Motivation in the word initial consonant onset.
Andrew McCrum




'Sound symbolism', here, is taken as a descriptive term for all kinds of non-arbitrary relationships between phonetic content and meaning. Sound symbolism occurs both language specifically and cross-linguistically. An example of a language specific sound symbolic item is the NE onset [skw], <squ->, representing the semantic concept 'compression' as in squ-eeze, squ-ash, squ-at (Rhodes and Lawler 1981: 332). Cross-linguistic sound symbolic items are the phones [w], [v], [f] representing the various concepts 'air', 'wind', 'gas' and 'flying', in many languges, often word initially.

1.1 Terms and definitions

'Sound symbolism' is conventionally recognized in the broad manner described above. Other terms for sound symbolism vary from lexical class (ideophone) to sign system category (icon, index). In between are a variety of terms used with varying interpretations. 'Onomatopoeia' is the term traditionally used, in English since the 16th century, for direct sound imitation through phonetic form: 'the coining of a word which attempts to represent a non-linguistic sound by a combination of appropriate segments selected from the ordinary phoneme inventory of the language' (Trask 1994: 247). But a range of interpretive angles are taken some of which are discussed below. 'Phonetic symbolism' was coined by Sapir (1929) and extensively used by American experimental researchers such as Newman (1933), Bentley and Varron (1933) and Brown et al (1955) from the 1920's to the 1970's. This term appeared in psychology journals such as the Journal of Abnormal and Social Psychology and Psychological Bulletin. More latterly in Europe Peterfalvi (1970) and Fischer-Jørgensen (1978), amongst others have used the term. It has now, by and large, been replaced by 'sound symbolism'.

'Phonetic symbolism' most often denotes the non-arbitrary representation of a phone by a range of specified semantic criteria. See 1.9.4 for a detailed description. These criteria were most often categorised by researchers into polar adjective pairs (as in large/small) and matched with the phone, or phone group ('voiced'/'voiceless plosives' in Peterfalvi 1977: 79-88) by subjects during experiments. The matching of adjective to phone or phone group is governed by the subjective impressions of the subjects to the phone relative to their native language. For example, the phones [i], [I] are matched with the pole 'small' in several English studies (Newman 1933; Crockett 1970). The same phone has also been identified with the term 'angular' (Bentley & Varron 1933; Weiss 1964a; Tarte & Barritt 1971). Matching of concept to phone is often made through the semantic areas of scale and geometrical configuration, for example square, triangle, pointed, flat (French 1977: 311) in addition to various other categories such as 'thin/thick', 'light/heavy' (Westermann 1933: 166). These are then sometimes correlated to produce larger conceptual domains, particularly 'potency', 'activity' and 'value' in Miron (1961) and others. So, the polar adjectival categories 'quiet/noisy' and 'passive/active', with others, correlate to form the conceptual domain 'activity'. Peterfalvi and others use the correlating categories grandeur, angularité, clarté ('size', 'angularity' and 'brightness'), (Peterfalvi 1970; Weiss 1964a). In experiments a phone or phone group, e.g.[n] , scoring highly in several categories with the same polar value, for example 'quiet' and 'passive', will therefore appear polarised on the concept domain, in this case having a low 'activity' score. To this group, 'phonetic symbolism' is less concerned with lexical or linguistic than cognitive properties of phones. This is firstly because in one common group of experiments the phones under study are presented in nonsense words, (e.g. CVC for Sapir 1929 and Bentley and Varron 1933; CC onset for Markell and Hamp 1960). Secondly, the correlated categories are often semantically superordinate to the meaning of the lexical form within which the category members, e.g. vowels, are present in the lexicon. So the category 'size' is superordinate to that of 'small' denoted by the terms 'little' and 'itsy-bitsy' in which the phoneme /I/ is present. Sapir also calls phonetic symbolism 'expressive symbolism' and 'a psychologically primary sort of symbolism' ([1929] 1985: 61-2).

Jespersen's understanding of the term 'symbolic' is in relation to the sound symbolic history and use of words, principally ones expressing a relation between the phone [I] and smallness (Jespersen 1933). Jespersen's 'symbolic value' of [I] is a socio-linguistic observation determined by its lexical representation and the semantic and pragmatic environment in which it is used. The socio-linguistic context is often etymologically and lexicographically illustrated as in 'c.f. NED Tom Tit' (1933: 289) and the list of cognate forms for nit 'egg of louse' (289) as in OE hnitu, Russ. gnida, Lett. gnída, ON gnit, G nisse. Other brief, but specific, details such as register and geographical range flesh out this skeletal comparative framework: for example the literary register of Shakespeare for 'Minikin 'little'' 1933: 287; the folk mythological register of nixie 'water-elf' (290); 'colloquial Dutch koppi 'little hill' (294) and slang 'nipper sl. 'boy'' (289). Or, more commonly, a colloquial or social register is implicit in an idiomatic or collocational phrase as in 'as little as ninepins' (287). Jespersen also covers a number of geographical linguistic distinctions within the confines of Indoeuropean giving uses of diminutive [I] in different languages as in 'Lat. minimus', different dialects, 'Norw. dial. pirre 'little'' (287) and isolects 'jittern 'nickel' local U.S.' (292). He also identifies value associations and pragmatic uses of sound symbolic i as in pejorative 'Nit ..'also contemptuously applied to a person'' (289), the 'emotional' content of NE 'little' (286) as in 'little marvel/fool', or 'chick 'a term of endearment''. And for hypocoristics, or diminutives of endearment, Jespersen makes the observation: 'that children will often add an -i at the end of words' citing American, English and German research (297). One proviso is that Jesperson refers to a square bracketed 'sound type' [i] as IPA [I] and gives I.P.A. [i] as [i:] but fails to make the distinction explicit in his lexical list.

Similarly dependent on the linguistic environment, but morphologically described, a 'submorpheme' (see 1.3) is recognised by Bolinger where monosyllabic or disyllabic words with a common onset or rhyme component show some consistency in meaning. So, MdnE cr- is the submorphemic component of the morphemes cr-ouch, cr-eep, cr-inge (Bolinger 1950: 120). The non-arbitrary relationship between the sub-morpheme cr- and the semantic component of the morphemes crouch etc. assigns to cr- the sound symbolic meaning 'bent'. Bolinger and Markell and Hamp (1961: 55) also use the term 'psychomorph' referring to a cognitive and inhherent relation between form and meaning.

Firth, who coined the term 'phonaestheme' (1930), Samuels (1972) and Wales (1990) regard the socio-linguistic development, 'phonetic habit', and history of certain 'phonaesthetic' forms as central to their meaning. Firth regards initial and final phone groups as affectively marked, for example the onset sl-. Sl- is present in words in the Germanic and Balto-Slavic languages many of which have a pejorative sense attached to them (see McCrum 2000): so MdnE slack 'loose; indolent, careless, remiss' from OE; MdnE sl-obber 'behave (e.g. feed) in a slovenly fashion' from ME; and following the Second Sound Shift of the Old High German period [s] > [S], <sch>; G. sch-lampen 'to be sloppy in one's work' from the 14th century. The expressive, subjective function of these words used in an appropriate social context allows the propagation of, here, a pejorative meaning of 'looseness', 'inactivity' and 'carelessness' in the common feature of the marked onset. (See 1. 6.4.2. for a fuller discussion) The pejorative meaning, in conjunction with these general semantic areas, then spreads through the respective languages by derivation, borrowing or semantic change through analogy; e.g ENE slipshod 'wearing slippers or very loose shoes' > NE 'untidy, slovenly' and ME slop 'muddy place'/slop 'loose fitting garments' > NE sloppy 'negligent'. Phonaesthemes need not have a pejorative meaning. The onset fl- represents 'rapid motion' in fl-icker, fl-utter, fl-are. But a pejorative sense, as in slut, slob, has a greater social, stigmatizing and therefore cognitive function than simply expressive words such as flutter. Stigma is interpersonal and memorable through name-calling, abuse and animosity. Phonaesthemes are 'accumulated associations' to Wales (1990: 348). The 'phonaesthetic habit' is formed from the accumulated affective use of onset and/or rhyme (Firth 1933: 184). The term 'phonaestheme' thus describes a phoneme or pattern of phonemes apparent below the morphemic level and used for certain 'expressive', that is aesthetically and sensorily, salient meanings (c.f. Wales 1994: 339).

The proponents of 'phonaesthemes' as habitual do not regard them as necesssarily violating the arbitrariness of form and meaning. Wales relates 'phonaesthesia' to two neologistic principles: i) 'salience' in pronunciation, where the more common combinations of phones are preferred, e.g. an onset C + l/r and short vowels in preference to long and, following Samuels (1977) and Firth; ii) analogy, whereby some forms come to have analagous meanings as a result of motivational derivation and borrowing from a 'nucleus of roots'. So the NE /sk/ onset 'suggests a frisky movement in ME skip (probably from Old Norse), skim (from Old French), scuttle (dialectal: c.f. scud(dle)), skid (seventeeth century), scurry (nineteenth century), scoot (nineteenth century, US) and skedaddle (nineteenth century)' (Wales 1994: 347-8). Some 'blends' are also regarded as 'phonaesthemes' where onset and rhyme have independant senses on a sound symbolic level (Jesperson 1922: 312-3).

'Synesthesia' is identified by Jakobson and Waugh (1979: 188) and Peterfalvi (1970: 153-157) as the cross sensory association between sound, principally vowels, and colour. Trask gives concisely 'the subjective sensation of a sense other than the one being stimulated' (Trask 1994: 346). In its clinical form it is an uncommon psychological condition. The idea of 'tone colour', or audition colorée, which is usually mentioned in the same breath as 'synesthesia', is a sensorily impoverished version and looser term for the former describing commonly held and perceptible relations between either i) acoustic pitch or ii) phonetic pitch (usually of vowels) and colour or various emotional criteria. The whole area of 'synesthesia' and 'tone colour' attracted the interest of the earliest sound symbolists such as Tolman (1887) and Binet (1892). But the psychological condition is often not distinguished from phonetic perceptions. Binet provides a simple and clear distinction (Binet 1892: 336). A more recent complication is where 'synesthesia' is increasingly used to refer to pervasive and 'universal' sound symbolism. Priestley regards 'synesthetic' as synonymous with universal remarking that: 'synesthetic (or: universal) sound symbolism...coexist in every language' (Priestley 1994: 237). Hinton, Nichols and Ohala describe synesthesia as the consistent representation of 'visual, tactile, or proprioreceptive properties of objects such as size or shape' by certain vowels, consonants and suprasegmentals (Hinton et al 1994: 4). As I see it sound symbolism can not be truly synesthetic for a number of reasons, principally because neither speech nor meaning are senses, that is either smell, touch, taste, hearing or vision.

'Ideophone' is probably the most widely used sound symbolic term, mostly for African, Austroasiatic and Australian languages (Samarin 1970, 1971; Alpher 1994: 164; Childs, 1994). Coined by C.M. Doke (1935) ideophones are sound symbolic lexemes with varied syntactical realisations. They typically 'express some particular and distinctive sound or movement' (Trask 1996: 176). Definitions have been the subject of debate for years but Childs' summary is one of the best. He identifies ideophones prototypically as phonologically marked, violating the prosody of the matrix language, as grammatically adverbial, syntactically isolated from the rest of the clause and with a high expressive semantic content with regard to colour, smell, action or manner. They are also prototypically morphologically stable, pragmatically functional (that is restricted according to social environment) and often share a paralinguistic category with gesture (Childs 1994: 180-1). See the relevant sections below for further discussion. Onomatopoeic ideophones sometimes show cross-linguistic sound symbolism. The association between labial fricatives and 'moving air' is common to Kisi, Hausa, ShiNzwani, Yoruba and Gbaya; so, Hausa fur , fur, fir-fièr-fir 'flutter of wings', Kisi foo 'wind whistling', ShiNzwani fwii 'sound of rapidly passing by' (Childs 1994: 190). Sometimes ideophones are language specifically sound symbolic as in Yoruba pétepète 'be muddy, soggy'. (Courtenay 1976: 19). Beyond these prototype definitions there are some wide variations making category membership open to extension by argument.

Smithers (1964) analyses various Germanic and English sound symbolic consonant clusters as ideophonic. These 'ideophones' are largely OE and ME words. They are principally verbs with intervocalic double consonants (-ff-, -bb- and so on). However, ideophones constitute a lexical class which, in addition to the more salient phonological and semantic peculiarities, are further characterised by the range of linguistic characteristics illustrated above. Phonological and semantic marking is evident in the English verbs Smithers cites (e.g. OE gaffetan 'to mock' > eMdnE gobble 'to swallow in large mouthfuls' (Smithers 1964: 104), but they are not as phonologically distinct as many ideophones. And the morphology of English verbs and their syntactic function is regular unlike most ideophones. Also, ideophones generally show some independance from the rest of the clause or utterance (Childs 1994: 187). So, Smithers' examples would be at the periphery of a prototypical core of ideophones and should not be regarded as typical. But his is a well argued case and even more relevant is the implication that many NE verbs derive from ideophones which were commonly in use in a less culturally developed and remote stage of Germanic and even Indoeuropean than now.(See 1.6).

In the early stages of research into ideophones Westermann coined the term lautbilder 'sound picture' for ideophones which qualify entities, 'the overall impression of a whole shape' as in Twi: kuhaa 'shaggy' (Westermann 1937: 169), or events 'the movement of the whole shape' as in Ewe: bOhObohO 'a fat man walking in a clumsy, wobbly manner' ('der Gesamteindruck, die ganze Gestalt, oder die Begwegung der ganzen Gestalt'), (Westermann 1937: 159).

Japanese sound symbolism has been traditionally separated by Japanese scholars into two or three classes: giseigo or giongo 'sound imitative words', pyuu 'whizzing'; gitaigo 'imitation of movement or manners of the external world', pera-pera 'fluently' and gizyoogo which symbolize mental conditions or states, ira-ira 'nervously'. (Shibatani [1990] 1996: 153-7). Matisoff lists Amanuma (1974) and Kakehi, H. et al (1981) for the lexicon (in Japanese). See Matisoff (1994: 119-20) or Shibatani for a terminological discussion.

Although classifiers are not generally regarded as sound symbolic Firth alludes to the classifying particles of Chinese, Burmese and Bantu as sound symbolically motivated (Firth 1933: 187). Also see n1 above. More recently, Rhodes & Lawler and Rhodes describe NE sound symbolic onset and rhyme elements as 'classifiers' (Rhodes and Lawler 1981: 327-336; Rhodes 1994). Rhodes and Lawler (1981) is a detailed attempt to equate sound symbolism in English with traditional classifier systems which are predominantly nominal and not sound symbolic. This revealing study gives many sound symbolic associations previously unaccounted for such as thr- 'constricted', dr-, 'liquid' and -oop 'a curved path'. Rhodesís recent study gives a more comprehensive list of ëliquidí ëclassifiersí as dr-, sl-, fl- and m-. (1994: 287). However, there are a couple of reservations with this definition. Firstly, 'classifier' categories are not satisfactorily defined in Rhodes and Lawlers' study but listed as onset modifiers of a rhyme head comprising a descriptive referent which is usually nominal or adjectival: so pl- 'layer' sp- 'cylinder' (Rhodes and Lawler 1981: 331). This is fine. 'Classifiers' largely denote the numerical or semantic category of a nominal head. But most English onset clusters listed by Rhodes and Lawler are grammatically and syntactically marked for verbs not nouns. McCrum makes the point that unaffixed NE cluster onsets account for a proportionally higher number of verbs per word than nouns according to the number of consonants in the onset (McCrum 1997). Compare the many sp- verbal sound symbolic onsets that is those 'expressing jet movement' (Marchand 1979: 406) e.g. 'spit' or movement from a location, SP2, or to a location, SP5, (McCrum 1997: Appendix 2). Statisticallly by type, using the Concise Oxford Dictionary, between 20% and 30% of lexemes with the sC- onset (30 % to 50% for sC(C)) are semantically event-denoting and grammatically verbal compared to 10% - 15 % for those with an sV-onset. This is a rough but consistently valid guide for English, depending to a small degree on the variables of the size of the lexicon and the difference between verb categorisation and grammatical use. So, whereas classifiers are nominally defining, those words defined by Rhodes and Lawler's 'classifiers' often function as verbs by type, as well as token in the case of spl- where verbal use-frequency = 4177 per 100 million contrastively compared to nominal use at 2677 (see British National Corpus). More significantly, these 'classifiers' are only evident as 'submorphemes' whereas conventional classifiers are morphemic (Allen 1977: 285). Classifiers also constitute a closed grammatical class and are almost universally syntactically restricted to noun phrases not full predicate forming verbs with a host of syntactical functions and valency options as the spray/load verbs attest; a group heavily weighted with #CC(C)- lexemes (Levin 1993: 117-9).

However, some atypical verb stem classifiers exist in the Athapaskan languages (Allen 1977: 287; Palmer 1996: 141). Moreover, reference to English sound symbolic onset and rhyme elements as classifiers seems to be gaining ground. Also, the work of Rhodes and Lawler makes a significant contribution to the perennial problem of classifying the fused sound symbolic submorphemes of English and provided much corroborative data for McCrum (1997). 'Classifier' to Rhodes and Lawler is then a generalised term for a number of classifier-like properties of onset and rhyme. Latterly, Rhodes has coined the terms 'wild' and 'tame' for, at each end of a scale, i) phonemically transcribed, as in [?mÑ"Ñ~: 545], and ii) lexically represented onomatopoeia as in moo (Rhodes 1994: 279).

Malkiel's exhaustive and enlightening studies have provided a variety of terms covering the relation between areas of non-arbitrary phonological and morphological representation. 'Phonosymbolism' and 'sound symbolism' are most commonly used. Both these terms relate to non-arbitrary sound meaning relations sometimes conforming to a lexical template which is diachronically productive giving a range of meanings which are often adjectival or adverbial (Malkiel 1990: 99-110). Malkiel also uses 'primary symbolism' for onomatopoiea and 'secondary symbolism' which is any other form of sound symbolism, often submorphemic. Crockett uses 'secondary onomatopoiea' for submorphemic sound symbolism (Crockett 1970). Rhodes and Lawler (1981, 1994), McCrum (1997) and others recognise the term 'phonosemantic' for studies concerning the semantic aspect of sound symbolic relationships. As a broad term for 'sound symbolism' this has had widespread recent use.

'Frequency code', named by Ohala (1982, 1994) and identified by Sapir (1911, 1927), Jesperson (1933), Newman (1933), and others is identified by Ohala as the widespread sound symbolic function of F0 and F2. (See 1.8 for a discussion). This relationship has two central sound symbolic functions: pragmatic and diminutive/augmentative. Pragmatically, human articulations and other animate vocalisations with a high pitch correspond to a regard for the goodwill of the receiver and sounds with a low pitch correspond to self-confidence and self-sufficiency (Ohala 1994: 343).

Lately, Sereno (1983, 1994) has used the term 'phonosyntactic' to refer to the grammatical realisation of sound symbolic vowel contrast. Here, English front vowels /i/, /I/, /e/, /E/ and /{/ in a stressed syllable are significantly more widely represented in verbs than nouns (62% versus 43%) and back stressed vowels, /@/, /A/, /V/, /O/, /o/, /U/, /u/, /aI/, /aU/, /OI/, in nouns (38% versus 57%). Figures are expressed as percentages of the first 200 high frequency nouns and verbs listed in Francis and Kucera (1982) representing words which have use-frequenices greater than 250 per million. Consequently, these findings were frequency dependant occurring only in high frequency words. Interestingly, few 'expressive' sound symbolic words, for example 'screech', would feature in this high frequency category.

Also, McCrum (1997) has used the term 'neurosemantic' for the direct relation of tactile, kinaesthetic and acoustic phenomena, through mechanoreceptors, to the use of consonants in English in words conveying semantics of action and space relations. The neurophysiological basis of this study centres around the presence of a number of organised neuron clusters, called mechano-receptors, in areas of high articulator activity. Mechano-receptors convey detailed information on dimension, pressure, velocity and time through the afferent nerves back to the central nervous system. McCrum claims that this accounts, for example, for the predominance in English of the /spl/ onset in dynamic verbs. 'Dynamic' involves the movement of something in time, with a degree of force or pressure from one location or situation to another. Whilst I refer to terms used in this study and one or two statistics This approach is discussed in more detail in 1.8. The term 'neurosemantics' is not to be confused with 'neurolinguistics' which is a term for the neural activity behind speech coordination in the central nervous system. See Catford (1977) for neurolinguistic programming in the phases of speech. 'Neurolinguistics' is also associated with linguistic pathology, for which see Luria (1976) and Steinberg (1993: ch.9).

'Icons', 'indexes' and 'symbols', are systems of signs identified and named by the American philosopher C.S. Peirce. These have been used extensively by natural morphologists for which see below 1.9. See a review of the use and meaning of this terminology by natural morphologists in Carstairs McCarthy (1992: 214-247) and a general description by Lyons (1977: 99-109). All are terms for the abstract criteria of sign systems. More specifically they comprise a signalling system linking the object of perception, the perceiver and the method by which she/he perceives. The three terms given above are one form of Peirce's divsion of signs. (There are ten different classes of sign in all, not necessarily intersecting). These can be simply conceived of on three different levels based upon the non-hierachical relationship between cognition and form. A 'symbol' is the cognitive reception of arbitrary signs, the mental effect which depends on the presence of an 'interpretant' (Lyons 1977: 100-6). An index is a sign which does not require the presence of an interpretant but does require the presence of its object or form. One example of an ëindexí, given by Pierce, is of a bullet hole as a sign for a bullet whether or not there were an interpreter to perceive the hole. An 'icon' 'is a sign which would possess the characater which renders it significant even though its object had no existence' (cited in Lyons 1977: 102). Peirce gives an example of an icon as a lead pencil streak which represents a geometrical line. According to Lyons 'iconicity' is dependant upon some natural resemblance, geometrical or functional, between the sign and its object' (Lyons 1977: 103). So, because of the closeness of the relationship between the straight lead mark and the cognitive model of a 1D geometrical construct the cognitive model can independantly and completely denote the mark and sign whether or not such a sign is present. Sign language is typically iconic, for example where the referent has a similar shape to the sign made by a hand gesture. 'Icon', 'iconism' and 'iconicity' is used by many recent authors for sound symbolic relationships which are felt to closely link semantic interpretation with phonological representation (e.g. Joseph 1987, 1994; Dressler 1987: 101; Wescott 1971, 1973; Anderson 1998).

2 Compare the Japanese classifier -wa, 'flying and hopping animals' with Berlin's ethnozoological study where /w/ is widely represented in bird names in Guaraní, the national language of Paraguay, (Berlin 1994). Berlin, however, does not refer to this fact. See also Alpher (1994: 163) and Childs (1988: 185, 1994) for some Australian and African languages containing labial fricatives and approximants in words for wind, vapour and gas. Oswalt gives similar examples from the Southern Californian Indian language Pomo (Oswalt 1971: 178). A range of Latinate and Germanic words also exists; see Chastaing (1965: 314-7) and Marchand (1979: 412).

3 See also Adams (1973), Bolinger (1950: 130) and Austerlitz (1994) among others. See Binet (1892), Karowski et al (1942), Masson (1951), Peterfalvi (1970: Ch's IV-VII), French (1977: 319),

4 Fischer-Jørgensen (1978: 88), Jakobson and Waugh (1979: 188) for observations on 'synesthesia'.

5 See Diffloth, (1972: 440) for a condensed and now, perhaps, dated but excellent world wide ideopheone bibliography.

6 This statement regards urbanisation, industrialization and communication technology as remote from both the history of individual cultures and the majority of world wide cultures.

7 Matisoff lists gitaigo as 'attitudinals' and does not give gizyoogo.

8 'Classifiers', recognized for at least 100 years (e.g.Pétrus Truo'ng Vinh K¦ 1883; Lee 1988: 225f), were classically categorized by Allan, (1977: 285) and defined as a morpheme which 'denotes some salient perceived or imputed characterisitic of the entity to which an associated noun refers'.