Ideophones in Arusa Maasai: Syntax, morphology, and phonetics

3.1 Syntax

Any ideophone may be used holophrastically in Arusa, thus functioning as an autonomous and non-elliptical utterance. For instance, in (1) below, a girl (speaker A) carries a jar of water. She suddenly stumbles and, as a result, drops the jar. To inform her friend that she has dropped the precious object, she employs the ideophone ɓúss. Her friend (speaker B) understands the message immediately and inquires about the reason of this misfortune, to which the girl responds that she tripped over a stone. In this example, the ideophone ɓúss constitutes a self-standing complete utterance that succinctly and effectively communicates the idea of dropping something, such that no other explanatory clause is needed to convey the intended message.

All ideophones may also be used within the frame of a larger sentence without entering into syntagmatic relationships with the elements of the core clause and without functioning as that clause’s structural components. In such cases that approximate the holophrastic usage discussed earlier, ideophones appear at the margin of the sentence, typically on its left edge, and are separated from the core clause by a notably audible pause. The core clause itself may be semantically equivalent to the ideophone, roughly constituting its synonym; it can comment on the ideophone; or it restricts the ideophone’s interpretation. Example (2) illustrates these properties: the ideophone pío, depicting a bad smell, fails to entertain any syntactic role in the nearly synonymous core clause eŋu ene wei náléŋ ‘this place is stinking a lot’; it appears in the left periphery of the sentence; and it is phonologically detached from that sentence’s remaining part.

Apart from the holophrastic and asyntagmatic uses described earlier, ideophones may also be employed as structural elements of core clauses. Three main types of syntagmatic functions are available to ideophones in Arusa: the predicate, part of a complex predicate, and the verbal modifier.

A limited number of ideophones – specifically, 17 tokens (approximately 25%) – may function as predicates. This usage is grammatical with it͡ʃúkut͡ʃúkɔ́ɔ, ɗɛ́ɗɛ́, ɗóssu, ikúm, kúrr, ɪ́kɔ́ɾɔ́ɾɔ́m, ŋám, ŋís, ŋób, ipíd, irús, isúŋusúŋ, isúrr, itíak, itús, itɔ́ktɔ́k, and wúap. When employed predicatively, ideophones exhibit a lexeme-specific argument structure as is typical of all verbs. They require a subject argument and, in the case of some lexemes (e.g., ɗɛ́ɗɛ́ and ít͡ʃúkut͡ʃúkɔ́ɔ), an object argument. Ideophonic predicates can thus be intransitive (3.a) or transitive (3.a–b). The subject and object arguments may be encoded through verbal inflections, e.g., 3sg. subject e- in (3.a) and the combined 3sg. subject and 3sg. object e- in (3.c).^[9] These arguments can also be expressed lexically – see the subject ɛnkɪ́nɛ ‘goat’ in (3.a) and olayíoni ‘(the) boy’ in (3.b), as well as the object olabóbóki ‘bark’ in (3.b) and eŋúdi ‘stick’ in (3.c). In all such instances, the position of the ideophonic predicate is analogous to the position occupied by genuine verbal predicates. That is, in unmarked predicate-focus contexts, the ideophone is placed clause-initially before any lexical subject and object as well as adjuncts (3.a–c).^[10]

In addition to being used as predicates, several ideophones may be introduced by quotative – speech or performative – verbs (so-called verba dicendi and facendi), e.g., ɉo ‘say’ (4.a) and nitejo ‘do’ (4.b). There are 37 ideophones of this type: 13 tokens are headed by ɉo ‘say’ (e.g., púɾkúʃ, rrí, píd, ɓúss, ɓúh, ʃíɾtít, ŋáɾɓáɓ, púɾúk, rrábal), while 24 tokens are headed by tejo ‘do’ (e.g., kwáh, tús, rús, táɓ, káɓ, kɛ́l, ŋís, ʃwáa, ɗóss, ɗɛ́ɗɛ́u, kwára, kúɾúɾúm, mílígít). In all such cases, ideophones occupy a position that is typical of adverbs, i.e., after the subject (e.g., enkóilíil ‘antelope’ in 8.a) and object (e.g., empíra ‘ball’ in 8.b), if these are expressed lexically. However, contrary to cases in which ideophones function adverbially as verbal modifiers, which will be described further below, an ideophone headed by one of the quotative verbs cannot be omitted. Indeed, its omission yields an ungrammatical or nonsensical utterance.^[11] Therefore, although the position of an ideophone and its quotative verb need not be adjacent, we analyze such constructions as a complex predicate similar to what is observed in Southern Bantu languages (e.g., Zulu and Xhosa, Andrason 2021a). As in Southern Bantu, the quotative verb is lexically empty, being used merely as the carrier of inflections. In contrast, the ideophone, which appears in its base form (see Section 3.2), specifies the semantic type of an event and determines the argument structure of the clause.

A single quotative verb may head more than one ideophone. In (5), the verb ɉo ‘say’ introduces two ideophones that express two events – slipping (ʃíɾtít) and falling (ɓúss) – that took place consecutively. Each ideophone forms a complex predicate with the quotative verb even though the verb itself is expressed only once. As a result, the sentence in (5) can be analyzed as the asyndetic coordination of two clauses, i.e., néjó ʃíɾtít and néjó ɓúss, with the ellipsis of the inflected verb ɉo before the second ideophone.

Finally, ideophones may be used as verbal modifiers. In such instances, ideophones accompany a predicate, usually a verb (although also a predicative ideophone), with their meanings overlapping to some extent. For example, in (6.a), the ideophone ɓúss, which depicts a thudding sound when falling, co-occurs with the verb urori, which similarly expresses the event of falling. Often, ideophonic modifiers specify the way in which the action expressed by the verb is performed, or they determine the intensity of that action (see again ɓúss, which indicates the force of the impact in 6.a). In the former function, ideophones correspond to adverbs of manner, such as réerée ‘quickly’ or sídai ‘nicely’; in the latter function, they correspond to adverbs of degree, such as naléŋ ‘intensely, very.’ In unmarked predicate-focus word order, ideophonic modifiers occupy a postverbal position. If the subject (e.g., entíto ‘girl’ in 6.a) or the object(s) (e.g., ɛnkíríŋo ‘meat’ in 6.b) are expressed lexically, the ideophone appears after all the arguments (see buss in 6.a and mílígít in 6.b). If adjuncts are used, especially adverbials indicating place and time (see ŋole ‘yesterday’ in 6.c), the ideophone is placed before them (see mílígít in 6.c). As a result, the placement of ideophonic modifiers is identical to the position occupied by adverbs of manner and degree. However, despite this word-order tendency, which is regular in colloquial speech, ideophones employed as modifiers may also appear in clause-initial position. This is grammatical – and indeed common – in narratives when telling a story (see rri in 6.d).

The use of ideophones as verbal modifiers is also grammatical with the 13 ideophones that can be employed as predicates (see earlier in this section). Accordingly, ideophones may not only modify a genuine verbal predicate with which they are semantically compatible, but may also accompany their own predicative counterpart. In such cases, both the predicate and the modifier draw on the same or similar ideophonic lexemes (see Section 3.2): the former specifies the type of action that takes place, while the latter usually conveys the nuance of intensity similar to an adverb of degree:

3.2 Morphology

The morphology of ideophones must be assessed separately at two levels: at the level of their bases and at the level of surface forms. As will be evident from the subsequent discussion, for ideophones that are used holophrastically and asyntagmatically, as well as for those that function as verbal modifiers and parts of complex predicates, the surface form is identical to the base. However, for ideophones that can be employed as predicates, the surface form differs from the base by containing (many) other morphological elements. Furthermore, a number of ideophones exhibit two distinct bases: one appears in all non-predicative functions – whether syntagmatic, asyntagmatic, or holophrastic – while the other is only attested in a predicative function.

Most ideophonic bases are monomorphemic and can neither synchronically nor (most likely) diachronically be divided into more elementary independent meaning-bearing units. The most evident cases of morphological simplicity and monomorphemicity are 43 “short” – monosyllabic – bases such as ɓúh, ɓúss, t͡ʃúm, kɪ́rr, pá, páa, pás, rrí, and júk.

Synchronically, the monomorphemic structure may also be attributed to some bases that are more robust and exhibit a bi- or trisyllabic structure. These types of bases might, however, have been more complex and pluri-morphemic from a diachronic perspective, i.e., at earlier stages of the language. Two classes of such ideophones can be distinguished. First, a small group of ideophones exhibits the element -C₁VC₁ in their second syllable: ŋáɾɓáɓ, púr(k)úk, ʃíɾtít, and tírrmám. This element is characterized by an “internal” consonantal harmony (i.e., the two consonants are identical) and an “external” vocalic harmony (the vowel of the element -C₁VC₁ usually coincides with the vowel of the first syllable of the word),^[12] and approximates the phenomenon of “recurrent partials” common in many ideophones in Southern Bantu languages, e.g., Zulu (Fivaz 1963) and Xhosa (Andrason 2020, 140, 152).^[13] Similar to the Arusa unit -C₁VC₁, recurrent partials attest to harmonious tendencies. They exhibit the pattern -C₁V₁C₁V₁, with V₁ usually coinciding with the vowel of the preceding syllable. Although the presence of -C₁VC₁ in Arusa ideophones may, as in Xhosa and Zulu, suggest some original derivative mechanism – i.e., this element may have been added to what now constitutes the first syllable – these ideophones are synchronically unanalyzable into two (or more) independent morphemes. Second, similar to ideophones containing the element -C₁VC₁, the robust form of some other lexemes (e.g., the bi- or trisyllabicity of t͡ʃámbul, kwára, mílígít, púɾkúʃ, and rrábal) and/or their obscure iconicity (e.g., the unclear structural relationship of tírrja and mút͡ʃa, with their respective depictions of quietness and tastelessness) may suggest the secondary status of such ideophones and their origin in other lexical classes. If this is the case, these ideophones could have contained, at least initially, two or more morphemes. Nevertheless, this diachronic relationship – if it ever existed – is now irrecoverable. All such words are synchronically indivisible into more fragmentary units and, thus, exhibit a monomorphemic structure.

Although synchronic monomorphemicity prevails, a group of ideophones exhibits bases that currently (not only historically) have a more complex structure. The most patent exponent of such complexity are reduplicative structures attested, for instance, in ɓúɾuɓúɾu, kílikíli, kíɾíkíɾí, kúlukúlu, láɓláɓ, ʔórʔór, téeté, and kúmukúm. Overall, 12 ideophones are reduplicated out of the possible 26, i.e., those that contain more than one syllable. This means that, while 17% of all ideophones exhibit some types of reduplication, for multisyllabic tokens, this percentage increases to 46%.^[14] Despite the fact that reduplication seems common in ideophones, it does not constitute a genuine morphological process but is rather an expressive and phonetic device. In other words, the reduplicated ideophones – except perhaps for kúrrkúrr which has a non-reduplicated predicative variant kúrr – cannot be divided into singletons that would have some meaning on their own, nor can they be extended by an additional singleton to yield triplicated or quadruplicated sequences. More generally, as far as multiplicative structures are concerned, neither triplication nor quadruplication is attested in ideophones in Arusa. This contrasts with the well-attested presence of multiplicative patterns, triplicated and quadruplicated, in ideophones in many other languages, e.g., in Southern Bantu languages such as Xhosa and Zulu (Andrason 2020, 155–6).

Crucially, ideophonic bases do not contain “ideophonizers” similar to the ideophonizing suffix -iyani/-iyane in Nguni languages (Zondo 1982, Msimang and Poulos 2001, 242–3, Andrason 2020, 154–5). In other words, no ideophone exhibits an affix that would mark ideophones overtly as members of the ideophonic category and/or derive them from other non-ideophonic lexical classes.

Ideophonic bases not only lack any ideophonizing derivations, they also usually fail to contain any morphemes that would mark them for one of the possible syntagmatic uses. This is always true of bases exhibited in ideophones functioning as parts of complex predicates and verbal modifiers. In addition, four bases exhibited in predicative ideophones (i.e., ŋám, ŋís, ŋób, and wúap) lack any element that could distinguish them morphologically as predicative.^[15] This behavior is, however, not particularly surprising: Arusa lacks adverbializers; apart from ideophones, complex predicates with quotative verbs are unattested and, thus, morphemes that would mark non-inflected parts of complex predicates are unknown; verbs are marked in a variety of manners, lacking a uniform verbalizing affix (compare with the verbalizer -a in Southern Bantu).^[16]

Unlike the aforementioned four predicative bases (i.e., ŋám, ŋís, ŋób, and wúap), which are identical to the bases used in all the other functions, 13 ideophones have marked predicative bases (this amounts to 76% of the 17 ideophones that allow for predicative uses; see Section 3.1). The overt marking of predicative bases thus constitutes a strong tendency in Arusa. The following pairs of non-predicative and predicative bases are found in the language: t͡ʃúkut͡ʃúkóo/it͡ʃúkut͡ʃúkóo, ɗɛ́ɗɛ́u/ɗɛ́ɗɛ́, dóss/dósu, kúm/ikúm, kúrrkúrr/kúrr, kúɾúɾúm/ikóɾóɾóm, píd/ipíd, rús/irús, súŋsúŋ/isúŋusúŋ, súrr/isúrr, tíak/itíak, tɔ́kɔ́tɔ́kɔ́/itɔ́ktɔ́k, and tús/itús. From this list, one may infer that the most common manner of marking a predicative base is with the initial vowel i-, attested in ten tokens: it͡ʃúkut͡ʃúkɔ́ɔ, ikúm, ikóɾóɾóm, ipíd, irús, isúŋusúŋ, isúrr, itíak, itɔ́ktɔ́k, and itús. The remaining means of marking predicative bases are erratic, with each one only being attested once. This involves: shortening of the diphthong (ɗɛ́ɗɛ́), shortening of the consonant and the presence of an additional vowel/syllable (dósu), absence of the final vowel/syllable of the reduplicated singleton (tɔ́ktɔ́k), absence of reduplication (kúrr), and change of vowel quality (ɪ́kɔ́ɾɔ́ɾɔ́m). Certainly, from a synchronic perspective, all such ideophones are unfragmentable into separate meaning carriers and can be viewed as monomorphemic. However, we propose that predicative bases are secondary, given that: they are the only bases that exhibit formal differences among ideophones; that in the set of predicative bases, the majority is formally distinguished from their non-predicative counterparts; that predicative bases tend to be more robust than non-predicative bases (except for ɗɛ́ɗɛ́u/ɗɛ́ɗɛ́, kúrrkúrr/kúrr and tɔ́kɔ́tɔ́kɔ́/itɔ́ktɔ́k); finally, that this formal distinction is quite regularly achieved through the presence of the initial i-, the most common vowel present in verbal class II (an inflectional class of genuine verbal bases beginning with a vowel; see Karani 2018, 34).^[17] These secondary predicative bases thus constitute analogical adaptations of primary non-predicative bases to predicative uses and their verbalization. In doing so, they depart from their original forms and approximate the forms of genuine verbal predicates (see Section 3.3 for further phonetic differences that confirm the distinctiveness of predicative bases among all ideophones).

Furthermore, contrary to all the other types, predicative ideophones exhibit surface forms that are always morphologically more complex than their bases. To be exact, predicative ideophones may contain all types of inflectional and derivative morphemes (i.e., stem extensions) that are available to verbs in Arusa. Regarding inflections, predicative ideophones are inflected for person (first, second, and third) and number (singular and plural) like the other verbs. This inflection concerns both the subject and the object. For example, the intransitive predicative ideophone ipíd ‘jump quickly’ exhibits the following subject-inflection forms: 1sg. á-ípíd ‘I will jump,’ 1pl. kí-ímpíd, 2sg. íímpíd and 2pl. ímpídipídi ‘you will jump,’^[18] and 3sg. and 3pl. é-ípíd ‘he/she/it/they will jump.’ Transitive predicative ideophones, such as íkúm ‘push with a bang,’ allow for overt subject and object inflections, e.g., 3sg. > 2sg. kí-inkúm ‘he will push you’ and 1sg. > 2sg. áá-ikúmo ‘I pushed you.’ Ideophonic predicates may also be inflected for the various tenses and aspects, e.g., for perfective (3sg. é-ipíd-o ‘s/he jumped’ and é-íkúm-o ‘s/he pushed it’) and imperfective (3sg. é-ípíd ‘s/he will jump’ and é-íkúm ‘s/he will push it’). In addition, predicative ideophones may contain derivative stem extensions, e.g., the dative (applicative) affix -ak in (8.a), the impersonal -i (8.a), and the reciprocal -ino (8.b). As a result, the surface forms of predicative ideophones do not differ morphologically from any genuine verb.

3.3 Phonetics

In Section 3.2, we explained that ideophonic bases appear as surface forms if ideophones are used holophrastically and asyntagmatically, and in two syntagmatic functions, i.e., those of verbal modifiers and parts of complex predicates. In the case of ideophones employed predicatively, a base is the form devoid of all inflections and derivations that typically accompany genuine verbal bases. We also noted that, for some ideophones, their predicative bases differ from a non-predicative variant. Given that the formal distinctiveness of predicative bases probably results from the analogical adjustment of non-predicative ideophonic bases to predicative uses and thus their verbalization, we will focus on non-predicative basic forms in our discussion of the phonetics of ideophones. Predicative bases (especially those that diverge from non-predicative ones) will be discussed only in the final paragraph of the present section. Furthermore, since inflectional and derivative affixes exhibited in the surface forms of predicative ideophones – whether those with specific predicative bases or those with bases that are identical to the bases found in the other functions – are not the property of ideophones but rather that of verbs, these affixes and the surface forms in which they appear will not be analyzed here.

The majority of non-predicative ideophonic bases – specifically, 43 tokens, which constitute more than 62% of all lexemes – have a monosyllabic phonetic structure, e.g., [ɓúh], [kúm], [pá], and [tɔ́ɔ(ː)]. Bisyllabic ideophones are less common, with 16 tokens (23%) being attested, e.g., [kóɾosː], [kwáɾa], and [mút͡ʃa]. A subgroup of the bisyllabic ideophones exhibits a -C₁VC₁ structure in the second syllable (e.g., [ŋáɾɓáɓ], [púɾkúk], [ʃíɾtít], and [tírːmám]), while some other tokens owe their bisyllabic nature to reduplication (e.g., [kúrːkúrː], [láɓláɓ], and [ʔóɾʔóɾ]). Trisyllabic and quadrisyllabic ideophones are much less frequent. The former are attested four times (nearly 6%) and the latter six times (nearly 9%). The tri- and quadrisyllabic composition of these ideophones almost exclusively results from replicative mechanisms, as illustrated by [kúmukúm] and [ɓúɾuɓúɾu]. The only non-replicative exception is the trisyllabic lexeme [mílígít] (see Section 3.2 regarding the element -C₁VC₁ and reduplicative mechanisms).

Non-predicative ideophonic bases invariably draw on both consonantal and vocalic materials. Conversely, ideophones that only consist of vowels and consonants are unattested. Nevertheless, the contribution of a consonantal component to the phonetic substance of ideophones is (very) significant, especially if one examines the base-initial onsets and base-final codas.

As far as onsets are concerned, 67 non-predicative ideophonic bases (97%) begin with a consonant that is almost invariably a simplex, e.g., [kírː] and [táɓ]. The initial consonant may be long, although this is only found with [r], e.g., [rːí]. In four cases, the consonantal onset may be additionally followed by an approximant: [w], e.g., [kwáh], and [j], e.g., [sjóo(ː)]. The most common onset consonants are voiceless plosives (37 examples): [k] – 15, [p] – 12, and [t] – 10. The affricate [t͡ʃ] appears six times and the two nasals eight times jointly: [ŋ] – 5 and [m] – 3. Two fricatives are attested: [s] and [ʃ], each appearing three times. The trill [r] is attested twice, while the tap [ɾ] and the lateral [l] each occurs once. The voiced stops – always implosive – are attested five times: [ɓ] – 3 and [ɗ] – 2. One lexeme exhibits [ʔ] as its initial consonant. In contrast, word-initial onsets that are not genuine consonants are much less common, being found in only two or three tokens (4%). Two ideophones begin with an approximant: [wúap] and [júk]. Only one lexeme, i.e., [óɾʔóɾ], may begin with a true vowel although it is more often realized with a consonantal onset, i.e., as [ʔórʔór].

Consonants are also a pervasive component in base-final codas. To be exact, genuine consonantal codas are found 47 times (68%). Like onsets, coda consonants are always simplexes. However, the frequencies of the specific consonants used in codas are distinct from those exhibited in onsets. Sonorants, if counted jointly, constitute the most common consonantal material in codas: [l], [m], and [rː] are attested six times each. The fricative [s] is also frequent, being found in seven tokens. The implosive [ɓ] and the plosive [k] appear five times each and the affricate [t͡ʃ] three times. The consonants [h], [ŋ], and [p] are used in the codas of two lexemes, while [d], [ɉ], [r], and [ʃ] feature once. In contrast, only 22 ideophones (32%) do not exhibit a genuine consonantal coda. Specifically, 17 tokens (25%) end in a pure vowel, e.g., [ɓúɾuɓúɾu], [rːí], [téːté], and [tírːja], while 5 tokens (7%) end in a diphthong, typically [-Vu], e.g., [ɗɛ́ɗɛ́u], [káu], and [píu], except for [pío], which ends in [-Vo].

Overall, as far as the syllabic structure of non-predicative ideophones is concerned, the pattern CVC is the most common, being found 19 times (28%). Taking into account forms with a long onset or coda consonant, i.e., CCVC and CVCC, the pattern CCVCC is attested 26 times (38%). Should one also include monosyllabic forms with diphthongized nuclei, i.e., CVV̯C (e.g., [kʊ́ɪk], [píal], and [tíak]), and forms with an approximant following the word-initial consonant, i.e., CAVC (e.g., [kwáh] and [sjóo(ː)]), the total number of C(A)V(V̯)C structures increases to 30 (43%). Any other pattern (e.g., CCVC, CVCV, CVCVV̯, CVCVC, VCVC, CVCCVC) is much less common.

Ideophonic bases – whether predicative or not – do not contain sounds or sound combinations that would be aberrant from the perspective of Arusa phonetics and phonology. In other words, all consonants, vowels, and approximants present in ideophones belong to the standard sound repertoire of the language. The only remarkable features are a base-medial [ʔ] (see [ʔórʔór])^[19] and potential extra-long consonants and vowels attested in the final position of non-predicative bases. In general vocabulary, the consonant [ʔ] appears only in a word-initial position immediately before a vowel. Extra-long consonants and vowels are absent in Arusa with the exception of interjections (Andrason and Karani forthcoming). The consonants that lend themselves to such exaggerated lengthening in ideophones are [s] and [r], illustrated by [ɗósː(ː)] and [t͡ʃúrː(ː)], respectively. As far as vowels are concerned, extra-long pronunciation is attested in some monosyllabic tokens containing [o], [ɔ], and [a]: see [sjóo(ː)], [tɔ́ɔ(ː)], and [páa(ː)]. Certainly, in all such cases, relevant consonants and vowels may also be realized as long and/or bimoraic. Nevertheless, at least in the aforementioned lexemes, the use of an exaggerated type of length – in the case of vowels, characterized by three morae – is by far the most preferred.

The immense majority of non-predicative bases exhibit high tone. For monosyllabic bases, the use of high tone constitutes an exceptionless rule. See, for instance, [ɓúsː], [kɛ́l], [táɓ], and [júk]. In the case of monosyllabic tokens containing diphthongs or long vowels, high tone is present in the vocalic nucleus (see [káu], [páu], and [kʊ́ɪk]) or the first vocalic morae (see [páa], [sjóo(ː)], and [tɔ́ɔ(ː)]), respectively. High tone also falls invariably on the first vowel of bisyllabic ideophones. The other syllable may carry a low tone, the lexeme thus exhibiting a HL tonal pattern, e.g., [t͡ʃámbul], [kóɾosː], and [tírːmam]. In the case of bisyllabic bases that draw on reduplication, high tone tends to be found in the second syllable, e.g., [kúrːkúrː], [óɾór], and [téːté]. This HH pattern is also attested with non-reduplicative ideophones – see, for instance, [ŋáɾɓáɓ] and [púɾkúk]. Similarly, high tone is typical of bases that are more complex from a phonetic perspective. In trisyllabic ideophones, high tone can be carried by all syllables (e.g., [kúmúkúm] and [kúɾúɾúm]) or the first and last ones (e.g., [mílígít] and [súŋúsúŋ]). This yields the possible tonal patterns of HHH and HLH. In quadrisyllabic ideophones, high tone is usually found in all syllables (e.g., [kíɾíkíɾí] and [tɔ́kɔ́tɔ́kɔ́]) or every second syllable (e.g., [ɓúɾuɓúɾu], [kílikíli], and [kúlukúlu]). In [t͡ʃúkut͡ʃúkɔ́ɔ], high tone falls on all the syllables but the second. For this class of ideophones, the attested patterns are thus HHHH, HLHL, and HLHH. This preference for high tone as well as the prevalent monosyllabicity and the use of consonantal onsets and codas, which were explained in the previous paragraphs, jointly suggest that the typical phonetic form of an Arusa non-predicative ideophonic base is C(A)V́(V̯)C, with the CV́C pattern constituting its most salient variant.

Non-predicative ideophonic bases that consist of more than one syllable tend to exhibit internal – i.e., operating within the word – harmonious configurations. The most patent expression of this is vocalic harmony, which is present in 21 of 26 (81%) multisyllabic ideophones.^[20] This can be illustrated by the bisyllabic bases [kwáɾa] and [kóɾosː], trisyllabic [kúɾúɾúm] and [míligít], and quadrisyllabic [ɓúɾuɓúɾu] and [tɔ́kɔ́tɔ́kɔ́]. The only exceptions are, in the case of bisyllabic bases, [t͡ʃámbul], [mút͡ʃa], [tírːmam], and [tírːja], and in the case of quadrisyllabic bases, [t͡ʃúkut͡ʃúkɔ́ɔ]. Given the common use of reduplication (see Section 3.2), many multisyllabic ideophones create the impression of rhymes.

Vocalic harmony and rhymes are also visible beyond the word structure of an ideophonic base itself, thus being external. That is, the vowel of many non-predicative ideophones, including those lexemes that are monosyllabic, often coincides with the vowel(s) present in the verbal stem with which a given ideophone tends to co-occur. See, for example, áibáʃa pá [pá], ápuku súrr [súrː], áitírrkim kíríkírí [kíɾíkíɾí], ábuku kúlukúlu [kúlukúlu], álutu cúrr [t͡ʃúrː(ː)], áud tús [tús], and metíi píp [píp].^[21] Similarly, harmony and rhymes emerge when a non-predicative ideophone co-occurs with its predicative variant, e.g., áŋis ŋís [ŋís], áwuap wúap [wúap], ádedu dédéu [ɗɛ́ɗɛ́u], áipid píd [píd], and áŋɔɓ ŋɔ́ɓ.

Finally, several non-predicative ideophones may be, and often are, uttered with (very) distinctive phonation and prosody. For instance, ɓúss [ɓúsː], which depicts the action of falling with a thud, tends to be pronounced with loud volume and particular speed to emphasize the unpleasantness, surprise, and brutality of an event. Similar articulatory force is common in the ideophones ɗóss [ɗósː] and púrkúʃ [púɾkúʃ], which respectively depict the action of uprooting something and the action of running fast. Exclamatory pronunciation in terms of loudness and speed is also typical of ideophones that have more abstract meanings. For example, the lexeme t͡ʃít͡ʃ, which depicts emptiness or the lack of something, is often produced in an excessively loud and markedly short manner. In fact, the louder and shorter it is, the more evident the idea of emptiness or absence seems to be. This articulatory markedness is particularly visible with ideophones used holophrastically and asyntagmatically, although it is also often attested in the two non-predicative syntagmatic uses, i.e., as verbal modifier and (perhaps slightly less so) as part of a complex predicate.

As explained in Section 3.2, there are 13 predicative bases that formally differ from bases used in non-predicative functions. If we include these divergent predicative bases in our phonetic analysis, the several generalizations we presented earlier need to be reformulated. First, monosyllabicity slightly decreases because six monosyllabic ideophones are bisyllabic in their predicative forms (see kúm/íkum, píd/ípid, rús/írus, súrr/ísurr, tíak/ítiak, tús/ítus, and dóss/ídosu).^[22] Similarly, predicative bases attest to two further cases of quadrisyllabic ideophones (kúrúrúm/íkɔ́ɾɔ́ɾɔ́m and súŋsúŋ/ísúŋusúŋ) and one case of a quintasyllabic ideophone, which is otherwise unattested in the language (t͡ʃúkut͡ʃúkóo/ít͡ʃúkut͡ʃúkóo). Overall, for predicative bases, bisyllabicity is the most common (nine occurrences), while monosyllabicity is only second-most common (five occurrences). Second, due to the presence of the initial i- in predicative bases, ten instances of onset-less ideophones may be added to the only one found in non-predicative bases (i.e., ít͡ʃúkút͡ʃúkóo, íkum, ɪ́kɔ́ɾɔ́ɾɔ́m, ípid, írus, ísúŋusúŋ, ísurr, ítíak, itɔktɔk, and itus). In fact, in predicative bases, forms beginning with a vowel are more common (ten occurrences) than those beginning with consonantal onsets (seven occurrences). In predicative bases, we also encounter an additional case of ideophones ending in a vowel (doss/dosu). As a result of the aforementioned observations, if predicative and non-predicative bases are considered jointly, the ideophonic pattern CVC/C(ː)VC(ː) becomes less pervasive. For predicative forms, this pattern is only attested 5 times of 17 instances (35%; compared with 43% in non-predicative forms). Third, most predicative bases with i- (see ícúkucúkóo, íkum, íkɔ́ɾɔ́ɾɔ́m, írus, ísúŋusúŋ, ísurr, ítɔ́ktɔ́k, and ítus) violate the tendency for vocalic harmony typical of ideophones (see also dosu in which the quality of the vowels does not coincide). Indeed, vocalic harmony is attested only three times in predicative bases (25% out of 12 possible instances). Fourth, in predicative bases, the visibility of low tone increases and, inversely, the preference for high tone is much less evident. Specifically, in their predicative forms, several monosyllabic high-tone ideophones exhibit an HL pattern: [íkum], [ípid], [írus], [ísurː], [ítiak], [ítus] (cf, non-predicative [kúm], [píd], [rús], [súrː], [tíak], [tús]). A similar preference for a low tone instead of high tones, at least in some syllables, is attested in [ít͡ʃúkut͡ʃúkɔ́ɔ] (cf. non-predicative [t͡ʃúkut͡ʃúkɔ́ɔ]), [íkɔ́ɾɔ́ɾɔ́m] (cf. non-predicative [kɔ́ɾɔ́ɾɔ́m]), and [ɗɛ́ɗɛ́] (cf. non-predicative [ɗɛ́ɗɛ́u]). Overall, there are only four predicative bases with exclusive high tone (kúrr, ŋám, ŋís, and ŋób). Finally, predicative ideophones – not only those with divergent bases but virtually all of them – tend to be realized with phonation that is typical of any other verb. Therefore, their intonation and prosody are usually unmarked.