Stanislav Voronin’s Universal Classification of Onomatopoeic Words: a Critical Approach (Part 1)

classification. Conclusion. Introduced half a century ago, Stanislav Voronin’s classification of onomatopoeic words still remains a useful tool of typological research. Critical additions and proposed changes do not lessen its impact on studies in linguistic iconicity. The first part of this paper is devoted to the description of the classification and to the discussion of its advantages and limitations. In the second part of the article some possible solutions to the detected problems

approach towards onomatopoeia used by the author was, undoubtedly, novel as he based his classification not on the semantic principle as it was widespread at that time (and still is) but on the principle of iconic relation of onomatopoeic words to their (psycho)acoustic denotata. Such approach allowed to classify all English words denoting sounds into five major groups (those denoting pulses, dissonances, non-pulses etc.) disregarding their specific semantic affiliation (sounds of nature, bird calls, human sounds, mechanical sounds etc.).
Almost immediately it became evident that such approach made the classification applicable not only to the material of the English language, but to other languages as well. The universality of the classification was tested on Indonesian [2,3], Bashkir [4], Estonian [5], Georgian [6] and Turkish [7]. The division of onomatopoeic words proposed by Voronin holds true even for invented languages [8].
However, the classification has been updated several times after it was first published in 1969. The majority of innovations were introduced by S. V. Voronin himself [9,10,11] and [12], some were added by other researchers [13]. The main aspect in which the current version of the classification [14, p. 44-66] differs from the original one is that in 1969 Voronin did not distinguish onomatopoeic and sound symbolic words. In the later version the author made a clear-cut distinction between acoustic imitation (onomatopoeia) and articulatory imitation (sound symbolism).
Currently, new synchronic [7,8] and diachronic [13] data calls for yet another update on the classification.

Universal classification of onomatopoeic words.
The main challenge of the 'Universal classification' was revealing the principal acoustic parameters (properties) of sound-denotata which define the choice of type of phonemes comprising an onomatopoeic word [14, p. 39]. Studying the (psycho)acoustic properties of various types of denotata allowed S. V. Voronin to create the classification of onomatopoeic words based on iconicity principles.

Parameters of acoustic denotata.
S. V. Voronin defines five main parameters of acoustic denotata of onomatopoeic words in the following way [14, p. 40]: -Parameter I -pitch (in a broad sense); reflects both the basic frequency and the specter of sound (psycho-acoustically, pitch as such and the tone quality of a sound).
-Parameter II -volume; the acoustic correlate of the volume of a sound and its intensity.
-Parameter IV -periodicity (the periodicity of sound waves); according to this parameter, sounds split into tones (tonal non-pulses) and noises (non-tonal non-pulses).
-Parameter V -dissonance; it is a type of sound when the pulse series is long enough to be perceived as a durative sound but is too rapid for the ear to distinguish the individual pulses.
These defined five main parameters, according to [14, p. 41], constitute the base for an objective evaluation of sound and revealing classes and types of sounds relevant to the onomatopoeic subsystem of the language: -In accordance with Parameter I (pitch), a sound of any type may be classified quantitatively as 'low' or 'high'.
-Parameter III (time) enables to outline two qualitatively different classes of sound: pulses and non-pulses. If applied to non-pulses, this parameter involves the distinction of short and long sounds (the parameter of duration).
-According to Parameter IV (periodicity), non-pulses split into two qualitatively distinct classes -tone and noise.
-The dissonance (pulse series) -together with its antithesis, the non-dissonance -is constituted in accordance with parameter V (dissonance).

Types and characteristics of acoustic denotata.
Upon analyzing these parameters S. V. Voronin [14, p. 42]  He distinguished three kinds of dissonance (pulse series): 1) dissonating pulse (quasi-pulse); 2) pure dissonance (pulse series proper), and 3) dissonating non-pulse (quasi-non-pulse); the latter may be either noise or tonal non-pulse, while noise non-pulse can be defined as pure noise or tonal noise.
Considering that non-dissonance includes pulse and non-pulse, while the latter is subdivided into noise and tone (the noise can be both pure noise and tonal noise), S. V. Voronin (Type III), and Pure Dissonance (Type VI) are simple sounds both acoustically and psycho-acoustically, as they are indivisible into elementary, simpler sounds with contrasting properties. We would suggest a sound of a hit as an example of a simple sound (Type I. Pulse). Other types are acoustically complex, decomposable into elementary simple types with contrasting properties. As an example of a complex sound the author gives the sound of buzzing: it is perceived as a simple sound, irreducible to other elements; nevertheless, this type of sound is acoustically complexit is a Tone-Noise Non-pulse having appreciable elements of the both tone and noise).

Types of onomatopoeic words.
According to Voronin [14, p. 44], 'the properties of the acoustic denotatum of an onomatopoeic word are defined by its characteristics and the properties of its source'. Thus, 'a classification of onomatopoeic words [should be done] according to their correlation with denotata' [ibid.].
To the main three classes of denotata defined in the previous section (A. Pulses; B. Non-Pulses; C. Dissonances (or a rapid series of pulses)) correspond three classes of onomatopoeic words: A The structural models for words of one onomatopoeic class differ from language to language. Voronin [14, p. 48] suggested that for languages with CONS + VOC syllabic structure, general model of instants shall be: PLOS/AFFR/CLICK + VǑC.

Class B. Continuants.
Continuants is a class of onomatopoeic words denoting non-pulse natural sounds, that is, sounds of 'prolonged' and 'coherent' duration (which are not divided into shorter segments).

Type II. Tonal continuants.
Tonal continuants render tonal non-pulse sounds or tones in their purest form. The examples of tonal continuants given in [14, p. 49 tonal characteristics -vowels, sonorants (especially nasal sonorants) as well as /j/ and /w/, and voiced consonants (especially voiced fricatives). Therefore, all of these phonemes can be used for rendering natural tones.

Type IV. Tone-noise continuants.
Tone-noise continuants imitate tone-noise non-pulses, which combine traits of pure noises together with noticeable tonal elements. Some examples of tone-noise continuants given in [14, p. 52] are: English buzz 'a rapidly vibrating humming sound, as that of a prolonged z or of a bee in flight'; whizz 'to make or cause to make a loud humming or buzzing sound', Bashkir syz 'sizzling of fat on a frying pan'. The structural model of the English tone-noise continuants (which also can be applied to Bashkir) is:

Class C. Frequentatives.
Frequentatives are a class of onomatopoeic words denoting a rapid series of pulses where each pulse is hardly perceived separately yet there is no complete fusion of pulses into one tone. Such rapid sequence of pulses is highly irritative for acoustic perception. Such sequences of pulses are perceived as dissonances; therefore, one can also name the class of frequentatives 'onomatopes-dissonances' [14, p. 53].

Type V. Frequentatives quasi-instants.
Frequentatives quasi-instants are onomatopoeic words denoting quasi-pulse sounds [14, p. 54]. Examples of words belonging to this type given in [14, p. 53] are: English crack 'to break or cause to break with a sudden sharp sound; snap'; chirp 'to make a short high-pitched sound'; Indonesian rik 'a sound of a broken twig or bone'.
The structural models for frequentatives quasi-instants in different languages differ. Thus, for Indonesian onomatopes Voronin suggests a model with R-formative (on RL-formatives see [15]) outside the root: Rf + PLOS + VǑC L/H, S/W + PLOS.

Type VI. Pure frequentatives.
Frequentatives denoting pure dissonances (sounds perceived as a rapid series of pulses, thrills) Voronin calls 'pure frequentatives ' 14, p. 55]. His examples of pure frequentatives are: English jar 'to make or cause to make a harsh discordant sound'; Bashkir byrr 'a noise made by the vibration of wings of small birds when they fly up'; Chuvash тырр [tyrr] 'sounds of movement, whirling', Indonesian rai 'a sound of multiple corns falling'.
According to Voronin [14, p. 55] dissonances are in-between pulses and non-pulses; consequently, pure frequentatives are in-between instants and continuants. The rapid alternation of pulses leads to irritation of hearing perception and such series of pulses are deciphered as dissonant sounds. That's why there is the phoneme /r/ in a phonetic structure of onomatopes-pure frequentatives. Structural model for pure frequentatives in English and Bashkir is the following [14, p. 55]: CONS + VǑC L/H, S/W + R.
It should be noted that in the present-day British English the model is only historical as /r/ is not encountered in the post-vocal position.
In Indonesian pure frequentatives have the following structural model: R + VǑC. There is also another structural model for Bashkir onomatopoeic wordswith an R-formative [14, p. 56]:

Type VII. Frequentatives tonal quasi-continuants.
Frequentatives tonal quasi-continuants denote tonal quasi-non-pulses, a type of natural sound denotata containing elements of pure dissonances and tonal non-pulses at the same time [14, p. 56]. The examples are: English screak '(dial.) to scream, to creak'; Bashkir lar(r) 'a roar'. The common structural model of the English and Bashkir onomatopes belonging to type VII is: An additional, specifically English, structural model is [14, p. 57]:

Type VIII. Frequentatives pure noise quasi-continuants.
Frequentatives pure noise continuants denote pure noise non-pulses which contain elements of both dissonances and pure noises [14, p. 57]. Voronin's examples for Type VIII onomatopoeic words are: English: whirr 'to fly, revolve, or move rapidly with a humming sound'; Bashkir šyptyr-šyptyr 'rustling of leaves, dry grass'; Indonesian ras 'imitation of the sound made by dry leaves touching each other'.
The structural model for frequentatives pure noise continuants of these three languages is [14, p. 58]: The author also gives a number of Bashkir and Indonesian models with R outside the root, where itis an R-formative [14, p. 58], for example:

Type X. Tonal 'post-pulse' instants-continuants.
Tonal post-pulse instants-continuants denote tonal 'post-pulse' sound sequences of a pulse followed by a non-pulse; that is, of a sound abruptly beginning with pulse and ending in a tonal non-pulse [ibid.]. Voronin distinguishes (a) short tonal post-pulse instants-continuants (English dump 'to throw down or out roughly'; plump 'to throw down or out roughly'; clank 'an abrupt harsh metallic sound'; Bashkir dömp 'a muffled sound of an abrupt hit') and (b) long tonal postpulse instants-continuants (English tang 'a loud, ringing sound; twang'; clang 'to emit a loud resonant ringing sound as of pieces of metal struck together'; Bashkir taŋ 'a sound of something wooden hitting metal surface').

Type XI. Pure noise 'post-pulse' instants-continuants.
Onomatopoeic words belonging to this type denote pulse-like sounds followed by a pure-

Type XII. Pure noise 'pre-pulse' instants-continuants.
Onomatopoeic words belonging to this class denote pulse-like sounds preceded by a pure noise [14, p. 61]. The examples of pure noise 'pre-pulse' instants-continuants are: English: flap 'to move or cause to move noisily back and forth or up and down'; whit 'a shrill abrupt sound, as a bird's chirp'; Bashkir sabuy 'to mow the grass'; Buryat шаб [shab] 'an imitation of swishing of a whip'. A structural model for Type XII onomatopoeic words in English is [14, p. 62]:

Type XIII. Tone-noise 'pre-pulse' instants-continuants.
Tone-noise 'pre-pulse' instants-continuants imitate pulse-like sounds preceded by a tone-noise non-pulse [14, p. 62]. The examples of Type XIII onomatopoeic words are: English zip 'a light sharp sound such as that produced by a bullet or other small or slender object passing rapidly through the air or through some obstacle'; Bashkir wyăt 'a whoosh of air accompanying a swift movement'. The model for the English tone-noise 'pre-pulse' instants-continuants is [14, p. 62]:

Type XIV. Pure noise-tonal 'pre-and post-pulse' instants-continuants.
Onomatopoeic words belonging to this type denote pulse-like sounds preceded by a pure noise and followed by a tone (which is a resonance 'ending' of the pulse) [14, p. 63]. Voronin distinguishes (a) short pure noise-tonal 'pre-and post-pulse' instants-continuants (English thump 'the sound of a heavy solid body hitting or pounding a comparatively soft surface', slump 'to sink or fall heavily and suddenly', whump 'a dull thud') and (b) long pure noise-tonal 'pre-and post-pulse' instants-continuants (whang 'to strike or be struck so as to cause a resounding noise'; whing 'a sharp high-pitched ringing sound'). The model for the English pure noise-tonal 'preand post-pulse' instants-continuants is [14, p.

Type XV. Tone-noise tonal 'pre-and post-pulse' instants-continuants.
Onomatopoeic words belonging to this type denote pulse-like sounds preceded by tone-noise non-pulses and followed by a tonal non-pulse [14, p. 63]. Voronin distinguishes (a) short (English zonk '(slang) an imitation of a short resonant blow') and (b) long (English sing '(slang.) to whoosh with a buzzing, whistling sound, usually said about bullets') subtypes. The onomatopoeic words of this type have the following structural model [14, p. 63]:

Hyper Class CAB. Frequentatives quasi-instants-continuants.
This hyper class of onomatopoeic words denotes complex natural sounds combining the traits of dissonances and pulses preceded or followed by non-pulses [14, p. 64].

Type XVIII. Frequentatives pure noise 'pre-pulse' quasi-instants-continuants.
Onomatopoeic words of this type denote a quasi-pulse preceded by a non-pulse; and this non-pulse is always a noise (English flirt 'to move in a jerky manner') [14, p. 66]. The onomatopoeic words of this type have the following structural model [ibid.]: The whole classification is schematically presented in figure 1 (after [11, supplement 2]).

Limitations of the classification.
The universal classification of onomatopoeic words (UCO), undoubtedly, was a breakthrough in the field of phonosemantic studies. However, in the course of work on [16] and, later, on [17] I came across several problems regarding its implementation (listed below from major to minor):

Problem 1. Place of frequentatives in the classification.
The frequentatives as a class of onomatopoeic words, according to the parameters of the UCO, render 'a rapid series of pulses where each pulse is hardly perceived separately yet there is no  Thus, Voronin arrives to the basic contrast of three 'simple' sound types to which correspond three major classes of onomatopoeic words:

AB. INSTANTS-CONTINUANTS CAB. FREQUENTATIVES-QUASI-INSTANTS-CONTINUANTS
A. Pulses ↔ A. Instants B. Non-pulses (tone/noise) ↔ B. Continuants (tone/noise) C. Dissonances ↔ C. Frequentatives These correspondences match (1) the 'high' points of the (psycho-)acoustic contrast division of the sounds perceived by a human ear and processed by a human brain and (2) the most contrast units of phonemic inventories of the languages which are chosen for the imitation of the perceived acoustic phenomena in onomatopoeic words. Thus, the groups three 'simple' sounds correspond to the groups of onomatopoeic words belonging to 'pure' classes.
But when it comes to frequentatives (see figure) the principle is not applied fully. Frequentatives are divided, according to [14, p. 53] into frequentatives quasi-instants; pure frequentatives; and frequentatives tonal quasi-continuants.
Thus, 'complex', 'mixed sounds' (frequentatives quasi-instants; and frequentatives tonal quasi-continuants) appear already on the same level of classification with the simple ones. Instants and continuants do not 'cross' with each other, their 'hybrid' is ascribed to a separate hyper-class -AB. Instants-Continuants. On the other hand, both the hybrids of (1) frequentatives and instants and (2) frequentatives and continuants do not form hyper-classes of their own, but are placed on the same level as 'pure' frequentatives [14, p. 53], which leads us to the second major problem of the UCO.

Problem 2. What are pure frequentatives?
'Pure' frequentatives apart from being placed on the same level with frequentatives of 'mixed' types themselves present a problem. According to [14, p. 55] they denote 'pure dissonances (sounds perceived as a rapid series of pulses, thrills)'. Thus, as instants, pure frequentatives should be sole representatives of the class and have no elements in their structure juxtaposed other 'pure' classes.
However, the examples of pure frequentatives given by the author are puzzling. He suggests such English words as chirr, birr, burr, jar(r) as well as Bashkir byrr 'a noise made by the vibration of wings' and Chuvash тырр [tyrr] 'sounds of movement, whirling' as representatives of pure frequentatives. Thus, he draws up a structural model with the core element bearing an imitative function (R) in auslaut: CONS + VǑC + R. This model describes both English and Bashkir onomatopoeic words from his sample. However, Voronin gives no arguments for this particular model with R in auslaut neither in [1], nor in [14] or in [11] editions of his UCO. Thus, a pure frequentative is an onomatopoeic word denoting a vibrant, harsh, dissonant sound which contains (1) any consonant apart from r; (2) any (short) vowel and (3) R of some quality. However, such a model for pure frequentatives presents several problems: (1) Initial consonantsin the given examples plosives or affricatesalso have some onomatopoeic function (rendering abrupt, pulse-like soundssee above), which is for some reason overlooked in pure frequentatives.
(2) Even in modern (British) English -VR is a prohibited combination of phonemes, thus the whole class 'pure frequentatives' is only hypothetical in modern synchrony (even if it was not in the 16 th century before the start of the regular sound change which made it impossible). Thus, as a candidate for a model describing the whole class in a universal classification the CONS + VǑC + R model is a poor candidate. It is not only inapplicable to English, but also to a number of languages with CV-syllable structure as well.
He attempts to do the same for Indonesian pure frequentatives: R + VǑC. Here, R unexpectedly (!) moves to the anlaut position. In English, however, frequentatives quasi-instants and frequentatives quasi-continuants and not pure frequentatives have R in anlaut position (see above).
(3) R itself (whether a trill, a tap or a retroflex) might not be a phoneme in certain languages, but an allophone, thus its second allophone or another consonant phoneme from a phonemic inventory might take its imitative function of conveying harsh, thrill-like dissonant sounds in onomatopoeic words. This hypothesis, however, requires verification as all languages subjected to the UCO so far, contained a rhotic consonant of some quality.
(4) Voronin introduces an R-formative to the models of pure frequentatives in Bashkir onomatopoeic words: CONS + VǑC + CONS + Rf. And this is our Problem 3 (see below).
Thus, pure frequentatives are not only placed on the same level as frequentatives of 'mixed types' in the classification, but also are not given a clear modeltheir core element (R) is (1) placed in the very unusual (for many languages, including the present-day British English) or even non-existent auslaut position; (2) is combined with core elements of other classes (e. g. plosives) or even (3) is placed outside the root (as an R-formative).
In his monograph Fundamentals of Phonosemantics Voronin devotes a whole chapter to the discussion of their origin, meanings and semantics [14, p. 111-118]. RL-formatives, according to Voronin (who followed the argumentation of [18, p. 273]), not only are encountered in imitative (especially, onomatopoeic) words, but also themselves are (1) of imitative origin [14, p. 111] and (2) render the meanings of plurality, iteration and repetition on par with their onomatopoeic function [14, p. 118]. (1) conveying the repetition of a denoted sound or action (cf. twit and twitter, chat and chatter); (2) indicating intensiveness of meaning (German klappern 'to make a rattling noise'); (3) conveying an iterative nature of a denotatum (Indonesian keretak 'the sound of footsteps on a wooden floor'); (4) designating the notion of plurality or multiplicity (Indonesian gerbak 'the sound of small fruit falling on the ground'; gelebak 'the sound of several small fruit or books falling').
Voronin, thus, on the one hand, describes RL-formatives as imitative affixes having a specific set of meanings of their own [14, p. 111-118] and, on the other hand, incorporates them into his UCO. Thus, following Voronin, English chat is an instant [14, p. 47]; but chatter is a frequentative-(quasi)-continuant. The majority of the Indonesian examples given in [14] for frequentatives of various types are examples with R-formatives.
One of the questions arising is: if R-formatives are included into structural models and accounted for in the classification, why L-formatives are not accounted for as well, as they have exactly the same four imitative functions listed above?
The other, more fundamental, question iswhy affixes are included into the classification at all?
More specifically, why are they accounted for the frequentatives only and not for other

classes?
Does it not undermine the main principle of the classificationthe presence of iconic correspondence onomatopoeic root / a type of natural sound? Why one should include affixes, even if they are (presumably) imitative in origin into the classification? And why these particular affixes? RL-formatives are not universal (e. g. they are not encountered in Slavonic languages). Also, there are other language-specific imitative affixes (e.g. English intensification prefixes ker-/ ka-/cha-as in ka-boom!). Should they be included into the models as well? And here we arrive at our fourth major problemstructural models.
The question is, what do these structural models actually reflect? In [13, p. 90] I arrived to the conclusion that these structural models reflect present-day phonotactic constraints of a language applied for one-syllable content onomatopoeic words.
This means, that (1) imitative interjections which violate phonotactic constraints of a language (a salient property of imitative interjections widely discussed e. g. in [20] or [21]) are left out; (2) polysyllabic and (3) borrowed words cannot be described with these models; (4) 'old' as well as borrowed onomatopoeic words also 'fall out' of the UCO.

Problem 5. Structural models change in diachrony.
One of the main conclusions of [13] is that structural models for onomatopoeic words change over time [13, p. 91]. All 'atypical' content monosyllabic native onomatopoeic words of the English language were coined earlier then the 17 th century [13, p. 92], and the number of 'atypical' onomatopoeic words increases with their 'age' [ibid.].
Thus, the structural models distinguished by Voronin are only applicable to SD-2 words (words on SD-3b are not accounted for in [14]). SD-2 words are words on the second stage of de-iconization [13, p. 126].
De-iconization is the gradual loss of iconicity caused by simultaneous acting of regular sound changes and regular sense development of the word [13, p. 120]. Altogether, I distinguish four stages of de-iconization: -An SD-1 word is an iconic interjection that may violate language's phonotactic constraints and vary in form (English zzz!, cling-clang!, ding-dong!).
-An SD-2 word is a content word with conventional form which hasn't undergone any regular sound changes and still retains its original meaning related to sound (to clap, a tap, to hoot).
-A word on SD-3 is a content word which has either undergone one or several regular sound changes (SD-3a) but retained its original meaning (laugh, chirp, knock); or it has an intact form (SD-3b) but has lost its original meaning (clip, cliché).
-An SD-4 word is a content word indistinguishable from the rest of the 'conventional' vocabulary, and the discovery of its original onomatopoeic nature requires an etymological analysis as both its form and its meaning have changed dramatically (e.g. gargoyle).
The reason why the structural models are valid only for SD-2 (and SD-3b) words id simple. As the models reflect the phonotactic constraints of a language in modern synchrony, any changes in phonemic inventories and phonotactic rules automatically lead to: (1) the change of these models in diachrony (e.g., the model for frequentatives-(quasi)instants has changed from (s)PLOS/AFFR + VǑC + R + PLOS to (S)PLOS/AFFR + VŌC + + PLOS for the British English and to (s)PLOS/AFFR + VOC R + PLOS for the American English [11, p. 85]; (2) the fact that single words which have undergone regular sound changes cease to 'fit into' existing models (e.g. knock /nɒk/ after the kn > n/#_ conditioned change does not currently fit into the structural model of instants which is (s)PLOS + (SON LAT/NAS/DENT ) / AFFR + VǑC + PLOS).

Problem 6. Unnecessary high number of types and structural models which complicates the classification.
According to the UCO, there are: for English: altogether three classes, two hyperclasses, 18 types of onomatopoeic words described by 30 structural models [11]; for Turkish: three classes, two hyperclasses and 16 types of onomatopoeic words [7]; -for Indonesian: three classes, two hyperclasses, 10 types and 21 structural models [2]; -for Georgian: three classes, two hyperclasses, 19 types of onomatopoeic words [6]; -for Bashkir: three classes, two hyperclasses, 15 types of onomatopoeic words [4]. The number of classes and hyperclasses remains stable in the studied languages and seems to be a language universal. They are applied to all onomatopoeic words of a language (including to polysyllabic if they are divided into segments). The number of types and models is different and reflects the present-day phonotactic constraints of a language applied for one-syllable content onomatopoeic words.
The structural models differ from language to language and describe only a part of an entire onomatopoeic lexicon of a language (69-86 % of the English onomatopoeic lexicon (see above), 68% of the Bashkir onomatopoeic lexicon [4]). They reflect language-specific structural characteristics of the studied languages.
The question isshould the UCO be based on the models (some of which describe only 2-3 words) or be only limited to the level of classes and hyper-classes? The structural models (the number of which is, in my opinion, unnecessarily large) do not reflect the peculiarities of onomatopoeic words -they merely describe the boundaries within which onomatopoeic imitation is possible at the synchrony of a particular language.
The six problems discussed above I consider the major problems of the classification. There is, however, a number of minor problems to be addressed:

Problem 7. The role of the affricates in the classification.
As known, affricates are a group of phonemes holding an intermediary position between stops and fricatives (they begin as stops and are released as fricatives). Their imitative function, according to Voronin [14, p. 67] equals that of stops. For this reason, they are included into the structural models of instants -(S) PLOS + (SON LAT/NAS/DENT ) / AFFR + VǑC + PLOS. Thus, such English words as chop or chuck are instants, according to Voronin. However, 'pure', easily definable sounds should not be rendered by phonemes of 'mixed' nature (according to the principles of the UCO).

Problem 8. The role of the sonants, laterals and approximants in the classification.
This problem is similar to the one pointed above. Sonants (m, n) are either included as 'extra' elements with no clear imitative function (see the models for instants above) or are ascribed the imitative function of rendering prolonged sounds (for example, in tonal post-pulse instants-continuants: PLOS/AFFR + VǑC L / H,S/W + SON NAS ). The same problem is with the approximant /w/ and its semivowel quality. It is not reflected in the models but is ascribed the function of rendering a tone.

Problem 9. The role of voice in the classification.
The models proposed in the UCO sometimes make a distinction between voiced and voiceless fricatives (e. g. cf. models for pure noise continuants (FRIC ʌ / (CONS) + VǑC L/H, S/W + + (CONS) / FRIC ʌ ) and tone-noise continuants (CONS + VǑC L/H, S/W + FRIC v )). However, voice is never a distinctive phonosemantic feature in plosives. The question is to what extent is the voiced: voiceless opposition is phonosemantically significant in onomatopoeic words and should it be included in the classification?
Another side-issue is the distinction of tone-noise continuants (whizz, buzz) which have the model CONS + VǑC L/H, S/W + FRIC v and pure noise 'post-pulse' instants-continuants (piff) with the model PLOS / AFFR + VǑC L / H, S/W + FRIC ʌ . We see that they only differ in the presence and absence of voice (the same refers to frequentatives (quasi-)instants-continuants and frequentatives tone-noise quasi-continuants). Pure tone and pure noise continuants (of which tone-noise continuants are supposed to be a cross), however, imply other means of imitation, which leads us to the next problem.

Problem 10. Vowel length as a distinctive feature of tonal continuants.
Vowel length is not a universal feature of phonetic systems of various languages (even for English it is now only historical), yet it is a core element for the English tonal continuants -[CONS (+SON LAT / LAB ) + ] VŌC L / H,S/W (+ PLOS). For the languages lacking the long: short vowel opposition it appears that any vowel is a core element of an onomatopoeic wordcontinuant.
The question is then, how to classify the [CONS]+VOC+[CONS] structures in such languages?
Also, Voronin [14, p. 49] states that apart from (long) vowels several other types of phonemes have tonal characteristicsvowels, sonorants as well as /j/ and /w/, and voiced consonants. However, these consonants are not always included in the models or included as 'extra' elementssee the discussion above.
The second part of the article is devoted to the possible solutions for the indicated problems.