Iconicity as the motivation for morphophonological metathesis and truncation in Nigerian Pidgin

3.1 Tone metathesis in all varieties of Naijá

Naijá expresses various semantic notions, which can be categorised as unconventional positive, by transposing the lexical tones of certain nouns. The pattern, which is illustrated with the examples in (5), is found in all varieties of Naijá.^[2]

As shown in (5), H-L nouns are realised as L-H and vice versa for the expression of unconventional positive. We used only one gloss for each aforementioned example, but the alternation can introduce more than one meaning. For instance, the form [mámà] ‘trendy elderly woman’ in (5b) can also mean ‘a powerful woman.’ ‘a female boss.’ etc. Regardless of the noun, the tonal alternation has a positive affective meaning, and compared to the underived nouns, the alternation also introduces the semantic sense of unconventionality. Thus, the semantic notions can be captured as unconventional positive. The tonal alternation mostly applies to and derives animate words, but the derivation of [mõ̀tó] ‘luxurious car’ from [ m o ˜ ́ t ò ] ‘car’ is the only example of an inanimate derived form. As shown in (6), the alternation can also derive animate forms from inanimate words.

We agree with an anonymous reviewer that the unconventional-positive form [bɔ̀tá] in (6) has the same meaning as the word [ a ̀ jɛ̀bɔ́t a ̀ ] (or [ a ̀ jɛ̀] in the Wafi variety), which is incorporated into Naijá from the derived Yorùbá word [ajɛbɔ́ t a ̀ ] with the same meaning, ‘someone raised with wealth.’ In the source language, the word is derived from the agentive prefix [a]- and the verb phrase [dʒɛ] ‘eat’ + [bɔ́tà] ‘butter’ (see Pulleyblank and Akinlabi 1988, for the phrasal morphology). The word is borrowed in whole into Naijá as a single word [àjɛ̀bɔ́tà] without its morphological subcomponents, given that the initial [a] is never used as a prefix in any known variety of Naijá. For example, the form *[àtʃɔ́bɔ́tà] ‘someone raised with wealth’ (expected meaning) from *[à] [tʃɔ́p] ‘eat’ + [bɔ́tà] ‘butter’ is not a word in any known variety of Naijá.

To the best of our knowledge, the tonal metathesis in Naijá only applies to the nouns cited in (5) and (6). Even though examples like [títʃà] ‘teacher’ and [bàká] ‘curvaceous buttock’ have H-L and L-H melodies like the nouns above, they are not morphologically marked for unconventional positive. That said, there is a purely expressive tonal metathesis, as illustrated in (7), that applies to all nouns. Depending on the context, this variant of the tonal metathesis can express either unconventional positive or negative. For example, in a situation where speakers A and B are talking about a person who is a teacher, and A says the sentence in (7a). To indicate that ‘the person is an unconventionally competent or incompetent teacher’ or ‘not a teacher at all’, Speaker B can realise the H-L tone of the word [títʃà] ‘teacher’ as L-H, as in (7b). In another context, where speakers A and B are talking about a curvaceous buttock, and A says the sentence in (8a). To indicate that the person has a bigger or smaller than usual [bàká] ‘curvaceous buttock.’ the person could say the sentence in (8b). The positive or negative reading depends on the speaker’s intention or context.

Unlike the unconventional-positive forms in (5) and (6), which can occur in isolation or any kind of construction, the purely expressive variant only appears in a contrastive negative construction, where it is compared to the base form. Another distinction is that the purely expressive variant is mostly used for exaggeration.

As noted by an anonymous reviewer, the expression of lexical or grammatical notions through tonal alternation as an attribute of Naijá is a property contributed by the indigenous or substrate languages. For example, the purely unconventional form in (7) and (8) has cognates in some indigenous languages of Nigeria (e.g. Yorùbá in Awoyale 1989). Beyond the purely expressive forms, the expression of lexical or grammatical notions through tonal alternation is a common attribute of the indigenous languages of Nigeria. However, to the best of our knowledge, no indigenous languages of Nigeria have been shown to derive new forms from existing lexical items through a pattern of tonal metathesis. In this sense, the tonal metathesis in Naijá is rare.

An anonymous reviewer also suggests that the unconventional positive is a property of youth language, slang or wordplay. While the unconventional positive is more productive among young speakers of Naijá, there is evidence to suggest that the reviewer’s characterisation of the form is inaccurate. It is known that both young and old speakers of Naijá use the unconventional-positive form for honorific purposes. For example, regardless of the context or interlocutors, speakers of Naijá refer to male and female pastors with the unconventional-positive forms [pápà] and [mámà], respectively, not the neutral forms [pàpá] ‘mother’ and [màmá] ‘father.’ The most recent and popular example of this usage involves a leaked phone conversation between the presidential candidate of the Labour Party, Peter Obi, and the founder of the Living Faith Church Worldwide, Bishop David Oyedepo (Folorunsho-Francis 2023). The former referred to the latter, using the unconventional-positive form [dàdí] (compared to the neutral form [dádì] ‘father’). Another piece of evidence suggesting that the unconventional-positive forms are not mere slang words is that researchers have long misclassified them as lexically contrastive, rather than derived from the neutral forms. If we consider that expressive morphology plays a key role in grammar (Dingemanse 2018), we can further argue that the unconventional positive is an aspect of Naijá grammar. In the next section, we present another pattern of morphological derivation that is similar to the tonal metathesis.

3.2 Syllable metathesis and onset elision in the Wafi variety of Naijá

This section focuses on a pattern of metathesis and truncation that is only found in the Wafi variety of Naijá. The pattern is found in the speech of middle-aged adults and most youths, but our datasets come from middle-aged adults. By transposing adjacent syllables or deleting a word-initial consonant, Wafi expresses inferiority and other notions that can be categorised as unconventional negative. Examples of the prosodic metathesis and deletion are presented in this section. For simplicity, only the inferior meaning of the unconventional negative is included in the gloss. The examples are grouped based on their syllables and tones.

The bisyllabic words in (9) show that the first syllable metathesises with the second syllable to express inferiority. While the syllables metathesise, the tones do not. For example in (10), the H and L tones of the diphthong in [ã́.drɔ́ìd] are associated with the second vowel of the diphthong and the word-final vowel in the unconventional-negative form [drɔ́íd.ã̀]. We had no examples of the metathesis occurring in words with three syllables, except for the form [àwóʃá] ‘cheap/inferior sex worker’ from [àʃáwó] ‘sex worker’ which was pointed out by an anonymous reviewer and confirmed by the second author. The use of this trisyllabic unconventional negative might have been popularised by the hit song Nwa Baby (Ashawo Remix) of the Nigerian artist Flavour (2011). That said, there are a few examples of four-syllable words with the metathesis, as shown in (11). In this case, the first two syllables of the words are transposed to the end of the word, while the linear order of tones remains unchanged.

The expression of unconventional negative is different in monosyllabic nouns. As shown in (12) and (13), the initial consonant of each monosyllabic noun is deleted.

As shown in (12), the initial consonant of each consonant cluster is deleted in the derivation of unconventional negative, regardless of whether the cluster involves two or more consonants. The initial-consonant deletion still applies even when a monosyllabic word has one onset consonant, as shown in (13). The onset deletion applies to all types of nouns including nonce words, as illustrated in (14). Similar to the actual words in the language, the onset deletion also introduces the notion of unconventional negative for the nonce words.

The expression of unconventional negative is not a secret language or argot, given that it occurs in natural conversations for open communication. The morphological derivation occurs in both expressive and non-expressive contexts. Regardless of the contexts or function, the morphophonology of the metathesis and truncation does not change. Compared to the unconventional positive, the unconventional negative is more productive. We find no instances where the unconventional negative applies to monosyllabic onsetless words, proper names or trisyllabic words, except [àwóʃá] ‘a cheap sex worker’ from [àʃáwó] ‘sex worker.’ Before turning to an analysis of the unconventional positive and negative, we present the descriptive generalisations in the next section.

3.3 Descriptive generalisations

The descriptive generalisations of the unconventional positive and negative are presented as follows:

The meanings of the unconventional positive and negative are comparable to those of diminutive and augmentative morphemes (good vs bad, original vs fake, rich vs poor, superior vs inferior, etc.), but they do not express the prototypical semantics of diminutive and augmentative crosslinguistically (e.g., size: big vs small; gender: masculine vs feminine, age: old vs young) (Jurafsky 1996, Kortvelyessy and Stekauer 2011). Thus, we call them unconventional positive and negative.

4.1 Morphophonological metathesis

We present an analysis of the unconventional positive and negative in this section. As a point of departure, we first focus on the expression of unconventional positive, which is found in all varieties of Naijá. Previous studies of Naijá suggest that the tone distinction between the nouns and the corresponding unconventional positive is lexically contrastive (Faraclas 1984, Faraclas 2002, Elugbe and Omamor 1991, Esizimetor and Egbokhare 2012), but the phonological and semantic features of the data are not consistent with the account. The first observation against the previous account is that the words in each pair share a core meaning. For example, the form [fàdá] ‘reverend father’ shares its core meaning with the form [fádà] ‘father’. Second, one of the words in each pair has an additional semantic meaning, which can be captured as unconventional positive, extraordinary or positive opposite. This suggests that the unconventional-positive forms are morphologically derived from the other phonologically and semantically related forms.

We now turn to the expression of unconventional negative. Compared to the expression of unconventional positive, the fact that the prosodic metathesis or onset deletion in Wafi involves a morphological derivation is fairly straightforward in the descriptive sense. To capture the fact that an onset-consonant deletion or syllable metathesis marks unconventional negative, we refer to the constituents of the prosodic word, following the unified version of the prosodic hierarchy (Selkirk 1986, 2011).

We assume that the realisation of the unconventional-negative morpheme involves altering the linear order of the elements in a prosodic word ω , but the re-ordering varies based on word size. Thus, the unconventional negative targets the foot Ft in ω with at least two feet, such that the order Ft₁ Ft₂ in (18a) surfaces as Ft₂ Ft₁. When the base of an unconventional negative lacks two feet, as in the structure shown in (18b), the metathesis occurs at the syllable level, resulting in σ ₂ σ ₁. Consequently, we refer to the morphophonological process in the unconventional negative as prosodic metathesis. In the absence of binarity at the level of foot and syllable, the unconventional negative is expressed by deleting the initial consonant of the syllable. The metathesis and truncation seem different on the surface, but the general goal is to change the precedence and linear order of the phonological elements in a word for the morphological derivation.

Before presenting a formal analysis of both unconventional positive and negative, we compare them to similar morphophonological processes across languages. Marking unconventional positive and negative with tone and prosodic metatheses, respectively, is consistent with descriptive patterns of morphophonological exchange. de Lacy (2020) describes morphophonological exchange as follows:

[T]wo pairs of forms, where each pair includes forms that differ in terms of a single morpheme M. One pair consists of a form that lacks M and contains [ α ] (a feature, segment, prosodic element, or substructure) and another form that contains M and has a corresponding [ β ]; the other pair has a form with [ β ] and an M form with a corresponding [ α ] (de Lacy 2020, 30)

The tone metathesis of unconventional positive is in line with a morphophonological exchange, given that H-L and L-H words surface with L-H and H-L, respectively. Similarly, transposing the initial prosodic unit with the second prosodic unit in the expression of unconventional negative (e.g. σ ₁ σ ₂   >   σ ₂ σ ₁) is unambiguously a kind of morphophonological exchange. Descriptive patterns with morphophonological exchange are rare but exist (Kouneli and Nie 2021 on Kipsigis, DiCanio et al. 2020 on Itunyoso Triqui). The most popular example is from DhoLuo (Okoth-Okombo 2008 cited in Alderete 1999, 129), where voicing polarity occurs in certain nouns when marked for plural.

Based on the arguments of de Lacy (2012) and Wunderlich (2012), DiCanio et al. (2020, 3) propose three criteria for identifying true morphophonological exchange: (i) ‘the alternation must not apply to a limited, closed set of roots in the language’; (ii) ‘the alternation must be the dominant morphological exponent for the morpheme in question’; (iii) ‘all roots undergoing the alternation must belong to the same morphosyntactic category.’ They named these criteria dominance, productivity and morphosyntactic uniformity, respectively. To show that the morphemic metatheses in Naijá are indeed morphophonological exchanges, we consider these criteria. As shown in Section 3, the metatheses and truncation are the only ways of expressing unconventional positive and negative, respectively. The morphemic metatheses apply only to nouns and always maintain their word class. While the onset truncation of the unconventional negative applies to all monosyllabic common nouns except onsetless ones, the prosodic metathesis only applies to a small set of bisyllabic and quadrisyllabic nouns. Similarly, the tonal metathesis only applies to a small set of bisyllabic nouns with H-L and L-H tones. These properties indicate that the morphemic metatheses and truncation are the dominant exponents of unconventional positive and negative; they are morphosyntactically uniform but less productive. In contact languages like pidgins or creoles, it is difficult to use productivity as a criterion considering that a linguistic pattern might apply to a subset of the lexicon at an early stage of creole development (Plag 2008, 2009). If we consider that Naijá has less affixes compared to its lexical sources (Babalola 2021), the morphophonological metatheses could be another stage of morphological development. The fact that the tone and syllable metatheses do not apply to three-syllable words also indicates that the unconventional positive and negative are still emerging. Thus, the morphemic metatheses are a true pattern of morphophonological exchange. As mentioned earlier, the unconventional positive has a purely expressive variant that is productive but restricted to the context of exaggeration and contrastive negation (see the examples in (7)). The unconventional positive can be considered a grammaticalised form of the purely expressive variant. The non-productivity of unconventional positive points to an ongoing grammaticalisation of the productive but restricted variant (see Section 5). That the restricted variant can express unconventional positive or negative also suggests that only the positive usage of the purely expressive tonal metathesis is grammaticalised.

The metatheses of the unconventional positive and negative are similar to the patterns found in reversal language games, such as French Verlan (Gotowski 2019), Yoruba Ena (Isola 1982), Dholuo Dhochi (Borowsky and Avery 2009), and others (Bagemihl 1995, Alidou 1997, Ozburn and Schellenberg 2019). Despite this similarity, there are certain differences. For example, monosyllabic forms in reversal language games, such as the examples in (20b), tend to undergo reversal, not truncation like the unconventional-negative form in the Wafi variety. Truncation language games are rare, but even in most documented cases, the segments at the right edge of the word are the usual target of deletion (Bagemihl 1988, Borowsky 2010). Unlike the expression of unconventional positive and negative in Naijá, language games targeting feet and tones are unattested (Bagemihl 1988, 308). While sociolinguistic factors, such as group identity and exclusion, motivate the distortion of linguistic patterns in language games (Isola 1982, Lefkowitz 1989), the metatheses and segment truncation in Naijá involve morphological derivation. Another difference is that language games rarely contribute any meaning to the base form, but the metatheses and truncation in Naijá make semantic contributions. Consequently, the metatheses and segment truncation in Naijá are not language games but rather morphological derivations that share the phonological properties of language games.

The unconventional negative in the Wafi variety of Naijá is structurally and historically similar to the Verlan sociolect, which has the French reversal language game with the same name as its source of grammar. Similar to Naijá, the sociolect emerged in a multiethnic population (Doran 2002). Thus, it is highly plausible that the unconventional negative in the Wafi variety is from a reversal language game. In the absence of strong evidence connecting the source of unconventional-negative form to a language game, we leave our analysis of the unconventional negative as an innovation of Wafi speakers open for further inquiry. Unlike language games, which have been analysed within various formal frameworks including Optimality Theory (OT), the patterns in Naijá had not been (properly) documented and described, let alone analysed until now. By drawing insights from the OT account of language games, we present an account of the unconventional positive and negative in the next section.

4.2 Transderivational analysis

We account for the realisation of the unconventional positive and negative morphemes by adopting transderivational correspondence theory (TCT) (Benua 2004, Alderete 2001), which is an extension of correspondence theory (McCarthy and Prince 1995). TCT is based on the idea that the correspondence relation does not only hold between input–output and base-reduplicant pairs but also between independent words (McCarthy and Prince 1994). The core proposal of TCT is that words in a paradigm are in a transderivational correspondence relation, whereby phonological similarity within the paradigm is regulated by O(utput)-O(utput) faithfulness constraints, similar to the constraints on base-reduplicant relations. As proposed in Benua (2004), a morphologically derived word and its base are in OO-faithfulness relations. Studies of language games have applied the OO-correspondence to the relations between the base and game form (argot) (Itô et al. 1996, Borowsky and Avery 2009, Borowsky 2010). The correspondence relations between the base and argot form are regulated by faithfulness constraints such as Anchor constraints in (21), which require any element at a specific edge of the base to have a correspondent at the same edge of the argot form. The reordering of base forms in language games violates the constraints. The metathesis and truncation in language games are considered the result of concatenating a word with a phonologically and semantically empty morpheme Argot (Bagemihl 1988, 1989, Itô et al. 1996, Borowsky and Avery 2009, Borowsky 2010). The assumption is that the violation of the faithfulness constraints is not driven by the Argot morpheme but by the constraint Morph-Real in (22), which assigns a violation to a morpheme in the input without a distinct phonetic realisation in the output (Borowsky and Avery 2009).

We adopt the account of language games sketched above for our analysis of the unconventional positive and negative in Naijá. Our proposal is that the unconventional positive and negative morphemes are phonologically empty, but unlike argot forms, they are semantically specified.

The unconventional positive and negative morphemes could have been realised without being phonetically distinct from the base, considering that they are phonetically empty. That the two morphemes have a phonetic realisation is considered the effect of the constraint Morph-Real, which requires every morpheme to have a unique phonetic realisation in the output. As demonstrated in the previous section, the realisation of the unconventional negative involves reversing the linear order of Ft or σ , or deleting a word-initial consonant. For a four-syllable word, the realisation of the unconventional negative can lead to a violation of the faithfulness constraint BNeg-Anchor _Ft -L in (23a). The constraint assigns no violation mark to a foot at the left edge of the base form, which has a correspondent at the left edge of the unconventional-negative form. If the leftmost foot Ft₁ of the base is realised in the unconventional-negative form but not at the left edge, the constraint BNeg-Anchor _Ft -L assigns a violation for each foot located to the left of Ft₁ within the unconventional-negative form. We assume that a right-edge instantiation of the constraint BNeg-Anchor _Ft -L exists, but it is redundant for the illustration of our analysis. For a bisyllabic form, the Morph-Real can force the violation of the constraint BNeg-Anchor σ -L_Ft in (23b), which requires the syllable at the left edge of the foot in the base to have a correspondent at the left edge of the foot in the unconventional-negative form. If the leftmost syllable σ ₁ of a foot in the base form is realised in the unconventional-negative form but not at the left edge, the constraint BNeg-Anchor σ -L_Ft assigns a violation for each syllable located to the left of σ ₁ within the foot of the unconventional form (right-edge version of this constraint is also assumed to exist in the grammar). Foot or syllable deletion, or metathesis can result in the violation of the anchor constraints.

We can account for the foot and syllable metatheses by ranking the constraint Morph-Real above BNeg-Anchor σ -L_Ft , which in turn is ranked above BNeg-Anchor _Ft -L. The constraint Morph-Real does not determine whether the unconventional negative is realised through tonal or syllable metathesis. The fact that the language expresses the unconventional negative through syllable metathesis, instead of tone metathesis, can be considered an effect of the constraint BNeg-Linearity-T, which requires the linear order of tones in the base to be preserved in the corresponding unconventional-negative form. The account is illustrated in tableau (24). The edges of Ft and σ are indicated with parentheses and dot ‘.,’ respectively.

The constraint Morph-Real rules out unconventional-negative forms that are not phonologically distinct from their base. For the four-syllable form, the candidate in (24a-iii) satisfies Morph-Real but is ruled out for violating the constraint BNeg-Anchor σ -L_Ft . The winning candidate in (24a-ii) satisfies Morph-Real to the detriment of BNeg-Anchor _Ft -L. The constraint ranking also accounts for the bisyllabic unconventional-negative form in (24b). For the words with two or four syllables, Morph-Real could have been satisfied by deleting a segment instead of resorting to metathesis. That the option of deletion is not utilised is considered an effect of the constraints BNeg-Anchor _seg - L σ and BNeg-Anchor _seg - R σ in (23c-d). The constraints assign no violation to the segments at the left and right edges of a syllable in the base form, when the segments have correspondents at the left and right edges of the syllable in the unconventional-negative form, respectively. If the leftmost segment S₁ of a syllable in the base form has a correspondent in the syllable of the unconventional-negative form but not at the left edge, the constraint BNeg-Anchor _seg - L σ assigns a violation for each segment located to the left of S₁ within the syllable of the unconventional-negative form. The same condition holds for the right-edge counterpart, BNeg-Anchor _seg - R σ . These constraints must be ranked above BNeg-Anchor _Ft -L and BNeg-Anchor σ -L_Ft in the language, but no evidence for the ranking of the segment anchoring constraints in relation to the constraint Morph-Real. The effects of these constraints are shown for the monosyllabic forms in (25), which vacuously satisfy BNeg-Anchor _Ft -L and BNeg-Anchor σ -L_Ft due to the absence of adjacent feet and syllables.

The possible ways for a monosyllabic form to satisfy Morph-Real are segment deletion or metathesis. In the case of a monosyllabic word with a consonant cluster, metathesising the clustered consonants or deleting the medial consonant of the cluster can result in the violation of BNeg-Contiguity SEG , σ , which requires contiguous segments of a syllable in the base form to be contiguous in the unconventional-negative form, as shown in (25aii-iii). The option of onset-coda metathesis can also result in a fatal violation of the constraints BNeg-Anchor _seg - L σ and BNeg-Anchor _seg - R σ , as the candidate in (25b-ii) does. While the candidate in (25a-iv) incurs a fatal violation of BNeg-Anchor _seg - R σ , the candidates in (25a-v) and (25b-iii) win despite satisfying Morph-Real to the detriment of BNeg-Anchor _seg - L σ .

As mentioned earlier, the unconventional positive is also a phonologically empty morpheme with semantic specification. The realisation of the unconventional-positive morpheme through base-tone metathesis results in the violation of the constraints BPos-Anchor _T-R and BPos-Anchor _T-L, which require the tones at the left and right edges of the unconventional-positive form to have correspondents at the left and right edges of the base form, respectively. By ranking these constraints below Morph-Real, we can account for the tonal metathesis. That the language expresses the unconventional positive through tonal metathesis, instead of syllable metathesis, can be considered an effect of the constraint BPos-Linearity- σ , which requires the linear order of syllables in the base to be preserved in the unconventional-positive form. The tableau in (27) illustrates this account.

We have shown that the realisation of the unconventional positive and negative involves violating OO-faithfulness constraints. The Morph-Real constraint drives these violations, determining the shape of the unconventional positive and negative. We assume that general faithfulness constraints are also violated in the realisation of the unconventional positive and negative, but the constraints are not included in the tableaux. We explore alternative analyses in the next section and compare them to the one in this section.

4.3 Alternative analyses

The proposal of TCT is that base-derivative relations are governed by faithfulness constraints (Benua 2004). Alderete (2001) expands upon this idea by proposing anti-faithfulness constraints, which are negated counterparts of faithfulness constraints. The constraints, such as the example in (28), only assign one violation.

The empirical bases of anti-faithfulness constraints are patterns of morphophonological exchange across languages (Alderete 1999, 2001, Gotowski 2019). Under this account, the ranking of anti-faithfulness constraints over faithfulness constraints drives morphophonological exchange. As we briefly illustrate with the unconventional-negative form in (29), the ranking can account for the morphophonological metathesis in Naijá. The account can also be extended to tone polarity and truncation.

The proposal that anti-faithfulness constraints are exchange-inducing mechanisms has been questioned, as various patterns of morphophonological exchange have been reanalysed and generated without the use of anti-faithfulness constraints (Kurisu 2001, Pulleyblank 2006, de Lacy 2012). Our account of the unconventional positive and negative in Naijá also points to the sufficiency of faithfulness constraints for the analysis of morphophonological exchange.

Another alternative analysis is that the unconventional positive is a morpheme with a floating L tone as its exponent (following Akinlabi 1996). Under this account, we have to assume that the lexical H tone occurs in a prosodically prominent position. The idea is uncontroversial, considering that high pitch is one of the phonetic correlates of stressed syllables in certain languages (Lehiste and Peterson 1959) and that stressed syllables tend to surface with H tone in loanword adaptation (Kenstowicz 2006). We can account for the realisation of the L tone with the constraint Unc.Pos-Anchor _prom , which requires the tone of the unconventional positive to have a correspondent in a prominent position. The realisation of the L tone in the prominent position will result in the violation of the constraint Max-H, which requires an H tone in the input to have a correspondent in the output (Zoll 2003). The Max-H constraint can be satisfied by relinking the delinked H tone to another TBU. Under this account, the tonal metathesis can be considered a by-product of equal ranking of Unc.Pos-Anchor _prom and Max-H. We illustrate the account in (30). The lexical tone and the exponent of the unconventional positive are indicated with numeral indexation. The association of the tones is indicated with parentheses, which are comparable to association line.

A reviewer suggests analysing the tonal metathesis of the unconventional-positive form as an effect of their final (or initial) syllable being extrametrical (as in Hayes 1982, Pulleyblank 1986). Under such an account, the H-tone syllables in Naijá will be considered stressed. The tonal metathesis will be considered a byproduct of an extrametrical morpheme which occupies the prominent position in the word and shifts the stress away from the extrametrical syllable. This account is similar to the floating tone account that is sketched in (30).

The alternative analyses sketched here are plausible but a combination of the accounts in (29) and (30) may not be able to capture the similarities between the unconventional positive and negative. We believe that our account, as presented in Section 4.2, not only unifies the prosodic and tonal metatheses with the truncation but is also consistent with the fact that the unconventional positive and negative share properties, such as reordering of phonological elements, semantic opposition (positive versus negative), expressiveness, and context sensitivity. In the next section, we discuss these similarities and their implications for form-meaning association.

4.4 Iconicity and expressiveness of unconventional positive and negative

This section focuses on the form-meaning association of morphophonological metatheses and truncation in Naijá. The emerging perspective on form-meaning mapping acknowledges the role of arbitrariness and iconicity in core grammar (Dingemanse 2018, Kawahara 2020), contrary to the Saussurean view that it is completely arbitrary (de Saussure 1916, Hockett 1960). Studies suggest that depictive mode of signification and expressive meaning distinguish iconic forms from arbitrary forms (Zwicky and Pullum 1987, Dingemanse 2015). We will argue that the form-meaning association of the unconventional positive and negative in Naijá involves iconicity. To this end, we refer to the properties of expressiveness and depictive signification.

Expressive morphology tends to involve evaluative meaning (e.g. positive vs negative, big vs small, good vs bad, inferior vs superior), vocal artistry and mimicry (Zwicky and Pullum 1987, Kita 1997, Dingemanse 2015, Akinbo 2021b, 2023). Morphosyntactically, most examples of expressive morphemes do not change the syntactic category of the base they apply to. Prosodically, expressive forms, such as ideophones, tend to have marked or unusual phonological patterns (Hinton et al. 1994, Dingemanse 2018). The other properties of expressive morphemes that are relevant to our discussion are semantic variability and non-displaceability (Potts 2007, Fortin 2011). The semantic variability of expressive morphology arises from its strong ties to the utterance situation. Potts (2007) refers to this property as non-displaceability.

The first expressive property of the tone metathesis and prosodic metathesis, with its truncation variant, in Naijá is their association with positive and negative affective meanings, respectively. In this case, we can consider the tone metathesis and the prosodic metathesis alongside its truncation variant to be two sides of the same affective expression. The semantic opposition between the unconventional positive and negative markers is responsible for their non-overlapping distribution. As mentioned earlier, the exact meaning contributed by the tone metathesis varies, but it can be captured only with the notion of unconventional positive. The multiple meanings of the tone metathesis are consistent with expressive morphology. To better understand the meanings introduced by tone metathesis, we have to consider their context of occurrence. For example, when the form [màmá] ‘mother’ is marked for unconventional positive, the derived form [mámà] can express multiple meanings, including beautiful, strong, charismatic, powerful, good, important, pious, etc. In some cases, female children that are believed to be reincarnations of their grandmothers are nicknamed using the unconventional-positive form [mámà] ‘a special woman’.^[3] The naming context suggests that the unconventional-positive forms can also be honorific. The unconventional negative is similar to the unconventional positive in this regard, as its meaning includes cheap, small, fake, bad, unremarkable, etc. The semantic variability of the unconventional positive and negative is from their semantic dependency on context and speaker’s intention.^[4]

We now turn to depiction as a property of expressive morphemes. As described by Dingemanse (2015, 950), “depictions are typically iconic, representing what they stand for in terms of structural…resemblances between form and meaning.” Crossmodal depiction is shared by various patterns of iconicity. For example, the association of low acoustic frequency (e.g. low F1 and F2) with high magnitude and vice versa in various patterns of iconicity depicts the natural acoustic principle, which is the inverse relationship between the length of a vibrating object (e.g. vocal cords) and the natural frequency at which the object vibrates (Sapir 1929, Titze 1989, Ohala 1994, Johansson and Zlatev 2013, Knoeferle et al. 2017). Another example of depiction is the association of repetition or intensity with morphophonological reduplication or repetition in languages around the world (Awoyale 1989, Hinton et al. 1994, Voeltz and Kilian-Hatz 2001, Hurch 2005). Depiction is robust in verbal arts such as poetry, singing and chants (Brown 1999, Dingemanse 2009, Agawu 2016). Expressing unconventionalities by changing the conventional linear order of linguistic units can be considered a kind of depiction. Associating the unconventional negative with truncation can be considered another pattern of iconicity, as deleting or subtracting a segment is like adding a negative. That said, consonant deletion exists as a form of non-iconic subtractive morphology, as observed in languages such as Tohono O’odham (Kosa 2008) and Hebrew (Bat-El 2002).

The expression of unconventional positive and negative through metathesis is comparable to the pattern of iconicity found in antonyms of languages such as Coptic and Demotic Egyptian ([mes-] ‘darkness’: [sem-] ‘to be become visible’), Lithuanian ([ker-as] ‘to be dumb, unable to speak’: [rek-ti] ‘to speak, talk’) and Russian ([rab-] ‘slave’: [bar-in] ‘lord, master’) (Croft 1978, 511). Gamkrelidze (1974) refers to these patterns of form-meaning mapping as horizontal iconicity. Considering that horizontal iconicity involves perceptual resemblance between word structure and meaning, it is similar to the expression of iteration through reduplication or lengthening. Dingemanse (2011a) classifies iconic reduplication and repetition as ‘gestalt diagrammatic iconicity,’ which he defines as ‘resemblance between a word structure and structure of perceived event’ (Dingemanse 2011a cited in Johansson et al. 2020, 258). Given that negation is conceptually a kind of reduction, the expression of unconventional negative by truncating or reducing a string of segments is another kind of iconicity, which is form-meaning resemblance. Consequently, the unconventional positive and negative in Naijá are consistent with gestalt diagrammatic iconicity.

The origin of the tonal metathesis can be understood if we consider that Naijá also exhibits an exaggerative unconventional marking, as shown in (7), which is similar to the unconventional-positive form in terms of meaning, expressiveness, and the use of tonal metathesis. Due to the similarity, the expression of unconventional positive with tonal metathesis can be considered a grammaticalisation of the exaggerative unconventional marking. As mentioned earlier, the foot and syllable metathesis are comparable to language games. However, it is difficult to tell whether the prosodic metathesis and truncation of the unconventional negative are from language games, linguistic innovations, or other morphophonological patterns in the substrate languages. Regardless of their origin, the morphophonological metatheses and truncation in Naijá contribute to the typology of non-segmental morphology in human language. Overall, the properties of depiction, expressiveness, semantic variability, and unusual morphophonological patterns strongly suggest that the unconventional positive and negative involve iconicity.