Voiceless and voiced plosives

Istanbul Judeo-Spanish has a contrast between voiceless /p t k/ and voiced /b d ɡ/. In utterance-initial position, /p t k/ are realized with somewhat longer VOT, on average, than in Peninsular (Castañeda Reference Castañeda1986) and Latin American Spanish (Williams Reference Williams1977), and sometimes are clearly aspirated (see the transcription of ‘The North Wind and the Sun’ below).Footnote ⁵ It should be noticed in this respect that in the main contact language, Turkish, in which all speakers are bilingual, prevocalic /p t k/ are aspirated (Kallestinova Reference Kallestinova2004, Göksel & Kerslake Reference Göksel and Kerslake2005: 35–36). The voiced stops /b d ɡ/ are generally realized with prevoicing in utterance-initial position.

Word-finally, we find /d/ in words of Spanish origin such as edad /eˈdad/ ‘age’, sivdad /sivˈdad/ ‘city’ (although not, for instance, in vedrá /veˈdra/ ‘truth’, where it has been lost, cf. MSp verdad), and the voiceless plosives /t/ and /k/ in borrowings: berit /beˈrit/ ‘circumcision’ (from Hebrew), pasuk /paˈsuk/ ‘religious verse’ (from Hebrew), torik /toˈrik/ ‘type of tuna’ (from Turkish).

In preconsonantal position, the contrast between voiced and voiceless plosives is neutralized. The voicing feature is determined by the following consonant: akseptar /aksepˈtar/ ‘to accept’, obtener /opteˈner/ ‘to obtain’, egzemplo /eɡˈzemplo/ ‘example’.

Labialization of labial and velar stops after /u/, as in asu[k^w]ar ‘sugar’ has been documented for Istanbul Judeo-Spanish (Quintana Rodríguez Reference Quintana Rodríguez2006: 38–39, 363, 462–472; see also Varol Reference Varol [Bornes]2008b). In the specific variety described here, this phenomenon only affects /ɡ/ [ɣ], as in djugar ~ djuguar [ʤuˈɣ^ʷar] ‘to play’, lugar ~ luguar [luˈɣ^ʷar] ‘place’.Footnote ⁶

We now turn to consider the phonetic realization of the voiced plosives /b d ɡ/ and the phonological status of quasi-homorganic voiced fricatives and approximants vis-à-vis [b d ɡ]. For the labials, there is a phonological contrast, in part inherited from Medieval Spanish, between /b/ and /v/, which is challenged at the phonetic level by the possibility of realizing both phonemes as approximants. On the other hand, for the dentals (denti-alveolars), and to a lesser extent for the velars, we find an incipient split between plosive and continuant realizations of what originally were merely allophones in complementary distribution.

Voiced labial plosive /b/: Contrast with /v/ and phonetic realization

Unlike Mainstream Spanish, Istanbul Judeo-Spanish has a phonological contrast between /b/ and /v/. This is a conservative trait, since Old Spanish also had a contrast between /b/ and, depending on the region, /β/ or /v/ (later lost in Mainstream Spanish by lenition of /b/). In word-initial position, the lexical distribution of these phonemes in words of Spanish origin is as in Old Spanish, e.g. boka ‘mouth’, bever ‘to drink’, vijitar ‘to visit’, ventana ‘window’. It is important to note that the word-initial contrast is usually maintained even after a vowel: la boka [laˈboka] ‘the mouth’ (see Figure 1) vs. la vida [laˈviða] ‘the life’.

Figure 1

la boka [laˈboka] ‘the mouth’

In intervocalic word-internal position there is also a phonemic contrast. Regarding the lexical distribution of these phonemes, however, it is clear that the Old Spanish lexical contrast has not been preserved here, unlike the situation in word-intial position, since in words of Spanish origin we find only /v/ word-internally between vowels, regardless of etymological origin. It appears that in Old Spanish (OSp) word-internal intervocalic /b/ was found in words where Latin (Lat) had -p-, e.g. OSp saber ‘to know’ < Lat sapēre and OSp cabeça ‘head’ < Lat capitia; whereas either /β/ or /v/, depending on the area, was found in words where Latin had -b- or -v-, e.g. OSp aver ‘to have’ < Lat habēre, OSp cavallo ‘horse’ < Lat caballum (see Penny Reference Penny2002: 72–74, 96–98). In Istanbul Judeo-Spanish we find /v/ in words of both origins: IJSp saver /saˈver/ ‘to know’(MSp saber /saˈber/ [saˈβer]). The phoneme /v/ is also found in onset consonant clusters palavra /paˈlavra/ ‘word’, puevlo /ˈpuevlo/ ‘people’, avlar /aˈvlar/ ‘to speak’ (MSp palabra, pueblo, hablar).

Old Spanish intervocalic /b/ has nevertheless been preserved after the prefix /a-/: abasho /aˈbaʃo/ ‘below’ (cf. /baˈʃiko/ ‘short’, MSp bajo ‘low; short’), aboltar /abolˈtar/ ‘to turn’(cf. OSp bolver ‘return’, MSp vuelta ‘turn’) (see Figure 2). The position after this productive suffix has thus been treated as word-initial for phonological purposes.

Figure 2

abasho [aˈbaʃo] ‘below’ Example of /b/ [b] in intervocalic position after a prefix.

The word-internal contrast between /b/ and /v/ has also been reinforced by borrowings containing intervocalic /b/. A minimal pair is haber ‘news’ (from Turkish) vs. haver ‘business associate’ (from Hebrew).

The two phonemes /b/ and /v/ also contrast in word-internal postconsonantal position (at least after some consonants): ambezar [ambeˈzar] ‘to learn, to teach’ vs. invyerno [iɱˈvjerno] ‘winter’; albondigás /albondiˈɡas/ [alβondiˈɣas] ‘meatballs’ vs. Kalvo [ˈkalvo] (a surname).

In addition, Istanbul Judeo-Spanish has preconsonantal /v/, on the one hand, in words like sivdad [sivˈdad] ‘city’ (OSp cibdad, MSp ciudad), kovdo [ˈkovdo] ‘elbow’ (OSp cobdo, MSp codo), devda [ˈdevda] ‘debt’, with preservation of an Old Spanish coda consonant, and, on the other hand, in words like Evropa [evˈropa] ‘Europe’, from the consonantization of labiovelar glides in falling diphthongs, an areal feature also found in Greek.

In spite of the fact that the phonemic contrast between /b/ and /v/ is stable (in the sense that speakers know which words have one phoneme and which have the other), it must be said that the phonetic distance between these two phonemes can be very small on occasion. This is because both plosive /b/ and fricative /v/ may optionally be realized as approximants in word-internal intervocalic position (and sometimes in other contexts). We notice in this respect that in Turkish, /v/ also has approximant allophones (Göksel & Kerslake Reference Göksel and Kerslake2005: 6–7). An example showing an approximant realization of both of these two phonemes, /b/ and /v/, is given in Figure 3.Footnote ⁷ The realization of /b/ as [β] in this example can be compared with the realization of the same phoneme in Figures 1 and 2, where it is realized as [b]. Notice the presence of formant structure in Figure 3 and its absence in the previous two figures.

Figure 3

Saves el haber? [ˈsaʋezelxaˈβe] ‘Do you know the news?’

Example containing both intervocalic /v/ and /b/ realized as approximants in this instance.

Voiced dental /d/: Incipient split between [d] and [ð]

Whereas in the case of /b/ and /v/ we find, to some extent, preservation of a historical contrast lost in Mainstream Spanish, an innovation in Istanbul Judeo-Spanish is a marginal phonemic split between [d] and [ð], two sounds that in Old Spanish were only allophones of a single phoneme (as they still are in Mainstream Spanish, e.g. MSp dama [ˈdama] ‘lady’, la dama [laˈðama] ‘the lady’). Unlike in Mainstream Spanish, where /d/ is consistently realized as an approximant in intervocalic position, both inside words and across word boundaries, in Istanbul Judeo-Spanish the general situation is the following: Word-internally, between vowels, we normally find [ð] (fricative or approximant), and not [d], as in sefaradí [sefaaˈði] ‘Sephardic’ (see Figure 4). Word-initially, on the other hand, we tend to find the sound [d], even after a vowel, as in el ayre del norte [eˈlaɾe delˈnort^he] ‘the north wind’ (in the transcribed text at the end of this article).

Figure 4

sefaradí [sefaaˈði] ‘Sephardic’

Example of word-internal intervocalic /d/ realized as a continuant [ð] with an intermediate degree of constriction.

A complication is that intervocalic [d] is also found after ‘strong’ morpheme boundaries, including after the prefix /a-/ (a context where, as we saw above, OSp [b] has also been maintained), as in adulsar [adulˈsar] ‘to sweeten’ (from dulse ‘sweet’), and in initial position in the suffix -dear, which is used productively to adapt borrowings from Turkish, as in boyadear [bojadeˈar] ‘to paint’ (< Turkish boyamak; see Perahya & Perahya Reference Perahya and Perahya1998: 34).Footnote ⁸ These affixes can thus be taken to introduce a phonological word boundary, in a possible morphophonological analysis. Except for these morphologically conditioned exceptions, word-internally we generally find [ð], as stated. Appeal to morphosyntactic boundaries would allow us to maintain a one-phoneme analysis, with allophones in complementary distribution where /d/ is realized as [ð] after a vowel provided that no word-boundary or strong morpheme boundary intervenes. A difficulty for this analysis is that in borrowings from Turkish, such as adá ‘island’, the consonant is consistently realized as [d], without lenition. There are also a few Romance words (where influence from French cognates may play a role), such as idea ‘idea’, in which intervocalic [d] also resists weakening to [ð] (see Perahya & Perahya Reference Perahya and Perahya1998: 34).

A second complication for the phonemic analysis of these sounds, this time against considering [d] and [ð] different phonemes, is that, in actual speech, a great degree of variation is found, even in the same context, so that the distribution of [d] and [ð] summarized above is only a matter of strong tendencies, at least for the speaker whose speech we are analyzing here. Compare the two examples of the word kada ‘each, every’, one in Figure 5, [ˈkada], where the intervocalic dental was realized as a stop, and the other in Figure 6, kada diya [ˈkaaˈðia] ‘every day’, where the same phoneme in the same word and in the same reading style was realized as almost completely vocalized. Notice, furthermore, that the word-initial consonant of diya ‘day’ was also spirantized in this example, even though, as mentioned above, word-initially the tendency is to produce [d].

Figure 5

kada [ˈkada] ‘each, every’

Example of constricted realization of intervocalic /d/ in a context where [ð] is the norm.

Figure 6

kada diya [ˈkaaˈðia] ‘every day’

Example of open realizations of intervocalic /d/. Notice that the /d/ of kada is especially open or vowellike in this example and that the word initial consonant of diya ‘day’ has also been realized as an approximant.

Intervocalic /d/ may also be realized as a flap [ɾ], causing neutralization with phonemic /r/ (although no examples of this phenomenon are included in the recordings for this article).

We may conclude that there is a quasi-phonemic, incipient, opposition between [d] and [ð]. The sounds are still in variation within the same lexical item, but there are words that resist lenition of /d/. Historical bilingualism in Greek, where /d/ and /ð/ are independent phonemes, would have increased awareness of the inherited allophony as involving an alternation between different sounds. This tendency would also have been reinforced by bilingualism in Turkish, which would have increased awareness of the difference in pronunciation between, say, kada and adá (on quasi-phonemic contrasts in Spanish, see Hualde Reference Hualde, Chand, Kelleher, Rodríguez and Schmeiser2004).Footnote ⁹

Voiced velar plosive: Incipient lexicalization of [ɡ] and [ɣ] in word-initial position

In general, the distribution of [ɡ] and [ɣ] is predictable and not contrastive. In intervocalic position, whether word-internal or word-initial, we usually find [ɣ], siguro [siˈɣuɾo] ‘safe’, dos amigos [ˈdozaˈmiɣos] ‘two friends’ (see Figure 7). Where we find a tendency to have a contrast is after pause. The contrast appears to be between words that admit variation between stop and approximant realizations in utterance-initial position, and words whose initial consonant is consistently realized as a stop in this position, as in gato [ˈɣato] ‘cat’, godro [ˈɣoðro] ‘fat’ vs. gerra [ˈɡerːa] ‘war’, with a stop. It appears that the quality of the following vowel may be a conditioning factor.

Figure 7

dos amigos [ˈdozaˈmiɣos] ‘two friends’ Example illustrating intervocalic /ɡ/ [ɣ]. Notice also phrase-initial /d/ (prevoiced stop).

Labiodental fricatives

One of the few clear non-Castilian features of Istanbul Judeo-Spanish is the preservation of Latin f- in some words like firido ‘wounded’ (MSp herido), fedor ‘stench’ (MSp hedor), fuyir ‘to flee’ (MSp huir), foya ‘hole’ (MSp hoya), forka ‘gallows’ (MSp horca), etc. In most words, however, we find loss of this consonant where it was lost (through aspiration) in Castilian, e.g. avlar ‘to speak’ (MSp hablar < Lat fabulāri), azer ‘to do, make’ (MSp hacer < Lat facere), ijo ‘son’ (MSp hijo < Lat fīlium), etc.Footnote ¹⁰ The presence of the group /fl-/ in flama ‘flame’(MSp llama) is also attributable to Eastern-Iberian (Aragonese or Catalan) influence (on the influence of Aragonese on Judeo-Spanish, see Quintana Rodríguez Reference Quintana Rodríguez2001).

For /v/ see above, section ‘Plosives and related consonants’.

Alveolar and postalveolar fricatives and affricates

As already mentioned, the most striking difference between Judeo-Spanish and both Modern Peninsular and Latin American Spanish is found in the sibilants. This is primarily because Judeo-Spanish has not undergone the devoicing of sibilants and velarization of postalveolars that affected other Spanish varieties. Regarding the lexical distribution of intervocalic IJSp /s/ and /z/, first, consider the following numbered correspondences:

Sibilants: Correspondences with Peninsular Spanish and Latin American Spanish

As shown in the examples above, in Istanbul Judeo-Spanish, the contrast between voiced and voiceless sibilants has been preserved, whereas the medieval contrast between fricatives and affricates (continued in Standard Peninsular Spanish as /s/ vs. /Ɵ/) has been lost. That is, Istanbul Judeo-Spanish has had the same evolution from the medieval Western Romance system of sibilants that we find in Portuguese and in Catalan (and in French). Interestingly, an exception to the correspondences above is found in the numerals dodje /ˈdoʤe/ ‘twelve’, tredje /ˈtreʤe/ ‘thirteen’. In these words the Judeo-Spanish of Istanbul (and Salonika, Nehama Reference Nehama1977) agrees more closely with Catalan than with Portuguese, e.g. Catalan dotze /ˈdoʣə/ vs. Portuguese doze /ˈdoze/, PenSp doce /ˈdoƟe/ (on these and other exceptions to the general correspondences, see Quintana Rodríguez Reference Quintana Rodríguez2006: 71–79).

As for the postalveolars, Judeo-Spanish has the three phonemes /ʃ/, /ʒ/ and /ʤ/ corresponding to the velar fricative /x/ of Mainstream Spanish. Old Spanish, on the other hand, had a simpler contrast between voiceless /ʃ/ and /ʒ/, which most likely had [ʤ] and [ʒ] as allophonic variants:

Postalveolar fricatives and voiced affricate: Correspondences with Old Spanish and modern Mainstream Spanish

What we find is that Old Spanish /ʒ/ continues as /ʒ/ word-medially, but has become /ʤ/ in word-initial position in Judeo-Spanish. In Old Spanish, most likely, word initially, fricative and affricate alternated in the same word, depending on the phrasal context: [ʤ]ente ‘people’, con [ʤ]ente ‘with people’, but la [ʒ]ente ‘the people’ (as in some present-day Catalan varieties).Footnote ¹¹

In Judeo-Spanish, [ʤ] and [ʒ] do not alternate. These are now independent phonemes, rather than allophones in complementary distribution. Word-initially, the affricate was generalized in all phrasal contexts, including after vowels.Footnote ¹² The phonemic split was caused by the introduction of both word-initial /ʒ/ and word-medial /ʤ/ in borrowings: jurnal /ʒurˈnal/ ‘newspaper’ (from French), kavedjí /kaveˈʤi/ ‘coffee-shop owner’ (from Turkish), with a likely later change /ʣ/ > /ʤ/ in the few native words like dodje /ˈdoʤe/ ‘twelve’ that had preserved the old affricate /ʣ/.

The affricate phoneme /ʧ/ does not present any important difference in its distribution with respect to Mainstream Spanish. In word-final position it is found only in a few words, including the old borrowing harach /xaˈraʧ/ ‘extorsion’ (historically ‘a tax paid by non-Muslims’), and recent borrowings like mach ‘match’.

The voice contrast in syllable-final and word-final fricatives tends to be neutralized. The allophonic distribution of [s] and [z] is essentially as in Catalan (see e.g. Wheeler Reference Wheeler2005: 145–149) and most likely as in Old Spanish as well. Word-internally, [s] is found before voiceless consonants, and [z] before voiced consonants. Word-finally, [s] is found before voiceless consonants and before pause, and [z] before voiced consonants and vowels (Sala Reference Sala1971, Bradley Reference Bradley, Bainbridge and Agbayani2007a, among others).

Neutralization of /s/ and /z/

• • Word-internal preconsonantal

  e[s]te ‘this’, mo[s]trar ‘to show’, di[s]persar ‘to scatter’

  mi[z]mo ‘same’, a[z]no ‘donkey’, pe[z]gado ‘heavy, boring’, e[z]vachear ‘to give up’

• • Word-final

  do[s] ‘two’, do[s] kavesa[s] ‘two heads’, ma[s] ‘more’

  do[z] gatos ‘two cats’, do[z] mansevikos ‘two boys’, do[z] ijos ‘two sons’, ma[z] agua ‘more water’

Final devoicing results in alternations like una ve[s] ‘once’, dos ve[z]es ‘twice’.

The same alternation is also found between word-final [ʃ] and [ʒ]: entende[ʃ] ‘you (pl) understand’, entende[ʒ] esto ‘you (pl) understand this’, entende[ʃ] todo ‘you (pl) understand everything’.

Nevertheless, phrase-final devoicing appears not to be completely categorical and may be lexically-conditioned to a certain extent. Before pause, some lexical items, such as nariz ‘nose’ and mesaj ‘message’, may be more likely to be realized with at least some voicing than other words.

A characteristic feature of Judeo-Spanish is the palatalization of /s/ before velars, as in moshka /ˈmoʃka/ ‘fly (n)’, bushkar /buʃˈkar/ ‘to search’, pishkado /piʃˈkado/ ‘fish’, cf. MSp mosca, buscar, pescado. This palatalization is not totally systematic, cf. eskola ‘school’. Note also the word eskenazí ‘Ashkenazi’, without palatalization. Word-finally /s/ has become /ʃ/ after a palatal glide, with absorption of the glide: seis > sesh /seʃ/ ‘six’, cantáis > kantásh /kanˈtaʃ/ ‘you (pl) sing’.

Voiceless velar fricative /x/

The voiceless velar fricative /x/ is found in borrowings from Hebrew and Turkish, including morphologically fully integrated ones, like aharvar /axarˈvar/ ‘to beat’ (Figure 8), and also in some words of Arabic origin that were current in Old Spanish such as haragan [xaɾaˈɣan] ‘lazy’ (cf. MSp haragán) and hazino ‘sick’.Footnote ¹³ In borrowings from Turkish, /x/ may be used to replace Turkish /h/ (in other borrowings of this origin /h/ is deleted).Footnote ¹⁴ Its actual place of articulation and the amount of friction with which it is produced appears to be variable, although a velar articulation seems to be most common.

Figure 8

aharvar /axarˈvar/ ‘to beat’