ton onomaton pephukenai) which is the same for all men, both Greeks and foreigners." Similarly Lucretius, a thoughtful and pragmatic investigator, rejected the arbitrariness of the origin of words:"putare aliquem tum nomina distribuisse desiperest. Nam cur hic posset cuncta notare vocibus et varios sonitus emittere linguae tempore eodem alii id non quisse putentur"(5: 1040-1045) - that is, people who think that some individual could, by himself, have invented words arbitrarily are talking nonsense.

The tradition that words symbolise their meanings has continued over the centuries. The polymath President De Brosses in the 18th century (the correspondent and opponent of Voltaire) argued for the physiological origin of words: "il y a de certains mouvements des organes appropriées à désigner une certaine classe de choses de même espéce ou de même qualité." In the 19th century the reality of the expressiveness of words was championed by Humboldt in Germany, and later by Grammont(1901) in France. Humboldt(1836) was certain that a connection between the sound of a word and its meaning exists; he distinguished between onomatopoeia and sound symbolism. The sound was not, in his view, a directly imitative sign but a sign which indicated a quality which the sign and the object have in common; to designate objects, language selected sounds which partly independently and partly in comparison with others produce an impression which to the ear is similar to that which the object makes upon the mind. Humboldt said that this kind of sign process (based upon the particular meaning of each individual letter and whole groups of letters) had undoubtedly exercised a prevailing, perhaps even exclusive, influence on primitive word formation resulting in a certain likeness of word-formation throughout all languages.

Otto Jespersen(1922) discussed at length the evidence for sound symbolism and concluded that it should be seen not simply as a force that influenced the initial formation of language but as one operating continually to make the words used more appropriate to their sense, that is, sound symbolism is a reality in the modern use and development of language. "Is there really much more logic in the opposite extreme which denies any kind of sound symbolism (apart from the small class of evident echoisms and 'onomatopoeia') and sees in our words only a collection of accidental and irrational associations of sound and meaning? .. Sounds may in some cases be symbolic of their sense even if they are not so in all words... There is no denying that there are words which we feel instinctively to be adequate to express the ideas they stand for."(397-398). He directly criticised Saussure's approach: "De Saussure gives as one of the main principles of our science that the tie between sound and sense is arbitrary and rather motiveless.. and to those that would object that onomatopoeic words are not arbitrary, he says that 'they are never organic elements of a linguistic system' ... Here we see one of the characteristics of modern linguistic science; it is so preoccupied with etymology that it pays much more attention to what words have come from than to what they have come to be .. Though some echo words may be very old, the great majority are not .. In the course of time, languages grow richer and richer in symbolic words .. Sound symbolism, we may say, makes some words much more fit to survive .. Echoism and related phenomena - these forces are vital to languages as we observe them day by day".(408-411)

Hormann, a more recent writer on psycholinguistics, quotes Stenzel who emphasised "the belief, deeply rooted in our natural feeling for language, that meaning lies directly in the sound of words; this belief is sustained by a peculiar feeling that it is self-evident, which certainly constitutes a very important experience in the mother-tongue and in any other language of which we have a reasonable understanding."(Hormann 1971: 215)

For this 'important experience in the mother-tongue', the results of Piagetian research seem relevant (though much neglected). Piaget and his assistants found that young children uniformly say that words are derived directly from the objects to which they relate. These findings were in agreement with what Piaget termed well-known theories according to which to a child's eye every object seems to possess a necessary and absolute name, a part of the object's very nature; children believe that they are not taught words for common things - the words originate within the child itself. Children's ideas of this kind were, Piaget thought, evidence of their lack of insight and understanding; they go on taking this sort of view until they have had several years of formal schooling and reach the age of about eleven when they come to accept that words are arbitrary and conventional. Piaget of course accepted the linguistic orthodoxy unquestioningly; he did not seek further for an explanation of the surprising uniformity with which children perceive a natural link between word and meaning, merely commenting: "This inability to dissociate names from things is very curious"(1973: 83). Perhaps it is more than curious. 'Out of the mouths of babes' there may be something which linguists ought to consider.

In spite of the accumulated evidence for sound symbolism, the unquestioning attachment of most linguists (and researchers in other disciplines) to the Saussurean dogma is puzzling. Perhaps Peterfalvi was not far from the truth: "Pourquoi est-il important pour la linguistique contemporaine que le signe linguistique soit arbitraire? .. Si en effet on partait du principe générale que le signifiant et le signifié d'un signe linguistique sont unis par des liens fondés sur l'analogie ou l'isomorphisme, toute l'analyse des langues menée à bien par les linguistes (à commencer par l'analyse phonologique) serait impossible sous sa forme actuelle"(1970: 77).

Most people probably have little idea of the weight of material bearing on sound symbolism and the natural expressiveness of words. The list of references for this paper(a partial bibliography for sound symbolism) gives some idea of the continuity of commentary and research and the range of languages and experimental work involved.

1. Evidence and experiment

The case for sound symbolism, its character, does not, of course, depend simply on authorities of the past. Since the 1920's and 1930's there has been a very considerable volume of increasingly sophisticated experiment to test and delimit the functioning of sound symbolism in modern languages. Some of the research related to what was described as 'phonetic symbolism', that is, the extent to which individual speech-sounds, phonemes, carried specific meanings; other research related to the appropriateness of words as a whole - 'morphosymbolism' (to adopt the term used by Malkiel(1978). The research can be divided, with some overlapping, into several categories:

Sound symbolism within a single language. For example, experiment and material indicating the extent to which in English individual phonemes are symbolic or the sound-structures of words otherwise appear to be sound symbolic or naturally appropriate to their meanings. Under this head, there is evidence that in English (and in other languages) there are systematic similarities between groups of words which refer to related percepts or have related aspects of meaning not explainable in terms of orthodox principles of etymology.

Sound symbolism between different languages, Typically, experiment under this head assesses how far speakers of one particular language appear to be able to use the sound-structure of words in another language (or in several other languages) to arrive at the meaning of the words in the other language (or languages). Under this head there is the wider question of the universality of sound symbolism (or, more narrowly, of phonetic symbolism). Also considered is evidence for sound symbolism in the form of correspondences in unrelated languages between the sound-structures of words used for particular percepts or ranges of percepts

Sound symbolism matching sounds or visual patterns with invented words This is the kind of experimentation associated with Kohler(1929).

2. Sound symbolism within a single language

There has been extensive consideration of 'sound symbolism', the natural expressiveness of word-sounds, in relation to the major European languages, The phenomenon that words are felt to be naturally appropriate to their meaning is as well-established for German, French and Spanish as it is for English. The disagreement has been over how the feeling of the natural suitability of words for their meanings is to be explained. In relation to the English language, even linguists such as Bloomfield and Firth,who proclaim the arbitrariness of language and deny the existence of sound symbolism, recognise that particular feelings of appropriateness are associated with particular words. Other authorities strongly support the reality of sound-symbolism in the English language. Jespersen(1922) said that there was no denying that there are words which we feel instinctively to be adequate to express the ideas they stand for. Roger Brown(1958), who dealt comprehensively with the issue of sound symbolism, concluded that speakers of a given language have similar notions of the semantic implications of various phonetic sequences.

Some of the most extensive work has been done in relation to German. Humboldt(1836) quoted words like 'wolke' 'wirren' and 'wunsch' as expressing the vacillating, wavering motion referred to. Hilmer(1914) compiled 170 pages of word-lists of expressive words in German. This was followed up in a wider context by other German researchers (described in the next section of this paper). For French, Sauvageot(1964), whilst accepting Saussure's doctrine of the arbitrariness of the sign, described the extent and importance of onomatopoeia and expressive words. More recently, Peterfalvi(1970) assembled evidence which convinced him of the reality of sound symbolism in French (and more generally); he repeated experiments similar to those conducted earlier by American and German researchers. In relation to Spanish, De Diego(1973) concluded that all words in varying degrees have a sensory or emotive value(69). As an example, he gave the Spanish name for the bird in English called a 'wagtail':"in the word PIMPIM we have not an acoustic representation of its song but an acoustic expression which translates a visual impression of the rapidity of movement characteristic of the bird".(67)

Unlimited examples could be provided from other European languages and from remoter languages for the existence of a felt relation between word-sounds and word-meanings. FIRTH collected examples of expressive words in Norwegian, Swedish and Dutch. In many examples, the syllables in a word apparently correspond to the number of distinct elements in the sound, object or action, for example, the word 'ongololo' meaning 'centipede' in Samoan (an example given by Brown 1958). Sound symbolism seems to be as much or more of a reality in remoter languages. Manchu is said to have been absolutely full of imitative formations. There are many examples of sound symbolism in African languages; in Ewe, high- tone words indicate small things and low-tone words large things; in certain Sudanese languages high-tone words are used to express long distances or high speed, and low-tone words to express proximity and slowness. These examples are of particular interest as showing that sound symbolism is a reality also for tonal languages. In Chinese, it is not irrelevant, in considering visual contour as a foundation for the natural feeling of words, that many of the Chinese classifiers (words indicating the category to which an object belongs) are based on shape; there are, for example, classifying particles which indicate long, flat and round objects, containers, pairs and sets (curiously similar to classifiers used in some American Indian languages). Studies of sound symbolism in a number of other languages are listed in the bibliography e.g. Korean (Martin 1962, Kim 1977), Chinese (Karlgren 1962), Japanese and Hindi (Koriat/Levy 1977), Semai (Diffloth 1976), Bini (Wescott 1973), Yuman (Langdon 1971), Coeur D'Alene (Reichard 1945), Azerbaijani (Householder 1962), N.W. American Indian (Haas 1970), Hungarian (Allport 1935) and various West African languages (Samarin 1967). Jakobson/Waugh(1987) refer also to studies of sound symbolism in Siouan and Dakota.

Research within single languages has cast light on the issue of 'phonetic' symbolism. Sapir(1929), on the basis of his experiment using two invented words MAL and MIL which subjects were asked to identify with 'a large table' or 'a small table', raised the issue whether phonemes in isolation are symbolic, for example, of differing size. Newman(1933), who was a student of Sapir's tested this for English by listing all words related to size, about 500. Comparing the small words with the large, he found no significant difference in the size implications of the vowels used, that is, no confirmation in natural language of the results found by Sapir. Roger Brown(1958) independently did a similar study and also found no evidence of symbolic representation of magnitude by the differentiated use of vowels. This seems very likely. Consider in English, for example, the words BIG and SMALL, MIGHTY and MITE. The conclusion seems unavoidable that the symbolic character is dependent on the whole word and not on any single speech-sound considered in isolation; if two words differ in only a single speech-sound, then the symbolic character may change but this does mean that in all words containing this opposition, the difference in symbolic effect will be the same.

3. Resemblance within a single language between etymologically unrelated words with related meanings

Many words, in English and other languages, which are similar in the sounds going to form them, derive their similarity from etymologically standard processes of word-formation and word- development. This is true of the conjugation and declension of words in inflected languages, of the formation of comparatives or plurals in English and other languages, of the composition of new words by the addition of prefixes or suffixes and so on. This kind of resemblance between words, in a systematic way, is at the same time taken for granted and fully natural. The more interesting kind of systematic resemblance is where words have sounds and meanings which resemble one another but etymologically the words are judged not to be related or derived from one another by any standard process of composition, prefixing etc. This is a phenomenon which has been observed in English and other languages and has been commented on, as an unexplained curiosity, by a number of linguists and others. Tylor(1871), the anthropologist, remarked on the way in which words, whilst preserving so to speak the same skeleton, may be made to follow the variation of sound, of force, of duration, which an imitative group will show:- CRICK CREAK CRACK CRUSH CRUNCH CRAUNCH SCRUNCH SCRAUNCH. Firth(1937), whilst disbelieving in sound symnbolism, assembled large collections of symbolic words and emphasised the systematic way they may be related to each other, quite separately from any etymological links; he acknowledged that we are appreciably affected by initial and final phoneme groups not ordinarily recognised as having any function. "Consider the following English words: SLACK SLOUCH SLUDGE SLIME SLOSH SLASH SLOPPY SLUG SLUGGARD SLATTERN SLUT SLANG SLY SLITHER SLOW SLOTH SLEEPY SLEET SLIP SLIPSHOD SLOPE SLIT SLAY SLEEK SLANT SLOVENLY SLAB SLAP SLOUGH SLUM SLUMP SLOBBER SLAVER SLUR SLOG SLATE ... A group of words such as the above has a cumulative suggestive value that cannot be overlooked in any consideration of our habits of speech. All the above words are in varying degrees pejorative".(Firth 1964: 184)

In English, there are many similar groupings of words where there is some underlying resemblance of meaning related to the surface resemblance of the sounds forming the words. Bloomfield(1933) quoted a number of these: examples are such groups as: FIRE FLAME FLARE FLASH FLICKER POINT POKE PIKE PEG PEAK PIERCE PRICK PROD PROBE PRONG HIT HACK HEW HATCHET HASH THROW THRUST THRASH THWACK THWART THUMP THROTTLE SWEEP SWAY SWING SWIRL SWERVE SWOOP SWISH SWITCH SWAT SWIPE SWAB WAG WAGGLE WEAVE WOBBLE WANDER WONDER WADDLE WAVER WAVE. The felt resemblances between these sets of words is apparent and must derive from symbolism in the sounds used - though not necessarily just from the initial letters of the words since one can easily find examples of other words with the same initial letters but belonging to completely different categories of meaning; the resemblance seems to derive from the whole structure of the words in each group.

Despite the examples he gave of expressive groups of words, Firth, as already noted, explained them as simply the result of habit - but even in his own terms, the question arises whether there is some universal or physiological basis for these habits, for the preferences which people demonstrate in their formation of words for particular sounds. There is no necessary conflict between saying that a particular mode of behaviour, in language or otherwise, is habitual and yet at the same time has a biological or physiological basis. Firth drew attention to the existence of parallel groups of expressive words in other languages. It is surprising that he did not feel it necessary to consider more closely how it is, if these are purely cultural phonaesthetic habits, they should be shared not only by speakers of the same language but by speakers of other distinct languages. The parallel development of similar phonetic habits in other communities makes it doubtful whether a purely cultural explanation is adequate.

Jespersen(1922) also presented detailed material bearing on sound symbolism: "the simplest case is the direct imitation of the sound: thus CLINK ... SPLASH ... BLEAT .. SNORT .. GRUNT ... But as our speech organs are not capable of giving a perfect imitation of all 'unarticulated' sound, the choice of speech- sounds is to a certain extent accidental and different nations have chosen different combinations, more or less conventionalised, for the same sounds, thus COCK-A-DOODLE-DO ... French COQUELICO .. and for whisper, French CHUCHOTER Spanish SUSURAR .. Next the echoic word designates the being that produces the sound (PEEWIT etc). Thirdly, as sound is always produced by some movement and is nothing but the impression which that movement makes on the ear, it is quite natural that movement itself may be expressed by the word for its sound: the two are in fact inseparable. Note for example such verbs as BUBBLE SPLASH CLASH CRACK PECK . Then we have words expressive of such movements as are not to the same extent characterised by large sounds; thus a great many words beginning with L combinations FL: FLOW ... SL: SLIDE ... SLIP ...GL: GLIDE . Sound and sight may originally have been combined in such expressions for an uncertain walk such as TOTTER DODDER .. but in cases of this kind, the audible element may be wanting and the word may come to be felt as symbolic of the movement as such. This is also the case with many expressions for the sudden rapid movement by which we take hold of something; as a short vowel suddenly interrupted by a stopped consonant serves to express the sound produced by a very rapid striking movement (PAT TAP ..). Similar sound combinations occur frequently for the more or less noiseless seizing of a thing ... SNAP ... CATCH .. There is also a natural connection between the action and the sound in the word TOT TICKLE (Latin: TITILLARE). There is some more or less obvious association of what is only visible with some sound or sounds ... We may also think of the word ZIGZAG as denoting movement in alternate turns here and there"(Jespersen 1922:Chapter XX)

The distinction between the lists of expressive words presented by Firth and those presented by Jespersen is that whereas Firth did not attempt to explain why the particular sounds should have been chosen, Jespersen looked for a systematic relation between the particular sounds composing expressive words and the action, sound, perception etc. to which the word refers. He was beginning to sketch out a detailed system of sound/meaning correspondences. This systematic relation of sound and meaning was developed further by a linguist opposed to the idea that there is any natural relation of sound and meaning, Bloomfield(1933) pointed out that English is especially rich in a type of intense forms, the symbolic forms which have a connotation of somehow illustrating the meaning more immediately than do ordinary speech-forms. "The explanation is a matter of grammatical structure ... to the speaker it seems as if the sounds were especially suited to the meaning". Bloomfield's promised explanation of symbolic forms in terms of grammatical structure never really arrived. The nearest he came to it was in discussing the concept of roots underlying vocabulary: "Primary words that do not contain any affix-like constituents (e.g. BOY RUN RED) are classed as primary root-words..(240) A root may appear in only one primary word, as is the case with ... MAN BOY CUT RED ... or it may appear in a whole series of primary words ...(241) In the Germanic languages, modification of the root occurs in words of symbolic connotation as FLAP FLIP FLOP. If we take FLAP as the basic form of this root, we shall describe FLIP FLOP as derivatives ...(242) However we find clearly-marked phonetic- semantic resemblances between elements which we view as different roots ... [In English symbolic word] we can distinguish, with varying degrees of clearness, and with doubtful cases on the borderline, a system of initial and final root-forming morphemes of vague signification. It is plain that the intense symbolic connotation is associated with this structure."(244-245)

This is a far from clear account of how the expressiveness of symbolic forms is explained as a matter of grammatical structure. If in fact the primary roots had any independent validity, if the so-called root-forming morphemes were the product of some grammatical necessity, he might have an adequate explanation - but his account was undermined by the admission that the roots have no particular historical reality: "Now and then one still hears the claim that the roots which we set up must once upon a time have been spoken as independent words. The reader needs scarcely to be told that this is utterly unjustified; the roots are merely units of partial resemblance between words".(240) This amounts to no more than saying that systematic resemblances can be observed between words which are said to be symbolic or expressive. It says nothing about the origin of the expressiveness.

4. Sound symbolism between languages

What has been quoted from Firth, Bloomfield and Jespersen represents an earlier stage of discussion of sound symbolism, based on rather scattered observation of apparent relations between sounds and meanings. More recently, there has been a more experimental and quantitative approach. The experimentation has been concerned with the different types of sound symbolism which may exist, symbolism of whole words (of the kind that Jespersen and Firth largely considered) and symbolism inherent in clusters of speech-sounds (as considered by Bloomfield) or in individual speech-sounds, consonants or vowels (as studied in Sapir's pioneering experiments). The research has not been confined to sound symbolism as perceived by English-speakers but has provided evidence for sound symbolism as a force operating in other language communities. The experiments counter the criticism that apparent appropriateness of sounds and meanings in any single language such as English may be a peculiarity of that language or in any event may simply show that single language communities tend (as Firth argued) to develop habits of conventional origin in the sounds they use for particular clusters of ideas, that sound symbolism is cultural rather than natural.

There has been as much, if not more, research into the extent to which cross-linguistic sound symbolism is a reality i.e. that speakers of one language can appreciate the expressive force of words in another language (possibly one not at all known to them). There seems no doubt that sound-symbolism does extend across languages but it is unsettled how far this is universally true or whether there are some languages or groups of languages for which the expressiveness of individual words in the languages cannot be appreciated by members of other language-communities. Certainly, experiment has shown cross-linguistic symbolism between a number of completely unrelated languages but some experiments involving Chinese (and other tonal languages) have been unsuccessful or unconvincing.

The sequence of experiments has developed consistently towards greater and greater rigour. At first experimental subjects were asked simply to say whether a selection of words drawn from various foreign languages sounded appropriate or not to their meanings; though the result was that the subjects were able to judge the appropriateness of the words, the experiment was not conclusive because it might have been biased by the selection of words by the experimenter. At the next stage, the experiment took the form (devised by Tsuru) of presenting 36 pairs of opposite words in Japanese (meaning 'hot-cold' and so on) to subjects with no knowledge of Japanese and asking them to match the Japanese words with the corresponding English words; the words were successfully matched more frequently than could result purely from chance and this finding suggested that the form or the sound of the Japanese word must give some clue to the meaning. It was then argued that the Japanese experimenter might by his selection of pairs of opposite words have unconsciously biased the result; to eliminate unconscious bias in a new series of experiments the pairs of opposite words in Japanese were translated into Hungarian and Polish; the subjects taking part in these new experiments did not, of course, speak or know Hungarian, Polish or Japanese but again the result was that the foreign words were matched with the corresponding English words more frequently than, statistically, could be explained as resulting from pure chance. The same sort of experiment was repeated with Czech, Hindi, Croatian and Hebrew and in each case added to the cumulative evidence for the existence of cross-linguistic symbolism between non-tonal languages. Experiment on somewhat different lines (Ertel/Dorst 1965) found evidence for expressive sound symbolism between languages for 25 languages.

What follows is a summary account of some of the more significant pieces of research (drawing where necessary on discussion of some of these experiments in Brown 1958, Hormann 1971, Peterfalvi 1970 and Jakobson/Waugh 1987):

Muller(1935) attempted to meet the difficulty that sound symbolism of natural words might be peculiar to a single language, a cultural artefact, by testing with children whether words drawn from remote languages (Samoan, Bantu, Eskimo) were felt to be appropriate. He chose such words as TUMBA (meaning swelling) and ONGOLOLO (centipede) where the appropriateness seemed apparent - but his results could have been vitiated by unconsciously biassed selection of the words used.

Tsuru/Fries(1935} followed a different plan which became the model for much subsequent experimentation. Tsuru compiled a list of 36 pairs of Japanese antonyms ('hot-cold' 'high-low' etc) and used as experimental subjects 57 native English-speakers with no knowledge of Japanese. The subjects were asked to match the English pairs of antonyms (spoken and written in Romanised form). A chance result would have been that they should guess correctly in 50% of the cases; they guessed correctly significantly more often than this and therefore must have been offered some clue to the right answers by the form or sound of the Japanese words. These results suggested (since Japanese is taken to be a language unrelated to English) that they must have been relying on some universal phonetic symbolism - not simply a conventional association of sound and meaning. Roger Brown(1958) commented that even this experiment is subject to the criticism that there may have been selection by the Japanese experimenter of words which happened to bear some relation to the corresponding English forms., or that there may in any case be coincidences between forms in Japanese and English which would lead to apparently better than chance results.

Allport(1935) sought to evade these criticisms, in experiment on similar lines to Tsuru, by translating Tsuru's set of antonyms into Hungarian (unrelated to Japanese or English) so that any unconscious selection would be nullified, Nevertheless, the Hungarian words were guessed by the experimental subjects with more than chance success. The experiment was repeated on the same pattern by Rich(1953) with pairs of Japanese and Polish words and again resulted in correct answers significantly above chance.

Brown/Black/Horowitz(1955) did an experiment with pairs of words translated into Chinese, Czech and Hindi which 85 experimental subjects had to guess. The results again were above chance. Roger Brown commented on these experiments that English-speaking subjects matching words with the Japanese, Hungarian, Chinese, Czech, Hindi and Croatian languages, were always right more than half the time, which suggests that there are resemblances between sound and meaning apparent to men everywhere and that these have played some part in the development of all natural languages with the result that semantic rules in totally unfamiliar languages do not seem to us to be quite arbitrary.

Brown/Black repeated the experiments with a group of Chinese- speakers. The results were somewhat discouraging. The Chinese matching words with Hindi and Czech, though still producing results somewhat above chance, did less well (particularly for Czech) than English-speakers. In particular, on the MAL/MIL contrast which was introduced into the test, the Chinese responded at the chance level. these results meant that Brown had to reach a much more tentative conclusion about universal phonetic symbolism than he had earlier proposed: "Totally problematical is the existence of a universal human phonetic symbolism. The phonetic symbolism of the English speaking community may be entirely a result of similar linguistic training... [The Chinese experiment] has shown that the answer to the questions of universal phonetic symbolism will not be easily obtained and when obtained is not likely to be a simple one. There may after all be some associations of sound and meaning that are universally known and others that are a cultural product"(Brown 1958:) On Brown's rather negative conclusion, one can comment that it is apparently only in relation to Chinese that the matching of pairs of words from different languages was less satisfactory. It may be relevant to note that for non-tonal languages, the appropriateness of sounds to meanings of words involves a relationship with the phonemes which form the common basis of these languages. In the case of tonal languages, like Chinese, a quite different principle is found to convey meaning, that is, the pattern of pitch of the voice in the speaking of the word. It would not be surprising if speakers of non-tonal languages should find more difficulty in observing the sound- appropriateness of words in tonal languages or if speakers of tonal languages found more difficulty in judging the phonetic appropriateness of words from non-tonal languages.

Maltzmann/Morrisett/Brooks(1956) argued that if the matching of English/Japanese and English/Croatian by English-speaking subjects succeeds with more than chance probability and this is really the result of universal phonetic symbolism, then it should not matter whether the experimental subjects know either of the two languages from which the words are drawn which are to be matched. It should therefore be possible to match the words of two languages unknown to the experimental subjects. However the experiment - with American subjects matching Croatian and Japanese word lists - produced results not above the chance level. The difference from the earlier successful experiments was that the subjects were not matching pairs of words but single words from each language.

Brackbill/Little(1957) Six groups of 40 English-speaking subjects were asked to judge equivalences of meaning for 3 sets of 50 pairs of words. The words were a random sample of high- frequency concepts and the three sets were pairs of words in English/Chinese, English/Japanese, English/Hebrew, Chinese/Hebrew. The subjects were able to give English equivalents of Hebrew words and to pair Japanese/Chinese and Japanese/Hebrew words at better than chance level. However, in pairing Chinese/Hebrew, correct guesses were significantly below chance. The authors interpreted the results of the study as contrary to the hypothesis of a universal phonetic symbolism: they did not find that subjects were able to guess the English meanings of Chinese words significantly above chance. But the process of matching pairs was facilitated by marked contrasts of meaning. They concluded that (apart from the Chinese/Hebrew failure) non-cognate languages may employ similar patterns of sounds (and length of words) to designate similar meanings - at least among frequently used words.

Miron(1961) undertook a cross-linguistic investigation of phonetic symbolism using invented words. Such invented words have expressive symbolic value according to their inherent phonetic content and not to any meanings they may have via real-word associates. These affective meanings were found to bear consistent lawful relations to the phonetic properties of sounds. He concluded that the laws governing phonetic symbolism may be of a universal character.

Taylor/Taylor(1962) studied phonetic symbolism between unrelated languages. Phonetic symbolism occurred under all conditions in the three experiments they conducted but the symbolism was not universal; their results pointed to language-dependent mechanisms - symbolism was not identical in the different languages used. So, for example, they found that in Korean the initial letters T and P in single syllable words are associated with very big magnitudes, whilst they are associated with very small magnitudes in English. The languages they used, English, Japanese, Korean and Tamil were assumed to be historically unrelated.

Weiss(1964) The process of matching words from different languages is facilitated if the context is restricted. If for example it is known that the topic is 'tables' correct matching is facilitated. The evidence is that the phonetic symbolism is not based on similarity of initial consonants.

Johnson/Suzuki/Ohls(1964) Certain letters and letter combinations occur more often than would be expected by chance in affectively pleasant words while others are over-represented in unpleasant words. Artificial words were constructed using these letters and combinations of letters.After hearing the invented words, the subjects were asked to match English words expressing pleasant or unpleasant feelings, attitudes etc. with one or other of two invented words. The appropriate matches were made significantly more often than chance expectancy.The experimenters concluded that their findings supported the notion that phonetic symbolism exists.

Aztet(1965) An experiment was run using 60 native speakers of the Navajo language who were asked to match Navajo antonyms with their Chinese and Hindi equivalents. The results were negative. Examples of the antonyms used: Beautiful ugly, Bright dark, Up down, Wet dry, Fast slow, Big Little, High low, Cool warm, Push pull, Thin fat. It is uncertain what weight should be attached to the results of this experiment. Navajo is also a tonal language but in a way very different from tonality in Chinese.

Ertel/Dorst(1965) using an approach similar to that of Weiss found confirmation for expressive sound symbolism in twenty five languages. They asked native speakers to make tape recordings of terms of emotion in the different languages. The subjects who listened to the tape recordings had to decide whether the sound sequences had a 'positive' aspect (good, happy etc) or a 'negative' aspect (bad, sad etc) i.e. a decision on feeling tone. Such matching succeeded in all languages with a probability in excess of chance.

Taylor(1965) Experiments matching antonyms do not answer the fundamental question, what sound has what meaning. For example, the Japanese words AKA 'red' MIDORI 'green' show some common features with the English words but which, if any, of these common features is identified with 'greenness' or 'redness'? The concept of phonetic symbolism has to be modified. A new hypothesis must be found that accounts not only for the fact that people associate certain sounds with certain meanings, but also the fact that people speaking different languages associate the same sounds with different meanings.

Peterfalvi(1970) repeated with French language-speakers experimental work which previously had been carried out in the United States and Germany. The experimental results suggested that the symbolism of a sequence of phonemes derives equally from the consonants and the vowels although the part played by the vowels is better established and more constant. Phonetic symbolism (or at least the feeling that it exists) is not restricted to artificial words invented for the purpose. The matching of words referring to movement is more adequate (in terms of the relation of sound to meaning) than for visual words. There is correspondence between the physical and symbolic character of sounds for three categories: between sounds and meanings, between visual patterns and meanings and between sounds and visual patterns.

Koriat/Levy(1977) examined the symbolic implications of vowels in Japanese and Hindi. They found that the sound-patterns carried cross-culturally consistent symbolic connotations.

Though the evidence is strong, on the basis of the research, for the view that sound symbolism is a reality within individual languages and for a sound symbolism extended to cover non-cognate languages (but not necessarily all languages - tonal languages such as Chinese look to be exceptions), little explanation is offered of the natural basis for phonetic symbolism (assuming that the cross-linguistic results rule out any simple cultural explanation).

5. Resemblance between words for the same percepts in different languages

If one examines the words in a large number of languages for a range of commonly encountered objects, actions or qualities (parts of the body, colours, simple actions, air, water, earth, pronouns, demonstratives), one finds that resemblances of sound and meaning spread across many languages, related and unrelated, including languages which geographically and in terms of language family are extremely remote from each other. The resemblances go beyond anything that can plausibly be explained as the result of chance or coincidence. Linguists have often noted the extent of these surprising resemblances, quoting particularly notorious examples such as 'have' in English and 'habere' in Latin (which etymologists say are unrelated to each other), 'path' and 'bad' in English and 'path' and 'bad' in Persian (also said to be unrelated), 'fire' and 'feuer' in English and German which are said to be unrelated to 'feu' in French, 'day' and 'dies' in Latin (not related), 'whole' and 'holos' in Greek (meaning 'whole' but unrelated to the English word).

Comparative linguists have reacted to these resemblances in two opposed ways: a few have argued that the extensive resemblances between distant languages in vocabulary can only be explained by assuming that all or nearly all languages, at some great distance in time, were related to a single parent language (Swadesh built up his procedure of glottochronology on the basis that the degree of resemblance between the words in different languages could be used to calculate how many centuries or thousands of years earlier the different languages developed from a common base). Other linguists, the orthodox majority, dismiss any idea that all languages are related or derived from a single proto-language. Over the centuries there has been a very great deal of effort put into looking for resemblances between particular languages (to suggest that Indo-European languages are related to Hebrew, or Basque to Georgian or to any number of other languages) and also some effort, though on a smaller scale, to support the contention that all languages are related and derive from one primeval language. Comparative linguists have not been convinced because they argue that similarity of vocabulary by itself proves nothing about the family relation of languages; what matters is systematic structural relationship.

Whether or not lexical resemblances indicate relationship, the fact remains that the resemblances do really exist and call for some explanation. A straightforward approach to the problem is to prepare a list of words referring to a range of ordinary objects, actions, qualities and relations, chosen on the basis that they are words most unlikely to have been borrowed into English and that for the most part they are likely to be acquired by children from direct experience rather than by explanation and instruction. This list of (English) words can then be set in parallel with the equivalent words in an extensive range of other languages (including particularly languages unrelated in any way to English). One can then examine what in fact is the extent of far resemblance is confined to related languages and how far it extends to unrelated languages and whether the words in other languages listed for a particular object etc. resemble one another even if they do not resemble the word found in English.

When a list prepared in this way (for some 40 objects and 23 languages) was examined, striking resemblances between remote languages were apparent. A few examples:

(1)CRAB: out of 23 words for 'crab', 14 began with K or C or had K as a prominent sound in the word; for the remainder of the words in the list (that is referring to objects etc. other than 'crab'), K as an initial sound or as a prominent sound was rare; similarities between some individual unrelated languages for 'crab' were Korean 'ke', Swahili 'kaa', Japanese 'kani', Telegu 'kappu', Latin 'cancer' Malayan 'kepiting'.

(2)CUT: out of 23 words for 'cut', 15 began with K (or C) or had K as a prominent sound ('cut' was the only other word, besides 'crab', out of the list of 40 to have K (or hard C) as an initial sound). Similarities between unrelated languages were Arabic 'kata', Chinese 'ko', Telegu 'koyu', Japanese 'kiru', Greek 'keiro', Hebrew 'karat', Spanish 'cortar'.

(3)LICK: 14 of the words began with L or had L as a prominent sound. Similarities: Hebrew 'likek', Basque 'milikatu', Greek 'likhmadzo', Hungarian 'nyal', Latin 'lingo' or 'lambo', Swahili 'lamba' (one Amerindian language had as its word for 'lick ' the word 'lambi').

(4)NAME: 12 of the words began with N whilst a further 8 out of the 23 had N or M as a prominent sound. Similarities: German 'name', Japanese 'namae', Turkish 'nam', Malay 'nama', Finnish 'nimi', (an Amerindian language had 'nami' and Chinese has 'ming').

(5)SAME: Resemblances: Finnish 'sama', Malay 'sama', Arabic 'sawa', Telegu 'sawa', Lozi 'swana', and Russian 'samyi'.

This is only a very small sample of resemblances to show the basis for the general conclusion reached that in the case of a good number of words for common objects etc. (where borrowing between languages or language families is unlikely) there are resemblances even between remote languages which can hardly be explained as the result of coincidence. Cross-language sound symbolism seems a more likely explanation.

If sound symbolism is operating in this way, then it may also explain the parallelisms between sound and meaning in groups of words for related percepts etc. In English, as mentioned earlier, there are groups of what Bloomfield described as intense forms, words with similar initial sound-structures which share some aspect of meaning although etymologically they are unrelated. The following matches one of these groups of English words with groups of words with similar meanings in Malay and French:
















It appears that Malay and French form imitative and expressive words for these related meanings through the use of sound- sequences similar to those familiar in English.

6. Sound symbolism matching sounds or visual shapes with invented words

The following account draws on Roger Brown's(1958) discussion of these forms of sound symbolism. Onomatopoeia in its limited sense (imitation of natural sounds) is a familiar and well- recorded form of sound symbolism:

Wissemann(1954) conducted an experiment to investigate the principles on which onomatopoeic words are formed. The pattern of the experiment was as follows: The subjects were German-speakers. They were asked to listen to 14 different artificial noises. Example: the noise of a wooden ball rolling down a plywood board to drop into a metal box. For each of the noises they were asked to invent an appropriate new word or to select an appropriate word from a list of words invented by the experimenter. In the outcome, the length of the words chosen or invented was found to be related systematically not to the duration of the noise but to the number of distinct divisions in the noise. Noises starting abruptly were matched with words starting with a consonant such as P T K; a noise beginning gradually was named by a word beginning with a consonant such as S or Z. Thus initial speech- sounds of onomatopoeic words reproduced the stimulus gradients of the noises to which the words referred. Tongue position in articulation was more important for phonetic symbolism than lip position. At the conclusion of the experiment, the subjects were asked to explain the basis on which they had chosen the onomatopoeic words. All of them, without discussion or instruction, said they had used similar principles in selecting the onomatopoeic words.

Wissemann's results were in line with what one might expect and with what can be seen in studying familiar onomatopoeic words. Rather more interesting and surprising has been experiment on the relation between visual shapes and words chosen to match individual shapes. There seems evidence from the experiments for a kind of visual onomatopoeia. Usnadze was perhaps the first to explore this though the experiments of Sapir and Kohler are better known and better tested.

Sapir(1929) An English-speaking subject is asked to consider the appropriate name for each of two tables, one large and one small. The word for one of the tables is MIL and for the other MAL. With five hundred subjects, Sapir reported that some 80% thought that MAL represented the larger table and MIL the smaller. Other researchers have repeated the experiment many times since, with similar results, using subjects speaking different languages and from countries with quite different cultures.

Newman(1933) extended Sapir's experimental pattern and was able to place the vowels on a scale of magnitude from small to large. In general larger magnitudes were associated by experimental subjects with a large oral cavity, a low vocalic resonance and the tongue towards the back of the mouth. However, as indicated earlier, Newman found no systematic confirmation for vowel- symbolism as such in natural English words.

Usnadze(1924) conducted the earliest exploration of 'visual onomatopoeia'. He presented his ten (German-speaking) subjects with a set of invented 'nonsense' drawings and asked them to select from a list of invented words appropriate names for them. In some instances the drawing was reminiscent of a familiar object and the subject chose an invented word similar to the conventional name. In other cases the subject chose a name with the same Gestalt or form as the picture. Roger Brown(1958), from whom this account is drawn, says that Usnadze did not attempt to explain the basis of the association between the words and the visual shapes. Brown suggested that the number of syllables in the words might correspond to the number of perceptual sub-wholes in the drawings. Usnadze's work has been continued by other Georgian psychologists (Natadze 1966) and by Fox(1935).

Kohler(1929) devised an experiment on rather similar lines to investigate the symbolic relation between visual patterns and the sound-structure of words. He made use of two arbitrary visual shapes, one rounded and the other angular and invented two words as names for them, TAKETE and MALUMA. Subjects were asked to associate with each shape one of the invented words. The overwhelming majority of subjects assigned MALUMA to the rounded shape and TAKETE to the angular one. His experiment has been repeated many times in many countries since then. The results have been equally striking and have been taken as demonstrating a so far unexplained parallelism between visual and auditory structures.

Whilst it is relatively easy to see how invented words might match invented sound-sequences (on the lines of Wissemann's experiment), it is rather harder to imagine the kind of relation that might exist between invented words and visual shapes (the kind of experiment conducted by Kohler and Usnadze). The following page illustrates what is involved. It shows Sapir's and Kohler's material and also a set of six arbitrary visual patterns to be matched with six invented word-forms. For the reader who wishes to attempt to identify the word matching each particular visual shape, a few notes of guidance:

1. The invented words are in alphabetical order, that is, not placed opposite the visual shapes to which they refer; 2. To judge the appropriateness of a word to any shape, articulate the word. Do not just look at the word or read it silently. Do not attempt only to match the spelling to the shape. What matters is the pattern of movement of the mouth, tongue etc, the feel of the word as it is spoken. 3. Consonants should be pronounced as in English; vowels should be given standard values. 4.It may be helpful to use gesture, that is, to match the contour of the visual shape by moving the arm. As the gesture is made, consider the appropriateness of any particular word to the particular shape.

7. CONCLUSION: Explanations for sound symbolism

The evidence for the reality of sound symbolism in various forms seems strong. That it operates within any single language can hardly be doubted. That it can operate between languages and language-communities to a considerable extent also seems to be established by extensive research. The existence of a universal sound-symbolism (that is, applying uniformly to all languages) seems less certain - tonal languages seem to form a distinct group from non-tonal languages in this respect. The operation of specific consonantal or vowel sound symbolism (phonetic symbolism in the narrow sense) has not been established; sound symbolism appears to be symbolism of the whole word (morphosymbolism) rather than of the individual phoneme. It seems probable that sound symbolism comprises not only traditional onomatopoeia (words imitating natural sounds) but also visual and auditory onomatopoeia (a relation between visual shapes, invented sound- patterns and invented words) as well as a symbolism related to less easily defined qualities (pleasant and unpleasant sensation, emotional values etc).

If sound symbolism is a reality, how is it to be explained? Most researchers have felt the need for some explanation but have not been able to suggest what form it might take. Hans Marchand(1959) devoted a detailed study to symbolism in English word-formation but confessed himself at a loss to find out what the symbolism is based upon. Peterfalvi noted that research had established the facts of sound symbolism but without interpreting them.

Nevertheless, the explanations offered by those few authors who have attempted to understand the basis of sound symbolism show remarkable convergence. Jakobson/Waugh(1987) say that sound symbolism derives from an inmost, natural similarity association between sound and meaning; sound symbolism is an undeniably objective reality founded on association between the different sensory modalities, particularly between vision and hearing.

De Diego(1973) Like other forms of gesture, articulatory gesticulation is directed towards expressing reality. The sense of the appropriateness of a word to its meaning is due to coincidence of the reality which it represents with our own internal organisation. Sound symbolism is associated with the interrelation of different modes of perception; stimulation of one form of perception automatically activates other modes of perception.

Peterfalvi(1970) There is a specific psychological basis for sound symbolism, Correspondence between the sound-structure of the word and its meaning is an aspect of synaesthesia, derived from the generalisation to other modalities of sensory stimuli in the particular modality to which a word refers. The sound-meaning matching for words referring to movement is more adequate than for visual words; a word which evokes impressions of movements may be thought more appropriate because it triggers off mechanisms similar to those which are involved in the actual source of sound symbolism. This seems not far removed from Sokolov's(1972) discussion of the relation of speech and thought (developing ideas originating with Sechenov). Sechenov stressed the enormous importance of the 'muscle sense' in objective- pictorial and in verbal-abstract thinking; owing to this 'muscle sense' (kinaesthetic perception), various sensory impressions are integrated into one complex whole.

These approaches fit well with the motor theory of language origin and function. This stresses the necessarily close relation in the central nervous system between motor control, perception in all its forms and language. Sound symbolism - a feeling of a natural relation between the sound-structure and meaning of words - is precisely what the motor theory would lead one to expect. This is synaesthesia (including Sechenov's 'muscle-sense') given a much more specific content and forming part of a larger theoretical system.