Silvia Cacchiani University of Modena and Reggio Emilia* UDK 811.131.r373.611 ON UNFAMILIAR ITALIAN LEXICAL BLENDS FROM NAMES AND NOUNS 1. INTRODUCTION Italian has recently witnessed a steady increase in the use of unfamiliar lexical blends from names and nouns with an identificatory and descriptive function (in the sense of Anderson 2007), e.g. Draquila < Dracula [Dracul] + L'Aquila, or docciabatta < doccia [shower] + ciabatta [slipper]. An important fact about blends is that they are created in extragrammatical morphology (Dressler 2000; Ronneberger-Sibold 2006, 2010). Unlike regular inflection and word formations - governed by rule-based grammatical competence -, they are consciously created to reduce morphotactic and mor-phosemantic transparency (Dressler et al. 1987; Dressler 1999). The resulting increase in processing time and effort makes the blend memorable (Lehrer 2003) in the interest of relevance (Sperber/Wilson 1986/1995) and effectiveness (maximum success chances) versus efficiency (minimum effort) (cf. de Beaugrande/Dressler 1981). More particularly, while intentional blends from names and nouns are coined with careful attention to the semantic concepts encoded by the individual source words (SWs), processing and understanding the blend depends on the decoder's direct or surrogate experience of the related reference. Whereas lexical blending has typically been dismissed as an elusive and peripheral word-formation process cross-linguistically, it is found to be underrepresented in Italian (as compared to English, German and French, cf. Bertinetto 2001). It is therefore not surprising that blending has been often neglected in Italian linguistics. As far as we are aware, the only exceptions are Migliorini (1949), Thornton (1993, 2004b) and Bertinetto (2001). And yet, it has received far less attention than it deserves. This paper is thus an attempt to redress the imbalance and contribute to the (by now growing) body of work on morphological blending. Our main concern relates most directly to the morphosyntactic and morphosemantic transparency -and, ultimately, conceptual motivation - of Italian blends from names and nouns as operations of extragrammatical morphology. The paper is structured as follows. Because the delimitation of the category is a matter of debate, in Section 2 we provide a working definition of lexical blends. At the same time, we touch upon the identificatory function of names. Section 3 is devoted to our research question. Specifically, we shall carry out a qualitative investigation into a restricted number of examples selected from a collection of 100 blends that we mainly gathered from the World Wide Web. They are used in media language and journalese, radio programmes, TV * Author's address: Universitá degli Studi di Modena e Reggio Emilia, Dipartimento di Scienze del Linguaggio e della Cultura, Largo Sant'Eufemia 19, 41121 Modena, Italia. Email: silvia.cacchiani@unimore.it programmes and movies, product advertising and marketing, and event advertising. While Section 3.1 concentrates on morphotactic transparency, Section 3.2 deals with morphose-mantic transparency and conceptual motivation. Naturalness Theory (Dressier et al. 1987, Dressier 1999) is extremely well-designed to analyze and categorize blends, with an eye at their categorical status within the radial category of subtractive word formations. We shall further touch upon conceptual motivation in the composite structure of individual morphological blends using concepts from Conceptual Metaphor Theory (CMT, cf. Lakoff/Johnson 1980) and Blending Theory (Ruiz de Mendoza 1998). This will enable us to provide some preliminary remarks on the meaning predictability of specific subtypes, and blends from personal names used in journalese and media language in particular (e.g. Berluscotti < Berlusconi + Bertinotti, Berlingotti < Berlinguer + Bertinotti), from the perspective of Cognitive Linguistics (cf. Langacker 1987, Kemmer 2003). 2. BLENDS FROM NAMES AND NOUNS On a narrow definition of the term (Bat-El 2006: 66), lexical blends are word formations in which the inner edges of the SWs are truncated, which excludes clipped compounds (sitcom < situation + comedy, mocamp < motor + camp). The most widely accepted view, summarized by Gries (in press), is that an intentional lexical blend is: an intentional fusion of typically two (but potentially more) words where a part of a first source word (sw^ - usually this part includes the beginning of sw1 [left base word] - is combined with a part of a second source word (sw2 [right base word]) - usually this part includes the end of sw2 - where at least one source word is shortened and/or the fusion may involve overlap of sw1 and sw2, which excludes speech error blends and clipped compounds (Gries, in press: complex clippings). Following Migliorini (1949), Italian blends are typically addressed under the heading parole macedonia (literally, Eng. fruit-salad words, cf. Thornton 1993: 153).1 Italian parole macedonia (Thornton 2004b) appear to be mainly coined juxtaposing SW beginning and SW, or beginning plus beginning. A few recent creations are: - splinter2 + SW: docuromanzo < documentario [documentary]/docu-3 + romanzo [novel] (possibly based on docufiction, a borrowing from Eng. docu-fiction < documentary + TV fiction, cf. GRADIT: docufiction); 1 More recent terms for blend are incrocio [intersection] and fusione [blending] (Bertinetto 2001: 62, footnote 1). 2 With Lehrer (2003), we define splinters as parts of truncated words that do not necessarily coincide with syllables or morphemes. 3 We specify 'combining form' for combing forms that started out in a blend and are now recorded in the Grande Dizionario Italiano dell'Uso (GRADIT) under separate entries, and/or are listed among Italian 'elementi formativi' [combining forms] in Grossman/Rainer (eds) (2004). Where the form is not recorded in the dictionary and yet it already appears to be forming new word families by analogical extension from the blend, the combining form follows the corresponding SW. - splinter + splinter: Edilplast < edilizia [building trade]/combining form -edil- + plastica [plastic]; enervit < energia [energy] + vitamina [vitamin/combining form -vit; Cerealvit < cereali [cereals] + vitamina [vitamin]/combining form -vit; colf < collaboratrice [assistant] + familiare [family, A]; Tipolito < tipografia [printers/combining form -lito- + litografia [litographer's]/combining form -tipo-). - SW + splinter: Mentadent < menta [mint] + denti [teeth]. By contrast, word-internal shortening and partial overlap would be features of English blends (e.g. brunch < breakfast + lunch, motel < motor + hotel, or smog < smoke +fog). While cross-linguistic comparison shows that these appear to be highly dispre-ferred techniques in Italian (Bertinetto 2001: 66), they are nevertheless used in a number of novel Italian blends as adjectives and, most often, nouns and names, e.g.: - with overlap: semplogica < semplice [simple/plain] + logica [logical]/combining form -logica; gastronauta < gastronomo [gourmet]/neoclassical compound gastro- + astronauta [astronaut]/combining form -nauta, oroschifo < oroscopo [horoscope] + schifo [disgust]; Sarlusconi < Sarkozy + Berlusconi; - without overlap: modtro < moderno [up-to-date] + rétro [retro]; radicalsauro < radicale [politically radical] + dinosauro [dinosaur]/combining form -sauro. Bauer (1983: 233) loosely defines blends as "new lexeme[s] formed from parts of two (or possibly more) other words in such a way that there is no transparent analysis into morphs", though "in many cases some kind of analysis can be made" and "at least one element is transparently recoverable". Blends thus comprise so-called portmanteau words, which exhibit truncation at inner boundaries and optional overlap at the switchpoint (chunnel < channel + tunnel, guestimate < guess + estimate, mimsy < miserable + flimsy). Less core items tend to shade off and merge into word formations such as clipped compounds or extended acronyms when: SW! is clipped and SW2 remains unaltered (cremains < cremate + remains, mocamp < motor + camp); the blend might be processed in terms of a neoclassical compound (arcology < architectural + ecology) or a combining form that started out in blends is present (e.g. -burger, -rama, -teria); or both SWs are truncated at the right edge (linac < LINear + ACcellerator) (all examples from Bauer 1983: 233-238). In a similar vein, Thornton (1993) distinguishes between prototypical blends, with truncation and overlap, and various types of partial blends - or, better, blends that merge into other categories - which are created in extragrammatical morphology via other techniques, whereby: the final open syllable of SW1 is deleted via hap-lology in case of overlap with the onset of the initial syllable of SW2 (eroicomico < eroico [heroic] + comico [funny]); SW2 remains unaltered while the truncated word has achieved morphological status as a quasi-prefix (palaghiaccio < combining form pala-/palazzo [palace/building] + ghiacco [ice]); both SWs are truncated at the right edge and the two splinters exceed one-syllable length (Interpol < internazionale [international] + police [polizia]/combining form -pol). Whereas on a strict definition of the term blends and neighbouring categories appear to represent a cluster of related phenomena which exhibit family resemblance (Brdar-Szabo/Brdar 2008) within the radial category of subtractive word formations (Lopez-Rua 2002: shortenings), on a broad definition the related phenomena we mentioned above can be seen as less core blends, which differ from more core items along specific scales. Accordingly, Ronneberger-Sibold (2006) regards blending as deliberate extragrammatical compounding and scales German blends and clipped compounds (in her terms, fragment blends) by their morphotactic transparency. Dealing as we do with Italian blends from names and nouns, we adopt Gries's (in press) definition of lexical blends for relatively more core items and allow for relatively less core items, also shading off into the neighbouring category of clipped compounds. The main emphasis, however, lies in the morphotactic transparency of relatively more core items, and, second, in morphosemantic transparency and conceptual motivation. 2.1. Identification and description The approach outlined in this paper proceeds from the observation that blends from nouns and names are variously created for identifying and descriptive purposes. Along with prototypical personal names, bynames, place names, product names, names of abstract notions and names of numbers, days, months, etc., nicknames belong to the category Name (vs. Noun). Unlike nouns, they serve a referential identi-ficatory function in referring to individuals when used as arguments in predications. Unlike nouns or pronouns, names have an indexical in their associated concept, thus exemplifying direct reference. Although names may undergo de-semantization and lose memory of their original descriptive, lexical meanings over time, they retain their original identificatory function (Anderson 2007). Blends from names (and nouns) as names may (but do not necessarily) have a descriptive meaning to them as a result of the etymology of their source names or, more simply, may be created or chosen for their seemingly nice sound effect (e.g. Josalind < Jocelyn + Rosalind, Tyrus < Tyrone + Cyrus, cf. Marchand 1969). In other cases (e.g. celebrity (nick-)names such as Brangelina < Brad (Pitt) + Angelina (Jolie)), however, the choice is determined by description and blends are coined with careful attention to the semantic concepts related to the individual SWs. Understanding the blend does not rest exclusively on the recognizability of the two SWs, but correlates with their discriminatory, classificatory and expressive/evaluative functions, which are ultimately responsible for their denotations and connotations. Description has a role in fixing the referent, though not in making reference (cf. Kripke 1972/1981). Knowledge of the referent of a name is through direct or surrogate experience of the reference, by ostension or description in its baptism (Kripke 1972/1981; Lyons 1977: 217-218: nomination). 3. DATA ANALYSIS Blending is not governed by rule-based grammar. Though the SWs are not fully nor immediately accessible to the decoder, the phonological make-up of the blend (size, syllable structure, segmental make-up, etc.) is not accidental: while SW1 and SW2 are still similar in various ways, similarity of blend and SWs (Gries, in press: recognizability) facilitates recognition. (For more on this point, see Bat-El 2006 and references therein; Gries, in press and previous work.) In Section 3.1, we'll be dealing with the recognizability of novel blends from nouns and names within the framework of Natural Morphology (Dressler et al. 1987; Dressler 1999). The markedness (as against naturalness) of lexemes illustrating different blending techniques will be assessed along the gradable semiotic parameter of morphotactic transparency (or morphological constituency of the morphs). Morphosemantic transparency (semantic compositionality of the constituent parts) (Dressler et al. 1987; Dressler 1999) and cognitive motivation will be touched upon in Section 3.2, where we provide some preliminary remarks on the role of knowledge in understanding this type of blends. 3.1. Morphotactic transparency Competing factors which facilitate SW recognition in various ways comprise, among others, length and currency of the truncated words, as well as (Gries, in press, and references therein) a tendency of SW1 to be more frequent than SW2. Blends undergo reduction after being generated as partially predictable novel compounds. Whereas coiners tend to shorten words at their psycholinguistically motivated uniqueness points (with truncation of SW1 nearly exactly at the selection point and truncation of SW2 half a phoneme too early in SW2, cf. Gries, in press) so as to facilitate analyzability, the semantic plausibility of the connective relation has a bearing on the morphosemantic transparency of the blend and favours identification of the non-head word. Thus, iconic blends based on a coordinative compound (tigone < tigre [tiger] + leone [lion], describing a hybrid (Thornton 1993), or, in our sample, Berlingotti < Berlinguer + Andreotti, describing a political coalition) are less marked than blends that instantiate other relations (Eataly < eat + Italy). Last, the presence of graphemic, phonological and morphemic overlap also works towards analyzability. Perceptual salience has a major role in determining morphotactic transparency. Broadly speaking, perceptual salience (Dressler 1987: 117) determines a preference for syllable-initial as against syllable-final consonants, as well as for suffixes over prefixes (which move roots/stems/lexical bases away from the beginning of the word). Additionally, word beginnings and word ends are more perceptually salient than word-internal segments, which excludes blends such as *torhot (as against motel < motor + hotel). If this is so, then: i. In blends whose size (number of syllables) is identical to one of the SWs: we expect SW2 to be longer (in terms of syllables) than SW1, which is more perceptually salient and thus more easily analyzable (Prodinotti < Prodi + Bertinotti, Happyfania < Eng. happy + Epifanía [Twelfth Night]). Among the factors which operate to move the shorter SW to the right are grammatical category of the base words and/or graphemic and phonemic similarity of segments within the blend. E.g.: Meditathe < medita/meditate [reflect, 2nd sing./pl. Imperative] + the [tea, variant from Fr. thé of It. te], which retains the stress pattern of SW1 (with stress on the antepenultimate syllable) and is based on a VN compound with instrument N. In polysyllabic blends which are identical to the size of both SWs: we expect SW1 not to be longer than SW2 (morbistenza /morbi'stentea/, where the onset of the stressed syllable is taken from SW2, vs. *morbidenza < morbidezza /morbi'dettea/ [softness] + resistenza /resi'stentea/ [strength]). ii. In polysyllabic overlap blends: we expect alignment at word edges (in Optimality Theory (OT), cf. Piñeros 2004: 215: ALIGN-MWd) and syllable boundaries. E.g., Happyfania /eppi'fanja/ results from superimposition of the most likely pronunciation for Eng. happy /'hspi/, It. /'eppi/ (GRADIT: happy)4 — with silent segment h, open-mid front vowel, and geminates — and epifania /epi'fanja/. iii. To facilitate recognition of the less perceptually salient SW2, we predict the similarity-motivated preservation of SW2's recognizability to dominate recognizability-motivated preservation of letters and phonemes. Specifically, in blends with word-internal truncation: a. If blend's size is identical to both SWs, the so-called size constraint (in OT, cf. BatEl/Cohen, in press: Faith Metrical Structure (FaithMS)), would predict that metrical structure (number of syllables and stress pattern) of the blend and SWs are identical (Berluscotti < Berlusconi + Bertinotti; mordistenza < morbidezza [softness] + resistenza [strength], both with stress on the penultimate syllable). Under the principle of saliency, however, the stressed syllable and stress pattern in the blend correspond to the stressed syllable and stress pattern of SW2 (Bat-El/Cohen, in press: position-based stress assignment. More technically: FaithHeadWr >> FaithHeadWl): gastro-nauta /gas'tronawta/ < gastronomo /gas'tronomo/ [gourmet]/neoclassical compound gastro- + astronauta /astro'nawta/ [astronaut]/combining form -nauta /'nawta/.5 b. If blend's size is identical to the size of one SW, the stress pattern of that SW ha priority over the other SW: britaliano /brita'ljano/ < British / briti */ (GRADIT: British) + italiano /ita'ljano/ [Italian]; oroschifo /o'roskifo/ < oroscopo /o'roscopo/ [horoscope] + schifo / skifo/ [disgust]. 4 All phonetic transcriptions of base nouns, English borrowings and names recorded in the dictionary, are taken from the Grande Dizionario Italiano dell'Uso (GRADIT). 5 Bat-El/Cohen (in press) show that there is no priority between size- and position-based stress assignment in English blends whose size is identical to the size of the left base word. Although our data do not suggest intra-word variation here, our sample is too small to exclude non-crucial ranking between FaithMS (size) and FaithHeadWr (position) in blends with longer SWj. When blend's size differs from both SWs, if SW2 is not truncated, SW2 dominates (Bat-El/Cohen, in press: FaithHeadWr >> FaithHeadWl), as in radicalsauro /radi-cal'sawro/ < radical /radi'cale/ [radical] + dinosauro /dino'sawro/ [dinosaur]/combin-ing form -sauro /'sawro/. c. 3.1.1. A typology of Italian blends Under the principle of saliency, Italian bars maximally opaque end-end concatenations. Other blending techniques are available to Italian. Framing our typology within Naturalness Theory and adapting Ronneberger-Sibold (2006) we distinguish three blending techniques, which range from relatively more morphotactically transparent to relatively opaque: i. complete blending (telescope blending and inclusive blending), or the most easily analyzable into constituents; ii. contour blending, where blend size is identical to one of the SWs; iii. semi-complete blending, with one truncated word and one full SW. Ronneberger-Sibold (2006) also devises another category, fragment blending (Gries, in press: complex clipping), which combines SW beginnings in clipped compounds. Based on insights from research in Optimality Theory (BatEl/Cohen, in press) and on their lack of transparency, we suggest viewing fragment blending as a neighbouring, though separate category along the continuum from relatively transparent to opaque (cf. Section 3.1.2). Broadly speaking, although overlap of graphemes and/or phonemes across words represents a marked option (while biuniqueness, or one-form one-meaning correspondence, is its opposite on the naturalness scale, cf. Dressler et al. 1987), the more the overlap, the more transparent the blend. Type ia. Telescopes Within complete blending, telescope (syntagmatic) blends (Algeo 1977) are created via overlap or truncation juxtaposing at converging edges SWs that are initially generated as subsequent words within a phrase. Telescopes are highly infrequent in the corpus, possibly because they represent the most transparent (and, thus, the less memorable and effective) type. Transparency correlates positively with overlap of graphic and sound shapes and, most importantly, morphemes, e.g.: RavennAntica /ra,vennan'tika/ < Ravenna /ra'venna/ + antica /an'tika/ [historical]; Cooptima /ko'op-tima/ < coop /ko'op/ [co-op, originally from It. cooperativa] + ottima /'ottima/ (also It. optimum + -a fem. sing., or Lat. optimus, -a, -um]; or the more peripheral member Futurauto < futura [next generation, A] + auto [car], based on the NP auto futura/del futuro [next generation car]. Type ib. Inclusive blends inclusive blends are a feature of written language. They are associative (or portmanteau cf. Algeo 1977) overlap blends in which one constituent includes the other phonologically though not graphemically (Ronneberger-Sibold 2006: 167). Examples from our collection only comprise blends with foreign monosyllabic words, mainly based on exocentric VN compounds, e.g.: EATaly < eat + italy for an Italian slow food market, restaurant and now restaurant brand; or the less transparent Meditathe /medi'tate/, trade mark for Italian literary coffee mugs, where the blend retains stress pattern and stressed syllable of SW1 to facilitate analyzability (in line with point iiib, Section 3.1). Another example is Happyfania, a reduction from AN compound, where phonological similarity replaces phonological identity in the pretonic segment (/eppi'fanja/ vs. /epi'fanja/ (cf. Section 3.1, point ii). Type ii. Contour blends This type comprises blends which retain stress pattern and stressed syllable of one SW (matrix word, cf. Ronneberger-Sibold 2006), which is not entirely present in the blend, while the non-matrix can be inserted in full. Generally speaking, adopting the stress pattern of the matrix word (most often SW2) facilitates recognition. Other factors that contribute transparency are identical size of blend and matrix word, maximization of SWrSW2 overlap, inclusion of the SW in the pretonic part of the blend or in the rhyme. Since overlap at word end is relatively marked, inclusion of material from SW2 in the onset of the stressed syllable might facilitate recognition.6 Some examples: - Blend's size and stress pattern are identical to the size and stress pattern of both SWs; size-based stress assignment: Draquila /'drakwila/ < Dracula /'drakwla/ [Dracul] + (L')Aquila /'lakwila/; morbistenza /morbi'stentea/ < morbidezza /morbi'dettea/ [softness] + resistenza /resi'stentea/ [strength]; Berluscotti /berlu'skotti/ < Berlusconi /berlu' skoni/ + Bertinotti /berti'notti/; Itagnolo /itaq'qolo/ < Italiano /ital'jano/ [Italian] + Spagnolo /spaq'qolo/ [spanish]. - Blend's size is identical to the size of both SWs; position-based stress assignment: gastronauta /gas' tronawta/ < gastronomo /gas'tronomo/ [gourmet]/neoclassical compound gastro- + astronauta /astro'nawta/ [astronaut]/combining form -nauta /'nawta/ (cf. Section 3.1, point iiia). - Left-hand alignment of SWs in blends whose size is identical to the size of one SW; size-based stress assignment: mocaccino /mokat'^ino/ < moca / moka/ + cappuccino /kapput'^ino/ (without overlap); frappuccino /frapput' ^ino/ < frappé /frap'pe*/ + cappuccino /kapput'^ino/ (with overlap).7 - Right-hand alignment of SWs in blends with overlap whose size is identical to the size of one SW; size-based stress assignment: Enterogelmini < Enterogelmina /.entero^el'mi- 6 Monosyllabic blends in Italian are loanwords from other languages, e.g. Eng. smog or brunch, which are not (or are not any longer) analyzed as complex words that combine the inner part of two SWs. Blend size is identical to the size of one SW in complete blends and of both SWs or one of the SWs in contour blends. When blend size differs from both words, the blend retains the stressed syllable of the matrix word, e.g. Biografilm /biogra'film/ < biografia [biography] /biogra'fia/ + film /'film/. 7 Because SWj is monosyllabic and aligns with the stressed syllable of SW2, the distinction between size-based and stress-based assignment is not relevant in Billary /'billari/ < Bill (Clinton) /'bil/ + Hillary (Rodham Clinton) /'illari/ ni/ < Enterogermina /.entero^er'mina/ + Gelmini /^el'mini/; oroschifo /o'roskifo/ < oroscopo /o'roscopo/ [horoscope] + schifo /'skifo/ [disgust].8 - Blends without overlap whose size is identical to the size of one SW; position-based stress assignment: Berlingotti /berlin'gotti/ < Berlinguer /berlin'gwer/ + Andreotti /andre ' otti/. Type 3. Semi-complete blends In semi-complete blends the truncated word precedes a full word which tends to determine the stress pattern of the blend, and follows a complete word in less transparent blends. Graphemic and phonemic overlap of splinter and SW, length and currency of the former, and stress pattern of the full word, facilitate recognition. Some examples: docuromanzo /dokuro 'man&o/ < documentario /dokumen'tarjo/ [documentary]/docu- + romanzo /ro 'man&o/;9 Edilcasa /.edilekasa/ < edile /' edile/, /e'dile/ [building trade, A]/edilizia /edi' litteja/ [building trade]/-edil- + casa /'kasa/; Devitalia /devi'talia/ < dev-(from Eng. development) + Italia /i'talia/; britaliano /brita'ljano/ < British / ' britif*/ (GRADIT: British) + italiano /ita'ljano/ [Italian]; Brangelina /bran^e'lina/ < Brad (Pitt) + Angelina (Jolie), a loanword. CLEANAP/klin ' ap/ can be seen as a less transparent blend which combines full word and splinter, if analyzed (with a bit of an effort) by the average Italian as clean / ' klin/ + Napoli /na'poli/ [Naples], as against CLEANAP /klin' ap/