Corela Cognition, représentation, langage

Transcription

1 Corela Cognition, représentation, langage Vol. 17, n 2 Daniel Roulland (dir.) Édition électronique URL : DOI : /corela.9234 ISSN : X Éditeur Cercle linguistique du Centre et de l'ouest - CerLICO Référence électronique Daniel Roulland (dir.),, «Vol. 17, n 2» [En ligne], mis en ligne le 26 novembre 2019, consulté le 21 janvier URL : ; DOI : /corela.9234 Ce document a été généré automatiquement le 21 janvier Corela cognition, représentation, langage est mis à disposition selon les termes de la licence Creative Commons Attribution - Pas d Utilisation Commerciale - Partage dans les Mêmes Conditions 4.0 International.

2 1 Avec le premier article de ce volume, Mori Edwige Traoré & Xiaolang Luo (Tagba Tone: a case of tier hierarchization) reviennent sur certains patrons tonals complexes de la langue tagba (groupe sénoufo, Afrique de l ouest), pour montrer que la hiérarchie classique de leurs représentations dans le modèle autosegmental fait des prédictions incorrectes. Elle doit être aménagée en associant directement le ton aux voyelles, ce qui conduit à abandonner les principes réputés universels d association avec des unités de rang supérieur ou de neutralisation complète. Boris Lefilliâtre (The cognitive motivation and purposes of playful blending in English) traite des amalgames par jeux de mots (playful blends) sous l angle de leurs motivations et des objectifs qui les sous-tendent. Il montre que, dans un domaine à la frontière de la sémantique et de la pragmatique, ces amalgames apparaissent cognitivement saillants dans leur co-texte, par une concision qui les distingue d emblée des mots environnants, un rapport interne spécial entre les éléments associés, et une forte créativité. Leur dimension pragmatique explique la grande variété des objectifs de leur production en contexte. Dans le cadre de la Théorie des Blocs Sémantiques de Marion Carel (2019), l article de Kohei Kida (De l interprétation argumentative) revient sur la notion de «Carte Argumentative du Lexique», le «décodage argumentatif» et l «interprétation argumentative» consistant tous deux, mais de manière inverse, à déterminer pour un énoncé donné le terme constitutif et par voie de conséquence le schéma argumentatif qu il signifie. Dans le cas de l interprétation, le terme constitutif est implicite, ce qui autorise pour M. Carel une certaine liberté d interprétation pour l auditeur/lecteur, quitte à devoir revenir sur un premier schéma. Kohei KIDA nuance cette position en soulignant la nécessaire prise en compte de l intention plus ou moins manifeste du locuteur qui contraint l interprétation. Démonstration est ainsi faite de l impossibilité de dissocier l organisation de la carte argumentative de son utilisation. Hélène Margerie et Philippe Muller (Excuse me vs (I'm) sorry as two contrastive markers of interlocutive relations) utilisent la Théorie de la Relation Interlocutive (C. Douay & D. Roulland 2014) comme cadre explicatif de la différence qu ils étudient entre les deux formes d excuse en anglais : Excuse me et (I m) sorry. La première, Excuse me, est caractérisée par un «désaccord» interlocutif entre les deux pôles émetteur et récepteur qui peut aller jusqu à la polémique, alors que la seconde, (I m) sorry, établit une «harmonie» interlocutive, avec une coopération de l allocutaire considérée comme acquise. Les effets pragmatiques polémiques d un côté ou apaisants de l autre relèvent des configurations systémiques C1 et C2 de la théorie. Loïc Renoud (La distinction accompli / inaccompli dans le récit en FLE: enjeux pour la conception de matériel pédagogique) aborde l apprentissage du Français Langue Etrangère avec une difficulté connue qui est celle de la relation entre le passé composé et l imparfait. Expliquant l approche aspectuelle du manuel Libre Echange (2001), il montre comment elle a été abandonnée au profit de l approche discursive centrée sur l organisation du récit en premier plan (passé composé) et arrière-plan (imparfait), sans que pour autant la question des choix énonciatifs soit résolue. L exploration de 17 manuels récents ( ) aboutit à Version Originale 2(2010) qui introduit une relation d observation (externe pour le passé composé avec «avancée» du récit, interne pour l imparfait avec «arrêt» du récit), relation qui permet de rendre compte à la fois des prototypies aspectuelles et des organisation discursives globales. Ceci amène l auteur à une perspective énactive (Bottineau 2010, 2014), prenant en compte plus largement les faits morphologiques, syntaxiques et prosodiques dans une progression

3 2 incarnée «d une conscience avançant dans le monde qu elle crée au fil des mots». Cette perspective permet de proposer un matériel pédagogique adapté, actuellement en cours d expérimentation, sur la construction et la progression du récit. Les propositions d articles pour le numéro 1 du volume 18 de la revue Corela sont à envoyer à Sylvester Osu avant le 15 avril 2020.

4 3 SOMMAIRE Tagba Tone: a case of tier hierarchization Mori Edwige TRAORÉ et Xiaoliang LUO The cognitive motivation and purposes of playful blending in English Boris Lefilliâtre De l interprétation argumentative Kohei Kida Excuse me vs. (I m) sorry as two contrasting markers of interlocutive relations Hélène Margerie et Philippe Muller La distinction accompli/inaccompli dans le récit en FLE : enjeux pour la conception de matériel pédagogique. Loïc Renoud

5 4 Tagba Tone: a case of tier hierarchization Mori Edwige TRAORÉ and Xiaoliang LUO We are grateful to the anonymous reviewers for their insightful comments. We also thank Shanti Ulfsbjorninn and Raymond Boyd for helpful corrections and comments. All errors and omissions remain, of course, the sole responsibility of the authors. 1. General features of Tagba 1.1. Generalities 1 Tagba is a Senufo language spoken in the West of Burkina Faso, in an area called Tagbara. It has about 35,000 speakers living in 28 villages. Westermann (1911), Tressan (1953), Greenberg (1970) and Naden (1989) class the Senufo languages among the Gur Languages of the Niger-Congo family. Tagba has a dialect cluster with various degrees of mutual intelligibility. 2 The data that we present in this paper have been collected in the village of Mahon. The data are based on a corpus resulting from long-term fieldwork by one of the authors between 2007 and It is composed of 1500 recordings of a group of native speakers (Traoré 2016). Traoré has obtained a phonological inventory which comes from the analysis of 621 basic roots (loanwords, compounds and derivations were excluded).

6 5 Figure 1. Senufo languages (source : Roulon 1968 : 57) Figure 2. The Tagbara (source: INSS/CNRST)

7 The phonological sketch of Tagba Inventory and contrastive features 3 Before launching into the tonology, in recognition of the fact that Tagba is a lesserknown language, we will introduce a sketch of its phonology. 4 The sound inventory is shown in (1) (1) Consonnants vowels p t c k i u b d ɟ g ɩ ʋ f s e ə o v z ɛ ɛ m n ɲ ŋ a a ɔ ɔ 5 As with many Niger-Congo languages, the same vowel height distributes into two series, [+ATR] and [ ATR]. 6 Tagba is a tone language with two underlying contrastive tones: high (H) and low (L). A minimal pair which illustrates their phonological status is shown in (2). 7 (2) tó grave tò close Phonological processes 8 Two phenomena are to be considered briefly before focusing on Tagba tonology: vowel harmony (VH) and intervocalic lenition. Vowel Harmony 9 Tagba has highly complex VH patterns, the analysis presented here follows Traoré (2016). It involves the features of Place, Height and ATR. 10 Place harmony affects suffixes with underlying form /-CV/. If V=/a/, then it occurs systematically, as shown in (3a); if V is another vowel, Place harmony is not systematic, (3b) shows cases where it occurs, (3c) those where it does not occur (certain words involve more than one kind of VH). (3) a. ɲɛ -ya [ɲɛ yɛ ] faces ɲɔ -ga [ɲɔ ɣɔ ] mouth buň-ga [bu ŋɔ ] terrace b. gó-tɩ [gbóró] scrape

8 7 ce -dɩ [ce de ] cut c. lʋǧ-tɩ [lʋ rɩ ] go up cɔ t-dɩ [cɔ dɩ ] press in 11 As for Height harmony, it occurs in some words, but not systematically. (4) a. ŋɛ -tɩ [ŋɛ rɛ ] heat kă-dɩ [kàlá] roast b. fɛ t-dɩ [fɛ dɩ ] jam fàd-dɩ [fàdɩ ] change 12 ATR harmony often goes with that of place and height, but not systematically either. Compare: (5) a. peď-bɛ [pe lɩ βé] size tuń-lɛ [tuńuĺé] squirre b. to d-gɔ [toĺo ɣɔ ] accompany jêt-lɛ [jérɩĺɛ ] tongue 13 To focus on the tonology, we will not enter into details of the VH processes. If in our data the vowel changes, it is due to VH. 14 Intervocalic lenition 15 Only a subset of consonants, oral and nasal stops 1, appears in intervocalic position, and the lenition affects oral stops systematically. The examples are shown in (6). (6) Lenition Example d>l caď-ɩ [ca lɩ ] flatten t>r bə t-ɔ [bə rɔ ] flow k>ʔ nɔ k [nɔ ʔɔ ] bottom g>ɣ mɛ -ga [mɛ ɣɛ ] name 16 Intervocalic lenition also occurs after a nasal vowel, shown in (7a). This differs from the vowel + nasal sequence in (7b).

9 8 (7) a. da -ga [da ɣa ] basket b. duň-ga [du ŋɔ ] chest 17 In (6) and (7a), the examples have only CV syllables. If the root has a CVC underlying form and the suffix CV form, this will give an underlying CVC-CV structure. Tagba structure is composed entirely of CVs at the surface level, codas are forbidden. There are two main strategies to avoid a surface coda, the first is vowel epenthesis, shown in (8) and the second is the fusion of two consonants, shown in (9). (8) Vowel epenthesis a. bɩ d-gɔ [bɩ lɩ ɣɔ ] respect b. ŋɔ t-ga [ŋɔ rɔ ɣɔ ] smoke c. saď-ja [sa lə ja ] laziness (9) Fusion of two consonants niň-gɛ [ni ŋɛ ] beef faď-dɩ [fa dɩ ] change tʋň-ja [tʋ ɲɔ ] caterpillar 18 As we see in (8ab), the epenthetic vowel is usually a copy of the vowel of the root. However, it does not necessarily have to be, as shown in (8c) where only part of the root vowel is copied. Interactions between consonants and vowels 19 Tagba also shows some correlations between consonants and vowels. Vowel features spread more easily across some consonants in vowel epenthesis and VH. Consider: (10) a. /d/ i. VE fed [felɩ] coincer ii. VH ku-dɩ [kulɩ] cried b. /g/ i. VE tɛg [tɛɣɛ] poser ii. VH do-ga [doɣo] smell

10 9 20 In (10a), the root vowel can t cross the consonant /d/, whether in VE or in VH. In (10b), it can cross /g/ both in VE and in VH. This is not a coincidence even though there are counter-examples: Traoré & Luo (2014) and Traoré (2016: Chap.1) propose a hierarchy of the strength of the consonants based on the phonotactics at the morphological boundary. /d/ is one of the strongest while /g/ one of the weakest. However, we will not develop this issue because of its extreme complexity and will focus on Tagba tone in this paper. 21 Now let s compare the two examples shown in (11): (11) a. /fɛ d/ [fɛ lɩ ] diluted b. /fɛ t-dɔ/[fɛ dɔ ] dilute, diluting 22 As (10a) shows, a phonological singleton consonant /d/ becomes a lenis [l] phonetically. On the other hand, in (11b), there is a fusion of two underlying coronal obstruents: / td/. This fused structure yields a [d]. This is analyzed as a mismatch between phonological form and phonetic interpretation. Phonologically, the fused structure is a geminate; however, it does not have the phonetic correlate of the increased closure duration. 23 This leads us to analyze it as a virtual geminate (VG) 2. As for (9b) with a fused /dd/ and (11b) with a fused /td/, it appears that they both result in a surface [d]. This is because Tagba never allows surface intervocalic voiceless obstruents. [d] as a voiced stop is already the strongest intervocalic consonant possible Morphological features 24 Tagba is a language with noun and verb classes. It has eight noun classes (NC) and ten verb classes (VC). 25 For nouns, the inflection indicates the NC and the number. A noun is associated with a single set of suffixes (or two if the class distinguishes genders). Every class has indefinite and definite suffixes. The eight NC are wɩ/pɩ, kɩ/yɩ, lɩ/ci, tɩ and pɩ. As for verbs, the root may have no suffix (the bare or unmarked form, which gives the zero aspect), or take a suffix for one of the ten VC, yielding the incompletive aspect. Examples already shown in (11) can illustrate these patterns. 26 A word is composed of root and suffix (except for compounds). The underlying form of a root is of CV or CVC types, with underlying tone(s). As for the suffixes, they are of CV or V forms, always toneless. 27 Like many languages of the same area, the basic nominal morphology pattern of Tagba consists of a root plus a classifier (CL), which gives the singular and the plural forms, illustrated by (12). (12) a. nɔ -ŋɛ [nɔ ŋɛ ] mother-cl mother b. nɔ -bi [nɔ bi ] mother-cl mothers

11 10 2. Tagba tonology 28 We analyze Tagba tone according to the general theoretical assumptions of Autosegmental Phonology (Goldsmith 1976). In this framework the skeleton, segments and tones are all on different tiers. 29 The underlying tone(s) of a root can be deduced from the contrast between forms with and without suffix. (13) a. ped [pèlè] swell-compl swelled ped-go [pèlɩ ɣɔ ] swell-incompl swelling b. gog [gbɔ ɣɔ ] gather-compl gathered gog-ji [gbɔ ɣɔ ji ] gather-incompl gathering 30 (13a) and (13b) have the same structure: a CVC root without suffix for the completive form. Both have two surface L tones. On the other hand, the CVC-CV suffixed form for the incompletive aspect gives different surface tones: LLH and LLL. We thus deduce that the root ped has underlying LH tones, while the root gog has underlying L tone. 31 At first glance, Tagba seems to be a perfect example of the tone association proposed by Goldsmith (1976). From left to right, tones and syllables are associated one to one, and a syllable without tone will receive the preceding tone. (14) a. kú-dɩ [kúlɩ ] cry- ACCOMP cried b. cěk-i [cèʔí] smile-unaccomp be smiling c. Ɉɩ t-da [Ɉɩ dɛ ] tongue-cl tongue 32 These examples can be illustrated as shown in (15): when a root has one underlying tone, it will spread onto vowel(s) without tonal association; when it has two underlying tones and two vowels, the first tone is associated with the first vowel, the second tone with the second vowel. 33 (15) 2.2. More complex patterns 34 But patterns are more complex in the following cases: 1) when there are more underlying tones than underlying vowels; 2) when there is vowel epenthesis; 3) when there is vowel deletion. Let s examine each of them.

12 11 35 If a root has two underlying tones and no suffix, the second tone will be floating instead of piling up on the vowel, as shown in (16): (16) a. cě [cè] trembled b. 36 If a CVC root with two underlying tones has no suffix, the second tone will also be floating. Let s compare (16) to (17): (17) a. cěk [cèʔè] smiled b. 37 The second vowel in (17) is an epenthetic one, the H tone floats just as in (16). 38 A tone, once floating, cannot cross the word boundary. For example, in the compound saď.fɔ d-ŋɛ > [sàlɩ fòlòŋɛ ], laziness.owner-cl, the floating H tone in a cannot spread to the next vowel of fɔ d owner. 39 If a CVC root takes a CV suffix, the root V will be copied entirely as an epenthetic vowel, as shown in (18), or partially as a single feature, as shown in (19). The epenthetic vowel does not have an underlying tone; it just copies the preceding tone. (18) a. ŋɔ t-ga [ŋɔ rɔ ɣɔ ] smoke-cl smoke b. (19) a. ka d-gɛ [kaĺɩ ɣɛ ] armpit-cl armpit b. 40 If we compare (18) and (19) to (17), we see that they are indeed the same pattern: the epenthetic vowel is transparent to lexical tone assignment. 41 Finally, if a CV root takes a V suffix, the two heteromorphemic vowels trigger processes of hiatus resolution. This appears as four strategies that are shown in (20).

13 12 (20) a. Deletion of the first vowel Ɉe -ɩ [Ɉɩ ] enter, entering lɔ -u [lu ] take, taking b. Deletion of the second vowel with tonal inversion ja -ɔ [ja ] son ɲa -ɔ [ɲa ] see, seeing c. Vowel assimilation to V2 Ɉě-i [Ɉi í] wash, washing Ɉa -i [Ɉií ] break, breaking d. Glide formation cɛ -u [cɛ w] know, knowing sɛ -u [sɛ w] produce, producing 42 In (20a), the first vowel is deleted, and the second one takes the second underlying tone; in (20b), the second vowel is deleted, but the first one also takes the second underlying tone; in (20c), the first vowel is assimilated to the second one; in (20d), the second vowel turns into a glide. 43 In a general way, high vowels dominate low ones and front vowels dominate back ones 3. The data in (20) are crucial for our analysis and we will come back to them in section A rule-based approach would solve the problem posed by these patterns with a series of ordered rules and to predict the data in (16) (19), the tone assignment must precede the epenthesis. (21) a. Tone assignment rule: b. Epenthesis rule:

14 13 c. Tone copy rule: 45 Similar rule ordering will predict the data in (20a): tone association must precede the vowel deletion. As for data in (20b-d), we will provide an analysis in section Tier hierarchization 46 As we have shown in 2.2, Tagba data do not fit the tone association convention perfectly. Let us look first at a canonic case of tonal behavior. (22) Ewe vowel coalescence (Clements & Ford 1979) mēkpɔ ètú > mēkp ɔ tú 47 In (22), to resolve the hiatus /ɔ è/, Ewe chooses to delete the second vowel /è/, but not its underlying L tone. This kind of phenomenon has been considered to be strong evidence since the inception of the autosegmental theories that the tonal tier is independent from the segmental tier. However, if we compare the Tagba data in (20a), we observe the opposite phenomenon: the tone drops with vowel deletion. If Tagba behaves differently, the organization of its tonal and segmental tiers should then be reconsidered Previous discussions on tier hierarchization 48 According to Yip (2002: xxi), the Tone-Bearing Unit (TBU) is the entity to which tones associate. It may be syllable, mora or perhaps vowel. However, this definition is far from exhaustive. Odden (1995) summarizes previous discussions on this question: for Goldsmith (1976), the TBU would be the vowel, i.e. tone is directly associated with a vowel; for Goldsmith (1995), it is the syllable. Odden agrees with Clements & Ford (1979) that the TBU is a higher prosodic unit. "There has been some ambiguity in previous uses of the term tone-bearing unit. It is maintained here that tones are not directly associated with vowels or other segments,

15 14 but rather with higher-level units such as the syllable or the syllable final... " Clements & Ford (1979: 181), 49 The higher-level units can also be the skeleton. This possibility is shown in Bao (1999: 7). However, although the skeleton is assumed by different autosegmental theories, i) it is not expressed in the same way, for example, x-skeleton (Archangeli 1985) and CVskeleton (Clements & Keyser 1983); ii) it is often implied but not directly expressed in autosegmental representations. The nature of the TBU implies different tier hierarchization. In this paper we will assume the x-skeleton. 50 Bakwiri slang (Hombert 1973), shown in (23), is another famous example of how tone is autonomous regarding segments. (23) lík w è k w élì to fall 51 If tones were associated to segments, tones would move with the vowels. Therefore, in an autosegmental representation, tones should not be associated to segments, but to higher prosodic units such as the mora, the syllable or the skeleton. In the latter case, the example in (23) could be represented as follows: (24) 52 This confirms Clements & Ford (1979) and Odden s (1995) point of view that tone should be associated to higher-level units but not directly to the vowels. 53 In the discussions above, the association of tone with higher prosodic units is relevant for analyses. However, Yip (2002) claims that "in the case of language with only monomoraic, open CV syllables, where each syllable bears exactly one tone, the TBU could be vowel, mora or syllable." 54 Accordingly, the following notations are equivalent for a language with only monomoraic open CV syllables

16 15 (25) 55 Since Tagba is a language with only CV syllables, we ll examine below if different notations of tonal association are really equivalent Analysis of Tagba 56 Tagba appears to be a counterexample for Yip s claim. Two cases: vowel epenthesis and vowel deletion, plead for an analysis of tone-to-vowel association. Let s compare the predictions of tone-skeleton association and tone-vowel association. 57 The two predictions are shown in (26): (26) Epenthesis: ŋɔ t-ga > [ŋɔ rɔ ɣɔ ] smoke a. Tone having access to skeleton b. Tone having access to vowels 58 In (26a), the association (arrowed lines) of the epenthetic vowel [ɔ] and that of the L tone are two separate operations. Once the vowel [ɔ] is copied to solve the CC cluster at the morphological boundary and thereby satisfy the CV syllable structure requirement, we have to suppose an additional tone copy rule to associate the L tone with the skeletal position with which [ɔ] is associated. 59 In (26b), in contrast, since L is already associated with [ɔ], the copy of this latter (arrowed line) will certainly also copy the L tone. This way, a single operation, vowel epenthesis, is needed to avoid the CC cluster. (26b) is more economical than (26a), in the sense that it demands a single operation instead of two, and should thus be preferred. 60 If for (26) the simpler one of the two solutions is the better one, for the vowel deletion in (20a), the hypothesis that tone has access to the skeleton in Tagba will give the

17 16 wrong prediction. Besides, the behavior of (20a) is exactly the opposite to (22), the classic example demonstrating the autonomy of tone in relation to vowels. Let s examine (27) below: (27) Vowel deletion: Ɉê-i > [Ɉi ] enter, entering a. Tone having access to skeleton b. Tone having access to vowels 61 In (27a), the L tone and the vowel [e] are associated to the same skeletal position, the deletion of [e] does not imply that of the H tone, which could give a similar case as in (22). In (27b), in contrast, since the tone is associated to the vowel [e] and this latter to the skeletal position, the deletion of [e] implies that of the H tone at the same time, which gives the attested [Ɉì], where the [i] has a L tone Analysis of other patterns 62 Our analysis covers almost all the vowel epenthesis in our corpus, although a few exceptions exist. They consist of a very small number of roots all of which have /k/ /g/ as C2: (28) a. se k-ga [se ʔe ɣi ] baskets b. cʋĝ-lɛ [cʋ ɣʋ lɔ ] pots 63 We don t have a principled explanation for these kinds of exceptions, but notice that these exceptions only occur with /k/ and /g/, which are the two weakest consonants among Tagba obstruent phonemes, in a CVC 2 -C 3 V word, if C 2 is /k/ or /g/ and C 3 another obstruent, C 2 will always drop. 64 For the strategies of vowel sequences shown in (20b-d), we should explain each of them. 65 Take (20bc) (repeated in 29), this example shows the deletion of the second vowel, with an apparent inversion of tone, that is, the first vowel is maintained, but it takes the second underlying tone.

18 17 (29) a. Deletion of the second vowel with tonal inversion ja -ɔ [ja ] son ɲa -ɔ [ɲa ] see, seeing b. Assimilation of the two vowels Ɉě-i [Ɉi í] wash, washing Ɉa -i [Ɉií ] break, breaking 66 In fact, the maintaining of vowel quantity is subject to variation, that is, the same phonological patterns can lead to different surface forms: (30) a. wa -ɔ [wa ] someone b. ta -ɔ [taá ] possess, possessing 67 Where (30a) corresponds to (29a), with the second vowel deleted, and (30b) to (29b), with the vowel assimilation. The fact that (30a) and (30b) have exactly the same phonological pattern leads us to analyze the data in (29a) not as an inversion of tone, but as the deletion of the first vowel after vowel assimilation. From a derivational viewpoint, we propose that the data in (29a) go through two intermediary steps: (31) a. underlying form: ja -ɔ b. vowel assimilation: ja -a c. tone assignment: jàa d. 1 st vowel deletion: [ja ] son 68 This analysis presents two advantages. Firstly, it unifies the tonal process in (14a) and (14b), thus tonal inversion is no longer needed; secondly, it accounts for the variation in (30): for (29a) and (30a), the surface form corresponds to (31d), yet for (29b) and (30b), the surface form corresponds to (31c). This view analyzes the variation not as a coincidence, but as part of the phonological structure already present at the underlying level. 69 As for (20d), the second vowel /u/ becomes a glide and thus cannot be a TBU Conclusion 70 In this paper we have presented data in Tagba and examined its tonal patterns. The problem that Tagba raises for autosegmental phonology is how the different tiers should be organized in this language. Previous studies either claim that tone is universally associated with higher-level units, or make the assumption that all tier

19 18 hierarchizations are equivalent. Tagba shows that whether tones are associated with the skeleton (positions, morae or other units according to different theories) or with segments (vowels) is relevant for the prediction of tonal patterns. Rather than a universal claim that they should be associated with higher-level units or that the type of association makes no difference, we propose that for Tagba, tone is associated directly to vowels. Archangeli, Diana (1985). Yokuts Harmony: Evidence for Coplanar Representation in Nonlinear Phonology. Linguistic Inquiry 16, Bao, Zhiming (1999), The structure of tone. New York: Oxford University Press. Clements, G.N. & K.C. Ford (1979), Kikuyu tone shift and its synchronic consequences. Linguistic Inquiry 10: Clements, George N. & Samuel Jay Keyser (1983), CV phonology, a generative theory of syllable. Cambridge : Mass. The MIT Press. Delafosse, Maurice (1904). Vocabulaires comparatifs de plus de 60 langues ou dialectes parlés à la Côte-d Ivoire et dans les régions limitrophes. Paris : Ernest Leroux. Delafosse, Maurice (1911). Les langues voltaïques (boucle du niger). Mémoire de la Société de Linguistique de Paris, 16, 6. Goldsmith, John (1976), Autosegmental phonology. PhD dissertation, MIT. Goldsmith, John (1995), Phonological Theory, in John A. Goldsmith (ed.), The handbook of phonological theory, Cambridge, MA: Blackwell. Greenberg, Joseph H. (1963). The Languages of Africa. Indiana University Press. Hyman, Larry M Tone: Is it different? In John Goldsmith, Jason Riggle & Alan C.L. Yu (eds), The handbook of phonological theory, 2nd ed., Blackwell. Hombert, Jean-Marie (1973). Speaking backwards in Bakwiri, SAfrL 4(3) - p Martinet, André (1970). Éléments de linguistique générale. Paris: Armand Colin. Naden, Anthony J. (1989). Gur. Dans John Bendor-Samuel & Rhonda L. Hartell, eds., The Niger-Congo Languages. A classification and description of Africa s largest language family. University Press of America, Lanham, New York, London. Odden (1995), Tone: African Languages. In John A. Goldsmith (ed.), The handbook of phonological theory, Cambridge, MA : Blackwell. Roncador, Manfred von & Miehe, Gudrun (1998). Les langues gur (voltaïques). Bibliographie commentée et inventaire des appelations des langues. Rüdiger Köppe Verlag, Köln. Roulon. P, (1968). Essai de phonologie du Tyembara (dialecte senoufo), Paris, SELAF, 57p. Traoré, Edwige Mori (2016). Étude ethnolinguistique du sicanɛ (chants de hochets des femmes senufo du Tagbara). PhD Thesis, Université d Orléans. Tressan, Lavergne de (1953). Inventaire linguistique de l Afrique occidentale française et du Togo. I.F.A.N., Dakar. Yip, Moira Tone in East Asian languages. In John A. Goldsmith (ed.), The handbook of phonological theory, Cambridge, MA: Blackwell. Yip, Moira (2002), Tone, Cambridge University Press.

20 19 NOTES 1. The only occurrences of approximants and fricatives in intervocalic position are in loanwords. 2. Virtual length (VL) is developed in works assuming the abstractness of phonology, in particular within the frameworks of Government Phonology and related theories. The essential idea is that phonological representations do not have to correspond immediately to a phonetic reality. In the Tagba case, the reason allowing us to consider phonetic [d] as a VG is that a simple intervocalic /t/ or /d/ always becomes sonorant. Readers interested in VG or VL in general are invited to consult Ségéral & Scheer (2001) for a survey of the debate since generative phonology. 3. As mentioned by one of the reviewers, Tagba appears to be one of the languages which challenge the primacy of structure on segments. Indeed, word structure depends on a competition between two segments in contact, here two vowels. By dominate we mean the primacy of high over low vowels and front ones over back. 4. An alternative solution, as pointed out by one of the reviewers, would be the mora: In a V1-V2 sequence, if tone is associated to mora, and the weight of the mora depends on the competition between /e/ and /i/, here the first μ falls, and both H tone and /e/ fall as well. 5. In Tagba, the glides cannot be vocalic, but must be consonantal, for this language avoids hiatus at all costs. Glides can of course be the second vowel of a diphthong in some languages such as the Mandarin Chinese, for they are moraic and clearly take part of the contour tone. See Yip (1995) for details. ABSTRACTS Tagba (Senufo) is a tone language that presents tonal patterns that appear quite regular at first glance. However, these are difficult to model under standard autosegmental hierarchical assumptions regarding the placement of skeleton, segments and tones. In this paper, we will present Tagba tonology and demonstrate that it can be accounted for by readjusting the tiers of hierarchization. From this, we reason that the proposed hierarchy is parametric and language specific. Le tagba (senufo) est une langue à tons qui présente des patrons tonals plutôt réguliers au premier abord, cependant, le point de vue traditionnel sur la hiérarchie des lignes de représentation dans le modèle autosegmental fait des prédictions qui s avèrent inappropriées sur certains patrons tonals dans cette langue. Dans cet article nous présenterons la tonologie du tagba et démontrerons qu un aménagement de la hiérarchisation des lignes de représentation

21 20 rendra compte de ses patrons tonals complexes, ainsi nous raisonnerons que cette hiérarchie, au lieu d être universelle, est paramétrique en fonction des langues. INDEX Mots-clés: tagba, ton, segments, squelette, lignes, hiérarchie Keywords: Tagba, tone, segments, skeleton, tiers, hierarchy AUTHORS MORI EDWIGE TRAORÉ Laboratoire Ligérien de Linguistique (UMR 7270) Université de Ouagadougou / Université d Orléans mori.sanou@univ-orleans.fr XIAOLIANG LUO Laboratoire Ligérien de Linguistique (UMR 7270) Université d Orléans Xiaoliang.luo@univ-orleans.fr

22 21 The cognitive motivation and purposes of playful blending in English Boris Lefilliâtre Introduction 1 A communicative situation typically displays an organisation of simplex and complex words whose formations are based on well-known, familiar morphemes. A simplex word like ball can be defined as a monomorphemic lexeme (Van Der Hulst 2008 : 233), as opposed to a complex word commonly referred to as a polymorphemic form (Fuster Márquez 2008 : 66). Complex words comprise derivatives, made up of a root to which at least one affix is added, for example with unmanliness in which the root man has been prefixed with un- and suffixed with -ly and -ness (OED Online 2018, s.v. unmanliness, n., unmanly, adj., manly, adj. ), and unclipped compounds, formed with unclipped words and/or combining forms, such as hotdog (Sanchez-Stockhammer 2018 : 27), hydroelectric (Mattiello 2013 : 35), hydrology (Mattiello 2013 : 35) or microscope (Baeskow 2004 : 72). In sharp contrast with its environing text, a playful blend stand out of the discourse here and there. 2 Blends are atypical words since the morphemic approach to complex and simplex words prevents us from defining them as complex words. A blend may imply a clipped segment called a splinter, thereby introducing a submorphemic, i.e. non-morphemic element in the combination (Bauer 1983 : 234 ; Mattiello 2013 : 34). 1 In most cases, more than one splinter is involved, as in camcorder (camera + recorder) 2 (Bassac 2012 : 182 ; Mattiello 2013 : 305). The created lexemes that combine back-clipped words, with at least two initial splinters, constitute clipped compounds and are excluded from the category of blends (Plag 2003 ; Gries 2004, 2006). These words include sitcom (situation + comedy) (Renner, Maniez, Arnaud 2012 : 7) or SoLoMo (social + local + mobile) (Beliaeva 2014 : 7). Moreover, this paper focuses on lexical blends, thereby excluding blending errors (Bertinetto 2001 : 2). Lexical blends are intentional and communicative,

23 22 whereas blending errors are accidental and caused by a derailment of the psycholinguistic system without any communicative function or purpose (Bertinetto 2001 : 18). These performance errors are also called slips or lapses (Cannon 1986 : 730, 750). Lexical blending may thus be summarily defined as follows: the intentional combination of two or more elements, shortened by the use of at least one splinter, or two or more splinters which are not all apocoped versions of their base words (as they would be called clipped compounds ). 3 3 Not only are blends outstanding in their context of utterance as they are surrounded by morphemically based words, most of which are simplex words or derivatives, but some specific blends stand out on account of their playfulness. 4 It is claimed that sometimes, blends are likely to be chosen partially at random (Bauer 1983 : 234). This arbitrariness deserves to be mentioned, but it is only partial (Bauer 1983 : 234), which means that the use of blending may be multifactorial. The arbitrary part will be left aside so as to concentrate on motivation and purpose. Motivations for blending include morphological transparency (Cannon 2000 : 954) to designate new referents in an explanatory or defining way, so as to fill a semantic void in the lexicon (Cannon 2000 : 953). They are serious coinings, often naming man-made objects, processes, and facilities (Cannon 1986 : 746). For the name of the company Groupon (group + coupon) (Eldridge 2011), the use of blending is motivated by the explanatory function of the source words, as the enterprise groups together local merchants and subscribers who collect money-off vouchers or coupons. A situation which has been much dealt with in the news is the withdrawal of the United Kingdom from the European Union, referred to with the blend Brexit (British + exit), right from the evocation of the idea in 2012, and especially with the referendum held in the UK on 23 June 2016 that favoured withdrawal from the EU (OED Online 2018, s.v. Brexit, n. ). Blending may also be motivated by a meliorative way of designating or describing a referent. For instance, the noun of the entertaining enterprise Funimation ( fun + animation) (The Freelance Studio 2016) which dubs and distributes foreign animated film, especially Japanese or Japanese-style animes, was produced with the help of the meliorative term fun, thereby enhancing the image of the company with its very name. Concision may be mentioned as another major motivation to blend words. Blending is a denser way of speaking or writing, which save(s) time and paper in communicating (Devereux 1984 : 210), according to the principles of least effort and cognitive economy (Mattiello 2013 : 27 ; Tournier 2004 : 162, 195), also referred to as linguistic economy (Tournier 2004 : ) regarding the process of clipping inherent to blending. Easiness of written or spoken performance with the help of concision, or communicative economy (Grzega, Schöner 2007 : 36 ; Miller 2014 : 84), also motivates blending. The phrase walking marathon is occasionally replaced by the blend walkathon (walk + marathon) (Bertinetto 2001 : 28 ; Bryant 1974 : 171), or the blend telecast is sometimes conveniently substituted to the longer formula television broadcast (Bertinetto 2001 : 28 ; Devereux 1984 : 213 ; Mattiello 2013 : 310). The economy may be significant in blends made up of more than two bases: turducken is produced with three etyma (turkey + duck + chicken) (Bassac 2012 : 169), like affluemza (affluence + influence + feminism) (Bassac 2012 : 169), while afflufemza is coined with four items (affluent + feminist + mothers + influenza) (Beliaeva 2014 : 7). 5 The motivation for blending under study in this paper is word play, 4 which will be explained in the first part of this paper. Studying playful blending is an efficient way of

24 23 clarifying the link and distinction between the notions of motivation, purpose, means and end underlying the mechanism of lexicogenesis. The means are embodied by the form of the created word, whereby the word-formation process is essential. The end is twofold: what is often merely called purpose corresponds to the intention of the coiner, and motivation is the purpose of the chosen form, i.e. the reasoned nonarbitrary choice to produce a form providing a certain meaning and lexical effect. In accordance with the literature, this paper will refer to the purpose of the chosen form as the motivation, and designate the purpose of the speaker or writer as such or with synonymous words like intention, goal or aim, keeping in mind that motivation is oriented to lexical forms, while purpose is oriented to the interlocutors. 6 The motivation and purposes of playful blends will be explicated and exemplified. Furthermore, this paper will examine whether the two concepts can be distinguished with the consideration that motivation would be context-free and purpose contextdependent. In order to inspect this question, different utterances of the same examples of playful blends will be examined, and synonymous forms inside and outside playful blending will be analysed. The blends are selected from a list of examples provided in the literature. This list includes blends which are playful or not, from which are selected playful blends according to the criteria of playfulness which are developed in the first part of this study. Utterances of these blends are found online, so as to illustrate combinations of various motivations in addition to word play, and manifold purposes justifying the use of an atypically non-morphemic playful form in the discourse. Usage variations among coreferential blended words are made explicit with the help of corpora and corpus-based dictionaries, in order to identify different motivations The motivation of playful blending 7 The motivation of playfulness in blending relies on the association of linguistically related source forms. With this definition of the playfulness of specific blends, all blends are not playful. If the outputs airmada (air + armada) (Bryant 1974 : 174) and airobics (air + aerobics) (Mattiello 2013 : 303) both play with the common diphthong /eə/ in their etyma, other blends, such as organule (organism + molecule) (Cannon 1986 : 731) or Heliport (helicopter + airport) (Connolly 2013 : 4) show no morphological or semantic link in their source words ; they are therefore not playful at all. 8 Blending is considered playful when it is etymologically playful, as well as employed playfully in communication. It excludes the words which are opacified as blends. Some blends are so well-integrated into standard English that people tend to be unaware of their status (O Grady, Dobrovolsky, Katamba 1996 : 158), as they hear or read them like simplex forms. In such cases, they are not (or no longer) playful, even if their source words are linguistically linked. These blends include: chortle (chuckle + snort) (Algeo 1977 : 51 ; Bertinetto 2001 : 27 ; Devereux 1984 : 214 ; Cannon 2000 : 955 ; Mattiello 2013 : 305 ; O Grady, Dobrovolsky, Katamba 1996 : 158) motel (motor + hotel) (Algeo 1977 : 52 ; Bat-El 1996 : 283 ; Bertinetto 2001 : 27 ; Cannon 2000 : 954 ; Connolly 2013 : 3 ; Devereux 1984 : 213 ; Gries 2004 : 415 ; Renner, Maniez, Arnaud 2012 : 3) bit (binary + digit) (Bertinetto 2001 : 27 ; Devereux 1984 : 213 ; O Grady, Dobrovolsky, Katamba 1996 : 158)

25 24 modem ( modulator + demodulator) (Mattiello 2013 : 308 ; O Grady, Dobrovolsky, Katamba 1996 : 158). 9 If a few of them used to be playful originally when they were created, nowadays they cannot be used for word play in so far as their source words are unknown by the users. However playful they can be etymologically, their opacification prevents them from being used playfully. 10 Word play in blends is a motivation taking root in a linguistic relatedness between the blended source words, but criteria need to be defined for this special relation between the input forms. Blended words from the English language are predictable according to their semantic links (i.e. their playfulness), as Swedish blends are (Ek 2018), and blend structure is ( ) constrained by phonological and semantic considerations (Lepic 2016), especially regarding playful blending. Mattiello (2014 : ) explains that the speaker tends to choose two source words which are similar to each other in terms of: 1) phonemic and/or graphemic length and stress pattern, 2) part of speech, and sometimes even 3) semantics (as in fantabulous), in the most intentional (coordinate) blends as well as in speech-error blends. According to the present definition of word play in blending, only the two lexical elements are retained: morphology and semantics. The syntactic criterion is discarded, on account of the existence of morphologically and/or semantically related source words of different parts of speech in jumbles, which are described as associating collocatable and semantically related but syntactically diverse bases (Algeo 1977 : 58). Thus, the blend numberous (noun number + adjective numerous) (Algeo 1977 : 58 ; Mattiello 2013 : 308) is playful as it associates bases with similar morphological segments, even if their bases belong to different parts of speech, and beermare (noun beer + noun nightmare) (Bertinetto 2001 : 27) is not playful as it associates bases without any morphological or semantic link, even if their bases belong to the same part of speech. 11 Morphological links of source words resulting in playful blends may occur with a common base morpheme, affix or any similarity of sound (Algeo 1977 : 57) or written form. Similarity between the source words fuels the playful character of blends (Gries 2004 : 417, 419 ; Kelly 1998 : ). It occurs that the similarity is such that it originates the formation of blends without the need to remove any part of the bases, as in predictionary (prediction + dictionary) (Beliaeva 2014 : 17). The base words may have a shared affix (Algeo 1977 : 57), for example with the prefix pre- in imperence (impertinence + impudence) (Bertinetto 2001 : 27 ; Cannon 1986 : 738), or the suffixes -ing in aggranoying (aggravating + annoying) (Mattiello 2013 : 303), -er in the competing forms scinter, scrinter and sprinter (scanner + printer) (Bauer, Lieber, Plag 2013 : 462) or -ly in prezactly (precisely + exactly) (Arndt-Lappe, Plag 2012 : 1 ; Bauer, Lieber, Plag 2013 : 459). Any similarity of sound may be a factor of playful blends. The forms torrible (terrible + horrible) (Gries 2004 : 425 ; Mattiello 2013 : 310) and slickery (slicker + trickery) (Bryant 1974 : 171) indicate that an overlapping morphology between pre-existing words prompts the formation of blends. Sometimes, the playfulness does not rely on a similarity between the bases, but with another source word which is homophonous with the output form, as in the pun Maxim-Eyes (maximum + eyes) (Bryant 1974 : 166) whose pronunciation is the same as the verb maximize. In puns 6, is it the graphic and/or the phonological similarity between the source words that prevail(s)? 12 In puns with a double meaning mixing two perfect homonyms, as in grave meaning both serious and tomb (Algeo 1977 : 50), neither the graphic nor the phonological

26 25 similarity prevails. For these words, both the graphic and phonemic forms motivate the blending and it is unclear which of the spoken or written form matters most. 13 By contrast, in other playful blends, either the written or the spoken form is of particular importance. In many puns where there is a double meaning, the source words are perfect homophones but imperfect homographs, so that the written form indicates the use of blending and the presence of a double meaning. This is what happens in the following puns: in-sin-uation (insinuation + sin) (Algeo 1977 : 50) racqueteer (racquet + racketeer) (Algeo 1977 : 50) sinema (sin + cinema) (Algeo 1977 : 50 ; Cannon 1986 : 746) cellebrity (cell + celebrity) (Algeo 1977 : 50) millionheiress (millionairess + heiress) (Algeo 1977 : 50) pessimystic (pessimist + mystic) (Mattiello 2013 : 308) 14 In such examples, written forms prevail, sometimes merely through one letter (for sinema) or typography (hence the hyphenation in in-sin-uation ). Conversely, it transpires that sound similarity is a motivation for playfully blended words such as peekture (peek + picture) (Bertinetto 2001 : 27 ; Bryant 1974 : 181), as it is the proximity in the vowel space positions between short lax /ɪ/ and long tense /iː/ that triggers off the blended form. The same holds true for rendezwoo (rendezvous + woo) (Bryant 1974 : 183 ; Mattiello 2013 : 309), with the common /u:/ phoneme written differently in the bases. 15 Regarding a semantic relation, blends may unite synonymous words. 7 Many of them are near-synonymous or synonymous adjectival blends (Bassac 2012 : ), like fantabulous (fantastic + fabulous) (Arndt-Lappe, Plag 2012 : 2 ; Bassac 2012 : 190 ; Bertinetto 2001 : 27 ; Mattiello 2013 : 306) or confuzzled (confused + puzzled) (Bassac 2012 : 190 ; Mattiello 2013 : 305). There are also synonymous blends from other parts of speech, like the verbs bonk (bump + conk) and shill (shiver + chill), or the nouns stocks (stockings + socks) and needcessity (need + necessity) (Algeo 1977 : 57). Conversely, some blends are playful because of the oxymoronic relation between their bases. These blends include antonymic pairs, as in frenemy ( friend + enemy) (OED Online 2018, s.v. frenemy, n. ) or craptacular (crap + spectacular) (Mattiello 2013 : 305 ; Oxford Dictionaries 2019, s.v. craptacular ), but not all are playful: antonymic blends like immittance (impedance + admittance) (Cannon 1986 : 741), or near-antonymically related source words like compander ( compressor + expander) (Cannon 2000 : 955), are technically motivated, the former in the area of electricity and the latter in telecommunications and electronics. Antonymic bases create a playful blend only if they have an oxymoronic relation. Moreover, if the semantic link stems from a mere collocation, as is the case with certain jumbles (a specific sort of blend mentioned previously) such as dumbfound (adjective dumb + verb confound) (Algeo 1977 : 51, 58 ; Mattiello 2013 : 306), there is no word play. All input forms are indeed blended on account of a link between them, but playfulness does not occur for any sort of association. Furthermore, playfulness is limited to linguistic links. Other words are blended because they are linked extralinguistically. A relation may occur between two referents, whose corresponding source words are not related semantically, but pragmatically, in a given communicative situation. They are called dvandva blends (Algeo 1977 : 57) and the examples provided are gasid indigestion ( gas + acid indigestion) (Algeo 1977 : 58), Hungarican (Hungarian + American) (Algeo 1977 : 52 ; Mattiello 2013 : 307) and smog (smoke + fog) (Algeo 1977 : 50 ; Arndt-Lappe, Plag 2012 : 2 ; Bauer, Lieber, Plag 2013 : 459,

27 ; Bertinetto 2001 : 28 ; Cannon 2000 : 954 ; Devereux 1984 : 213 ; Mattiello 2013 : 309). Dvandva blends are here discarded from the category of playful blends. 16 It may be summarized that word play is circumscribed to source words with an oxymoronic effect or a linguistic similarity (either semantic or morphological); those which present an antonymic relation are playfully blended only if they are used for an oxymoronic effect. As discussed previously, the inputs may include another word than the etyma, when the output blend is homophonous with an existing word, which makes the blend playful. Exploiting recognizable (non-opacified) linguistic relatedness between attested words to blend them constitutes the motivation of the locutor to play on these words. The motivation of playfulness is cognitive as the minds of the interlocutors have to recognize this special relation between two or more source items among pre-existing words. 17 A cognitive remotivation of certain playful (i.e. linguistically related) blends as phonesthetic blends has recently been studied (Smith 2014). This study reinforces Bauer s suggestion (1983 : 235) that the preservation of initial consonant clusters, in other words initial phonesthemes, plays a role in determining the structure of short blend outputs (Smith 2014 : 12). In the area of sound symbolism or complex iconicity (Miller 2014 : 155), phonaesthesia is defined as the network-like association between words sharing partial form and meaning (cf. De Cuypere 2008 : 113 ; Miller 2014 : 155). This paper falls within this scope, and suggests that playful blends between similar input forms influence the formation of other blends adopting the previously attested blends consonant clusters to make them phonesthemes, i.e. sub-morphemic unit(s) that ha(ve) a predictable effect on the meaning of a word as a whole (Otis ; Sagi 2008 : 65). A case in point is the playful blend blotch (blot + splotch) (Smith 2014 : 23, 25, 26-28, 38), which probably triggered the coinage of blurt (blow + blare) (Smith 2014 : 23, 27, 38, 41), thus establishing the existence of the phonestheme bl- containing the idea of breathing or swelling (Drellishak 2006 ; Smith 2014 : 23 ; Tournier 1985). In a similar way, slosh (slop + slush) (Smith 2014 : 15-16, 23, 39, 41) incited the formation of slithy (slimy + lithe) (Smith 2014 : 23, 39, 42), on account of a sl- phonestheme bringing about a sense of slowness and sloth (Tournier 1985), or the idea of attack or sliding movement (Reay 2009). Likewise, glaze (gaze + glare) (Smith 2014 : 23, 38, 41) precedes the formation of glob (blob + gob) (Smith 2014 : 16, 23, 39, 42), adopting a gl- phonestheme related to the notions of light, vision, or light reflecting on a smooth surface (Drellishak 2006 ; Smith 2014 : 23 ; Tournier 1985). Thus, on the basis of a playful blend, phonesthetically motivated blend-words are formed afterwards, which engenders the remotivation of the previously coined blend as not only playful but also phonesthetic, in accordance with the cognitive and psycholinguistic theory that there is a structured sound symbolism upon which coiners rely to produce novel forms (Smith 2014 : 12). The motivation of playful blends may be bidirectional with the help of phonaesthesia. 18 Playfulness in blending is achieved with the help of a recognizable linguistic relatedness between the source lexemes, taking the form of a morphological similarity, or a semantic link with similar or oxymoronic source items 8. The motivation of word play is sometimes doubled afterwards, when speakers or writers associate semantic features to a consonant cluster to coin other blends, thereby remotivating the original playful blend with phonesthesia. As the aim of the form of playful blends has been explored, this paper then moves on to the aim(s) of coiners playing with existing words through blending.