Lexical and syntactic gemination in Italian consonants -- Does a geminate Italian consonant consist of a repeated or a strengthened consonant?
Maria Gabriella Di Benedetto, Stefanie Shattuck-Hufnagel, Luca De Nardis, Sara Budoni, Javier Arango, Ian Chan, Alec DeCaprio
LLexical and syntactic gemination in Italianconsonants - Does a geminate Italian consonantconsist of a repeated or a strengthened consonant?
M.-G. Di Benedetto, Stefanie Shattuck-Hufnagel, Luca De Nardis, Sara Budoni, Javier Arango, Ian Chan, and Alec DeCaprio Massachusetts Institute of Technology (MIT), Cambridge, MA, United States,DIET Department, Sapienza University of Rome, Rome, Italy Massachusetts Institute of Technology (MIT), Cambridge, MA, United States DIET Department, Sapienza University of Rome, Rome, Italy Radcliffe Institute for Advanced Study, Harvard University, Cambridge, MA, United States (Dated: 8 February 2021)
Two types of consonant gemination characterize Italian: lexical and syntactic. Italian lexicalgemination is contrastive, so that two words may differ by only one geminated consonant. Incontrast, syntactic gemination occurs across word boundaries, and affects the initial conso-nant of a word in specific contexts, such as the presence of a monosyllabic morpheme beforethe word. This study investigates the acoustic correlates of Italian lexical and syntactic gem-ination, asking if the correlates for the two types are similar in the case of stop consonants.Results confirmed previous studies showing that duration is a prominent gemination cue,with a lengthened consonant closure and a shortened pre-consonant vowel for both types.Results also revealed the presence, in about 10-12% of instances, of a double stop-releaseburst, providing strong support for the biphonematic nature of Italian geminated stop con-sonants. Moreover, the timing of these bursts suggests a different planning process for lexicalvs. syntactic geminates. The second burst, when present, is accommodated within the closureinterval in syntactic geminates, while lexical geminates are lengthened by the extra burst.This suggests that syntactic gemination occurs during a post-lexical phase of productionplanning, after timing has already been established. © [https://doi.org(DOI number)][XYZ] Pages: 1–12 I. INTRODUCTION
Consonant gemination is the process by which aconsonant is produced, as the word “gemination” hints,as “doubled”, that is, as two consecutive occurrences ofthe same phoneme, or, under a different interpretation,as a stronger, longer, or intense, consonant. Geminateconsonants are present in several languages such asItalian, Japanese, Arabic, Russian, and Persian, to cite afew. In these languages, gemination is contrastive, thatis, the lexicon of these languages includes minimal wordpairs in which the meaning of one word is distinguishedfrom its minimal pair counterpart on the sole basis ofconsonant gemination; in Italian this contrast is verywidely used and numerous minimal pairs are present inthe lexicon, as, for instance, pala vs. palla (shovel vs.ball) or pena vs. penna (pain vs. pen).In Italian - see the examples above - when a gemi-nate consonant appears within a word, it is usually or-thographically transcribed as two consecutive graphemesof the same consonant. This is the case in Italian formost consonants: stop consonants as well as a subsetof nasals and fricatives. As a matter of fact, most Ital- ian consonants can be geminated in intervocalic position,with the exception of a few such as /z/, although differ-ent experts of Italian phonology hold contrasting viewsregarding a particular subset of five consonants /ts, dz , S , ñ , ń / ((Porru, 1939) vs. (Muljacic, 1972)). Through-out this study, we will characterize gemination propertiesin agreement with what is proposed by Muljacic (1972),and, in particular, we will assume that all Italian conso-nants except /z/ can be geminated, although the abovefive consonants do have a special status, in that these par-ticular consonants are always geminated in intervocalicposition and there exist no minimal pairs based on thecontrastive gemination effect. For these five consonants,the orthographic transcription makes use of the presenceof either one or two graphemes, as in the words azione (/ats’tsjone/) (action) vs. polizza (/’politstsa/) (policy),for instance, although /ts/ is acoustically geminated inboth words. Table I shows a list of the Italian consonantsand of their geminate counterparts, and summarizes thespecific properties of the different consonants.Although gemination is found in many languages,in Italian it has a peculiar property that distinguishes itfrom many others. In Italian, in fact, gemination mayreflect a kind of assimilation across word boundaries J. Acoust. Soc. Am. / 8 February 2021 1 a r X i v : . [ ee ss . A S ] J a n ABLE I. List of Italian phonemes and their gemination behavior. For each phoneme an example of a word containing it, theIPA phonemic and geminate transcriptions, and typical properties of occurrence are given.
Grapheme Example ofWord IPA PhonemicTranscription IPA PhonemicGeminateTranscription Occurrencen nonna /n/ /nn/ single and geminated form intervocalicallyr ragazzi /r/ /rr/ single and geminated form intervocalicallyt teoria /t/ /tt/ single and geminated form intervocalicallyd digitale /d/ /dd/ single and geminated form intervocalicallyl lavoro /l/ /ll/ single and geminated form intervocalicallys sorelle /s/ /ss/ single and geminated form intervocalically(c) cugino /k/ /kk/ single and geminated form intervocalicallyp parole /p/ /pp/ single and geminated form intervocalicallym mattino /m/ /mm/ single and geminated form intervocalicallyv vacanza /v/ /vv/ single and geminated form intervocalically(i) piedi /j/ N/A never in geminated form(u) questa /w/ N/A never in geminated form(ci, ce) citt`a / Ù / / ÙÙ / single and geminated form intervocalicallyf fiamme /f/ /ff/ single and geminated form intervocalicallyg gatto /g/ /gg/ single and geminated form intervocalicallyb bambino /b/ /bb/ single and geminated form intervocalicallygi giardino / Ã / / ÃÃ / single and geminated form intervocalically(z) zitto / ţ / / ţţ / only in geminated form intervocalicallygl figlio / ń / / ńń / only in geminated form intervocalically(sci) scienzato / S / / SS / only in geminated form intervocalically(z) zoo / dz / / dzdz / only in geminated form intervocalically(s) svetta /z/ N/A never in geminated formgn gnomi / ñ / / ññ / only in geminated form intervocalically in particular circumstances, giving rise to the so-calledsyntactic gemination effect (in Italian RaddoppiamentoSintattico , RS). This phenomenon is widely used inItalian compared to the very few other languages thatshow a similar effect, such as Finnish and in some wayMaltese for Italian and Sicilian imported words. In RS,the initial consonant of a word, that in standard Italianis always a single consonant, becomes geminated whenthat word is preceded by a monosyllabic morpheme,for example a function word, or if the preceding wordhas its lexical accent on the last syllable. For example,in the group of words a piedi (by foot), the initialconsonant of piedi /p/ becomes geminated, so that thephonemic transcription of the post-lexical word groupis /ap’pj E di/. Although it is not within the scope ofthis paper to describe in detail all the specific cases inwhich syntactic gemination may take place in Italian(for a comprehensive analysis see (Camilli, 1965), pp.133-154), but rather to introduce the phenomenonin order to include it in the study, it is interestingto note that syntactic gemination in Italian, and toour knowledge in Italian only, can be contrastive withrespect of pairs of word groups; To better explain thiseffect an interesting example is the case of the group ofwords tra monti (among mountains) in which /m/ isgeminated and the post-lexical word group is transcribedas /tram’monti/ vs. the word tramonti (sunsets) that istranscribed as /tra’monti/. The gemination of the /m/consonant distinguishes a post-lexical word group froma word of the lexicon.To sum up, Italian is characterized by two types of gem-ination: lexical and syntactic. In standard Italian lexicalgeminate consonants only occur within words (that is, never in initial position), while initial consonantsof words may become geminated due to the syntacticgemination effect. It should be noted that in somedialects of southern and central Italy consonants may begeminated also in word initial position, independentlyof syntactic gemination ((Bertinetto and Loporcaro,1999); (Bonucci, 2011))), but in this case the effect isnot contrastive and seems to be more of a phenomenoninvolving junction and adjustment between consonantsoccurring at word boundaries. In this instantiation,gemination seems to resemble the typical phenomenonof liaison in French, which is known to occur more oftenwith words that co-occur frequently but does not prove,when realized, to favor lexical recognition of linkedwords (Fougeron et al. , 2001). This result leads to aninterpretation by which non-contrastive germination,like liaison , may be considered as an expressive sandhiphenomenon.As mentioned above, consonant gemination isusually described either as the doubling of a consonant,or as the production of a single consonant as strongeror more intense and typically characterized by longerduration. Whether a geminate consonant is representedin the mind of the speaker as a single longer or strongerconsonant vs. a double consonant has been debatedfor several decades (Swadesh, 1937), and, at least forthe Italian language, is still under discussion. Thesetwo opposite views, by which a geminate consonant isinterpreted as either one /C:/ or two /CC/ phonemes,may lead to two different ways of considering thesyllabic structure of /CC/, either heterosyllabic /C.C/or homosyllabic /C:/. These two different views have ossible relevant consequences for the extent of coar-ticulation effects between syllables. It is interestingto note that Latin, from which Italian derives and towhich is the closest one among the Romance languages,had contrastive lexical consonant gemination, as wellas expressive consonant gemination, and that there isan almost unanimous consensus that in Latin geminateconsonants were actually two consonants (Giannini andMarotta, 1989). We will come back to this issue later inthe paper.The search for acoustic correlates of geminationin Italian, and the verification of their perceptualrelevance, has been the object of a longstanding project,the Gemination project GEMMA (Di Benedetto, 2000),(Di Benedetto, 2019). This project began at SapienzaUniversity of Rome in 1992, with the goal of analyz-ing gemination in Italian consonants, based on theanalysis of VCV vs. VCCV words. Results for stops(Esposito and Di Benedetto, 1999), nasals and liquids(Di Benedetto and De Nardis, 2020b), and fricativesand affricates (Di Benedetto and De Nardis, 2020a),showed a general tendency to shorten the pre-consonantvowel and to lengthen the word-medial consonant in ageminate word. No significant effects of gemination wereobserved on other acoustic parameters, such as energy-and frequency-related measurements. The generalconclusion was that consonant duration is a primarycue to gemination, and pre-consonant vowel duration asecondary cue.This article addresses the problems of characterizinglexical and syntactic gemination in Italian in terms ofacoustic manifestations and acoustic correlates, andof understanding whether these correlates are similarbetween lexical and syntactic geminates. The analysiswas carried out on spoken sentences in which both lex-ical and syntactic gemination occurred. The questionswere: a) Do geminates in running speech manifest,as observed in previous studies, mostly by varyingtime-based parameters? One goal was thus to determinewhether it was possible to disentangle the questionabout the biphonematic vs. monophonematic nature ofItalian geminate consonants, by observing the acousticmanifestation of gemination in running speech; b) Dolexical and syntactic geminates organize the temporaldistribution among segments in a similar way?; c) Howdo the findings impact cluster syllabification? In otherwords, does germination sometimes result in a longersingle consonant but in other cases in a doubled conso-nant? And if so, when the geminated consonant consistsof two consecutive consonants C (1) C (2) rather than onestronger and longer consonant, do C (1) and C (2) differacoustically, and if so, is C (2) stronger and more stablethan C (1) ? A positive answer may lead to the hypothesisthat the sequence C (1) C (2) is heterosyllabic, C (1) beinga coda consonant and C (2) an onset consonant. Theanswer to this specific question may lead to improved understanding of the production planning process.The paper is organized as follows. Section II containsthe description of the database and of the experiment.Section III contains the results of the acoustic analysis.Section IV contains a general discussion of results andthe proposed interpretation, as well as the conclusions. II. EXPERIMENTATION: GEMINATION IN SPOKEN SEN-TENCES
As mentioned in the Introduction, previous studies ofconsonant gemination in Italian VCV and VCCV words,showed that the contrast between singleton vs. gemi-nate consonants was durational in nature. In particular,these studies indicated that the duration of the conso-nant and the duration of the pre-consonant vowel are thetwo parameters that are significantly different for the twoconsonant categories. An example of the impact of gem-ination on the parameters is presented in Figure 1 wherethe waveforms associated with the words fato vs. fatto aspronounced in running speech in sentences
Il fato ancora vs.
Il fatto ancora .Based on these previous studies, acoustic analysiscarried out in this experiment aimed at measuring thesespecific parameters for both lexical and syntactic gemi-nates.In the experiment design, stop consonants were selectedas the consonants to be measured, following an approachthat was also adopted in the GEMMA project, sincestops are the most frequent geminated consonants in Ital-ian and are also the most informative; Not only are theseconsonants easier to measure with fine detail thanks toa clear presence of closure and release phases, but alsoin stops there would be a possibility that if the biphone-matic hypothesis were true one could eventually find ev-idence of two bursts, two closures and two releases. Thespeech material on which the analysis is carried out con-sists of 100 Italian spoken sentences forming the LaMITdatabase (Di Benedetto et al. , 2020). The set of sentencesis reported in Table II.The sentences were designed to include all thephonemes of the Italian language and the geminateversions of the consonants (except for /z/, since thisconsonant is not geminated, as mentioned in the Intro-duction). Take for instance the first sentence of TableII, in which both lexical and syntactic geminates occur.Highlighting lexical geminates in cyan vs. syntacticgeminates in yellow, the first sentence
Il gatto dellavicina `e bianco peloso e pazzo is transcribed as follows:/ ' il ' gatto ' della vi ' Ù ina ' E b ' bjanko pe ' loso ' ep ' pa ţţ o /The LaMIT database was designed so to reflect thetypical frequency of occurrence of the different phonemesin the Italian language, as suggested by a recent study(Arango et al. , 2020a), (Arango et al. , 2020b) thatprovides updated values of the phonemic frequencies, J. Acoust. Soc. Am. / 8 February 2021 3
20 40 60 80 100 120 140 160 180 200
Time [ms] -0.4-0.3-0.2-0.100.10.20.30.40 20 40 60 80 100 120 140 160 180 200
Time [ms] -0.4-0.3-0.2-0.100.10.20.30.4 /t//tt//a//a//f//f/ /o/ /o/
FIG. 1. Singleton stop /t/ vs. geminated stop /tt/ in words fato vs. fatto as pronounced in running speech in sentences
Il fato ancora vs.
Il fatto ancora , highlighting the lengthening of consonant and shortening of preceding vowel associated togemination, as described in (Esposito and Di Benedetto, 1999). assuming the existence of 30 phonemes as identified byMuljacic (1972), that is: 1) seven vowels, /a/, /i/, /u/,/e/, / E /, /o/, / O /; 2) twenty-one consonants: /p/, /b/,/f/, /v/, /t/, /d/, / ţ /, / dz /, /s/, /z/, /k/, /g/, / Ã /,/ Ù /, / S /, /m/, /n/, / ñ /, /l/, / ń /, /r/; 3) two glides: /j/and /w/. Allophones are excluded, consistent with thetheoretical framework provided by Muljacic (1972).Speech materials were recorded in the Speech Lab-oratory of the DIET Department at the University ofRome ’La Sapienza’, Rome, Italy, using professional equipment, in a sound-treated room under the super-vision of an acoustically trained person. The speakerswere Italian native speakers, raised and living in Rome(Italy), pronunciation defectless and free of evident di-alectal inflexions. As suggested in (Payne, 2006), theRoman accent, although quite distinctive, is phonolog-ically very close to Standard Italian. The entire set ofsentences was recorded twice in two different recordingsessions, leading to two repetitions for each sentence andfor each speaker. In case of evident mispronunciations,the speaker was asked to repeat the sentence. ABLE II. Sentences of the LaMIT database.
1. Il gatto della vicina `e bianco peloso e pazzo 51. Sono belli i programmi decisi all’ultimo momento2. Il giardino di mio cugino `e pieno di gladioli e di gnomi 52. Con Cristiana pratico yoga ogni mercoled`ı3. L’universit`a italiana `e un’istituzione pubblica dello stato 53. Pensieri e parole cantava la diva con voce suadente4. Passeggerei volentieri a piedi nudi nella citt`a vecchia 54. Aprile si esaurisce mentre arriva carico di promesse il mese dimaggio5. Pietro non scappa fugge a gambe levate con il cuore in fiamme 55. Abbiamo trasmesso il giornale radio del mattino6. Cosa ne penseresti di alzarti presto e salutare il sole 56. Il tempo previsto sull’Italia per questa sera non prevedetemperature in aumento7. Quando Maria `e in vacanza compra volentieri la settimanaenigmistica 57. Pensavo che tu volessi fare solo uno spuntino8. Lo schermo del tuo cellulare `e graffiato e opacizzato 58. Assicurati che non si dimentichino di scrivere alla zia9. All’imbrunire la cattedrale svetta nel cielo basso e uggioso 59. Per salvarci dobbiamo restare uniti10. Alcuni studenti dell’anno accademico corrente potrannolaurearsi a luglio 60. Il mondo `e nelle nostre mani11. Due sorelle si aiutano se vanno d’amore e d’accordo 61. Comportati educatamente a tavola12. La struttura precaria resse malgrado il forte vento 62. Pare che sia rimasto solo per un colpo di testa13. Mandare cartoline da citt`a remote non `e pi`u di moda 63. La piccola peste vuole il ciuccio per calmarsi14. Discendi il Monte Bianco con gli sci e vivi un’esperienza unicae indimenticabile 64. Non mordere la spalla della nonna15. Prima o poi dovrai pur deciderti a leggere le opere di Niccol`oMachiavelli 65. La carta non si mangia se non sei una capra16. Non potendo fare a meno del cioccolato pens`o bene di privarsidella panna montata 66. Il pavone becca le foglie sul viale dello zoo17. Che avventura meravigliosa quella di guardare gattonare unbeb`e 67. All’improvviso si ud`ı l’urlo del barbagianni18. “E pur si muove” disse il famoso scienziato rivolgendosi agliinquisitori 68. Basta con i fanatismi esagerati19. Oggi piove a dirotto governo ladro 69. Non smettere di fantasticare ad occhi aperti20. Riporre tanti sogni nel cassetto rinforza la fantasia del poeta 70. Col vento in poppa attraversarono il Mediterraneo in un soffio21. Vent’anni di allenamento non furono sufficienti a chiudere lapinza 71. Voltati e renditi conto di quanta strada hai percorso22. Senti un po’ di musica e vedi che ti passa la nostalgiadell’inverno 72. Una tazza di te verde al giorno rinfresca la mente23. I clienti della Banca devono attenersi alle regole stabilite dalcontratto 73. Scriver`o questa lettera con la penna a sfera24. Apponi la firma in calce perch´e `e necessario per rendere validala transazione 74. Che ne pensi di una fetta di torta25. I ragazzi della scuola religiosa fisseranno un appuntamento conil sindaco ateo 75. Il cestino per la carta sta sotto la scrivania26. Se prendi in prestito un libro alla biblioteca godi del vantaggiodi non dover acquistarlo 76. Una vacanza in agriturismo in Toscana ha un costo ridotto27. La rappresentazione digitale delle immagini ha rivoluzionato lafotografia 77. Sul pavimento del salone giace un tappeto persiano28. Addio all’imperatore giapponese abdicher`a oggi in favore di suofiglio 78. L’albero di cedro `e simbolo del Libano29. Uno sciame di api invest`ı il bambino biondo costringendolo abuttarsi gi`u dall’albero 79. Un biglietto di auguri accompagna il regalo30. La ferrovia si snoda lungo il fiume seguendo un tracciatotortuoso 80. Torneresti a casa a piedi31. Dopo avere letto molti libri Luca si rimise a studiare ancoraper un po’ 81. Il grano saraceno non contiene glutine32. Se arrivi all’alba a Capri butta l’ancora e prosegui a nuoto 82. Il pane lievita quando la luna `e piena33. Giorgio ha deciso di prendere i voti ma prima ha dovutobattezzarsi 83. Stendi il bucato al sole e risparmi energia34. Che ne farai dei quaderni di storia 84. Sotto la piazza giace un tesoro35. Chiedi pure a tuo padre cosa ne pensa dell’anguria 85. La balena blu nuota in solitario36. Mamma e pap`a ti vogliono bene 86. Servono nuovi dirigenti per rilanciare le aziende37. Non poggiare il bicchiere colmo d’acqua sul pianoforte 87. Creare lavoro `e un dovere costituzionale38. Con la bicicletta elettrica le salite sono una passeggiata 88. Ma questa `e un’altra storia su cui si indagher`a39. Saluta la signora e fai l’inchino 89. Si al regolamento che impone limiti alla stupidit`a40. La teoria dei numeri `e una branca della matematica 90. La ballerina indossa un costume rosa fragola41. Mio nipote ama trovare soluzioni a problemi complessi 91. L’autore si muove con scioltezza nella palude delle parole42. Aguzza l’ingegno e progetta una radio intelligente 92. La fascetta giusta dovrebbe essere alienazione43. Il giornalaio vende e invia riviste e oggetti turistici 93. Resti in collegamento che risponder`a il primo operatore libero44. Il cane corse forsennatamente verso il padrone calpestando leaiuole 94. Vediamo se la risposta `e quella giusta45. Il dolore sorgeva mentre la luna non era ancora tramontata 95. L’avocado cresce nei paesi tropicali46. Puoi accendere la radio a caso e sintonizzarti su qualsiasifrequenza 96. Pesce fritto e insalata mista grazie47. Poi ci sono i rimedi naturali che sono pi`u efficaci di tantiprodotti presenti in farmacia 97. Favorisce un caff`e dopo cena col digestivo48. Impariamo a meditare giornalmente 98. La folla era impazzita alla vista dell’assassino49. Si perde cos`ı tanto tempo a discutere del niente 99. Lei col maglione rosso si stia zitto50. Stasera andremo al cinema a vedere un film francese 100. Mangerebbe volentieri un filetto di baccal`a con le olive
The speech materials recorded by one male speaker (MS)and one female speaker (FS), formed the object of thepresent analysis.
III. ACOUSTIC ANALYSIS
The acoustic analysis was conducted manually by ex-amining the signal and the spectrogram of all sentences,for both repetitions and speakers. Speech signals wereanalyzed using the xkl software, part of the set of soft-ware tools developed by Dennis Klatt ((Klatt, 1984) pp. 73-82). All sentences were also checked by listening toconfirm the presence of gemination.A substantial number of instances of double closures andbursts were observed for both speakers MS and FS, pro-viding evidence for the presence of two consecutive conso-nants C (1) and C (2) . Double bursts were found in about12% of instances of lexical geminates and 10% of syntac-tic geminates. Table III shows the distribution of singlevs. double bursts for lexical vs. syntactic geminationof both speakers. Although a similar number of doublebursts was found for the two speakers, in both syntactic
J. Acoust. Soc. Am. / 8 February 2021 5 nd lexical forms, the observed double bursts occurredin different sentences and for different consonants. TableIV shows where a double burst was present – in whichsentences and which words. A typical example of anobserved double burst is presented in Fig. 2, showingthe waveform and spectrogram of the geminate [t] in theword filetto of sentence n.100 of speaker MS, version 1.As shown in Fig. 2, a first burst appears at time ∼ ∼ ∼ ∼ P burst = 1 N N (cid:88) i =1 x i , (1)where x i is the i -th speech sample amplitude and N is the number of speech samples in the burst. ThreeANOVA univariate tests on the P burst parameter, wherethe fixed factor was first burst vs. second burst werethen carried out, one test for each gemination form,i.e. lexical vs. syntactic, and one test with lexical andsyntactic forms merged. Given the numerosity of theavailable samples (30 cases for lexical and 16 casesfor syntactic) the threshold for significance was set at p ∗ = 0 .
05. Results are presented on Fig. 3, showingthat the second burst was significantly stronger thanthe first, for both forms separately, and also combined.Very similar results were obtained for each speakerindividually.In terms of durational parameters, the two consecu-tive consonants in both lexical and syntactic geminateswith double bursts had similar duration. Although webelieve that durational parameters refer to intervals be-tween acoustic events rather than to duration of segmentsthemselves, we will from now on refer, for the sake of sim-plicity, to durational parameters as segment durations. Aunivariate ANOVA test on consonant duration (durationof closure+duration of burst) with threshold p ∗ = 0 . p = 0 . > p ∗ = 0 . p = 0 . > p ∗ = 0 .
05. Op-positely, burst durations were significantly different inthe two consonants for both forms, with C (2) burst be-ing much longer than C (1) burst: for lexical geminates p < . < p ∗ = 0 .
05 and for syntactic geminates p = 0 . < p ∗ = 0 .
05, suggesting a time compensationbetween burst and closure, so to keep consonant dura-tion constant. A univariate ANOVA test on closure du-ration confirmed this prediction for syntactic geminatesin which closure duration of C (1) was significantly higherthan C (2) ( p = 0 . < p ∗ = 0 .
05) but not for lexicalgeminates ( p = 0 . < p ∗ = 0 .
05) although also in this case we observed a systematic longer closure for the firstconsonant. The above results are summarized in Fig. 4(lexical form in Fig. 4(a) and syntactic form in Fig. 4(b))that shows the average values of consonant duration, clo-sure duration, and burst duration. As shown in the fig-ure, average values were: a) for lexical geminates, C (1) duration = 61.41 ms, C (1) closure duration = 48.91 ms,C (1) burst duration = 12.49 ms, C (2) duration = 67.66ms, C (2) closure duration = 38.97 ms, C (2) burst dura-tion = 28.7 ms; b) for syntactic geminates, C (1) duration= 46.21 ms, C (1) closure duration = 35.29 ms, C (1) burstduration = 10.92 ms, C (2) duration = 50.94 ms, C (2) clo-sure duration = 22.38 ms, C (2) burst duration= 28.56ms. Very similar results were obtained for each speakerindividually.In summary, evidence for the presence of two consecu-tive consonants C (1) and C (2) was found for both gemina-tion forms. In both forms, C (1) burst power and durationwere significantly weaker (lower power and shorter dura-tion) than in C (2) bursts. C (1) and C (2) durations weresimilar for both forms. Closure duration was significantlylonger in C (1) than in C (2) in syntactic geminates, but notin lexical geminates, although the observed tendency fol-lowed that same trend. The two consonants seemed thussimilar overall in terms of duration but inter-event timingwithin and power measurements support the hypothesisthat the second consonant is stronger than the first.A related question of interest is whether double burstconsonants behave similarly to single burst consonants,in terms of durational parameters. A first parameter ofcomparison was burst duration. A different burst dura-tion was observed above for C (1) vs. C (2) , but the addi-tional question was whether, on average, burst durationwas similar in double vs. single burst consonants. Theanalysis was therefore based on computing the averageof the duration of the bursts of C (1) and C (2) and com-paring this average to the duration of the single burst ofsingle burst consonants. Results of a univariate ANOVAtest with fixed factor being single burst vs. double burstconsonant showed no significant difference between thetwo groups ( p = 0 . > p ∗ = 0 . (1) and C (2) was similar tothe duration of a single burst in a single burst consonant.The average burst duration for double burst consonantswas 20.3 ms and for single burst of single burst conso-nants 23.3 ms. Also in terms of consonant duration andclosure duration, results of statistical analyses indicateda similar trend, where closure duration for double burstconsonants was computed as the sum of the two closures.In particular, two univariate ANOVA tests on these twoparameters with fixed factor single vs. double burst con-sonant indicated no significant difference between the twogroups: a) for consonant duration p = 0 . > p ∗ = 0 . p = 0 . > p ∗ = 0 .
01. Averagedurations for single burst consonants were: consonantduration = 110.04 ms and closure duration = 86.71 ms,while for double burst consonants: consonant duration= 118.43 ms and closure duration = 77.81 ms. However,we note that this last result blurs a difference in behav-
ABLE III. Number of single burst vs. double burst geminates in both lexical and syntactic forms and for each speaker.
Lexical gemination Syntactic geminationsingle burst double burst total single burst double burst totalspeaker MS 105 15 120 69 7 76speaker FS 105 15 120 68 8 76Total 210 30 240 137 15 152
TABLE IV. Inventory of words and word groups, and corresponding LaMIT database sentence number and version, in whichdouble burst lexical and syntactic geminates were found.
Lexical geminationSpeaker FS Speaker MSsentence number version word sentence number version word1 1 ga tt o 4 1 ve cc hia9 1 sve tt a 12 1 stru tt ura13 1 ci tt `a 15 1 Ni cc ol`o15 1 Ni cc ol`o 23 1 contra tt o20 1 casse tt o 38 1 ele tt rica63 1 pi cc ola 63 1 pi cc ola100 1 file tt o 69 1 sme tt ere3 2 pu bb lica 69 1 o cc hi4 2 ci tt `a 100 1 file tt o5 2 sca pp a 4 2 ve cc hia7 2 se tt imana 9 2 sve tt a19 2 diro tt o 15 2 Ni cc ol`o69 2 sme tt ere 19 2 diro tt o92 2 fasce tt a 32 2 bu tt a99 2 zi tt o 27 2 bi cc hiereSyntactic geminationSpeaker FS Speaker MSsentence number version word sentence number version word19 1 a d irotto 5 1 a g ambe21 1 a c hiudere 8 1 `e g raffiato33 1 ha d eciso 21 1 a c hiudere35 1 a t uo 33 1 ha d eciso80 1 a c asa 94 1 `e q uella21 2 a c hiudere 33 2 ha d eciso22 2 po’ d i 94 2 `e q uella32 2 a c apri - - - ior that was highlighted when lexical vs. syntactic gemi-nates were considered separately. While the observationson the burst holds for both gemination forms, the same isnot true for consonant duration and closure duration. Inparticular, consonant duration increased significantly inlexical geminates when a double burst was present, butthis was not the case in syntactic geminates. In termsof closure duration, the opposite was observed; closureduration was not significantly changed by the presenceof a double burst in lexical geminates, while it decreasedsignificantly in syntactic geminates when a double burstoccurred. The univariate ANOVA test for consonant du-ration with fixed factor single vs. double burst conso-nant provided the following quantitative data: a) for lex-ical gemination p = 0 . < p ∗ = 0 .
01; b) for syntacticgemination p = 0 . > p ∗ = 0 .
01. Average durationswere: a) for lexical single burst geminated consonants= 114.77 ms and double burst geminated consonants=129.07 ms; b) for syntactic single burst geminated conso-nants = 102.8 ms and double burst geminated consonants= 97.15 ms. ANOVA univariate tests for closure durationwith fixed factor single vs. double burst consonant pro-vided the following quantitative data: a) for lexical gem- ination p = 0 . > p ∗ = 0 .
01; b) for syntactic gemination p < . < p ∗ = 0 .
01. Average durations were: a) forlexical single burst geminated consonants = 89.09 ms anddouble burst geminated consonants = 87.88 ms; b) forsyntactic single burst geminated consonants = 83.04 msand double burst geminated consonants = 57.67 ms. Tosummarize, the average duration values are reported inTable V. In summary, single vs. double burst consonantswere not significantly different in terms of burst duration,as well as consonant and closure durations. When lexi-cal and syntactic geminates were considered separately,however, the analysis showed a different timing organiza-tion. In particular, in lexical geminates, closure durationwas stable in average, but consonant duration increasedfor double burst consonants, i.e., the addition of a sec-ond burst had the effect of increasing consonant duration,since closure duration was stable. In syntactic geminates,the organizational pattern was different; consonant dura-tion was stable, while closure duration was significantlydecreased in double burst instances, suggesting that thesecond burst was inserted at the expense of closure du-ration. Figure 5 summarizes the above comments andpresents an overall comparison of the durations of con-
J. Acoust. Soc. Am. / 8 February 2021 7 a)(b)
FIG. 2. Example of double burst: the geminate [t] in the word filetto of sentence n.100 of speaker MS, version 1. Waveform(Fig. 2(a)) and spectrogram (Fig. 2(b)) show a first burst at time ∼ ∼ ∼ ∼ sonant and of the preceding vowel, that highlights thedifference between lexical and syntactic gemination or-ganizational pattern discussed above.The above analysis highlighted the possibility of a dif-ferent time planning in lexical vs. syntactic gemination,when accommodating for a second burst. But was thisthe only behavioral difference of these two geminationforms? To come back to the initial question, were lexicalgeminates also acoustically different from syntactic gem- inates in other respects?As mentioned in the Introduction, in previous studies,pre-consonant vowel duration was shown to be a relevantparameter in the acoustic manifestation of geminationof Italian consonants in general and in stops in partic-ular (Esposito and Di Benedetto, 1999). Those studiesshowed that pre-consonant vowel duration was shortenedwhile consonant duration was lengthened in geminate vs.singleton instances, and moreover that the ratio between exical Syntactic Total00.511.5 P bu r s t [ W ] -4 Burst 1Burst 2 p<0.001 p<0.001p=0.008
FIG. 3. Power of the first vs. the second burst, for lexical and syntactic groups, and for both groups combined indicated onfigure as Total. The power of the burst was computed as the energy of the burst divided by the number of samples composingit. Values of p in bold indicate that p < p ∗ = 0 .
05, i.e., a statistical significance of the parameter.TABLE V. Average duration in ms of time parameters for geminate stop consonants, divided by gemination type and singlevs. double burst.
Gemination Vd Cd C1d C2d Cld Cl1d Cl2d Bd B1d B2dLexical SB 70.9 114.8 - - 89.1 - - 25.7 - -DB 78.8 129.1 61.4 67.7 87.9 48.9 39.0 41.2 12.5 28.7Total 71.9 116.6 - - 88.9 - - 27.6 - -Syntactic SB 56.3 102.8 - - 83.0 - - 19.8 - -DB 59.7 97.2 46.2 50.9 57.7 35.3 22.4 39.5 10.9 28.6Total 56.6 102.2 - - 80.5 - - 21.7 - - consonant durations and pre-consonant vowel durationswas a good predictor of the presence of gemination; gem-inates stops were typically characterized by a ratio ofabout 1.64 in VCCV words (Di Benedetto and De Nardis,2020a). The ratio is usually much smaller for singletonconsonants; in singleton VCV stops it is about 0.62. Inthe present study, the analysis of the ratio between con-sonant duration and vowel duration was extended by con- sidering syntactic gemination, that can only be presentin running speech, and by evaluating it both in combi-nation with, and in comparison to, lexical gemination.To this aim, durational parameters, consonant durationto vowel duration and closure duration to vowel durationratios were also measured for singleton stops; results arepresented in Table VI, together with data for geminatesaveraged over gemination type.
J. Acoust. Soc. Am. / 8 February 2021 9 exical Gemination DB (Cons. 1 vs. Cons. 2)
Consonant Closure Burst010203040506070 D u r a t i on [ m s ] First consonantSecond consonantp=0.378 p=0.171 p<0.001 (a)
Syntactic Gemination DB (Cons. 1 vs. Cons. 2)
Consonant Closure Burst0102030405060 D u r a t i on [ m s ] First consonantSecond consonantp=0.573 p=0.041 p=0.001 (b)
FIG. 4. Durations of consonant, closure, and burst, of first (blue) and second (red) consonants, in lexical geminates (Fig. 4(a))and syntactic geminates (Fig. 4(b)). p-values on figure indicate significance level obtained with ANOVA tests. Bold valuesindicate significant difference, where the significance threshold was set at p ∗ = 0 . The comparison between the ratios for singleton stopsand for geminates confirms the results reported in theprevious works for VCV vs. VCCV words. The pres-ence of gemination leads to a shorter vowel and a longerconsonant, and thus to a ratio that increases from 0.75in singletons to 1.84 in geminates; a threshold set to 1for the ratio between consonant duration and vowel du-ration is therefore confirmed as a reliable discriminantfor the presence of gemination. Moving to the com-parison between gemination types, we found that pre-consonant vowel and consonant durations were both sig-nificantly different in lexical vs. syntactic gemination.Univariate ANOVA tests with lexical vs. syntactic gem-ination as a fixed factor indicated a significantly longervowel in the lexical case ( p < . < p ∗ = 0 .
01) andalso a significantly longer consonant for the lexical set( p < . < p ∗ = 0 . p = 0 . < p ∗ = 0 .
01. This result highlights an ad-ditional difference in the acoustic manifestation of lexi-cal vs. syntactic gemination. Was the ratio, however,stable among geminates of the two forms? Interestinglyenough, results of statistical analyses showed that withineach gemination form, the ratio was stable when con-sidering single vs. double bursts consonants. In lexicalgemination, the univariate ANOVA test with fixed vari-able single vs. double burst indicated a non-significantdifference in the ratio ( p = 0 . > p ∗ = 0 . p = 0 . > p ∗ = 0 .
01, average ratio for single burst con-sonants = 1.99 and average ratio for double burst con-sonants = 1.73. The ratio between pre-consonant voweland consonant durations proved therefore to be a crucialindicator of gemination in both gemination forms, witha higher ratio observed in syntactic gemination.
IV. DISCUSSION OF RESULTS AND CONCLUSIONS
Results of acoustic analyses showed that the acousticmanifestation of geminate consonants includes the pres-ence, in some instances, of two consecutive consonantsC (1) and C (2) . This was observed in both lexical and syn-tactic gemination. This finding provides strong supportfor a biphonematic nature of Italian geminated stop con-sonants and an answer to our first question: the acousticmanifestation of gemination in running speech was foundto be not only related to durational parameters but wascomplemented by the discovery, in the acoustic signal,of double bursts and double closures, explicitly signalingthe presence of a geminate, i.e. we may say a doubleconsonant.Furthermore, it was shown that the two consonants C (1) and C (2) did not differ in total duration but they typicallydid so in terms of the power and duration of the burst,as well as in closure duration, with a compensation effectbetween closure and burst durations. Since the first con-sonant C (1) is characterized by a weaker burst (weakerpower and shorter duration) than the second C (2) it isplausible to say that the first consonant is less strongthan the second. This observation supports the hypoth-esis that C (1) is a coda consonant and C (2) is an onset
10 J. Acoust. Soc. Am. / 8 February 2021 exical, SBLexical, DBSyntactic, SBSyntactic, DB Cl1d Cl2d B2dVd B1dVd Cld Bd Vd Cld BdVd Cl1d B2dCl2dB1d
FIG. 5. Overall comparison of durations for consonant and preceding vowel in Lexical vs. Syntactic geminated stops, dividedin Single Burst (SB) vs. Double Burst (DB). All values are expressed in ms.TABLE VI. Average values of time-related parameters for singleton stop consonants vs. geminate ones in the LaMIT databasefor the two speakers considered in this work. Durations are expressed in ms.
Vd Cd Cld Bd Cld/Vd Cd/VdSingleton 85.07 55.48 35.9 19.58 0.48 0.75Geminate 65.98 111.01 85.69 25.32 1.42 1.84 consonant, and therefore that C (1) C (2) form an heterosyl-labic sequence. This finding answers our third question,on cluster syllabification. This observation also providesadditional evidence that relates to the planning of tim-ing of the production process. In syntactic gemination,the presence of the second burst only impacted closureduration; it did not influence consonant duration. Thatis, the extra burst was accommodated in the closure timeinterval. But this was not the case in lexical gemination,for which closure duration was kept stable when the ex-tra burst was included, by simply making the consonantlonger. This finding answered our second question, andindicated that the two gemination forms, lexical vs. syn-tactic, may arise at two different points during the pro-duction planning process. In syntactic gemination in par-ticular the phenomenon must arise after words have al-ready been planned, since the phenomenon occurs acrosswords; this may explain why syntactic gemination maynot alter the duration of the onset consonant, which hasalready been specified. In contrast, lexical geminationhappens within a word, and, as such, timing elements in the word may still find room for adjustments.Finally, the ratio between consonant and pre-consonantvowel durations was analyzed since this ratio was shownin previous studies focusing on VCV vs. VCCV words tobe a good indicator for the presence of gemination. Thisresult was confirmed for the running speech material pro-vided in the LaMIT database: a ratio above 1 typicallyindicates the presence of gemination.In the present study, results of the analysis showed thatthe above ratio was stable across single burst vs. dou-ble burst groups for both lexical and syntactic geminates,with ratio values higher than in VCV words, i.e. in thedirection of reinforced gemination, manifested by mul-tiple cues that may reveal the presence of the secondconsonant; This concept may recall the concept of en-hancement and leads to a model by which the geminatedconsonant is made of two consonants and this appears inthe acoustic signal as a longer consonant (because theyare two consonants), an additional durational cue beingshortening of the pre-consonant vowel (because the signalcontaining that vowel is a closed syllable) and the pres- J. Acoust. Soc. Am. / 8 February 2021 11 nce of the burst. This finding indicates that the ratio is astable parameter and that the insertion of a second burstdoes not alter the rhythmic structure of a word (lexicalgemination) or the rhythmic structure across words (syn-tactic gemination). The finding that an additional burstis not always visible in geminated consonants paves theway to an interesting research question. Is a missingextra burst the result of occasional articulatory failurein introducing an extra cue reflecting the intention ofthe speaker? The fact that an extra burst is not alwaysvisible in the repetitions of a same word seems to sup-port this interpretation. On the other hand, some stopsseem to be produced with this additional cue more oftenthan others when geminated (/k/ vs. /d/), as if somearticulatory gestures are more prone to reduction. Fu-ture research will also focus on this aspect, together withgathering articulatory data.The analysis also showed an additional difference be-tween the two gemination forms: the ratio was signifi-cantly different for lexical vs. syntactic gemination and inparticular it was higher for syntactic gemination, mainlydue to a shorter pre-consonant vowel duration.The above difference may be due to the different natureof the two geminations, lexical gemination being alwayscontrastive while syntactic gemination is almost alwaysnot. Is the expressive nature of syntactic gemination thereason for a shorter pre-consonant vowel leading to a re-inforced ratio? The interpretation of this finding mayrequire additional experiments focused on this specificaspect, possibly addressing those rare but existing caseswhere in Italian, as mentioned in the introduction, syn-tactic gemination becomes contrastive.Beyond addressing the natural extension of the analy-sis of lexical vs. syntactic gemination in other conso-nant classes, future work will further focus on the effectof the proposed heterosyllabic structure /C (1) . C (2) / oncoarticulation, and analyze in particular whether the ex-tent of coarticulation between different speech segmentsis different across geminate vs. singleton consonants, assuggested by previous studies that address the effect ofconsonant cluster syllabification on vowel-to-vowel coar-ticulation (Mok, 2012) . ACKNOWLEDGMENTS
This work was supported in part by the Radcliffe In-stitute for Advanced Study at Harvard University and bySapienza University of Rome within the research project“Towards Speech Recognition of the Italian LanguageBased on Detection of Landmarks and Other AcousticCues to Features”, grant
Arango, J., DeCaprio, A., Baik, S., De Nardis, L., Shattuck-Hufnagel, S., and Di Benedetto, M.-G. ( a). “Estimation of the Frequency of Occurrence of Italian Phonemes in Text,”Speech Communication (submitted) .Arango, J., Yao, S., DeCaprio, A., Baik, S., Shattuck-Hufnagel,S., and Di Benedetto, M.-G. ( b). “Estimation of the Fre-quency of Occurrence of Italian Phonemes,” in , Chicago, Illinois, USA.Bertinetto, P., and Loporcaro, M. ( ). “Geminate distintive inposizione iniziale: uno studio percettivo sul dialetto di Altamura(Bari),” Annali della Scuola Normale Superiore di Pisa VI (2),305–322.Bonucci, P. ( ). Consonanti geminate iniziali in Perugino , 9–38 (Olschky).Camilli, A. ( ). Pronuncia e grafia dell’italiano (Sansoni Edi-tore, Florence, Italy).Di Benedetto, M.-G. ( ). “Gemination in Italian: the GEMMAproject,” The Journal of the Acoustical Society of America (5), 2507–2507.Di Benedetto, M.-G. ( ). “GEMMA project webpage” http://acts.ing.uniroma1.it/speech_research_projects.php .Di Benedetto, M.-G., Choi, J.-Y., Shattuck-Hufnagel, S.,De Nardis, L., Budoni, S., Vivaldi, J., Arango, J., DeCaprio,A., and Yao, S. ( ). “Speech recognition of spoken Italianbased on detection of landmarks and other acoustic cues to dis-tinctive features,” in , Chicago, Illinois, USA.Di Benedetto, M.-G., and De Nardis, L. ( a). “Gemination inItalian: the affricate and fricative case,” Speech Communication(under major revision) .Di Benedetto, M.-G., and De Nardis, L. ( b). “Geminationin Italian: the nasal and liquid case,” Speech Communication(under major revision) .Esposito, A., and Di Benedetto, M.-G. ( ). “Acoustical andperceptual study of gemination in Italian stops,” The Journal ofthe Acoustical Society of America (4), 2051–2062.Fougeron, C., Goldman, J.-P., and Frauenfelder, U. H. ( ). “Li-aison and schwa deletion in French: An effect of lexical frequencyand competition?,” in , Aalborg, Denmark,pp. 639–642.Giannini, S., and Marotta, G. ( ). Fra grammatica e pragmat-ica: la geminazione consonantica in latino (Giardini editori estampatori in Pisa, Pisa, Italy).Klatt, D. ( ). “The New MIT Speechvax Computer Facility,”Speech Communication Group Working Papers IV.Mok, P. P. ( ). “Effects of consonant cluster syllabification onvowel-to-vowel coarticulation in English,” Speech Communica-tion (8), 946–956.Muljacic, Z. ( ). Fonologia della lingua italiana (Societ`a ed-itrice il Mulino, Bologna, Italy).Payne, E. M. ( ). “Non-durational indices in Italian geminateconsonants,” Journal of the International Phonetic Association (1), 83–95.Porru, G. ( ). “Anmerkungen ¨uber die Phonologie des Italienis-ches,” Travaux du Cercle Linguistique de Prague VIII , 187–208.Swadesh, M. ( ). “The Phonemic Interpretation of Long Con-sonants,” Language (1), 1.(1), 1.