1. The Latin system
2. Consonantization of Latin v and i
3. Palatalization of Latin /k/ and /g/
4. Palatalization of the coronal consonants
5. Semivocalization of syllable-final velars and /-l/
6. Initial consonants and groups
7. Lenition
8. Adjustments due to vowel syncope
9. Prothetic /e/
10. Final consonants
11. Devoicing of the sibilants
12. References

1. The Latin system
To understand how the modern Spanish consonantal system came into being, one has to go back to the pronunciation of consonants in Latin. As a first approximation, it can be noted that the core Latin alphabet had fifteen consonantal letters (this excludes Greek z which occurred only in certain learnèd words that played no role in the development of the Spanish phonology). These fifteen letters are shown below using modern lower-case characters, which were not themselves used by the Romans, who used either capitals (for inscriptions) or the handwritten script known as Latin cursive (for letters, accounts and many other types of document).

b, c, d, f, g, h, l, m, n, p, q, r, s, t, x

The above consonants are all thought to have been originally pronounced, but the letter h came to be silent, as did final -m in all words other than monosyllables. The following points should also be noted:

1. The letter x represented a cluster, /ks/, rather than a single consonant.
2. The letter q had the same value as c, i.e. /k/, although it was confined in principle to the sequence qu, pronounced [kw].
3. The letter f in the Latin of north-central Iberia – the variety from which Spanish descends – probably had a bilabial articulation (symbol: [ɸ]), as opposed to the labiodental pronunciation, [f], found in many other Latin-speaking territories.

In light of the foregoing, the starting point for the Spanish consonants can be idealized as in Table 1 below.

In addition to the consonants in Table 1, we need to take into account the labial-velar and palatal semivowels [w] and [j], corresponding to the letters v and i in initial and intervocalic positions. Even though these sounds were not originally consonantal in Latin, they subsequently became so in those positions, a development which is addressed in Section 2 below.

A further point to note is that, apart from /ɸ/ and the voiced stops, the consonants shown in Table 1 routinely appeared in geminate form, i.e. doubled, as in the examples below (the transcriptions of which assume the spoken Latin vowel system described here).

/ˈgotta/ gŭttam (> gota ‘drop’)
/ˈbokka/ bŭccam (> boca ‘mouth’)
/ˈɔsso/ ŏssum (> hueso ‘bone’)
/ˈgɛmma/ gĕmmam (> yema ‘yolk’)
/kaˈballo/ caballum (> caballo ‘horse’)
/ˈanno/ annum (> año ‘year’)

Apart from /ɸ/, which was displaced by /f/ in the Late Middle Ages, all the consonants in Table 1 survive into Modern Spanish, but their distribution across the lexicon has changed in many cases. In addition, a number of new consonants have been added to the system; most obviously, the voiceless dental fricative /θ/, the alveolar trill /r/, the palato-alveolar affricate /tʃ/, the palatal nasal /ɲ/, the palatal lateral /ʎ/ and the voiceless velar fricative /x/. The emergence of these sounds, as well as the other major developments that have affected consonants over the course of the history of Spanish, is described in 2 to 11 below.

2. Consonantization of Latin v and i
In initial and intervocalic positions written v and i appear to have originally represented labial-velar and palatal semivowels respectively, viz. [w] and [j]. However, their reflexes in medieval and modern Spanish are full consonants, implying that they hardened into consonants in later spoken Latin.

2.1 Latin v > Old Spanish /β/ > Mod. Spanish /b/
In the Iberian Peninsula, the labial semivowel [w] corresponding to Latin v must have developed into a labial fricative or approximant, specifically the bilabial continuant /β/ (cf. French, where the relevant fricative is labiodental /v/, as in voisin < vecīnum). In initial position, this /β/ contrasted phonemically with the bilabial stop /b/ corresponding to Latin b. This contrast appears to have continued into Old Spanish, given that medieval scribes distinguished fairly consistently between words that began with u or v (= /β/) and words which began with b (= /b/):

uaca/vaca ‘cow’ (< vaccam)
ualle/valle ‘valley’ (< vallem)
vassura ‘rubbish’ (< versūram)
boca ‘mouth’ (< bŭccam)
bueno ‘good’ (< bŏnum)

Old Sp. auer [aˈβeɾ] ‘to have’ (< Lat. habēre [aˈbeɾe])
Old Sp. seruir [seɾˈβiɾ] ‘to serve’ (< Lat. sĕrvīre [seɾˈwiɾe])

iŏcum [ˈʝɔko] > Old Sp. iuego [ˈʒweɣo] > [ˈʃweɣo] > [ˈxweɣo] juego ‘game’
iūnctum [ˈʝunkto] > [ˈʒunto] > [ˈʃunto] > [ˈxunto] junto ‘joined’
iŭdaeum [ʝoˈdεo] > Old Sp. iudio [ʒuˈð̞io] > [ʃuˈð̞io] > [xuˈð̞io] judío ‘jew(ish)’

In intervocalic position, the phonetic correlate of i appears to have been a geminate, perhaps [jj] to begin with but subsequently /ʝʝ/. For example, māium (accusative singular of māius ‘May’) probably came to be pronounced [ˈmaʝʝo]. This geminate /ʝʝ/ usually has simple /ʝ/ as its reflex in modern Spanish, a result that implies simplification as part of the generalized process of lenition or ‘weakening’ which, according to Penny (2002:74), began ‘in the last centuries of the Empire and [continued] through the Dark Ages’:

māium [ˈmajjo] > [ˈmaʝʝo] > [ˈmaʝo] mayo ‘May’

3. Palatalization of Latin syllable-initial /k/ and /g/
In syllable-initial position, the Latin velars /k/ and /g/ each underwent phonemic split, yielding different phonemes in medieval and modern Spanish according as the following segment was a front vowel or a non-front vowel. When followed by a front vowel, both /k/ and /g/ were palatalized, /k/ in this context being one of the main sources of /θ/ in modern Castilian Spanish and /g/ usually resulting in /ʝ/ or disappearing altogether, but occasionally having /θ/ as its reflex. In contrast, when the following sound was a back vowel or the central vowel /a/, /k/ generally survives as a velar consonant (albeit voiced in some contexts), while /g/ either survives intact or it was lost altogether.

3.1 Voiceless /k/
The voiceless velar stop /k/ corresponds mainly to written c in Latin, but it is represented by the letter q in the unit qu (= [kw]) and by the digraph ch in certain Greek loanwords.

Before front vowels
Before front vowels, Latin /k/ is likely to have had a fronted or palatal articulation, which can tentatively be represented as [c] (Menéndez Pidal (1987) uses the ad hoc symbol ć). In the International Phonetic Alphabet, [c] stands for the voiceless palatal stop, the sound which occurs, for example, at the beginning of modern French qui [ci] ‘who’. This sound can be used to reconstruct the pronunciation, in later spoken Latin, of words such as the following:

cĭrcā [ˈceɾka] (> Sp. cerca ‘near’)
pācem [ˈpace] (> Sp. paz ‘peace’)
caelum [ˈcεlo] (> Sp. cielo ‘sky’)
VL vĭcīnum (for vīcīnum) [βeˈcino] (> Sp. vecino ‘neighbour’)

It should also be noted that when the front vowel following Latin /k/ (realized as [c]) was unstressed and preceded another vowel, it evolved into the palatal semivowel [j], following the loss of the hiatus:

ērīcium [e.ˈɾi.ce.o] > [eˈɾicjo] (> Sp. erizo ‘hedgehog’)
bracchium [ˈbɾac.ce.o] > [ˈbɾaccjo] (> Sp. brazo ‘arm’)

The foregoing palatal allophone [c] of /k/, with or without a following [j], gives the same result in Old Spanish as the sequence [tj], suggesting merger of all three items in the late spoken Latin of central northern Iberia. Probably, spoken Latin [c] when not followed by [j] developed an epenthetic [j], so that the c of pācem, for example, came to be pronounced identically to the ci of words like ērīcium; that is, as [cj]. This latter sequence is auditorily very similar to [tj] and, consequently, the two appear to have been analysed by speakers (in central northern Iberia) as being instances of the same phonetic string, viz. [tj]. The [j] in this string subsequently assibilated, creating the voiceless dental or denti-alveolar affricate /ts/ which was a characteristic feature of many varieties of early Romance.

In the Spanish case, /ts/ split into a voiceless phoneme (= /ts/), often written as ç, and a voiced one /dz/, written as z. In principle, the voiceless item /ts/ was found in the following contexts: in initial and postconsonantal positions; in words like peçes (< pĭscēs), where the Latin source had been postconsonantal but the relevant consonant (usually /s/) had subsequently been lost; and in words like braço (bracchium), where the original Latin [c] had had a geminate occurrence. Conversely, /dz/ was found in intervocalic contexts and can thus be assumed to be the reflex of an earlier /ts/ which became voiced as part of the series of changes known collectively as lenition. Illustrative examples are given below:

cerca/çerca [ˈtseɾka] ‘near’ (< cĭrcā [ˈceɾka])
peçes [ˈpetses] ‘fish’ (< pĭscēs [ˈpesces]
uezino [βeˈdzino] ‘neighbour’ (< VL vĭcīnum [βeˈcino])

Note that, given the above assumption of a three-way merger of [c], [cj] and [tj] into [tj], with the [j] element in the latter string subsequently assibilating, the [j] of Latin [cj] sequences is not expected to survive into Old Spanish. This result is illustrated by the two examples below:

bracchium [ˈbɾaccjo] > braço [ˈbɾatso] ‘arm’
ērīcium [eˈɾicjo] > erizo [eˈɾidzo] ‘hedgehog’

At first glance, the word laçio ‘lank’ appears to be an exception to the foregoing principle. However, in view of the phonetic shape of the Latin etymon flaccĭ(d)um, this particular case is slightly different from the two examples just given. As can be seen, the Latin vowel from which the [j] of laçio developed, viz. ĭ (= /e/), was originally preconsonantal and hence could not semivocalize to [j] until after its following consonant had been lost. Now this loss was a product of the series of changes known collectively as lenition, which took place some time after the formation of /ts/. Accordingly, by the time the [j] of laçio came into existence, the original Latin [c] had long since evolved into /ts/, implying that the [j] (= i) in this word was not implicated in the formation of its immediately preceding /ts/. The overall derivation is shown below:

flaccĭ(d)um [ˈɸlaccedo] > [ˈʎattsedo] > laçio [ˈlatsjo] ‘lank’

In the late Middle Ages, the two dental affricates /ts/ and /dz/ became deaffricated, evolving over the course of the fifteenth century into the dental fricatives /θ/ and /ð/ (for a full discussion of these developments, see Mackenzie 2022). Then, over the course of the sixteenth century, the voiced member of this latter pair was lost in the standard language, although it appears to have survived as late as the twentieth century in certain rural dialects, notably in Extremadura. For example, writing in the early twentieth century, Menéndez Pidal (1987: 116) reported that the dialect used in certain villages in the north of Cáceres province distinguished between /θ/ and /ð/. Thus while the word cazar ‘to hunt’, for example, was pronounced [kaˈθal], the word hacer ‘to do’ was pronounced [haˈðel], a pronunciation captured locally by the spelling jadel. Apart from these localized relics, however, /ð/ only occurs in the modern language as a phonologically conditioned allophone of /θ/. The overall diachronic process culminating in /θ/ is schematized in Table 2 below.

Examples illustrating the transition from Old to Modern Spanish are as follows:

Old Sp. calça [ˈkaltsa] > Mod. Sp. [ˈkalθa] calza ‘stocking’
Old Sp. peçes [ˈpetses] > Mod. Sp. [ˈpeθes] peces ‘fish’
Old Sp. laçio [ˈlatsjo] > Mod. Sp. [ˈlaθjo] lacio ‘lank’
Old Sp. uezino [βeˈdzino] > [βeˈðino] > Mod. Sp. [beˈθino] vecino ‘neighbour’

In much of Andalusia, the deaffricated reflexes of earlier /ts/ and /dz/ merged with and displaced the fricative sibilants /s/ and /z/, which stemmed from Latin /ss/ (as in passum ‘step’) and Latin /s/ (as in casam ‘cottage’) respectively (/s/ having become voiced in intervocalic position, as in Old Spanish [ˈkaza] casa ‘house’). As in Castilian Spanish, the voiced member of the pair /θ/ and /ð/ was subsequently lost, leaving just /θ/ where previously four sounds had existed, viz. /ts, dz, s, z/. In the relevant Andalusian varieties, associated with Huelva, Seville, Cadiz and Malaga provinces, the words casa ‘house’ and vecino ‘neighbour’, for example, are both pronounced with /θ/, i.e. as [ˈkaθa] and [beˈθino]. This phenomenon, characterized by the use of the non-sibilant /θ/ in contexts in which the standard language has the sibilant /s/, is known as ceceo.

In the related phenomenon of seseo, found in Latin America, Seville City and those Andalusian provinces that abut New Castile, the sibilant /s/ occurs in place of the non-sibilant /θ/ that is characteristic of ceceo. The /s/ of seseante dialects is typically a laminal (denti-alveolar) sound, transcribable as [s̪], unlike the Castilian /s/, which is apical. This laminal or denti-alveolar /s/ is essentially the same sound as the /s/ of modern French, found in words like façade and poisson ‘fish’. In dialects that exhibit seseo, words which in medieval Spanish were spelled with ç or z, have /s̪/ in the modern language rather than /θ/, as do words which were spelled with ss or s:

Old Sp. calça [ˈkaltsa] > [ˈkals̪a] calza ‘stocking’
Old Sp. uezino [βeˈdzino] > [beˈs̪ino] vecino ‘neighbour’
Old Sp. passo [ˈpaso] > [ˈpas̪o] paso ‘step’
Old Sp. casa [ˈkaza] > [ˈkas̪a] casa ‘house’

Before non-front vowels
In contrast to the foregoing developments, if /k/ was followed by a back vowel or by /a/ in Latin, it survives as a velar into modern Spanish but lenites to /g/ unless it is preceded by a consonant:

arcum [ˈaɾko] > arco ‘arch’
acūtum [aˈkuto] > [aˈgudo] > [aˈɣuð̞o] agudo ‘sharp’
aquam [ˈakwa] > [ˈagwa] > [ˈaɣwa] agua ‘water’

3.2 Voiced /g/
Like its voiceless counterpart /k/, the voiced velar phoneme /g/ must also have had a palatal articulation before front vowels. The palatal articulation of /g/ is represented here using the IPA symbol [ɟ], which stands for the voiced palatal stop, the sound that occurs, for example, in modern French [ɟi] gui ‘mistletoe’. When not preceded by a consonant, this [ɟ] < Latin /g/ either has the palatal fricative /ʝ/ as its reflex in modern Spanish or it is lost altogether. The details according to context are given below.

Intervocalic [ɟ]
In intervocalic position, the /ʝ/ outcome is found in words in which the vowel immediately following [ɟ] was semivocalized to [j] in spoken Latin, due to loss of the hiatus. The survival in this context of a residue of [ɟ] suggests that the spoken Latin sequence [ɟj] developed into a geminate which was subsequently simplified:

fāgeam ‘of beechwood’ [ˈhaɟja] > [ˈhaʝʝa] > [ˈaʝa] haya ‘beech tree’

magĭstrum [maˈɟestɾo] > [maˈʝestɾo] > [maˈestɾo] maestro ‘teacher/master’

cŏrrĭgiam ‘shoelace’ [koɾˈɾeɟja] > [koɾˈɾeʝʝa] > [koˈreʝa] > [koˈrea] correa ‘leather strap’

gĕrmānum [ɟeɾˈmano] > [eɾˈmano] hermano ‘brother’
gypsum [ˈɟepso] > [ˈɟʝeso] yeso ‘plaster’

gĕmmam [ˈɟɛmma] > [ˈɟjemma] > [ˈɟʝema] yema ‘yolk’

gĭngīvam [ɟeɲˈɟiβa] > [enˈdzia] Old Sp. enzia > [enˈθia] encía ‘gum’ ([enˈs̪ia] in L.A.)
argillam [aɾˈɟilla] > [aɾˈdziʎa] Old Sp. arzilla > [aɾˈθiʎa] arcilla ‘clay’ ([aɾˈs̪iʝa] in L.A.)

gaudium [ˈgawdjo] > [ˈgotso] > [ˈgodzo] > [ˈgoθo] gozo ‘joy’
lēgālem [leˈgale] > [leˈɣal] > [leˈal] leal ‘loyal’
negāre [neˈgaɾe] > [neˈɣaɾ] negar ‘to deny’

4. Palatalization of coronal consonants
With the exception of /ɾ/ and /s/ the Latin coronal consonants were all palatalized when they occurred immediately before [j], corresponding to an earlier front vowel in hiatus.

4.1 /t/ and /d/
In the case of the stops /t/ and /d/, the outcome in medieval Spanish of the palatalization process is identical with the outcome of the palatalization of /k/ and /g/. Accordingly, both [tj] and [dj] yield /ts/ after a consonant, while in the intervocalic context [tj] yields /dz/ and [dj] yields /ʝ/, the latter presumably resulting from the simplification of an earlier geminate /ʝʝ/:

Old Sp. caçar [kaˈtsaɾ] ‘to hunt’ (< Lat. captiāre [kapˈtjaɾe])
Old Sp. março [ˈmaɾtso] ‘March’ (< Lat. martium [ˈmaɾtjo])
Old Sp. orçuelo [oɾˈtswelo] ‘stye’ (< Lat. hordeŏlum [oɾˈdjɔlo])
Old Sp. pozo [ˈpodzo] ‘well’ (< Lat. pŭteum [ˈpotjo])
Old Sp. razon [raˈdzon] ‘reason’ (< Lat. ratiōnem [ɾaˈtjone])
Old Sp. poyo [ˈpoʝo] ‘stone bench/ledge’ (< Latin pŏdium [ˈpɔdjo])

As was discussed in 3.1 above, medieval Spanish /ts/ and /dz/ were later deaffricated and then merged to produce modern Castilian Spanish /θ/:

Old Sp. caçar [kaˈtsaɾ] ‘to hunt’ > mod. Sp. [kaˈθaɾ] cazar
Old Sp. março [ˈmaɾtso] ‘March’ > mod. Sp. [ˈmaɾθo] cazar
Old Sp. orçuelo [oɾˈtswelo] ‘stye’ > mod. Sp. [oɾˈθwelo] orzuelo
Old Sp. pozo [ˈpodzo] ‘well’ > [ˈpoðo] > mod. Sp. [ˈpoθo] pozo
Old Sp. razon [raˈdzon] ‘reason’ > [raˈðon] > mod. Sp. [raˈθon] razón

As was noted in 3.1, in Andalusian and Latin American Spanish, the deaffricated reflexes of /ts/ and /dz/ displaced the alveolar fricatives /s/ and /z/ which descended from Latin /ss/ and /s/, with the result that those dialects now have just a single phoneme in place of the two phonemes /θ/ and /s/ exhibited by Castilian Spanish. In Latin America and northern Andalusia, this single phoneme is typically pronounced as a laminal or denti-alveolar sibilant [s̪], while in much of southern Andalusia it is realized as the non-sibilant [θ].

4.2 The sonorants
Turning now to the sonorants /n/ and /l/, palatalization in this case produced the palatal nasal /ɲ/ and the voiced palato-alveolar fricative /ʒ/ in medieval Spanish, the latter sound stemming from an earlier palatal lateral /ʎ/ (which in fact is retained in Portuguese words such as alho ‘garlic’, mulher ‘woman’ etc.). The palatal nasal /ɲ/ continues unmodified into modern Spanish, while /ʒ/ was first of all devoiced to /ʃ/ and later had its place of articulation retracted to the velar area, producing modern Spanish /x/:

arāneam [aˈɾanja] > [aˈɾaɲa] araña ‘spider’
ālium [ˈaljo] > [ˈaʎo] > Old Sp. aio [ˈaʒo] > [ˈaʃo] > [ˈaxo] ajo ‘garlic’
fīlium [ˈhiljo] > [ˈhiʎo] > Old Sp. fijo [ˈhiʒo] > [ˈiʃo] > [ˈixo] hijo ‘son’

The development of early medieval /ʎ/ in the above, intervocalic position into /ʒ/ implies convergence with the reflex of Latin initial i before back vowels, which also delivers /ʒ/ in Old Spanish. Given that this latter /ʒ/ was originally a mid-palatal, presumably /ʝ/, this convergence is analogous to the modern phenomenon of žeísmo which exists in River Plate Spanish. The essence of žeísmo is a historical merger of the Castilian phonemes /ʎ/ and /ʝ/, but with /ʒ/ as the result, rather than the /ʝ/ which has this role in so-called yeísta dialects.

The derivation of modern cuchara ‘spoon’ (< cochleāre), offers a variant on the ‘[lj] to /ʎ/’ change, given that in this case a voiceless stop, viz. /k/, was positioned immediately before the lateral in the Latin etymon. Presumably the voiceless property of this /k/ caused the following lateral, once palatalized, to evolve directly to /ʃ/ rather than /ʒ/. The [kʃ] sequence must then have coalesced into the palato-alveolar affricate /tʃ/:

cochleāre [koˈkljaɾe] > [kuˈkʎaɾe] > [kuˈkʃaɾe] > [kuˈtʃaɾ] Old Sp. cuchar, later cuchara (due to gender hypercharacterization).

(The change /o/ to /u/ in the first syllable is due to metaphony, induced by the [j] of the [lj] sequence before this latter item coalesced into /ʎ/.)

5. Semivocalization of syllable-final velars and /-l/

5.1 Syllable-final velars
In syllable-final position (i.e. before any consonant other than /ɾ/), the Latin velar consonants were semivocalized to [j], the latter sound then becoming the trigger for palatalization of the following consonant. Thus the /k/ or /g/ at the beginning of the groups represented in the orthography by -ct-, -x-, -c(V)l-, -g(V)l- and -gn- is assumed to have been weakened to [j]. This [j] must then have exercised an assimilatory effect on the consonant it preceded, as the latter’s place of articulation was generally retracted to the palatal or palato-alveolar region, with [j] itself being lost in the process. These developments are illustrated in Table 3 below.

Old Sp. [aˈβeʒa] abeja/abeia > [aˈβeʃa] > [aˈβexa] abeja
Old Sp. [ˈdiʃe] dixe > [ˈdixe] dije

Note that the palatalization process just described failed to occur if the ‘[-j] + consonant’ sequence resulting from semivocalization was itself followed by a consonant, this case usually arising through intertonic vowel syncope:

pĕct(i)nem [ˈpεktine] > [ˈpejtne] > [ˈpejne] peine ‘comb’

The same failure to palatalize is also apparent when the ‘[-j] + consonant’ sequence was word-final, as in the example below:

sĕx [sεks] > [sejs] seis ‘six’

5.2 Syllable final /-l/
The /-l/ of the Latin group represented by -ŭlt- was in all probably articulated as a ‘dark l’. This is the kind of l that occurs in syllable-final position in most varieties of British English – e.g. in feel or belt – and its phonetic symbol is [ɫ]. What gives the sound [ɫ] its distinctive ‘dark’ quality is its secondary, velar articulation. Due to this velar feature, Latin /-l/ in the -ŭlt- sequence developed in the same way as the syllable-final velar consonants considered in 5.1. That is, it semivocalized and then caused the following consonant to palatalize:

cŭltĕllum [koɫˈtεllo] > [kojˈtεllo] > [kuˈtʃjeʎo] > [kuˈtʃiʎo] cuchillo ‘knife’
mŭltum [ˈmoɫto] > [ˈmojto] > [ˈmutʃo] mucho ‘much’

As with the velar consonants discussed in 5.1, the palatalizing effect of syllable-final [-j] (< /-l/) on the following consonant was generally blocked if the consonant in question was itself followed by a consonant:

vŭlt(u)rem [ˈβoɫtɾe] > [ˈβojtɾe] > [ˈβujtɾe] > [ˈbwitɾe] buitre ‘vulture’

In word-internal positions, the reflex of Latin f is the bilabial approximant /β/, which reflects voicing as part of the generalized lenition process. This can be seen, for example, in prōfectum > [pɾoˈβetʃo] provecho ‘benefit’.

This latter fact is quite suggestive, as Latin f is likely to have originally been a voiceless bilabial sound, viz. /ɸ/, which which came to be replaced in the Latin of Rome (and other areas) by labiodental /f/. Conceivably, then, the (often conservative) Latin of the area in which Spanish emerged may have retained an archaic, bilabial pronunciation of f. Under that hypothesis, the labiodental sound /f/ that now occurs word-initially in Spanish in words like fuerte and frío would reflect medieval adjustment to an original /ɸ/ that was inherited directly from the local dialect of Latin. Many regional varieties of modern Spanish, e.g. Andean Spanish, actually use [ɸ] in place of [f], and it is not implausible to suppose that these non-standard articulations are simply the continuation of an earlier, more generalized situation.

Given the dual outcome in Spanish of Latin f in initial position, the underlying phoneme /ɸ/ presumably had (or came to have) two distinct articulations when it occurred at the beginning of a word, viz. [ɸ] before liquids or the semivowel [w], and [h] before all other sounds. Under this assumption, pronunciations such as the following can be reconstructed for north Iberian Latin, at a relatively late period:

[ˈɸɾito] frīctum ‘fried’
[ˈɸweɾte] fŏrtem ‘strong’
[haˈɾina] farīnam ‘flour’
[ˈhoɾno] fŭrnum ‘oven’

The reason why the allophones were distributed in the specific way implied by the examples above is impossible to identify with certainty at this historical distance. One possible explanation (see Penny 2002: 92) is that the [h] allophone arose before the labial vowels /o/ and /u/, due to dissimilation, and subsequently generalized, coming to appear before any syllabic vowel.

Modulo the palatalization of the specific sequence [ɸl] into /ʎ/, the situation just described can be assumed to have persisted into Old Spanish, the letter f continuing to be used to represent the underlying phoneme regardless of its realization. This latter point can be seen from Old Spanish spellings such as frio ‘cold’, fijo ‘son’ and fazer ‘to do/make’, which make no distinction between the allophones [ɸ] and [h].

In the Late Middle Ages, the bilabial allophone [ɸ] was apparently modified to [f], a development which may be linked to an influx of French speakers that occurred in the twelfth and thirteenth centuries (see Penny 2002: 92–93). This turn of events resulted in a situation in which, on the one hand, words such as farina and fijo continued to be pronounced with an initial [h], while on the other hand words like fuerte and frío were now pronounced with initial [f].

In addition, from the Middle Ages onwards, learnèd words in which [f-] appeared before sounds other than [ɾ] and [w] were borrowed into Old Spanish from Latin, Old French and Occitan. The accumulation of such words in the Spanish lexicon ultimately brought [f] and [h] into contrast in initial position, resulting in a phonemic split. For example, the homographs forma ‘form’ and forma ‘shoemaker’s last’ constituted a minimal pair, being pronounced [ˈfoɾma] and [ˈhoɾma] respectively (phonemically: /ˈfoRma/ versus /ˈhoRma/).

Until the 15th century both [f-] and [h-] were written as f. However, with the emergence of the phonemic distinction just alluded to, Spanish orthography began to reserve f for the first sound and h for the second. On the other hand, by the 16th century, /h/ was on the way to being eliminated from the phonemic inventory of Castilian Spanish, although the letter h was retained in the spelling system (and remains there, a curious anomaly in an otherwise largely phonemic system).

The developments overall can be exemplified as follows:

fŏrtem [ˈɸɔɾte] > [ˈɸweɾte] > [ˈfweɾte] fuerte ‘strong’
fab(u)lārī/e [haˈβlaɾe] > [haˈβlaɾ] > [aˈβlaɾ] hablar (previously fablar) ‘to speak’

6.2 Latin initial r
Word-initial /ɾ-/ became trilled in all cases, analogously to the intervocalic geminate /-ɾɾ-/:

rēgem [ˈɾege] > [rey] rey ‘king’
rŭptum [ˈɾopto] > [ˈroto] roto ‘broken’

plānum [ˈpʎano] > [ˈʎano] llano ‘flat’
flammam [ˈɸʎama] > [ˈʎama] llama ‘flame’
clāvem [ˈkʎawe] > [ˈʎaβe] llave ‘key’

There are however some exceptions to this pattern, such as plateam > plaza ‘town square’, flōrem > flor ‘flower’ and clavīc(u)lam > clavija ‘peg’. These may or may not reflect learnèd adjustment during the Middle Ages.

The development of the clusters /pl/, /ɸl/ and /kl/ in postconsonantal position is analogous to their evolution in initial position, in the sense that the final outcome is also a (broadly) palatal sound, viz. the voiceless affricate /tʃ/:

amplum > ancho ‘wide’
inflāre > hinchar ‘to swell’
masc(u)lum > macho ‘male’

Given the palatal outcome, it seems reasonable to assume that the /l/ in these word-internal clusters was realized as [ʎ], as it was when the cluster came at the beginning of a word. The voiceless nature of the reflex can then be attributed to the clusters’ postconsonantal occurrence, while, for the postnasal context at least, the affrication is arguably predictable, given that palatal phonemes in Spanish are always affricated after a nasal.

6.4 Initial Latin qu-
The sequence [kw-] (corresponding to written qu-) has quite a high frequency in the lexical stock of Latin but it was generally reduced to [k] before the Old Spanish period whenever it was followed by a vowel other than /a/:

quōmo(d)o [ˈkwomo] > [ˈkomo] como ‘how’
quaerō [ˈkwεɾo] > [ˈkεɾo] > [ˈkjeɾo] quiero ‘I want’
quīnd(e)cim [ˈkwind(e)ce] > Old Sp. quinze [ˈkindze] > [ˈkinθe] quince ‘fifteen’

In the last two examples, the loss of [w] exposes the initial /k/ to a following front vowel, viz. /ε/ or /i/. Given that in these cases the initial /k/ self evidently did not develop into /ts/ (and ultimately /θ/), as it normally did before a front vowel, it can be surmised that the change whereby Latin /k/ was assibilated to [ts] before front vowels had come to an end by the time [w] was lost in the qu sequence.

Before /a/ the reflex of initial [kw-] varies between [kw] and [k]:

VL quattor (for CL quattŭŏr) [ˈkwattoɾ] > [ˈkwatɾo] cuatro ‘four’
VL quattord(e)cim [kwatˈtoɾd(e)ce] > Old Sp. catorze [kaˈtoɾdze] > [kaˈtoɾθe] catorce ‘fourteen’

cŭppam [ˈkoppa] > [ˈkopa] copa ‘cup/goblet’
gŭttam [ˈgotta] > [ˈgota] gota ‘drop’
bŭccam [ˈbokka] > [ˈboka] boca ‘mouth’
massam [ˈmassa] > [ˈmasa] masa ‘dough’

ĭpse [ˈepse] > [ˈesse] > [ˈese] ese ‘that’
ŭrsum [ˈoɾso] > [ˈosso] > [ˈoso] oso ‘bear’
captāre [kapˈtaɾe] > [katˈtaɾe] > [kaˈtaɾ] catar ‘taste/look at’

flammam [ˈɸlamma] > [ˈʎama] llama ‘flame’
pŭllum [ˈpollo] > [ˈpoʎo] pollo ‘chicken’
annum [ˈanno] > [ˈaɲo] año ‘year’
tĕrram [ˈtεɾɾa] > [ˈtjera] tierra ‘land’

This development also affects the Latin groups /mb/ and /mn/, which, due to assimilation, merged with the geminates /mm/ and /nn/ respectively:

lŭmbum [ˈlombo] > [ˈlommo] > [ˈlomo] lomo ‘back’
dŏm(i)num [ˈdɔmno] > [ˈdɔnno] > [dweɲo] dueño ‘owner’

7.3 Simple voiceless consonants
Among the voiceless consonants, the effect of lenition was voicing, although in the case of /s/, such voicing was subsequently reversed as part of the generalized devoicing of the sibilants.

Intervocalic examples are as follows:

casam [ˈkasa] > [ˈkaza] > [ˈkasa] casa ‘house’
prōfectum [pɾoˈɸekto] > [pɾoˈβetʃo] provecho ‘benefit’
prātum [ˈpɾato] > [ˈpɾado] > [ˈprað̞o] prado ‘meadow’
acūtum [aˈkuto] > [aˈgudo] > [aˈɣuð̞o] agudo ‘sharp’
lŭpum [ˈlopo] > [ˈlobo] > [ˈloβo] lobo ‘wolf’

Pre-liquid examples include the following:

aprīcum [aˈpɾiko] > [aˈbɾigo] > [aˈβɾiɣo] abrigo ‘shelter’
dŭplum [ˈdoplo] > [ˈdoblo] > [ˈdoβle] doble ‘double’
patrem [ˈpatɾe] > [ˈpadɾe] > [ˈpað̞ɾe] padre ‘father’
sŏcrum [ˈsɔkɾo] > [ˈswegɾo] > [ˈsweɣɾo] suegro ‘father-in-law’

As can be seen from the examples above, when lenition produced a voiced stop, this was as a rule approximantized: Modern Spanish disallows voiced stops in precisely those positions in which the diachronic process of lenition occurred. In the case of /b/ < Latin /p/, approximantization can be assumed to have occurred relatively late, given that the Old Spanish reflexes of words which in Latin had intervocalic /p/ were consistently spelled with -b- rather than -v-, the latter (in all probability) designating the approximant articulation [β] and the former the stop [b].

A further point to note is that lenition predates the generalized loss of the intertonic vowel (although not the initial, phonologically conditioned phase which produced clusters such as /kl/ and /gl/). Syncope of the intertonic vowel often caused an originally intervocalic consonant to become adjacent to a consonant, a position in which a voiceless consonant would not be expected to become voiced (except in the pre-liquid case alluded to above). Accordingly, in those cases in which an originally voiceless consonant became voiced, despite ending up in pre- or post-consonantal position due to the loss of an adjacent vowel, the voicing must have occurred prior to the loss of the vowel. This is illustrated by examples such as the following:

bŏn(ĭ)tātem [bon(e)ˈtate] > [bon(e)ˈdade] > [bonˈdað̞] bondad ‘goodness’
dŏmīn(ĭ)cum [doˈmin(e)ko] > [doˈmin(e)go] > [doˈmiŋgo] domingo ‘sunday’
VL. pŏrtāt(ĭ)cum [poɾˈtat(e)ko] > [poɾˈtad(e)go] > [poɾˈtaðɣo] portazgo ‘toll’ (previously portadgo)

7.4 Simple voiced consonants
As regards those (simple) consonants that were already voiced in Latin, only the obstruents were affected by lenition, specifically the stops /b/, /d/ and /g/. Like the voiced stops produced by the lenition process itself, these were all approximantized in intervocalic and pre-liquid positions, with subsequent loss being common in the case of /d/ and /g/:

habēre [aˈbeɾe] > [aˈβeɾ] haber ‘have’ (auxiliary)
quadrāgĭntā [kwadɾaˈɟenta] > [kwað̞ɾaˈʝenta] > [kwaɾaˈenta] > [kwaˈɾenta] cuarenta ‘forty’
quadrum [ˈkwadɾo] > [ˈkwað̞ɾo] cuadro ‘picture/square’
rēgālem [ɾeˈgale] > [reˈɣal] > [reˈal] real ‘royal’
pĭgrĭtiam [peˈgɾetja] > [peˈɣɾedza] > [peˈɾeða] > [peˈɾeθa] pereza ‘laziness’
nĭgrum [ˈnegɾo] > [ˈneɣɾo] negro ‘black’

8. Adjustments due to vowel syncope
The elision of Latin intertonic vowels in many cases resulted in sequences of phonetic segments that were (or came to be) disallowed by either the phonotactics or the allophonic rules of Old Spanish. A variety of developments took place to regularize these cases. To fully understand these, it needs to be borne in mind that the generalized loss of the intertonic vowel occurred after rather than before the consonantal voicing referred to in Section 7.3.

Often the loss of an intertonic vowel necessitated merely a change in the place of articulation of a nasal (e.g. sēm(i)tam > senda ‘path’), in accordance with the principle that pre-consonantal nasals assimilate to the following consonant. Equally predictable are (i) the simplification of geminate sonorants when they became pre-consonantal (e.g. gall(i)cum > galgo ‘racing dog’) and (ii) the trilling of [ɾ] > [r] after a nasal, as in hon(ō)rāre [onoˈɾaɾe] > [onˈraɾ] honrar ‘to honour’.

Other developments have more of an ad hoc character, being specific to particular words or groups of words. Some of the commoner processes are described below.

In the first place, groups of three consonants were reduced to two in a variety of ways:

cŏmp(u)tāre > contar ‘to count’
hŏsp(i)tālem > hostal ‘boarding house’

Secondly, there was frequent dissimilation (the converse of assimilation) between nasals, usually resulting in a nasal becoming a lateral or /ɾ/:

an(i)mam > [ˈanma] > [ˈalma ] alma ‘soul’
sang(ui)nem > [ˈsaŋgne] sangne > [ˈsaŋgɾe] sangre ‘blood’

Thirdly, a so-called epenthetic consonant was often inserted between a nasal and a following /ɾ/ (representing an alternative solution to the trilling illustrated by the hon(ō)rāre case above):

hŏm(i)nem > [ˈomne] omne > [ˈomɾe] omre > [ˈombɾe] hombre ‘man’

Fourthly, two consonants were sometimes transposed in a word (a phenomenon known as metathesis):

cat(ē)nātum > [kað̞ˈnað̞o] > [kanˈdað̞o] candado ‘padlock’

Finally, syllable-final /b/ (realized as [β]) was generally semivocalized to [w], although this latter sound was later lost through assimilation if it followed a back vowel (see the codo example below):

dēb(i)ta > [ˈdeβð̞a] debda > [ˈdewð̞a] deuda ‘debt’
cap(i)tālem > [kaβˈð̞al] cabdal > [kawˈð̞al] caudal ‘money, river flow’
cŭb(i)tum > [ˈkoβð̞o] cobdo > [ˈkowð̞o] > [ˈkoð̞o] codo ‘elbow’

An analogous development is found with /-l/ when this consonant came after /a/, as can be seen from the historical trajectory of the words for ‘willow’ (sauce) and ‘river bed’ (cauce):

sal(i)cem > [ˈsaldze] salze > [ˈsawðe] > [ˈsawθe] sauce ‘willow’
cal(i)cem > [ˈkaldze] calze > [ˈkawðe] > [ˈkawθe] cauce ‘river channel’

9. Prothetic /e/
A well-known feature of Spanish phonotactics is a rule that forbids the sequence /s/ + consonant at the beginning of a word. Similar rules operate or have in the past operated in other Romance languages and the tendency seems to predate the Middle Ages. In terms of etymology, the Spanish rule against word-initial /s/ + consonant has meant that a so-called prothetic [e] has been added to the beginning of those words which in Latin began with the prohibited sequence:

stāre > estar ‘to be’
spērāre > esperar ‘to wait’

The term prosthetic can also be used in this sense, although ‘prothetic’, without the s, is more common in linguistics.

10. Final Consonants
Of the consonants that in Latin appeared in final position only /-l/, /-s/ and /-n/ survive into Spanish. Latin /-ɾ/ was often transposed with the preceding vowel, as in ĭnter > entre ‘between’, quattŭŏr > cuatro ‘four’ etc. Final /t/ was eliminated in Vulgar Latin, except as the marker of the 3rd person singular in verbs, in which role it survived until about the 12th century. The remaining Latin final consonants had been lost by the time Spanish began to written down. Apocope of final /e/ in the Middle Ages has ensured that some Spanish words end in consonants other than the /-l/, /-s/ and /-n/ that were directly inherited from Latin.
 

11. Devoicing of the sibilants
The existence of both a voiceless and a voiced series of sibilants is a salient feature of Old Spanish and one that is firmly reflected in the old orthography. In initial and intervocalic positions, the graphs ç (or c when followed by a front vowel), ss and x represented the voiceless series, viz. dental /ts/, alveolar /s/ and palato-alveolar /ʃ/ respectively. The graphs z, s and j/i (or g when followed by a front vowel) represented their voiced counterparts /dz/, /z/ and /ʒ/. This system is illustrated in the table below.

In final and preconsonantal positions, the above voicing distinction appears not to have been operative. In those positions, the sibilant subsystem was represented orthographically by just three graphs, viz. z (dental), s (alveolar) and x (palato-alveolar).

From the beginning of the sixteenth century, the voiced sounds illustrated in Table 4 – the dentals by now having been deaffricated – were lost in northern Iberian dialects and this innovation spread southwards, becoming the Peninsular norm by the beginning of the seventeenth century (for possible causes of this change, see Rost Bagudanch 2022). The much quoted observation below provides a nice snapshot of the variability of the time, suggesting that the devoicing isogloss was then located somewhere between Toledo and the southern frontier of Old Castile:

Los de Castilla la Vieja dizen hacer y en Toledo hazer, y dizen xugar, y en Toledo dizen jugar.
‘The Old Castilians say hacer and in Toldeo they say hazer, and they say xugar and in Toldeo it’s jugar.’
(Fray Juan de Córdoba, Arte en lengua zapoteca, 1578)

In contrast to what happened in Iberian and Latin American Spanish, the devoicing of the sibilants did not occur in Judaeo-Spanish. For the latter variety had become separated from developments in the Iberian Peninsula as a consequence of the Expulsion of the Jews in 1492. Thus the modern Judaeo-Spanish phonology retains four of the sounds shown in Table 4, viz. /s/, /z/ /ʃ/ and /ʒ/. Contemporary varieties apparently do not exhibit a distinctive reflex of the dental series (originally /ts/ and /dz/), which presumably merged with the alveolar series comprised by /s/ and /z/, as also happened in Andalusian and Latin American dialects of Spanish. Interestingly, some varieties of Judaeo-Spanish appear to have manifested a residue of the old dental affricates until relatively recently, given that Menéndez Pidal, writing in the early twentieth century, made the following observation:

Hoy los judíos españoles de Bulgaria y Marruecos conservan restos de la pronunciación africada en la sonora de ciertas palabras como podsu ‘pozo’, tedsu ‘tieso’ (con s sonora).
‘Today the Sephardic Jews in Bulgaria and Morocco conserve remains of the affricated pronunciation in the voiced consonant of certain words such as podsu (for pozo ‘well’) and tedsu (for tieso ‘stiff’) with voiced s.’
(Menéndez Pidal 1987: 112).

As regards the tedsu case, the affrication does not come from any of the usual sources, such as a palatalized velar or the [tj] of spoken Latin. The Latin etymon is in fact tēnsum, which gives teso in Old Spanish generally. Presumably the affricated articulation reported by Menéndez Pidal in the specified Judaeo-Spanish varieties reflects epenthesis of a dental or alveolar stop between the etymological /n/ and the following /s/, as indeed can happen in the English form of this word, tense being capabable of being homophonous with tents.

12. References