De Sentmenat’s Vocabulary as a window on the phonological past of Maltese

Andrei Avram (University of Bucharest)

Abstract

The paper discusses the relevance of the 18th-century Catalan-Maltese Vocabulary by Marquis de Sentmenat to a number of issues pertaining to the historical phonology of Maltese. The topics addressed are the following: word-medial imāla of */a:/; deletion of short unstressed vowels in word-initial open syllables; auxiliary vowels; interdental fricatives; regressive voicing assimilation in obstruent clusters; word-final obstruent devoicing. The accuracy of the data in de Sentmenat’s Vocabulary is evaluated in light of corroborating evidence adduced mainly from other 18th-century records of Maltese.

10.5281/zenodo.16933084

1 Introduction

Marquis Francesc de Sentmenat-Torrelles i d’Agulló (1697-1762), a philologist, historian and poet, was one of the founding members of the Reial Acadèmia de Bones Lletres in Barcelona (Queraltó Bartrés 2003: 5). His unpublished legacy, preserved in the Biblioteca de Catalunya, includes notebook 5 of manuscript 1185, under the heading “Vocabulari i gramàtica de la llengua maltesa”. It is not known whether de Sentmenat ever visited Malta and the source of his Maltese data is unknown.

The manuscript is believed to have been written at the beginning of the second half of the 18th century and was transcribed and edited in 2003 by Alexandre Queraltó Bartrés. It consists of a Catalan-Maltese Vocabulary (Queraltó Bartrés 2003: 21-42); a new [= second] Catalan-Maltese Vocabulary (Queraltó Bartrés 2003: 43-64); sentences in Catalan and Maltese (Queraltó Bartrés 2003: 65-68); a prayer to Virgin Mary in Italian and Maltese (Queraltó Bartrés 2003: 69); a dialogue in Catan and Maltese (Queraltó Bartrés 2003: 70-71); a Catalan-Maltese thematic Vocabulary, arranged into 36 semantic fields (Queraltó Bartrés 2003: 71-84); a morphological sketch of Maltese (Queraltó Bartrés 85-101); a Maltese Catechism (Queraltó Bartrés 2003: 102-103). In what follows de Sentmenat’s work will be referred to as Vocabulary.

According to Queraltó Bartrés (2003: 2), the Vocabulary is “una obra importantíssima des del punt de vista del maltès” [= an extremely important work from the point of view of Maltese], given that “no es coneixen obres de contingut lingüístic maltès anteriors al segle XVIII” [= no works with Maltese linguistic content prior to the 18th century are known]. The latter statement is factually not true: Queraltó Bartrès was presumably not aware of Thezan’s Regole per la lingua maltese, written by 1647, and edited by Cassola (1992). On the other hand, de Sentmenat’s Vocabulary is indeed an important work, relevant to the history of Maltese, including, as will be shown, to the historical phonology of the language.

The paper is structured as follows. Section 2 presents the corpus and the methodology. Section 3 examines word-final imāla of */a:/. Section 4 is concerned with the preservation of short unstressed vowels in word-initial position. Section 5 considers auxiliary vowels. Section 6 deals with the preservation of interdental fricatives. Section 7 focuses on regressive voicing assimilation. Section 8 looks at word-final obstruent devoicing. Section 9 discusses some implications of the findings.

2 Corpus and methodology

Most of the data from de Sentmenat are taken from the Catalan-Maltese Vocabulary and the Catalan-Maltese thematic Vocabulary.

Reference is also made to several pre-18th-century records of Maltese: Megiser’s word list collected in 1588 (Megiser 1606), Thezan’s Regole per la lingua maltese (Cassola 1992), the place names in Abela (1647), Skippon’s word list collected in 1664 (Skippon 1732), Bonamico’s Sonnet dated c. 1672, and archival records of the Roman Inquisition in Malta (Cassar 2005).

In the evaluation of the forms (and their variants) attested in de Sentmenat’s Vocabulary, corroborating evidence is adduced from other 18th-century records of Maltese, such as Mifsud’s sermons (1739–1746), de Soldanis’s (1750) grammar, de Solanis’s dialogues (Id-Djalogi ta’ de Soldanis n.d.), Wzzino’s (1752) translation of the Christian Catechism and the so-called Mezzo Vocabolario, written by 1775 (Cassola 1996). Reference is also made to metalinguistic comments by contemporary authors.

Since the analysis of the forms (and their variants) in de Sentmenat’s Vocabulary crucially depends on the author’s system of transcription, several remarks are in order here. As noted by Queraltó Bartrès (2003: 12, and 13-16 for examples) there are numerous inconsistencies in the transcription of reflexes of */k/, */q/, */χ/, */ɣ/, */h/, */ħ/, and */ʕ/. For instance, both <k> and <q> are used to transcribe reflexes of */k/ and */q/.¹ Reflexes of */ʕ/ and */ɣ/ are both transcribed with <h>, <hh> and even <hhh>. As for reflexes of */χ/ and */ħ/, de Sentmenat transcribes both of them with <ch>, <chh>, <hh> or <hhh>. Moreover, <h> and <hhh> are sometimes used superfluously, since they represent no consonant whatsoever. Obviously, this makes it difficult and at times impossible to exploit the relevant forms in de Sentmenat for a discussion of the fate of the consonants at issue in 18th-century Maltese. Other inconsistencies and erroneous transcriptions include the following: the use of <f>, but also of the digraph <ph> for [f]; the use of <t>, but in a few cases of the digraph <th> as well, for [t]; the occasional confusion of <t> and <r>; the occasional use of the digraph <tz>² for [s]. In all other respects, de Sentmenat’s orthography is essentially based on that of Catalan. This explains the use of <x> to render [ʃ] and of the digraph <tx> to represent [ʧ]. The fact that Catalan orthography itself was not yet standardized at the time also accounts for the oscillating and ambiguous use of <s> and <z>.

The number of examples has been kept to a reasonable minimum. Necessary details are provided, where necessary, with respect to the system of transcription or orthography used in sources other than de Sentmenat’s Vocabulary. The original glosses in Catalan, German, Italian, or Latin in the sources have all been translated into English.

3 Word-medial imāla of */a:/

Cowan (1975: 7) claims that “by the time of Megiser in 1588 […] the plain non-final e had become diphthongized to /ie/”. In support of this claim, Cowan (1975) adduces forms found in Megiser’s word list, such as the following:

1. 1. chtieb ‘book’ 1588 (Megiser 1606: 11)
      cf. Ar. kitāb
   2. mnieher ‘nose’ 1588 (Megiser 1606: 10)
      cf. Ar manāḫir ‘nostrils’, Mod. M. mnieħer ‘nose’

In fact, however, the monophthongal reflex /e:/ of word-medial */a:/ continues to be attested even later. Consider first early 17th century examples, occurring in, among others, archival records of the Roman Inquisition in Malta:

2. 1. bebi ‘my door’ 1602 (Cassar 2005: 77)
      cf. Ar. bāb
   2. chen ‘[he] was’ 1602 (Cassar 2005: 77)
      cf. Ar. kān
   3. seha ‘hour’ 1602 (Cassar 2005: 78)
      cf. Ar. sāʕa

Consider next Thezan’s dictionary.³ The most frequently attested reflex of word-medial */a:/ is the diphthong /ie/. However, its monophthongal counterpart is still found in a few forms:

3. 1. moﻫereث ‘plough’ (Cassola 1992: 165v)
      cf. Ar. miḥrāṯ, Mod. M. moħriet
   2. حamen⁴ ‘bath’ (Cassola 1992: 145v)
      cf. Ar. ḥammām

Moreover, there is also evidence of variation, i.e. */a:/ > /e:/ ~ /ie/, in a number of words. For instance, a few forms with <ie> by Thezan are transcribed with <e> by Abela (1647):

4. 1. Binet ‘girls’ (Abela 1647: 65)
      cf. Ar. banāt
   2. Tigegi ‘hens’ (Abela 1647: 80)
      cf. Ar. daǧāǧ

The data examined show that the change /e:/ > /ie/ was still in progress in the 17th century, contra Cowan (1975), but was nearly complete. Indeed, in both Thezan (by 1647) and Abela (1647) forms with the diphthongal reflex /ie/ outnumber those with its monophthongal counterpart /e:/. Also, the occurrence of variation /e:/ ~ /ie/ constitutes evidence of lexical diffusion in the change */a:/ > /e:/ > /ie/. Continuing to track the change /e:/ > /ie/, one notices that all the relevant forms in Skippon’s word list collected in 1664 (Skippon 1732: 624-626) contain /ie/ as the reflex of word-media */a:/, as illustrated by the selected examples under (5), with a single exception (6), which exhibits the reflex /e:/:

5. 1. lisien ‘tongue’ (Skippon 1732: 625)
      cf. Ar. lisān
   2. sinien ‘teeth’ (Skippon 1732: 625)
      cf. Ar. asnān
   3. tieh⁵ ‘bottom’ (Skippon 1732: 626)
      cf. Ar. qāʕ
      
6. temenia ‘eight’ (Skippon 1732: 625)
   cf. Ar. ṯamāniya

In Bonamico’s (1672) Sonnet, again all relevant forms display /ie/ as the reflex of word-medial */a:/ (7), with the exception of the form in (8) with /e:/:

7. 1. gebiel ‘hills’ (Cachia 2000: 18)
      cf. Ar. ǧibāl ‘mountains’
   2. kien ‘[he] was’ (Cachia 2000: 18)
      cf. Ar. kān
   3. kiensce ‘[he] was not’ (Cachia 2000: 18)
      cf. Mod. M. kienx
      
8. kecu ‘if’ (Cachia 2000: 18)
   cf. Mod. M. kieku

Whereas in the second half of the 17th century /e:/ still occurred occasionally, pre-1750 records of Maltese suggest that the change /e:/ > /ie/ must have essentially run its course. In Ignazio Saverio Mifsud’s sermons, most relevant forms exhibit /ie/, as illustrated by the examples reproduced below:

9. 1. bliet ‘town’ 1739 (Ġabra tal-Malti Qadim n.d.)
      cf. Ar. bilād ‘countries’
   2. ħbieb ‘friends’ 1746 (Ġabra tal-Malti Qadim n.d.)
      cf. Ar. aḥbāb
   3. lsien ‘tongue’ 1741 (Ġabra tal-Malti Qadim n.d.)
      cf. Ar. lisān

De Soldanis’s grammar and dictionary (1750) also includes forms which mostly exhibit /ie/ as the reflex of word-medial */a:/. Consider the following examples:

10. 1. bniet ‘girls’ (de Soldanis 1750: 125)
       cf. Ar. banāt
    2. rmied ‘ashes’ (de Soldanis 1750: 187)
       cf. Ar. ramād
    3. qtiep ‘book’ (de Soldanis 1750: 170)
       cf. Ar. kitāb

However, the reflex /e:/ still occurs in a few forms:

11. 1. seqia ‘aqueduct’ (de Soldanis 1750: 176)
       cf. Ar. sāqiya
    2. tmegnà ‘eight’ (de Soldanis 1750: 106)
       cf. Ar. ṯamāniya

As for de Sentmenat’s Vocabulary,⁶ there are over 60 relevant forms with the reflex /ie/, exemplified in (12):

12. 1. bieb ‘door’ (QB 2003: 37)
       cf. Ar. bāb
    2. girien ‘neighbour’ (QB 2003: 41)
       cf. Ar. ǧirān ‘neighbours’
    3. mbièreck ‘blessed.M’ (QB 2003: 96)
       cf. Ar. mubārak
    4. koddiessa ‘mass’ (QB 2003: 33)
       cf. Mod. M. quddiesa
    5. zmien ‘time’ (QB 2003: 40)
       cf. Ar. zamān

In light of the remarks on de Sentmenat’s transcription system in section 2, there is no reason whatsoever to consider that <ie> stands for anything but a diphthongal reflex, i.e. [ie], of word-medial */a:/. There is a single form which appears to illustrate the preservation of the monophthongal reflex /e:/ of word-medial */a:/:

13. saÿer⁷/saÿhet ‘fisherman’ (QB 2003: 81)
    cf. Ar. ṣayyād ‘hunter, fisherman’, Mod. M. sajjied ‘fisherman’

It is confusing, but not surprising, given the many imperfections and inconsistencies of the transcriptions, to find that the diaeresis on <y> is used in different ways. This is found in 17 forms in de Sentmenat’s Vocabulary, and in 7 of these⁸ it corresponds to [jj]. These consist of four nouns (14a-d), a present participle (14e) and two past tense forms of verbs (14f-g):

14. 1. hhaÿa ‘life’ (QB 2003: 42)
       cf. Mod. M. ħajja
    2. hhaÿat ‘tailor’ (QB 2003: 39)
       cf. Ar. ḫayyāṭ
    3. maÿhal ‘boar’ (QB 2003: 39)
       cf. Mod. M. majjal
    4. thaÿar ‘cotton’ (QB 2003: 26)
       cf. Mod. M. tajjar
    5. seÿer ‘going’ (QB 2003: 90)
       cf. Mod. M. sejjer
    6. saÿhiartx ‘did [you] cook?’ (QB 2003: 66)
       cf. Mod. M. sajjartx
    7. seÿhhhatlu ‘[I] called him’ (QB 2003: 67)
       cf. Mod. M. sejjaħtlu

On the evidence provided by the forms above, the word in (13) would read [sajje:t] and is therefore one of the last linguistic residues in 18th-century Maltese still exhibiting the reflex /e:/ of word-medial */a:/.

In addition to /ie/ and /e:/, in a few forms the reflex of word-medial */a:/ is transcribed with <i>, which presumably stands for [i:]. For those under (15) this is the only form attested:

15. 1. menhihhherr⁹ ‘nose’ (QB 2003: 34)
       cf. Ar. manāḫir ‘nostrils’, Mod. M. mnieħer ‘nose’
    2. muphtihhe ‘key’ (QB 2003: 25)
       cf. Ar. muftāḥ

Two other cases consist of variants with <i> or <ie>:

16. 1. henina/chanina (QB 2003: 69, 102) ~ hhhniena/cheniena (QB 2003: 65, 36) ‘mercy’
       cf. Ar. ḥanāna
    2. tziahh ~ tsiehhh ‘hour’ (QB 2003: 31)
       cf. Ar. sāʕa

The forms with <i> should not be dismissed as mere errors of transcription. The reflex /i:/, spelled <i>, is attested both before and after de Sentmenat’s Vocabulary. Consider the form for ‘hour’, cf. Ar. sāʕa, in three sources predating (17a-c) and one dated after de Sentmenat (17d):

17. 1. siaﻉ by 1647 (Cassola 1996: 175r)
    2. siha 1739, 1740, 1741 (Mifsud, in Ġabra tal-Malti Qadim. n.d.)
    3. sigha 1750 (de Soldanis 1750: 111)
    4. siaa by 1775 (Cassola 1996: 146)

Summing up, by around the half of the 18th century the change /e:/ > /ie/ appears to have been essentially completed. That the process is yet another manifestation of the phenomenon of lexical diffusion is further proved by the data set out in Table 1 below. As can be seen, before approximately 1750, both /e:/ and /ie/ as well as variation /e:/ ~ /ie/ are attested in the five words which all have /ie/ in de Sentmenat’s Vocabulary:

To the best of my knowledge, after de Sentmenat’s Vocabulary, forms still displaying the monophthongal reflex /e:/ of word-medial */a:/ are extremely poorly represented. The following is one such rare example, from Mezzo Vocabolario:

18. hammem ‘bath’ by 1775 (Cassola 1996: 61)
    cf. Ar. ḥammām

However, both the reflex /e:/ and the reflex /i:/ continue to be found in Maltese dialects (see e.g. Borg 1976: 196, Table 2.011). This accords with the scenario in Avram (2016), whereby the reflex /e:/ further develops into /ie/ or /i:/. Also, the co-occurrence of two reflexes, /ie/ and /i:/, of word-medial */a:/ is unsurprising from the perspective of lexical diffusion theory. Finally, the co-occurrence of /ie/ and /i:/ is reported for some Maltese dialects (Borg 1976: 196, Table 2.011).

4 Short unstressed vowels in word-initial open syllables

Cowan (1975: 9) concludes his discussion of the deletion of short unstressed /i, a, u/ in word-initial open syllables by writing that “when we compare this material [i.e. in Caxaro’s Cantilena] with that of Megiser, we see that all the vowels in question had been elided”.

Actually, as shown in Avram (2023), short unstressed vowels in word-initial open syllables continue to be attested in Maltese even after Megiser’s time, in a number of sources. Consider first Cowan’s (1975) examples from Megiser (1611), under (19), and their counterparts in Thezan, under (20):

19. 1. chtieb ‘book’
       cf. Ar. kitāb
    2. embit ‘wine’
       cf. Ar. nabīḏ
    3. flus ‘money’
       cf. Ar. fulūs
    4. guir¹⁰ ‘big.M’
       cf. Ar. kabīr
    5. mnieher ‘nose’
       cf. Ar. manāḫir ‘nostrils’, Mod. M. mnieħer ‘nose’
    6. fliesan ‘three’
       cf. Ar. ṯalāṯa
    7. sfniema ‘eight’
       cf. Ar. ṯamāniya

As can be seen below, 5 of the above forms exhibit a short unstressed vowel in the word-initial syllable.

20. 1. ketieb ‘book’ (Cassola 1992: 153v)
    2. kebir ‘big.M’ (Cassola 1992: 152v)
    3. menieﺡar ‘nose’ (Cassola 1992: 71r)
    4. ﺛelieta ‘three’ (Cassola 1992: 181v)
    5. ﺛemienia ‘eight’ (Cassola 1992: 181v)

The vowel at issue is transcribed <e>. However, as already noted by Hull (1994: 394), in Thezan’s dictionary “the Arabic short vowel i (tonic and atonic) is regularly noted /e/”. A similar observation is made more recently by van Putten (2020: 62, n. 3) who writes that “in the 17th century dictionary of Thezan, modern Maltese i is often spelled with <e>”. Therefore, Thezan’s <e> frequently represents [ɪ]. Other entries in Thezan’s dictionary, such as those in (21), exhibit the vowel [a]:

21. 1. ﺡamar ‘donkey’ (Cassola 1992: 145v)
       cf. Ar. ḥimār
    2. ﻕarīb ‘close’ (Cassola 1992: 151r)
       cf. Ar. qarīb

Moreover, in Thezan forms with a short unstressed vowel in word-initial open syllables outnumber those which the vowel at issue is deleted (Avram 2023).

Forms exhibiting short unstressed vowels in word-initial open syllables are also found in Abela’s (1647) place names and Skippon’s word list collected in 1664 (Skippon 1732):

22. 1. Hamàr ‘donkey’ (Abela 1647: 251)
       cf. Ar. ḥimār
    2. Kibir ‘big.M’ (Abela 1647: 66)
       cf. Ar. kabīr
       
23. chaniena ‘mercy’ 1664 (Skippon 1732: 626)
    cf. Ar. ḥanāna

On the strength of the evidence provided by such sources, Avram (2023) concludes that the change word-initial unstressed */i, a, u/ > Ø was still in progress in the 17th century, i.e. after the time of Megiser, contra Cowan (1975). This is confirmed by the fact that several forms recorded by de Sentmenat show that short unstressed vowels in word-initial open syllables still occurred even in the 18th century. Consider the first set of examples below, consisting mostly of nouns (24a-f):

24. 1. hhhemara ‘female donkey’ (QB 2003: 23)
       cf. Ar. ḥimāra
    2. kenisia ‘church’ (QB 2003: 31)
       cf. Ar. kanīsa, Mod. M. knisja
    3. menhihhherr ‘nose’ (QB 2003: 34)
       cf. Ar. manāḫir ‘nostrils’, Mod. M. mnieħer ‘nose’
    4. sahabahhh ‘cloud’ (QB 2003: 35)
       cf. Ar. saḥāba
    5. denub ‘sin’ (QB 2003: 36)
       cf. Ar. ḏunūb ‘sins’
    6. zeniet ‘flint’ (QB 2003: 36)
       cf. Ar. zinād ‘fire steel’, Mod. M. znied ‘flint’
    7. petit ‘a little’ (QB 2003: 37)
       cf. Mod. M. ftit¹¹

A second set consists of cardinal numerals¹² (25a-h) and the word for ‘Monday’ (25i):

25. 1. tenhax ‘twelve’ (QB 2003: 27)
       cf. Mod. M. tnax
    2. timintax ‘eighteen’ (QB 2003: 27
       cf. Mod. M. tmintax
    3. tinein ‘two’ (QB 2003: 27)
       cf. Mod. M. tnejn
    4. hhemistax ‘fifteen’ (QB 2003: 38)
       cf. Mod. M. ħmistax
    5. tilieta ‘three’ (QB 2003: 41)
       cf. Ar. ṯalāṯa
    6. tilitax ‘thirteen’ (QB 2003: 41)
       cf. Mod. M. tlettax
    7. thimienia ‘eight’ (QB 2003: 42)
       cf. Ar. ṯamāniya
    8. theminìn ‘eighty’ (QB 2003: 42)
       cf. Mod. M. tmenin
    9. ettenein ‘Monday’ (QB 2003: 27)
       cf. Ar. al-iṯnayn, Mod. M. it-Tnejn

In two other forms there is variation between the occurrence of the short unstressed vowel and its absence:

26. 1. kabir ‘big.M’ ~ kbira ‘big-F’ (QB 2003: 103, 30)
       cf. Ar. kabīr ‘big.M’, kabīra ‘big-F’
    2. cheniena/henina/chanina vs. hhhniena ‘mercy’ (QB 2003: 65, 69, 102, 36)
       cf. Ar. ḥanāna

That the short unstressed vowels in word-initial position in the forms reproduced in (24)-(26) do not represent errors of transcription is also shown by the fact that such vowels are still sporadically attested in other 18th-century sources. Consider first two pre-1750 examples, from one of Mifsud’s sermons (27) and from the archival records of the Roman Inquisition in Malta (28), respectively:

27. 1. chibir ‘big.M’ 1739 (Zammit Ciantar 1985: 193)
       cf. Ar. kabīr
    2. timiena ‘eight’ 1741 (Ġabra tal-Malti Qadim n.d.)
       cf. Ar. ṯamāniya
       
28. miniel ‘shod’ 1743 (Cassar 2005: 71)
    cf. Mod. M. mniegħel

Post-1750 examples include entries in the Mezzo Vocabolario (Cassola 1996), such as:

29. 1. kêtieb ‘book’ by 1775 (Cassola 1996: 87)
       cf. Ar. kitāb
    2. semin ‘fat’ by 1775 (Cassola 1996: 144)
       cf. Ar. samīn

Finally, forms containing unstressed short vowels in word-initial syllables are attested even towards the end of the 18th century (Zammit 2009–2010: 30). The next set of examples is from the archival records of the Roman Inquisition in Malta (30) and from a sermon (31):

30. dimuh ‘tears’ 1788 (Cassar 2005: 79)
    cf. Ar. dumūʕ

31. 1. chibar ‘big.PL’ 1791 (Zammit 2009–2010: 30)
       cf. Ar. kubār
    2. chinisia ‘church’ 1791 (Zammit 2009–2010: 30)
       cf. Ar. kanīsa, Mod. M. knisja
    3. hilejac ‘creatures’ 1791 (Zammit 2009–2010: 30)
       cf. Ar. ḫalāʔiq, Mod. M. ħlejjaq

The survival of reflexes of short unstressed /i, a, u/ in word-initial position in a number of entries in de Sentmenat’s Vocabulary as well as the occurrence of variation between forms with or without a short unstressed vowel in this position constitute evidence of lexical diffusion in the change word-initial unstressed */i, a, u/ > Ø. This is confirmed by the data in Table 2, which compares forms found in de Sentmenat’s Vocabulary attested both in pre-1750 records of Maltese and in later 18th-century ones.¹³ As can be seen, after de Sentmenat there is a gradual decrease in the number of forms which still exhibit unstressed short vowels in word-initial position:

Finally, the preservation of word-initial unstressed short vowels may arguably be said to reflect the occurrence in late 18th-century Maltese of what Cantineau (1960: 110) calls “voyelles ultra-brèves” (see also Avram 2016).

5 Auxiliary vowels

Borg (1978) illustrates, among others, the occurrence in 15th- and 16th-century Maltese of auxiliary vowels¹⁴ epenthesized into word-medial consonant clusters, which include reflexes of the Arabic patterns aCCaC, maCCaC, maCCūC and CVCCa. With respect to the quality of the auxiliary vowel, Borg (1978: 223-24) argues that it is phonologically conditioned: [a] in a back environment, i.e. after /q, χ, ʁ, h, ʔ, ħ, ʕ/ and [ɪ] elsewhere. On this analysis, then, the two auxiliary vowels are in complementary distribution.

In Avram (2023, 2024b) it is shown that auxiliary vowels are attested in the 17th-century as well. This is particularly true of the entries in Thezan’s dictionary. These include nouns (32a), cardinal numerals (32b), adjectives (32c), comparative forms of adjectives (32d), past participles of the 1^st form (32e) and imperatives (32f):

32. 1. meseraﻫ ‘plain’ (Cassola 1992: 163v)
       cf. Ar. masraḥ
    2. oﻉﺵerin ‘twenty’ (Cassola 1992: 5r)
       cf. Mod. M. għoxrin
    3. neserani ‘Christian’ (Cassola 1992: 168r)
       cf. Ar. naṣrānī
    4. ekereﺡ ‘uglier’ (Cassola 1992: 130v)
       cf. Ar. akrah
    5. maﺡruﻕ ‘burnt.M’ (Cassola 1992: 159r)
       cf. Ar. maḥrūq
    6. eﺶirob ‘drink!’ (Cassola 1992: 137v)
       cf. Ar. ʔišrab

While in all the forms above the distribution of the auxiliary vowels [a] and [ɪ] is in accordance with the phonological condition posited by Borg (1978), there are also exceptions. For instance, in several forms such as those in (33) the auxiliary vowel [ɪ] is inserted into a non-back environment:

33. 1. moﻫereث¹⁵ ‘plough’ (Cassola 1992: 165v)
       cf. Ar. miḥrāṯ, Mod M. moħriet
    2. aﻍelaﻕ ‘close!’ (Cassola 1992: 116r)
       cf. Ar. ʔiġliq

Also, a number of forms display the auxiliary vowel [o]:

34. 1. ﺛomona ‘unit of measurement for grains’ (Cassola 1992: 182v)
       cf. Ar. ṯumna
    2. moﻉoti ‘given.M’ (Cassola 1992: 165v)
       cf. Ar. muʕṭ-
    3. oﻕoros ‘pinch!’ (Cassola 1992: 169v)
       cf. Ar. ʔuqruṣ

The auxiliary vowel [o] in the forms under (34) can only be accounted for by assuming vowel harmony of the left-to-right (progressive) vowel-copying type, whereby the first [o], the reflex of the Ar. stem vowel /u/, triggers the insertion of [o] into the word-medial consonant cluster.

Two other 17th-century sources, the place names in Abela (1647) and Skippon’s 1664 word list (Skippon 1732), provide evidence for the occurrence of auxiliary vowels in word-medial consonant clusters. Consider the following examples:

35. Machadar ‘Assembly’ (Abela 1647: 258)
    cf. Ar. maḥḍar
    
36. 1. chagiara ‘stone’ 1664 (Skippon 1732: 624)
       cf. Ar. ḥaǧra, Mod. M. ħaġra
    2. tizira ‘to plant’ 1664 (Skippon 1732: 626)
       cf. Mod M. tiżra’

Note that the [a] in (36a), in a non-back environment, violates Borg’s (1978) conditions on the occurrence of this auxiliary vowel.

Auxiliary vowels are still attested in de Sentmenat’ Vocabulary. The four forms below illustrate the insertion of the auxiliary vowel [a] in a back environment:

37. 1. takara ‘you.SG read’ (QB 2003: 31)
       cf. Ar. taqraʔ
    2. xaharein ‘two months’ (QB 2003: 33)
       cf. Ar. šahrayn
    3. bahhhàla ‘female mule’ (QB 2003: 34)
       cf. Ar. baʕla
    4. ràhhhada ‘thunder’ (QB 2003: 41)
       cf. Ar. raʕda

Two other forms illustrate variation, between the occurrence and the absence of the auxiliary vowel [a] in a back environment:

38. 1. mahhalùb ~ mahhhlùb ‘thin.M’ (QB 2003: 32, 79)
       cf. Mod. M. magħlub
    2. sahatar ~ sahtar ‘wild thyme’ (QB 2003: 31, 74)
       cf. Ar. saʕtar/ṣaʕtar

As for non-back environments, judging by de Sentmenat’s transcriptions, the auxiliary vowel is transcribed with <e>. The insertion of this auxiliary vowel into a variety of etymological word-medial clusters, in a non-back environment, is better attested. Consider the forms below:

39. 1. ebbeni ‘my son’ (QB 2003: 29)
       cf. Ar. ibnī
    2. tiphela ‘girl’ (QB 2003: 34)
       cf. Ar. ṭifla
    3. ettènein ‘Monday’ (QB 2003: 27)
       cf. Ar. al-iṯnayn, Mod. M. it-Tnejn
    4. bàssela ‘onion’ (QB 2003: 39)
       cf. Ar. baṣla
    5. kasseria/kesseria/kosserìa ‘flowerpot’ (QB 2003: 40, 77, 80)
       cf. Ar. qaṣriyya Mod. M. qasrijja
    6. nehhmehla ‘ant’ (QB 2003: 29)
       cf. Ar namla
    7. hamera ‘red.F’ (QB 2003: 37)
       cf. Ar. ḥamrāʔ
    8. thomena ‘unit of measurement for grains’ (QB 2003: 33)
       cf. Ar. ṯumna

Also, one form appears to attest to the occurrence of variation, i.e. occurring either without or with an auxiliary vowel:

40. hhhoxrin ~ hhhoxérin ‘twenty’ (QB 2003: 63)
    cf. Mod. M. għoxrin

Some forms in de Sentmenat’s Vocabulary show again that the phonological conditioning posited by Borg (1978) is less straightforward, given the occurrence of forms violating it. In the following forms, for instance, the vowel transcribed with <e> is inserted in a back environment:

41. 1. hhhebiebi ‘friends (QB 2003: 21)
       cf. Ar. aḥbāb, Mod. M. ħbieb
    2. hackkeli ‘[he] fries’ (QB 2003: 29)
       cf. Mod. M. jaqli
    3. iochhelom/ihòchelom ‘[he] dreams’ (QB 2003: 39, 68)
       cf. Ar. yaḥlum
    4. machelut ‘mixed grain’ (QB 2003: 33)
       cf. Mod. M. maħlut
    5. achena ‘we’ (QB 2003: 102)
       cf. Mod. M. aħna

There are two other forms which exhibit an auxiliary vowel which runs counter to Borg’s (1978) phonological conditioning. As can be seen below, the vowel inserted into the etymological word-medial cluster is [o], in a back environment:

42. 1. nohhhochrotx¹⁶ ‘[I] get out’ (QB 2003: 67)
       cf. Mod. M. noħroġ
    2. sohhhòla ‘cough’ (QB 2003: 41)
       cf. Ar. suʕla, Mod. M. sogħla

The auxiliary vowel [o] in these forms is the outcome of vowel harmony of the left-to-right (progressive) vowel copying type, whereby the first [o] triggers the insertion of [o] into the word-medial consonant cluster.

De Sentmenat’s transcriptions of words with an auxiliary vowel appear to be accurate, given that a number of these (including different forms in the paradigm of a word) are found in two 17th-century records of Maltese, as can be seen in Table 3, even though the quality of the vowel inserted into the word-medial cluster is not always the same:

Most relevant entries in de Sentmenat’s Vocabulary do not display auxiliary vowels. Also, as seen, several ones have variants with or without an auxiliary vowel. This suggests that by the time of de Sentmenat deletion of the former auxiliary vowel was spreading through the lexicon, with only some residual forms still exhibiting it. Again, this is in line with a lexical-diffusion approach to sound change.

6 Interdental fricatives

As shown in Avram (2014, 2021, 2022b, 2023) the first convincing evidence for the occurrence of fricative interdentals in earlier Maltese is provided by the entries in Thezan’s dictionary. The use of the Arabic letters <ﺛ>, <ﺙ> to render the voiceless interdental fricative /θ/ and of <ﺫ> and, very rarely, of the digraph <dh>, for the voiced interdental fricative /ð/ leaves room for no doubt as to their occurrence in 17th-century Maltese.

In Thezan’s dictionary <ﺛ> or <ﺙ> correspond to */θ/, in word-initial, word-medial and word-final position:

43. 1. ثaقab ‘to perforate’ (Cassola 1992: 137v)
       cf. Ar. ṯaqab
    2. uﺙiﻕ ‘strong.M’ (Cassola 1992: 184r)
       cf. Ar. waṯīq
    3. حarraﺙ ‘ploughman’ (Cassola 1992: 146r)
       cf. Ar. ḥarraṯ

As for <ﺫ>, it corresponds to three etymological interdental fricatives */ð/, */dˁ/ and */ðˁ/. In a few forms the reflexes of */ð/ and */ðˁ/ are transcribed with the digraph <dh>. Reflexes of */ð/ occur word-initially, word-medially and word-finally:

44. 1. ﺬrieﻫ ‘arm’ (Cassola 1992: 128v)
       cf. Ar. ḏirāʕ
    2. barذa/bradha¹⁷ ‘pack-saddle’ (Cassola 1992: 121v)
       cf. A barḏaʕa
    3. menfuذ ‘transpierced.M’ (Cassola 1992: 162v)
       cf. Ar. manfūḏ

The letter <ﺬ> corresponding to */dˁ/ is attested in all three syllabic positions:

45. 1. ﺫarab ‘to wound’ (Cassola 1992: 127r)
       cf. Ar. ḍarab
    2. feذذa ‘silver’ (Cassola 1992: 139r)
       cf. Ar. fiḍḍa
    3. mariذ ‘sick.M’ (Cassola 1992: 160r)
       cf. Ar. marīḍ

Finally, the grapheme <ﺫ> and the digraph <dh> in reflexes */ðˁ/ are found in word-initial and word-medial position:

46. 1. ﺫefer ‘nail’ (Cassola 1992: 127v)
       cf. Ar. ẓifr
    2. aعdhem ‘bone’ (Cassola 1992: 115v)
       cf. Ar. ʕaẓm

A number of forms occurring in Abela’s (1647) place names contain both voiceless and voiced interdental fricatives, transcribed with the digraph <th> and <dh> respectively; the latter corresponds to reflexes of */ð/, as in (47b), and */dˁ/, as in (47c):

47. 1. MethKub¹⁸ ‘perforated’ (Abela 1647: 22)
       cf. Ar. maṯqūb
    2. dheeb ‘gold’ (Abela 1647: 107)
       cf. Ar. ḏahab
    3. baidha ‘white’ (Abela 1647: 25)
       cf. Ar. baiḍāʔ

Interdental fricatives are also found in several works by de Soldanis’s, such as his grammar (1750) and dialogues (Id-Djalogi ta’ de Soldanis n.d.):¹⁹ /θ/ is transcribed with <th> or <ṫ>, while /ð/ is rendered by <dh> or <ḋ>. As in Thezan’s dictionary, /ð/ corresponds to */ð/, */dˁ/ and */ðˁ/, as shown in (48b), (48c) and (48d), respectively:

48. 1. thielet ‘third’ (de Soldanis 1750: 71)
       cf. Ar. ṯāliṯ
    2. ḋeep ‘gold’ (de Soldanis 1750: 72-73)
       cf. Ar. ḏahab
    3. ḋaul ‘light’ (Id-djalogi ta’ de Soldanis n.d.)
       cf. Ar. ḍauʔ, Mod. M. dawl
    4. ḋlam ‘darkness’ (Id-Djalogi ta’ de Soldanis n.d.)
       cf. Ar. ẓalām

As for de Sentmenat’s Vocabulary, there is evidence pointing to the preservation in some lexical items of both voiceless and voiced interdental fricatives. However, unlike in Thezan’s dictionary, the number of such forms is very small and interdental fricatives are attested only in word-initial onsets.

Consider first reflexes of etymological voiceless interdental fricatives. The following forms, two nouns and a cardinal numeral, occur only with what appears to be the voiceless interdental fricative [θ], transcribed with the digraph <th>:

49. 1. thehuma/theuma ‘garlic’ (QB 2003: 21, 43)
       cf. Ar. ṯūm
    2. thomena ‘unit of measurement for grains’ (QB 2003: 33)
       cf. Ar. ṯumna
    3. theminìn ‘eighty’ (QB 2003: 42)
       cf. Mod. M. tmenin

One other noun could contain a voiceless interdental fricative:

50. themòntox ‘tow’ (QB 2003: 24)
    cf. Mod. M. tmontox²⁰

While no Arabic etymon with */θ/ has been suggested, note that the noun also appears in Thezan’s dictionary, where it is transcribed ﺛemontoﺵ, i.e. with a word-initial voiceless interdental fricative.

The transcriptions of seven other forms are suggestive of variation in the phonetic realization of the etymological voiceless interdental fricative. As shown below, these forms are transcribed either with the digraph <th> or with <t>:

51. 1. etthenein vs. ettenein ‘Monday’ (QB 2003: 73, 27)
       cf. Ar. al-iṯnayn, Mod. M. it-Tnejn
    2. thielet vs. ittielet ‘third’ (QB 2003: 102, 62)
       cf. Ar. ṯaliṯ
    3. thimienia vs. temienia ‘eight’ (QB 2003: 42, 102)
       cf. Ar. ṯamāniya
    4. thenàx vs. tenhax ‘twelve’ (QB 2003: 84, 27)
       cf. Mod. M. tnax
    5. thilieta vs. tilieta ‘three’ (QB 2003: 84, 41)
       cf. Ar. ṯalāṯa
    6. thilitàx vs. tilitax ‘thirteen’ (QB 2003: 84, 41)
       cf. Mod. M. tlettax
    7. thinein vs. tinein ‘two’ (QB 2003: 84, 27)
       cf. Mod. M. tnejn

However, three forms are only attested with a voiceless stop reflex of the etymological voiceless interdental fricative:

52. 1. itlieta ‘Tuesday’ (QB 2003: 27)
       cf. Ar. al-ṯulāṯāʔ, Mod. M. it-Tlieta
    2. tletin ‘thirty’ (QB 2003: 62)
       cf. Ar. ṯalāṯīn
    3. timintax ‘eighteen’ (QB 2003: 27)
       cf. Mod. M. tmintax

The voiced counterpart of /θ/ is attested only in two nouns, in which the digraph <dh> renders what appears to be the voiced interdental fricative [ð]:

53. 1. dhehhheb ‘gold’ (QB 2003: 35)
       cf. Ar. ḏahab
    2. dheil ‘shirt’ (QB 2003: 24)
       cf. Ar. ḏayl ‘hem of a garment, train of a skirt’, Mod. M. dejl ‘skirt’

The fact that <th> is erroneously used in a few cases, as mentioned in section 2, raises the question of the accuracy of de Sentmenat’s transcriptions. However, the digraph <dh> only appears twice, in words derived from etyma with the voiced interdental fricative /ð/. Further, all the forms under (49)-(51) and (53), i.e. with what appear to be interdental fricatives, are also attested in 17th-century records as well as, more importantly, in de Soldanis, a contemporary of de Sentmenat’s. Consider the data set out in Table 3:

As also shown in Avram (2022b), no interdental fricatives are found in other 18th-century records such as Mifsud’s (1739–1746) sermons, Wzzino’s (1752) translation of the Christian Catechism or the Mezzo Vocabolario (by 1775). This also accounts for the fact that the number of such forms in de Sentmenat’s Vocabulary is rather low, as expected from the perspective of lexical diffusion.

Metalinguistic comments by contemporary writers point to the same conclusion. In a letter to de Soldanis, dated September 27, 1749, a certain Giovanni Cachia writes that “nella pronuncia della lettera t e th, non vi è gran differenza” [= there is not a great difference between the pronunciation of the letter t and th] and that “per distinguerli bisogna che uno sudi” [= to distinguish them one must sweat] (Cassola 1993: 217). This indicates that by that time, there already was confusion of the interdental fricatives and the corresponding stops.

The clearest evidence of the gradual replacement of the interdental fricatives by the stops /t/ and /d/, both in the speech community and through the lexicon, is provided by Vassalli (1796). In the “Discorso preliminare” to his dictionary, Vassalli (1796: XXIX, f.n. 11) writes that “gli Arabi descrivono il suono T con tre lettere, cioè ﺕ Te, ﺙ The bleso, e ﻁ Ta; quello della S con due, ﺱ Sin, e ﺹ Sad; e quello del D con quattro, ﺩ Dal, ﺫ Dhal, ﺽ Dad, e ﻅ Da” [= the Arabs describe the sound T with three letters, i.e. ﺕ Te, ﺙ with a lisp, and ﻁ Ta; that of S with two, ﺱ Sin, e ﺹ Sad; and that of D with four, ﺩ Dal, ﺫ Dhal, ﺽ Dad, e ﻅ Da] and that “egli pare che distinguano od almeno debbano distinguere tai suoni, che noi non distin-guiamo, fuori del The e del Dhal” [= they seem to distinguish or at least should distinguish such sounds, which we do not distinguish, except for The and Dhal]. While ṯ and ḏ were still distinguished, Vassalli (1796: XXIX, f.n. 11) adds that “a’ quali suoni non ò assegnato due cifre a parte per essere molto raramente e tra pochi usati” [= I did not assign two separate letters since they are used very rarely and by few people], i.e. the occurrence of interdental fricatives was confined to few speakers. Evidence for the gradual replacement of the interdental fricatives by stops through the lexicon of Maltese is also provided; consider Vassalli’s (1796: 62) comments De littera T: “Ħossha hu bħal fyl-Latĭn u fyt-Taljân, għâd xi drâbi tkûn mtemtma: kĭf nystgħu narav fl’aħħar ta dân yl ktŷb, fejn jynsâbu yl kelmŷt kollha mtemtmĭn” [= Its sound is as in Latin and in Italian, yet sometimes it is pronounced with a lisp: as can be seen at the end of this book, where all the words pronounced with a lisp are found].²¹

7 Regressive vocing assimilation

Thezan’s dictionary is the first record of Maltese in which regressive voicing assimilation is attested in a significant number of forms, though in the minority of potential forms (Avram 2022a, 2023, 2025). Regressive voicing assimilation is rather well represented in onset clusters (54a-b), but it is also attested in word-medial (54c-d) and word-final clusters (54e-f)

54. 1. ptala ‘holiday’ (Cassola 1992l 171v)
       cf. Ar. baṭāla/biṭāla
    2. pچ_ei_چa ‘small piece’ (Cassola 1992: 121v)
       cf. Sic. piccia, Mod. M. bċejċa (diminutive of biċċa ‘piece’)
    3. ﺵepka ‘net’ (Cassola 1992: 178r)
       cf. Ar. šabaka
    4. etfen ‘bury!’ (Cassola 1992: 137r)
       cf. Ar. idfin
    5. sept ‘Saturday’ (Cassola 1992: 174v)
       cf. Ar. sabt
    6. nepﻫ ‘barking’ (Cassola 1992: 167v)
       cf. Ar. nabḥ

Ignazio Saverio Mifsud’s collections of sermons, covering the period 1739–1746, is the first record of Maltese in which forms exhibiting regressive voicing assimilation outnumber those which do not. Regressive voicing assimilation is attested in all types of obstruent clusters, i.e. word-initial (55a-b), word-medial (55c-d) and word-final (55e-f):

55. 1. p’tant ‘with so much’ 1741 (Ġabra tal-Malti Qadim)
       cf. Sic. tantu, It. tanto, Mod. M. b’tant
    2. thalt ‘[I] entered’ 1746 (Ġabra tal-Malti Qadim)
       cf. Ar. daḫalt
    3. darptein ‘twice’ 1741
       cf. Mod. M. darbtejn
    4. ithac ‘[he] laughs’ 1741
       cf. Ar. yaḍḥak
    5. hops ‘bread’ 1746
       cf. Ar. ḫubz
    6. insapsc ‘[he] was not found’ 1739
       cf. Mod. M. insabx

Consider next several records of Maltese dating from approximately the same period when de Sentmenat’s Vocabulary was written. In Wzzino’s (1752) translation of the Christian Catechism, regressive voicing assimilation is attested in onset (56a-b), word-medial (56c-d) and, less frequently, in word-final clusters (56e):

56. 1. ptajel ‘holidays’ (Wzzino 1752: 49)
       cf. Mod M. btajjel
    2. pḣhal²² ‘like’ (Wzzino 1752: 45)
       cf. Mod. Mal. bħal
    3. saptu ‘[she] found him’ (Wzzino 1752: 87)
       cf. Mod. M. sabthu
    4. jipk̇a²³ ‘[he] remains’ (Wzzino 1752: 78)
       cf. Ar. yabqā
    5. hhsipt ‘[you.SG] thought’ (Wzinno 1750: 107)
       cf. Mod. M. hsibt

Illustrated below are examples of forms displaying regressive voicing assimilation from de Soldanis’s grammar (1750) and dialogues (Id-Djalogi ta’ de Soldanis. n.d.). These are attested in word-initial (57a-b), word-medial (57c-d) and word-final clusters (57e-f):

57. 1. sfin ‘dancing’ (de Soldanis 1750: 93)
       cf. Mod. M. żfin
    2. dbeddel ‘[it] changed’ (Id-Djalogi ta’ de Soldanis n.d.)
       cf. Mod. M. tbiddel
    3. iepka ‘[he] remains’ (Id-Djalogi ta’ de Soldanis n.d.)
       cf. Ar. yabqā
    4. ietchol ‘[he] enters’ (Id-Djalogi ta’ de Soldanis n.d.)
       cf. Ar. yadḫul
    5. hhsept ‘[I] thought’ (Id-Djalogi ta’ de Soldanis n.d.)
       cf. Ar. ḥasibt
    6. ghagest ‘[I] have grown old’ (Id-Djalogi ta’ de Soldanis n.d.)
       cf. Mod. Mal. għaġiżt

Turning to de Sentmenat’s Vocabulary, the picture that emerges is a relatively similar one. While regressive voicing assimilation is attested in word-initial, word-medial and word-final clusters, there are also examples of variation as well as of forms in which obstruent clusters do not agree in voicing. The following are examples of onset clusters which agree in voicing:

58. 1. pkaila ‘spinach’ (QB 2003: 38)
       cf. Ar. buqayla
    2. ptaiel ‘holidays’ (QB 2003: 103)
       cf. Mod. M. btajjel
    3. phal ‘like’ (QB 2003: 102)
       cf. Mod. M. bħal

The next set of examples illustrates the occurrence of regressive voicing assimilation in obstruent clusters in word-medial position:

59. 1. ar kupteik ‘on your knees’ (QB 2003: 67)
       cf. Mod. M. għarkobbtejk
    2. ipki ‘[he] cries’ (QB 2003: 37)
       cf. Ar. yabkī
    3. ethhhhól ‘enter!’ (QB 2003: 28)
       cf. Ar. udḫul
    4. sipsi ‘pipe’ (QB 2003: 70)
       cf. Alg. Ar. sebsi, Mor. Ar. sebsi
    5. isphenu²⁴ ‘[they] dance’ (QB 2003: 91)
       cf. Mod. M. jiżfnu

The last set of examples contains coda clusters which exhibit regressive voicing assimilation:

60. 1. gipps ‘plaster’ (QB 2003: 30)
       cf. Ar. ǧibs
    2. hops ‘bread’ (QB 2003: 30)
       cf. Ar. ḫubz
    3. marritx ‘[I] don’t want’ (QB 2003: 42)
       cf. Mod. M. ma rridx

In a few forms there is alternation between variants which do not agree in voicing and those displaying regressive voicing assimilation:

61. 1. libsa ~ lipsa ‘clothes’ (QB 2003: 70, 42) ‘clothes’
    2. tridx ~ tritx (QB 2003: 97, 67) ‘don’t [you.SG] play?’

Finally, in several forms the obstruent clusters consistently do not agree in voicing:

62. 1. kbira ‘big-F’ (QB 2003: 30)
       cf. Ar. kbira
    2. issabt/issèbt ‘Saturday’ (QB 2003: 28, 73)
       cf. Ar. as-sabt
    3. titlobx ‘[you.SG] don’t ask (QB 2003: 49)
       cf. Mod. M. titlobx
    4. tilhhhabx ‘[you.SG] don’t play’ (QB 2003: 53)
       cf. Mod. M. tilgħabx

As is well known, Catalan displays regressive voicing assimilation which affects all classes of obstruents.²⁵ According to Prieto (2004: 217), “en tots els dialectes catalans, quan una obstruent està en posició de coda sil·làbica sempre s’assimila en sonoritat a la consonant que segueix” [= in all Catalan dialects, when an obstruent is a syllabic coda position it always assimilates in voicing to the consonant which follows]. Wheeler (2005: 2005) writes that “in all other environments than in initial or medial onsets there is neutralization of voice contrasts in obstruents”. The question therefore arises whether de Sentmenat’s transcriptions may have been influenced by his mother tongue. However, while forms without regressive voicing asssimilation still occur in 18th-century records of Maltese by native speakers, forms which do exhibit it gradually become the majority (Avram 2022a). Also, in de Sentmenat’s Vocabulary forms which exhibit regressive voicing assimilation outnumber those which do not by a ratio of 2 to 1. This tallies rather well with the ratio of more than 2 to 1 in the translation of the Christian Catechism by Wzzino, who was a native speaker of Maltese writing in approximately the same period. Examination of other later 18th-century records of Maltese shows an ever-increasing number of forms with regressive voicing assimilation (Avram 2022a, 2024a, 2025), in accordance with the tenets of lexical diffusion theory. Also, as can be seen in Table 5, some of the forms with regressive voicing assimilation in de Sentmenat’s Vocabulary are found in 18th-century records of Maltese by native speakers, while the expected variation is also attested:

As shown in Avram (2022a, 2024a, 2025), it is towards the end of the 18th century that regressive voicing assimilation of obstruents essentially became a rule of Maltese phonology. This is confirmed by metalinguistic comments from this period. Vassalli (1791: 80) states about the letter “Be Bb” that “ante T K Q S & Š […] sonat P” [= before T K Q S & Š … it sounds like P]. Similarly, Vassalli (1791: 90) notes the case of “Zajn Zz”, which “ante Q, T […] sonat S” [= before Q, T […] sounds like S]. Also, Vassalli (1796: 140) states about the “letter D” that “ante litteras F X K & Q legitur T” [= before the letters F X K & Q it is read T]. Consider also the more detailed remarks in the second edition of Vassalli’s grammar (Vassali 1827). About “B b” it is mentioned that “ogni qual volta incontrasi immediatamente con una delle seguenti lettere, c, k, q, ɦ, ɧ, s, щ, t, si sentirà profferire come si fosse p” [= whenever it immediately meets with one of the following letters, c, k, q, ɦ, ɧ, s, щ, t, will be heard pronounced […] as if it were p] (Vassalli 1827: 6).²⁶ Next, Vassalli (1827: 6-7) states about D d that “si sente però pronunciato to, quante volte sarà seguito da f, ɦ, ɧ, k, q” [= it is heard however pronounced t, whenever it is immediately followed by f, ɦ, ɧ, k, q]. Also discussed is the case of “the Z” which “avanti q […] sentesi pronunciare come se fosse un s” [= before q […] is heard pronounced as if it were an s]” (Vassalli 1827: 9).

To sum up, de Sentmenat’s Vocabulary captures regressive voicing assimilation while it was still spreading through the lexicon, i.e. before it became a phonological rule of Maltese. This accounts for the fact that a sizable minority of forms, as transcribed by de Sentmenat, do not exhibit regressive voicing assimilation.

8 Word-final obstruent devoicing

As in the case of regressive voicing assimilation, Thezan’s dictionary is the first record of Maltese convincingly documenting the occurrence of word-final obstruent devoicing (Avram 2023). This notwithstanding, the fact remains that forms which exhibit word-final obstruent devoicing are vastly outnumbered by those which still have voiced obstruents word-finally (see Avram 2020: 49, Table 1). The following are examples of entries in Thezan’s dictionary displaying word-final obstruent devoicing

63. 1. ﻍarp ‘sunset’ (Cassola 1992: 141v)
       cf. Ar. ġarb
    2. ﻕabes ‘[he] jumped’ (Cassola 1992: 117r)
       cf. Mod. M. qabeż
    3. aعaﭺ ‘ivory’ (Cassola 1992: 115v)
       cf. Ar. ʕāǧ

The picture that emerges upon examination of 18th-century records of Maltese preceding de Sentmenat’s Vocabulary is a rather mixed one (Avram 2017, 2020, 2024a, 2025). As shown below, the spread through the lexicon of word-final obstruent devoicing is not a smooth one, with considerable variation even in the transcriptions by native speakers.

In Mifsud’s (1739–1746) sermons, forms with word-final devoiced obstruents, illustrated in (64), by far outnumber those still having word-final voiced obstruents, which are very few in number:

64. 1. deep ‘gold’ 1739 (Ġabra tal-Malti Qadim n.d.)
       cf. Ar. ḏahab
    2. iocot ‘[he] sits’ 1741 (Ġabra tal-Malti Qadim n.d.)
       cf. Mod. M. joqgħod
    3. uic ‘face’ 1740 (Ġabra tal-Malti Qadim n.d.)
       cf. Ar. waǧh, Mod. M. wiċċ

Most of Wzzino’s (1752) relevant forms have word-final voiced obstruents. However, a number of words appear only with devoiced ones. In addition, there is variation in several forms. Illustrative examples of word-final devoiced obstruents are given below:

65. 1. itġhammet²⁷ ‘[he] is baptized’ (Wzzino 1752: 98)
       cf. Mod. M. jitgħammed
    2. ġhażis ‘beloved.M’ (Wzzino 1752: 37)
       cf. Ar. ʕazīz
    3. ḣhueieċ ‘things’ (Wzzino 1752: 50)
       cf. Mod. M. ħwejjeġ

Moving on to de Soldanis’s (1750) grammar and dialogues (Id-Djalogi ta’ de Soldanis n.d.), forms displaying word-final obstruent devoicing are again in the majority.

66. 1. l’art ‘earth’ (de Soldanis 1750: 79)
       cf. Ar. al-arḍ
    2. supperf ‘rigid.M’ (de Soldanis 1750: 194)
       cf. Mod. M. supperv ‘haughty’
    3. uec ‘face’ (Id-Djalogi ta’ de Soldanis n.d.)
       cf. Ar. waǧh, Mod. M. wiċċ

Since de Sentmenat’s Vocabulary is the major record of 18th-century Maltese written by a foreigner, it is instructive to see how it fits into the picture as far as word-final obstruent devoicing is concerned. To start with, it should be noted that forms with word-final devoiced obstruents and those with word-final voiced obstruents are almost equal in number (see Avram 2020: 49, Table 1). As shown below, word-final devoicing affects all classes of obstruents, i.e. stops (67a-d), fricatives (67e-f) and affricates (67g-h):

67. 1. demp ‘tail’ (QB 2003: 26)
       cf. Ar. ḏanab
    2. hhhenep ‘grapes’ (QB 2003: 38)
       cf. Ar. ʕinab
    3. hhhadit ‘iron’ (QB 2003: 29)
       cf. Ar. ħadīd
    4. marrit ‘sick.M’ (QB 2003: 32)
       cf. Ar. marīḍ
    5. geuss ‘nuts’ (QB 2003: 88)
       cf. Ar. ǧawz
    6. txerf ‘deer’ (QB 2003: 25)
       cf. Sic. cervu
    7. mizeuhuetx ‘married.M’ (QB 2003: 25)
       cf. Mod. M. miżżewweġ
    8. seltx ‘snow’ (QB 2003: 34)
       cf. Ar. ṯalǧ

On the other hand, as already mentioned, forms with voiced obstruents in word-final position are also found. As seen below, these include only instances of word-final voiced stops:

68. 1. hhhabib ‘friend’ (QB 2003: 21)
       cf. Ar. ḥabīb
    2. lohhhob ‘game’ (QB 2003: 31)
       cf. Ar. laʕb
    3. ard ‘earth’ (QB 2003: 40)
       cf. Ar. arḏ
    4. smid ‘semolina’ (QB 2003: 39)
       cf. Ar. samīd

Finally, the following forms have variants with a voiced and respectively a devoiced obstruent in word-final position:

69. 1. Ilhhheid ~ Ilhhheit ‘Easter’ (QB 2003: 36, 83)
       cf. Mod. M. Għid [il-Kbir]
    2. irrid ~ irrit ‘[I] want’ (QB 2003: 97, 66)
       cf. Mod. M. irrid
    3. salib ~ salip ‘cross’ (QB 2003: 102, 26)
       cf. Ar. ṣalīb
    4. thahhhàmed ~ thahhhammet ‘to be baptized’ (QB 2003: 96, 23)
       cf. Mod. M. tgħammed
    5. trid ~ trit ‘[you.SG] want’ (QB 2003: 97, 65)
       cf. Ar. turīd
    6. thhheid ~ thhheit ‘what do [you.SG] say?’ (QB 2003: 27, 66)
       cf. Mod. M. x’tgħid

De Sentmenat’s native language, Catalan, belongs to the languages in which obstruents do not contrast word-finally. Prieto (2004: 209) writes that “les consonants obstruents (és a dir, totes les oclusives, fricatives i africades) […] són sempre sordes quan es troben en posició final de mot” [= the obstruent consonants, that is to say, all the stops, fricatives and affricates […] are always voiceless when they are found in word-final position] and Wheeler (2005: 147) illustrates the word-final neutralization of voicing in all classes of obstruents. As with regressive voicing assimilation, then, this may raise reasonable doubts as to the accuracy of de Sentmenat’s forms displaying word-final obstruent devoicing. In this case again, there are no obvious differences, however, between de Sentmenat’s transcriptions and those by other contemporary authors who were native speakers of Maltese.²⁸ Consider the data set out in Table 6 below:

De Sentmenat’s transcriptions therefore appear to be accurate and to reflect a stage in the history of Maltese when there was still fluctuation between forms with word-final voiced and respectively devoiced obstruents. This accord with the fact that it is only towards the end of the 18th-century century that the phonological process of word-final obstruent devoicing is first mentioned, by Vassalli (1791). Several passages in Vassalli’s (1791) grammar are relevant in this regard: the letter Be “in fine dictionum sonat P” [= at the end of the word sounds like P] (Vassalli 1791: 80); the letter Dal “in fine dictionis T profertur” [= at the end of the word is pronounced T] (Vassalli 1791: 81); the letter Zajn “in fine dictionis sonat S” [= at the end of the word sounds like S] (Vassalli 1791: 90). The same work (Vassalli 1791: 5) also includes the first categorical formulation of the rule of word-final obstruent devoicing in Maltese “litterae B D Ǧ […] Z in fine dictionum sonos proprios amittunt & pronunciantur ac si essent P T Č […] & S” [= the letters B D Ǧ … Z at the end of words lose their own sounds and are pronounced as if they were P T Č … & S].

9 Conclusions

In spite of the inconsistent system of transcription used and of erroneous transcriptions, de Sentmenat’s Vocabulary is in many ways a useful source for the study of the historical phonology of Maltese, shedding light on a number of issues.

Consider first the Maltese reflexes of */a:/ in word-medial position. Borg (1976: 200) writes that “basing oneself on such early sources of Maltese as are available, one is inclined to conclude that historically, the monophthongal and diphthongal renderings of the high imaala vowel in Maltese are both very old”. It appears that this statement needs to be amended: the monophthongal reflex /e:/ chronologically precedes both the diphthongal one /ie/ – which becomes entrenched after 1750, as shown by evidence from de Sentmenat’s Vocabulary – and the monophthongal reflex /i:/.

The data from de Sentmenat document the occurrence of short unstressed vowels in word-initial position more than a century and a half later than claimed by Cowan (1975: 9) These data are interesting not only per se, but also when placed within the larger framework of Maghrebi dialects of Arabic. Cantineau (1960: 110) concludes that “la chute des voyelles brèves en syllable ouverte ne doit remonter dans la plupart des parlers maghrébins qu’à deux ou trois siècles”. In confirmation of Cantineau’s conclusion, this is, then, an independent development shared by Maltese with other Maghrebi Arabic dialects.

De Sentmenat provides further evidence in support of the claims in Avram (2023, 2024b: 44) that earlier Maltese resorted to three auxiliary vowels to break up word-medial consonant clusters – [ɪ], [a], and [o] – and that, in addition to the word-medial consonant-cluster type (Borg 1978: 21), vowel harmony of the vowel-copying type also plays a role in determining the quality of the auxiliary vowel. Also, de Sentmenat’s Vocabulary appears to be the last record of earlier Maltese containing forms with word-medial clusters broken up by auxiliary vowels, which are virtually non-existent in later 18th-century records of Maltese. It is therefore a terminus post quem for the disappearance of the auxiliary vowels.

De Sentmenat’s Vocabulary is one of the last sources attesting to the fact that interdental fricatives still occurred in the second half of the 18th century in some varieties of Maltese. As for regressive voicing assimilation and word-final obstruent devoicing, the evidence provided by de Sentmenat confirms the fact that both are rather late developments in the historical phonology of Maltese.

10 Abbreviations

11 References

Abela, Giovanni Francesco. 1647. Della descrittione di Malta isola nel mare siciliano. Con le sue antichita, ed altre notitie. Malta: Paolo Bonacota.

Avram, Andrei A. 2014. “The fate of the interdental fricatives in Maltese.” Romano-Arabica XIV. 19-32.

Avram, Andrei A. 2016. Evidence from onomastics for the diachrony of Maltese vowels. Mediterranean Language Review 23. 163-176.

Avram, Andrei A. 2017. “Word-final obstruent devoicing in Maltese: Inherited, internal development or contact-induced?” Academic Journal of Modern Philology 6. 23-37.

Avram, Andrei A. 2020. The diachrony of word-final obstruent devoicing in Maltese. In Slavomír Čéplö & Jaroslav Drobný (eds.), Maltese Linguistics on the Danube, 27-57. Berlin: De Gruyter Mouton.

Avram, Andrei A. 2021. Fortition in the historical phonology of Maltese: Two case studies. In Mihaela Tănase-Dogaru, Alina Mihaela Tigău & Mihaela Zamfirescu (eds.), Deconstructing Language Structure and Meaning: Studies on Syntax, Semantics, Language Acquisition, and Phonology, 198-219. Newcastle upon Tyne: Cambridge Scholars Publishing.

Avram, Andrei A. 2022a. Regressive voicing assimilation in Maltese: Diachrony and typology”. In Przemysław Turek & Julia Nintemann (eds.), Maltese. Contemporary Changes and Historical Innovations, 231-261. Berlin: De Gruyter Mouton.

Avram, Andrei A. 2022b. “From interdental fricatives to stops in Maltese: A case of lexical diffusion.” In Guram Chikovani & Zviad Tskhvediani (eds.), Studies on Arabic Dialectology and Sociolinguistics. Proceedings of the 13th AIDA International Conference, 21-33. Kutaisi: Akaki Tsereteli State University. Digital edition.

Avram, Andrei A. 2023. “Thezan’s Le regole per la lingua maltese and the historical phonology of Maltese.” Paper presented at the 8th International Conference on Maltese Linguistics, 18–20 September 2023, Universität Bremen.

Avram, Andrei A. 2024a. “Voicing and devoicing in the history of Maltese.” In Carmen Berlinches Ramos, Jairo Guerrero & Montserrat Benítez Fernández (eds.), AIDA Granada: A Pomegranate of Arabic Varieties, 35-46. Zaragoza: Prensas de la Universidad de Zaragoza.

Avram. Andrei A. 2024b. “Auxiliary vowels in word-medial consonant clusters in the history of Maltese.” Bucharest Working Papers in Linguistics XXVI (2). 29-45.

Avram, Andrei A. 2025. “Contextual neutralisation of voicing in Maltese obstruents: A historical perspective.” In Anthony J. Frendo & Kurstin Gatt (eds.), Arabic in Context: Language, Dialects, and Culture-Essays in Honour of Martin R. Zammit, 33-73. Cambridge: University of Cambridge and Open Book Publishers.

Borg, Alexander. 1976. “The imaala in Maltese.” Israel Oriental Studies VI. 191-223

Borg, Alexander. 1978. “Historical aspects of auxiliary vowels in Maltese.” Wiener Zeitschrift für die Kunde des Morgenlandes 70. 15-34.

Cachia, Lawrenz. 2000. Ħabbew l-ilsien Malti [= They loved the Maltese language]. Zabbar: Veritas Press.

Cantineau, Jean. 1960. Cours de phonétique arabe. Paris: Librairie C. Klincksieck.

Cassar, Mario. 2005. “Nella lingua nostra nativa: l-użu tal-Malti fil processi ta’ l-Inkwiżizzjoni Rumana” [= Nella lingua nostra nativa: The use of Maltese in the trials of the Roman Inquisition]. Symposia Melitensia 2. 57-84.

Cassola, Arnold. 1992. The Biblioteca Vallicelliana Regole per la lingua Maltese. Valletta, Said International.

Cassola, Arnold. 1993. “A mixed orthography of the Maltese language: The Latin-Arabic alphabet”, Collected Papers Published on the Occasion of the Collegium Melitense Quatercentenary Celebrations (1592 – 1992), 203-219. Msida: University of Malta.

Cassola, Arnold. 1996. Il “Mezzo Vocabolario” Maltese-Italiano del ’700. Valletta: Said International.

Cowan, William. 1964. “An early Maltese wordlist.” Journal of Maltese Studies 2. 217-225.

Cowan, William. 1975. Caxaros’ Cantilena: A checkpoint for change in Maltese. Journal of Maltese Studies 10. 4-10.

Ġabra tal-Malti Qadim. n.d. https://mlrs.research.um.edu.mt/resources/gabrastorika.

Hull, Geoffrey. 1994. Review of Arnold Cassola, The Biblioteca Vallicelliana: Regole per la lingua maltese. The Earliest Extant Grammar and Dictionary of the Maltese Language. Journal of Semitic Studies XXXIX (2). 391-395.

Id-Djalogi ta’ de Soldanis. n.d. http://malti.skola.edu.mt/kotba/djalogi_desoldanis.pdf.

Megiser, Hieronymus. 1606. Propugnaculum Europae. Wahrhaffte/eigentliche und ausführliche beschreibung der viel und weitberühmten Africanischen Insul Malta. Leipzig: In Verlag Henning Grossen des Jüngern.

Prieto, Pilar. 2004. Fonètica i fonologia. Els sons del català. Barcelona: Editorial UOC.

van Putten, Marijn. 2020. “The history of the Maltese short vowels.” In Slavomír Čéplö & Jaroslav Drobný (eds.), Maltese Linguistics on the Danube, 59-89. Berlin: De Gruyter Mouton.

Queraltó Bartrés, Alexandre. 2003. Un vocabulari català – maltès manuscript del segle XVIII. Edició, transcripció i estudi. Barcelona: PPU.

Skippon, Philip. 1732. An Account of a Journey Made thro’ Part of the Low-Countries, Germany, Italy, and France. London.

de Soldanis, Giovanni Pietro Francesco Agius. 1750. Nuova scuola di grammatica per agevolmente apprendere la lingua punica maltese. Rome: Salomoni.

Vassalli, Michaelis Antonii 1791. Mylsen Phoenico-Punicum sive Grammatica Melitensis. Rome: Antonio Fulgoni.

Vassalli, Mikiel Anton. 1796. Ktŷb yl Klŷm Mâlti. Mfysser byl-Latin u byt-Taljan. Lexicon Melitense-Latin-Italum. Vocabolario maltese. Recato nelle lingue latina e italiana. Rome: Antonio Fulgoni.

Vassalli, Michelantonio. 1827. Grammatica della lingua maltese. Rome: The Author.

Wheeler, Max. 2005. The Phonology of Catalan. Oxford: Oxford University Press.

Wzzino, Francesco. 1752. Taġhlim nisrani [= Christian Catechism]. Rome: Arcangelo Casaletti.