Pronunciation guide for medieval Galician-Portuguese

See my Pronunciation bibliography for the sources used in creating this guide.

I make full use of the International Phonetic Alphabet on this page. Whilst this may be unfamiliar to some readers, it is without a doubt the only way to achieve any sort of precision, and for anyone who is likely to make a habit of singing in medieval languages, it is worth investing a bit of study time so that you can then make sense of other guides and resources. To assist with this, each of the symbols in the Sound (IPA) column is linked where possible to the Wikipedia page that explains it. In all cases though, I have tried to relate the sounds of medieval Galician-Portuguese to those of modern English and Spanish (Castilian) or, where it seems useful, to other major modern European languages.

I'll start off with a few general Dos and Don'ts, many of which apply to the singing of medieval texts in any language, not just the Cantigas.

General things to avoid doing

  1. Don't take advantage of your audience's naïvety and enthusiasm. The attitude that “they will never know” and “it's just a bit of fun” has no place in an historically informed performance. A slip of the tongue is just as bad as a wrong note, and systematically poor pronunciation should no more be tolerated than an out-of-tune or badly-played instrument. So think about what you can achieve, not what you can get away with.
  2. Don't trust your own intuition and guesswork to see you through. Inform yourself.
  3. Don't rely entirely on the intuition of a modern Galician or Portuguese speaker (let alone Spanish) who has not personally studied the medieval language of the Cantigas. A lot has changed in the past 700 years.
  4. Don't assume that the concerts and recordings you hear have correct pronunciation, however prestigious the performer, however accomplished the musical delivery. Sadly, a large proportion of the recordings of the Cantigas that you will find online are riddled with bad pronunciation and garbled words. Copying their mistakes cannot be excused as ‘oral tradition’. Instead, aim to identify those mistakes and correct them.
  5. Don't take differences of opinion between published authorities on the pronunciation of the Cantigas as a licence to ignore them all. If you don't know which is right, and have no solid evidence of your own, then at least stay consistent with one of them—preferably the one that seems to you to be the best researched and cites the most sources of its own.
  6. Don't forget which language you're dealing with. Correspondences between spelling and pronunciation that you've learnt from other languages might not apply to medieval Galician-Portuguese. English speakers in particular should remember that there's no single, unified ‘foreign’ orthography out there. If you pronounce ü as in German, é as in French or Catalan, or ch as in Italian, it will not be correct for the Cantigas.

General things that you should always do

  1. Do take the time to read through the pronunciation guide for the edition you're using, whether it's this website or something else. These guides can undoubtedly be intimidating at first, and there's no way you will remember everything in one go. But you will nonetheless get a good idea of what you need to work on, and which bits to come back to when you're ready to absorb the details.
  2. Do be cautious in applying the pronunciation guidelines from one edition of the Cantigas to the spelling found in a different edition. Every editor is compelled to make a variety of choices when it comes to regularizing the spellings found in the manuscripts, with some aiming at philological or palaeographical fidelity, and others at readability and convenience for the performer. Results inevitably vary and aren't always compatible.
  3. Do approach the Cantigas de Santa Maria, first and foremost, as miracle tales. If you can begin by putting aside the music—and even the metrics—and learning to speak the written words fluidly, as a compelling story, your eventual sung performance will be immeasurably better. In particular, learn the natural stress of the words, and don't let that get completely smothered by a strict musical rhythm.
  4. Do use the metrical schemes as a check on pronunciation. This simple but immensely useful rule is so often forgotten, but if at any point you find yourself singing a line that doesn't rhyme with any other line, then it's very likely that you are doing something wrong.

And now for the specifics.

Word stress

In this edition, the position of the stress in any word of two or more syllables can be identified from the following simple rules, in descending order of precedence:

  • Words with an acute or circumflex accent have the stress on the marked syllable, e.g. ángeos [ˈaŋʤe.os], porê [poˈɾeŋ], duocima, ú [ni.ˈuŋ] apren, razô, diz
  • Words ending in a consonant other than m, n or s are stressed on the final syllable, e.g. aquel, trobar, seor, perdiz
  • Words ending in any of the ‘falling’ diphthongs ai, ei, eu, iu, oi, ou and ui, or these diphthongs plus -s, are also stressed on the final syllable, e.g. mamou [maˈmou̯], cooceu [koɲoˈʦeu̯], guariu [gu̯aˈɾiu̯], babous, sandeus, demais.
  • All other words are stressed on the syllable before last, e.g. hoge, bẽeita [beˈŋei̯.ta], frores, devemos, começaro.

Hiatus or diphthong?

When the main table below states that a particular pair of vowels can form a diphthong, this does not imply that they do so in every word. In cases where two vowels look as if they might be a diphthong but are in fact in separate syllables—i.e. they are in ‘hiatus’—the text of this edition always marks the weak vowel (i or u) with an acute accent if it is stressed (e.g. María, rúas) or a trema if it is unstressed (e.g. pïadoso, nïú, Génüa). For maximum clarity, this use of the trema is also extended to words like Gabrïél and nïente even though ie is not a diphthong in medieval Galician-Portuguese, since speakers of Spanish (for example) might easily read them incorrectly otherwise. If you are ever in any doubt—which is perhaps most likely with words like bestia that follow the rules but are unmarked—you can of course always click the button to switch on syllable bullets for the definitive answer (in this case bestia with a diphthong).

Please remember that the trema is not like a German umlaut and doesn't change the quality of the vowel sound!

The tilde

Before going into details of individual vowel sounds, it will be useful to make some common observations regarding vowels written with the tilde, i.e. ã, ĩ, õ and ũ. These only appear in my text immediately before another vowel.

The most important thing here is to recognize that the purpose of the tilde is not to indicate a nasal vowel, as might be expected from modern Portuguese usage. Rather, it indicates that the marked vowel is followed by the velar nasal consonant [ŋ], the sound of ng in English singer, or of nh in modern Galician unha (official spelling¹). So each vowel with a tilde actually represents two separate phonemes: ã = /aŋ/,  = /eŋ/, ĩ = /iŋ/, õ = /oŋ/ and ũ = /uŋ/. This nasal consonant may then cause nasalization of the preceding vowel, through regressive assimilation, such that the actual sounds are [ãŋ], [ẽŋ], [ĩŋ], [õŋ] and [ũŋ]. But the vowel nasality itself is not phonemic, only a secondary effect; there can be no nasal vowel without the nasal consonant. This is no different from what happens when a vowel is followed by syllable-final m or n: the same secondary nasalization happens automatically, though it has no phonemic value. Therefore, in all cases my phonetic transcriptions of the texts omit any explicit marking of the nasal vowel, as it is completely predictable. So for example mão is transcribed simply as [maŋo], tẽer as [teŋeɾ], davan as [daβaŋ] and ambos as [ambos], even though the natural realizations are [mãŋo], [tẽŋeɾ], [daβãŋ] and [ãmbos].

In this interpretation of the tilde I am adopting the ‘working hypothesis’ of José-Martinho Montero Santalha as described in [JMMS2002], p. 135, footnote 21, essentially to the effect that in the medieval language all vowels were phonologically oral, though they may be realized phonetically more or less nasalised through contact with a nasal consonant. This is true for modern Galician, and probably always has been; whether it is still the case for modern Portuguese, or genuine nasal vowel phonemes have developed there, is more open to debate. Either way, here is not the place for my own detailed defense of Montero Santalha's proposal, but I will just say that I find it to be the simplest and most elegant theory, and agree whole-heartedly with him that it is more consistent with the available data than alternative theories that assume complete loss of original intervocalic [n] leaving only vowel nasalization.²

For certain pairs of vowels in the manuscripts, the tilde indicates not the velar nasal [ŋ] but the palatal nasal [ɲ]. However, you do not need to worry about identifying these cases, as my spelling system always writes out the [ɲ] sound explicitly, either as the default -nn- (see below) or as -nh- / -ñ- if you have selected different spelling options. For example, reỹa → reínna / reínha / reíña and menĩez → meninnez etc. See points 13 and 14 on the Spelling page for more details.

Finally, and for good measure, I'll just state that it is never correct to read a tilde as if it were a plain n in medieval Galician-Portuguese: to pronounce mão as [mano] instead of [maŋo] is simply wrong. This is one of the most common mistakes to be heard in recordings of the cantigas, right up there with ignoring tildes altogether.

[1] Don't confuse official Galician nh [ŋ] with Portuguese nh, which has a different sound [ɲ].

[2] Contrast, for example, the trouble that John M. Lipski had to go to in [JML1975] when trying to explain the ‘insertion’ of [ŋ] into medieval ũa—taken by him, without question, to represent [ũa]—in order to give modern Galician unha [uŋa]. It is surely much simpler to assume that original intervocalic [n] became [ŋ] and was then never lost in this particular word due to its high functional load. (Portuguese didn't lose the consonant either, but changed the anomalous [ŋ] to [m] giving uma.)

Vowels

Important: Where the table below gives ‘English’ equivalents of vowels, these should be understood to refer to RP, the ‘Received Pronunciation’ of British English, and even then, they are only intended to provide the most basic orientation. Modern Galician and Spanish equivalents are much closer to the medieval Galician-Portuguese vowels, but only the IPA representation itself should really be considered to give a precise target for the sound required in each case.

Spelling Sound (IPA) Comments To be avoided
a [a] As in Spanish ama or English tap.

Do not over-lengthen into the sound of a [ɑː] in English father. Also, the vowel must keep its clear [a] quality even in unstressed syllables, in contrast to modern Portuguese, so do not let it drop to a ‘schwa’ [ə] (like the final a in English Anna).

There are two diphthongs with a as the nucleus, both containing the semivowel [i̯]:
  • ai = [ai̯] as in Spanish hay or English eye.
  • ia = [i̯a] as ia in Spanish bestia or ya in English yak
[ɑː], [ə]
ã [aŋ][ãŋ] As a followed by the velar nasal consonant [ŋ] (see The tilde above).

Do not pronounce as French an which has a different underlying vowel [ɑ̃].

See also Change 14 on the Spelling page for details of the cases where my spelling system replaces an original ã with the sequence -ann-.
[an], [aɲ], [ɑ̃]
e [e], [ɛ] Unstressed e is always pronounced as close [e], like the e in Spanish and Galician madre. Do not allow this sound to close even further to [ɪ] or [i]—Teyssier [PT1993] pp. 56–63 dates such pronunciations in Portuguese to no earlier than 17th century for pre-tonic vowels, and the 18th century in other unstressed positions.

When stressed, e can represent either close (high-mid) [e] or open (low-mid) [ɛ], depending on the Latin vowel from which it originated. However, in this edition the distinction is unambiguously marked in the texts: close [e] is always written e or ê and open [ɛ] is always written é (occasionally è). The [e] / [ɛ] contrast is also fully observed in the IPA transcriptions and the rhyme tables.

It is important to be aware that these two sounds [e] and [ɛ] are different phonemes, being as distinct from each other in native speakers' minds as either is from, say, [a] or [i]. This was true in the period of the Cantigas and remains so in modern Portuguese and Galician, as well as in other Romance languages such as Catalan and Italian: all of these languages more or less preserve the Vulgar Latin system of seven stressed vowels [i e ɛ a ɔ o u]. In Castilian Spanish, on the other hand, just five distinct vowels survive, due to [ɛ] and [ɔ] having evolved into the diphthongs [ie] and [ue].

Nevertheless, meaning rarely if ever depends solely on the distinction between [e] and [ɛ] (the contrast has what linguists call a ‘low functional yield’). Minimal pairs like poder (infinitive) and podér (future subjunctive) are very rare and would easily be disambiguated in context. So if as a performer you find the distinction too burdensome to learn and maintain, I'd suggest the following simple approach to dealing with e:
  • Native speakers of Portuguese and Galician should use their intuition: if you know the modern counterpart of a medieval word, using the corresponding choice of [e] or [ɛ] should serve you pretty well. (It won't be perfect though—both metaphony and analogy had a considerable effect on vowel height after the period of the Cantigas, and in fact the modern languages often disagree with each other.)
  • Other performers can just use whichever of [e] or [ɛ] comes more naturally according to their native language. In losing the distinction you will most likely be doing no worse than the average medieval Castilian visitor to Galicia.
There are three diphthongs with e as the nucleus:
  • ei = [ei̯] approximately as ey in Spanish rey or ay in English hay.
  • eu = [eu̯] and éu = [ɛu̯] both similar to Spanish deuda or Europa, but the first with close [e] and the second with open [ɛ]. There are no easy English equivalents, but do not let either diphthong turn into the sound of eu in Europe or ew [ju] in few.
[ɪ], [i]
[ju] for eu
[eŋ][ẽŋ] As close e [e] followed by the velar nasal consonant [ŋ] (see The tilde above).

Note that open [ɛŋ] does not occur in the language of the Cantigas (despite its presence in modern Galician): all the evidence seems to suggest that the nasalization ‘neutralized’ the open/close contrast, so anywhere you see -ẽ- (or syllable-final -en or -em) you can safely pronounce the vowel as close.

Do not pronounce as in modern French en which has a different underlying vowel [ɑ̃]. You should also resist any modern Portuguese-influenced tendency to diphthongize this sound into [ei̯ŋ] or [ẽĩ̯ŋ]. (I have seen the latter pronunciation recommended elsewhere, but do not know of any evidence for it being correct in the 13th century.)

See also Change 14 on the Spelling page for details of the cases where my spelling system replaces an original with the sequence -enn-.
[e], [en], [ɛŋ], [ɑ̃], [ẽĩ̯]
i [i] As in Spanish bicho or English machine.

One diphthong occurs with i as the nucleus:
  • iu = [iu̯]
This sound is quite common as the 3rd personal singular preterite verb ending (conjugation III), e.g. pariu, viu. Try not to turn it from a falling diphthong into a rising one: the pronunciation [i̯u] or [ju] (= English you) is technically incorrect, even when the diphthong follows another vowel as in saiu, which is [sa·iu̯] not [sa·i̯u] or [sa·ju]. Having said that, if this is the worst mistake you make, you will be doing very well indeed!
 
ĩ [iŋ][ı̃ŋ] As i followed by the velar nasal consonant [ŋ] (see The tilde above).

Do not pronounce [ɛ̃] as in standard French vin which has a quite different underlying vowel.

See also Change 13 on the Spelling page for details of the special cases where my spelling system replaces an original ĩ (or ) with the sequence -inn-.
[in], [ɛ̃]
o [o], [ɔ] Unstressed o is always pronounced as close [o], like the o in Spanish vino, Galician viño. Do not allow this sound to close even further to [u] under the influence of modern Portuguese—as with unstressed e becoming [ɪ] or [i], Teyssier [PT1993] pp. 56–63 dates this pronunciation to the 17th century at the earliest.

When stressed, o has two possible values, paralleling e above: it can represent either close (high-mid) [o] or open (low-mid) [ɔ], depending on the original Latin vowel. As with e, this contrast is fully marked in lyric texts: close [o] is always written o or ô and open [ɔ] is always written ó (occasionally ò).

Like [e] and [ɛ], [o] and [ɔ] are distinct phonemes. In this case, it is a bit easier to find minimal pairs where meaning depends clearly on making the distinction, such as the 3rd person verb forms póde [pɔde] = ‘can’ vs. pode [pode] = ‘could’; also fóra adv. ‘out’ vs. fora vb. ‘had gone’. But even these examples are rare in the scheme of things, and you should not let making the distinction become too much of a burden if it does not come naturally; there are more important things to get right in a performance of the Cantigas.

There are three diphthongs with o as the nucleus:
  • oi = [oi̯] as oy in Spanish hoy or English toy.
  • ou = [ou̯] as in English (RP) mould or the Galician city Ourense.
  • io = [i̯o] as io in Spanish ocio, nación or yo in English yoghurt.
[u]
[au] for ou
[u] for ou
õ [oŋ][õŋ], As close o followed by the velar nasal consonant [ŋ] (see The tilde above).

Like open [ɛŋ], open [ɔŋ] does not occur, so the vowel is always close where you read -õ- (or syllable-final -on or -om).

See also Change 14 on the Spelling page for details of the cases where my spelling system replaces an original õ with the sequence -onn-.
[on], [ɔŋ]
u [u] As in Spanish mucho, or intermediate in length between the vowels in English put and lunar.

One diphthong occurs with u as the nucleus:
  • ui = [ui̯]
Like iu, this is a falling diphthong, not a rising one, so try to avoid the pronunciation [u̯i] or [wi].
 
ũ [uŋ][ũŋ] As u followed by the velar nasal consonant [ŋ] (see The tilde above).

Speakers of modern Galician might like to note that their indefinite articles un and unha (in official spelling) sound exactly like their medieval ancestors un [uŋ] and ũa (or hũa) [u.ŋa].

Consonants

Where opt. appears in the first column below, this refers to options available via the Spelling preferences page.

Spelling Sound (IPA) Comments To be avoided
b [b] As in English bat or Spanish nombre.

See also v below for more on the distinction between b and v.
[β]
c + a, o, u
qu + e, i
[k] As in Spanish casa and queso, or the unaspirated c in Eng. scan. Try if possible to avoid pronunciation as the aspirated c in Eng. cat, which is followed by an extra puff of breath.

Note that the spelling qu' is used consistently in this edition to indicate the [k] sound when a final -ca or -co elides its vowel before a word beginning with e- or i-, for example riqu' e poderoso (from rico+e). This is [rike], never [rikwe] or [riku e]. Similarly, biqu' inchado is [bikinʧado].
[kh], [kw]
qua [kwa] As in Portuguese quadro, English quack or the cua- in Spanish cuando. Avoid the sound [kwɒ] as in Eng. quad or quantity.

(The sequence quo never occurs in the Cantigas.)
[ka], [kwɒ]
c + e, i
ç + any vowel
[ʦ] As ts in Eng. bits, or the German z in Zimmer.

Beware that this is not the sound of c before e or i in modern Spanish or Galician ([θ], [s] = Eng. th or s), nor Portuguese [s], and definitely not Italian [ʧ] (= Eng. ch).
[θ], [s], [ʧ]
ch [ʧ] As in Spanish chico or English chips.

Do not pronounce ch as sh [ʃ] in English ships. The latter sound is written x (see below) and the two letters must be carefully distinguished. Teyssier in [PT1993] p. 53 dates the origin of the modern Portuguese merger of ch and x to no earlier than the beginning of the 17th century, and it never occurred in Galician.
[ʃ]
d [d̪] As Spanish d in mundo or molde, or English d in dog.

I am not aware of any evidence that -d- between vowels was already pronounced in medieval Galician-Portuguese as a fricative [ð] (= Eng. th in rather), but I couldn't claim that it is wrong to do so. For most modern Galician and Spanish performers a [ð] in this position will come naturally, and therefore I would consider it quite acceptable, especially as it cannot be confused with any other sound.
 
f [f] As in Spanish fama or English fire.  
g + a, o
gu + e, i
[g] As g in Spanish galgo and guía, or English goal.

Paralleling qu' above, the spelling gu' is used consistently in this edition to indicate the hard [g] sound when a final -ga or -go elides its vowel before a word beginning with e- or i-, for example logu' entó (from logo+entó). Do not pronounce this as [logwenton] or [logu enton]. Similarly, fogu' infernal (from fogo infernal) is never [fogwinfernal].
[gw]
gua [gwa] As in Spanish lengua, or the gw in the name Gwen followed by a.

(Like quo, the sequence guo never occurs in the Cantigas.)
[ga]
g + e, i
j + any vowel
[ʤ], [ʒ] As the affricate [ʤ] in English gem or jam (never as in get), or the fricative [ʒ] in Portuguese gelo, joelho or French gîte, jouer.

Sources vary on the question of how early the sound change [ʤ] → [ʒ] happened in medieval Galician-Portuguese, so I would consider either pronunciation to be acceptable. My own preference for [ʤ] (which appears in the IPA transcriptions) is motivated by consistency with the affricate pronunciations of ch, ç (= c + e, i) and z. Either way, the important thing is to pick one of the two sounds and stick with it throughout, and not to vacillate between them.

Do not pronounce as Spanish g or j [x] in the same position (e.g. general, elige, joven), or as German j (= [j]) in jeder. Cognate words in modern standard Galician usually have x [ʃ] in the same position (e.g. gente → xente) but this sound should also be avoided here as it is a distinct phoneme in the medieval language (see x below).
[g], [j], [x], [ʃ]
h Silent, as in modern Galician, Portuguese and Spanish.

See the entries on ch, ll (opt. lh) and nn (opt. ñ, nh) for the use of this letter in digraphs.
[h]
k [k] As in Spanish kilo or English kettle.

This letter is used exactly once in all of the Cantigas de Santa Maria, in the Greek phrase kirieleison.
 
l [l] As in Spanish leche or the 'clear' l in English leek.

This sound does not change when the l is at the end of a syllable, so take care not to pronounce it as the velarized 'dark' l [ɫ] in English cool.
[ɫ]
ll
(opt. lh)
[ʎ] As the ‘standard’ (but increasingly rare) pronunciation of Spanish ll in calle or llegar, the lh in Portuguese ovelha, or the ll in Catalan lluna. More or less the same as lli in English million, especially when this word is pronounced with just two syllables.

Do not pronounce as a long or doubled l, or split the ll across two syllables. Spanish-speaking performers should avoid ‘yeísmo’ which merges ll and y into [ʝ] in the modern language, and has recently begun to encroach on modern Galician.

Warning: as of October 2012, the sound sample on the Wikipedia page Palatal lateral approximant is clearly wrong.
[l], [l·l], [ʝ]
m- syll. initial
-m syll. final before labial (bmp)
[m] As in Spanish madre and tiempo, or English mouse.  
n- syll. initial
-n syll. final before dental (dt, çz)
[n] As in Spanish nombre and quando, or English name.  
-n syll. final elsewhere and word-final
(opt. -m word-final)
[ŋ] As the ng in English sing.

In this edition, n never appears in syllable-final position before b or p (see spelling change 10), but is replaced by m to indicate the correct pronunciation.
[n]
nn
(opt. ñ, nh)
[ɲ] As ñ in Spanish señor.

Do not pronounce as a long or doubled n, or split the nn across two syllables.
[n], [nː], [n·n]
p [p] As Spanish p in padre, or the unaspirated p in Eng. span. Try to avoid pronunciation as the aspirated p in Eng. pat. [ph]
q See above under c.
r word-initial
rr any pos.
[r] As the strongly trilled (rolled) r in Spanish roble or perro.

Note that only word-initial single r has this pronunciation: other syllable-initial single r's are pronounced [ɾ] as indicated below. Do not weaken to an English approximant r sound [ɹ].
[ɾ], [ɹ]
r non-word-init. [ɾ] As the tapped r in Spanish abre or pero.

Again, do not weaken to an English approximant r sound [ɹ].
[r], [ɹ]
s word-initial
s final or before a consonant
ss any position
[s] Unvoiced, as in Spanish sabor or English sing.

Do not pronounce ss as a long or doubled s, or split it across two syllables. Also, avoid the influence of modern Portuguese and Galician dialects and do not pronounce s as [ʃ] (Eng. sh) or [ʒ] at the end of a syllable (or indeed anywhere else). According to Teyssier, [PT1993] p. 54, evidence of the latter pronunciation does not appear until the mid-18th century.
[ʃ], [ʒ], [z], [sː], [s·s]
s between vowels in word [z] Voiced, like the s in English rose or the z in zebra.
[s], [ʃ], [ʒ]
t [t̪] As Spanish t in tener, or the unaspirated t in Eng. star. Try to avoid pronunciation as the aspirated t in Eng. tar. [th]
v [β], [v] Either as the bilabial fricative v [β] in Spanish lavar, or as the labiodental fricative v [v] in Portuguese vaca, provar or English very.

(As with [ʤ] and [ʒ], choose either [β] or [v] and use it consistently.)

It's important to at least be aware that b and v represent different phonemes in the medieval language, as they do in modern Portuguese, even though in modern standard Galician they have merged into a single phoneme, just as in Spanish (Castilian). Modern Portuguese speakers will naturally pronounce b and v in the cantigas as [b] and [v], and have no difficulty keeping the phonemes separate. Native speakers of other languages that make the same distinction with the same sounds, such as English or Italian, should have no problems either.

For native speakers of modern standard Galician and Spanish, distinguishing b and v may take a bit of extra effort: for instance, when v is at the beginning of a word, or falls after an n, they will tend to pronounce it as [b], and when b falls between vowels they will tend to pronounce it as [β]. My advice would be no more than to give it a try. If you can maintain the distinction without a disproportionate effort, your pronunciation will be all the more authentic. But if it turns out to be too burdensome, put it aside and concentrate on things that are more important to performance.
[b], [w]
x [ʃ] In all positions, as x in modern Portuguese xarope or modern Galician xunta, or the sh in English ship.

There seem to be no clear-cut cases where x should be pronounced as [ks] in the Cantigas. The only possible candidates are the words Ale(i)xandria and Aleixi (= Roman emperor Alexius), but even here I would assume a nativized pronunciation of x as [ʃ] in conjunction with the nativized diphthong ei.
[ks]
z [ʣ] As dz in Eng. adze, or the ds in fads. This is the voiced counterpart of the voiceless ç (= [ʦ]) and must be carefully distinguished from it.

Beware that this letter does not have the sound of z in modern Spanish and Galician (= [θ] or [s]), nor that of English z in zebra. Also, as for s, avoid all modern Portuguese-influenced pronunciations of syllable-final z as [ʃ] or [ʒ].
[ʦ], [s], [θ], [z], [ʃ], [ʒ]