Go to previous page Go to main page Go to next page

Putting It All Together

It is now time to learn how the Greek script relates to the Greek sounds.


Even though Greek is an almost WYSIWYR language (in the sense that the sound of each letter or combination of letters can almost always be determined in a unique manner), there is not a one-to-one correspondence between the letters and the sounds: some Greek letters have more than one associated sounds (the appropriate one being, in each case, determined by the context) and some letter combinations produce different sounds than one would expect (if one were to pronounce the constituent letters with their "normal" values).

Letter Values

The "normal" (primary and {secondary}) values of the letters of the Greek alphabet are:

Letter Value Remarks
Α/α *a*
Β/β *v*
Γ/γ *γ* {*γ*, *n*, **}
*γ* only before *e* and *i*; see note on assimilation
** before velars (*k*, *g*, *γ*, *χ*); see note on velar nasal
*n* before palatals (*k*, *g*, *γ*, *χ*); see note on assimilation
Δ/δ *δ*
Ε/ε *e*
Ζ/ζ *z*
Η/η *i*
Θ/θ *θ*
Ι/ι *i* {*γ*} for *γ* see note on consonantisation (also for its effect on preceding consonants)
Κ/κ *k* {*k*} *k* only before *e* and *i*; see note on assimilation
Λ/λ *l*
Μ/μ *m*
Ν/ν *n*
Ξ/ξ *ks*
Ο/ο *o*
Π/π *p*
Ρ/ρ *r*
Σ/σ, ς *s* {*z*} for *z* see note on assimilation
Τ/τ *t*
Υ/υ *i* {*v*, *f*} *v* and *f* only when preceded by Α/α or Ε/ε or Η/η; see note on digraphs
Φ/φ *f*
Χ/χ *χ* (*χ*) *χ* only before *e* and *i*; see note on assimilation
Ψ/ψ *ps*
Ω/ω *o*


Some letter combinations represent a sound different from the sound one would expect. Since these are always represented by an ensemble of two letters, they are referred to as digraphs.

The following vowel digraphs have since antiquity been termed "δίφθογγοι|diphthongs, i.e., double sounds" even though at least some of them already stood for monophthongs (cf., e.g., ALLE87, p. 177-178).

DigraphΑΙ, αιΕΙ, ειΟΙ, οιΥΙ, υι ΟΥ, ου ΑΥ, αυΕΥ, ευΗΥ, ηυ
*i* *u* *av*, *af**ev*, *ef**iv*, *if*

There are also six consonant digraphs:

Value*b*, *mb* {*mp*}*d*, *nd* {*nt*}*g*, *g* {*k*, *g*, *ng*, *nk*}
*g* {*γ*, **, *ng*} *ts* *dz*


A particular case of digraphs are the "doubled" consonants, namely pairs of identical consonants. We have already seen one doubled consonant, ΓΓ/γγ; this, however, is not a true doubled consonant, since the first Γ/γ is always a nasal (** or *n*) and the second a non-nasal velar or palatal (*g*, *γ*, *g* or *γ*) and the "doubling" of the same consonant is a mere "optical illusion". All other consonants, save for the double consonants Ξ/ξ and Ψ/ψ, may feature in some words as (truly) doubled (in writing), without, however, having any effect on pronunciation. In other words, the second of a pair of identical consonants is not pronounced. In linguistic terms, there is no gemination in (at least mainstream) Greek. Thus, θάλασσα|sea=*θálasa*qalassa, εκκλησία|church=*eklisía*ekklhsia, Σάββατο|Sabbath, Saturday=*sávato*, αλληγορία|allegory=*aliγoría*, πρόγραμμα|program(me)=*próγrama*, etc. This also extends to combinations of (optically) different, but homophonous letters; for instance, ευφορία|euphoria and εφορία|tax office sound the same (=*eforía*eforia).


The accent corresponds to a raise of the intensity of the voice; in other words, the accented vowel (or its syllable) is "stressed", i.e., pronounced louder than the other vowels (if any) of the same word, exactly like English, Spanish, German, etc.; hence, accent in Greek is stress accent.


Theory: Diphthongs are usually the result of trying to accommodate two neighbouring vowels in one syllable. So, whenever two vowels (that is, vowel sounds) are found next to each other in a word, they may either be pronounced separately (that is, in separate syllables, e.g., koala≈*ko-á-la*) or together (that is, in the same syllable, e.g., Beirut≈*bei-rút*); the former manner is called diaeresis or hiatus and the result of the latter is a diphthong. Although in theory any two vowels can be pronounced diphthongally, in practice diphthongs almost always involve one close vowel (*i* or *u*), e.g., clown≈*kláun*, voice≈*vóis*, etc.I use ALPAG representations, instead of IPA, as I do not intend to provide all possible "flavours" of the English vowels. A consequence is, of course, that the provided pronunciations are only approximations.

Greek practice: In general, Greek does not have any diphthongs, at least in the strict sense (that the vowels of a diphthong always belong to the same syllable). When a word comprises two consecutive vowel sounds, it is up to the speaker, whether they will be pronounced in the same or in two syllables. This is more evident in poetry, where two consecutive vowels may be pronounced together or separately depending on the meter. In the following iambic verse, the same word ("καίει|he/she/it burns") is pronounced as one syllable (*kéi*kaii) in the first hemistich and as two syllables (*ké-i*kaiei) in the second without having a different meaning, both pronunciations being equally right.

Greek: Αγάπη καίει τα σωθικά, μου καίει την καρδιά μου
ALPAG: *a-γá-pi-kéi-ta-só-θi-ká|mu-ké-i-tín-kar-δγá-mu* In the phonetic representation, the dashes ("-") separate the syllables and the vertical bar ("|") separates the hemistichs; the stress on the syllables *só* and *tín*, which are unaccented in normal speech, is only observed during recitation according to the metric pattern; diphtongs are represented as connected digraphs (*éi*).
English: Love burns (my) inside, it burns my heart.

The vowel pairs that are almost always pronounced as diphthongs are *ai* and *oi* (but only when their second constituent, *i*, is unstressed) and to a lesser extend *ei* (as in the previous example). The diphthongs may be either stressed at their first constituent (e.g., νεράιδα|fairy=*neráiδa*, κορόιδο|sucker, mug=*koróiδo*) or not stressed at all (e.g., καημός|grief, longing=*kaimós*kahmos, αηδόνι|nightingale=*aiδóni*ahdoni, οϊμέ|woe to me! alas!=*oimé*oime). There are a few cases, in which the first constituent is accented and the vowel pair can only be pronounced as a diphthong because of the rule of "trisyllabotony"; typical examples (from SIAM88, §2,273, p. 389) are: χάιδεμα|caress, stroke=*χáiδema*xaidema, γάιδαρος|donkey=*γáiδaros*gaidaros, κορόιδεψα|I fooled=*koróiδepsa*koroideya.


There are several phonological phenomena in Greek that affect the pronunciation under particular circumstances. In principle, learning the values of the letters and digraphs described above is enough for an acceptable pronunciation. However, to achieve relative fluency one needs to master all aspects of the pronunciation, the most subtle (as described here) being the most important.


Assimilation is a phenomenon where a particular sound within a word affects its neighbouring sounds, in the sense that at least one property of the neighbouring sound is altered to match the same property of the particular sound. The reason for it is a need to alleviate discontinuities that occur inside words, due to juxtaposition of dissimilar sounds; it thus involves the modification of one of the dissimilar sounds to resemble the other one, so as to achieve a smoother transition.

A typical example of assimilation in English is the prefix "con-", which, apart from its usual form (e.g., "contract"), is also encountered as "col-" (e.g., "collaborate"), "cor-" (e.g., "corrupt") and "com-" (e.g., "compassion") depending on the following letter being "l", "r" or a labial respectively; in the first two cases (con→col, cor), the manner of articulation (liquid) spreads from the following letter ("l", "r") to the preceding ("n"); in the last case (con→com), the place of articulation (labial) is "diffused" from the following letter ("p") to the preceding ("n"), thus converting "n" (=nasality@teeth/gums) to "m" (=nasality@lips)A similar assimilation, which is concealed by orthography, is the conversion of "n" to the velar nasal (which is also denoted by "n") before a velar consonant (e.g., in "congress", the "n" does not have the dental/alveolar sound of "n" in "contact", but rather that of "n" in "finger").. I will refer to the first kind as assimilation of manner and to the second as assimilation of place. A third kind is assimilation of voice, wherein the voicing of one sound spreads to its neighbouring one(s), a typical example in English being the pronunciation of the plural suffix "-s" as voiced (e.g., "boys", "pads") or voiceless (e.g., "cats", "caps"), depending on the voicing of the preceding sound.

When the property of the affecting sound travels backwards to affect a preceding sound, we define this as regressive assimilation (as is the case in the "con-" examples above), whereas the transmission of the property forwards to affect a following sound is referred to as progressive assimilation (as is the case in the "-s" examples above).

Assimilation of Voice

ΜΠ/μπ, ΝΤ/ντ, ΓΚ/γκ
We have already seen a case of progressive assimilation of voice in the consonant clusters ΜΠ/μπ, ΝΤ/ντ, ΓΚ/γκ, wherein the voice of the nasal (Μ/μ, Ν/ν, Γ/γ) affects the voicing of the following stop (Π/π→*b*, Τ/τ→*d*, Κ/κ→*g*/*g*).

The pronunciation of ΤΖ/τζ as *dz* could be seen as a case of regressive assimilation of voice (namely the voice of Ζ/ζ affecting the voicing of Τ/τ), but in this case Τ/τ is not an original sound that undergoes assimilation (as is the case of "n" in "con-"), but it is a mere orthographic convention to use the digraph ΤΖ/τζ for representing the affricate *dz*.

A clear case of regressive assimilation of voice is the pronunciation of Σ/σ, which is usually pronounced as *s*, but before the voiced consonants Β/β, Γ/γ, Δ/δ, Μ/μ (or, equivalently, the voiced sounds *v*,*γ*, * γ*, *δ*, *m*) it is voiced to *z*; thus, σβήνω|I erase=*zvíno*sbhnw, σγουρός|curly=*zγurós*sgouros, προσδοκία|expectation=*prozdokía*prosdokia, κόσμος|cosmos, world, universe=*kózmos*kosmos. For the remaining voiced sounds, the behaviour of Σ/σ is not consistent. Σ/σ is not to be found before *b*, *g*, *g*, *d*, *z*, *dz* or **At least I cannot think of any word that contains a combination of Σ/σ with one of those consonants; if there is one, then Σ/σ should be pronounced as *z*.. In front of vowels, which are by default (if not by definition) voiced, Σ/σ retains its voicelessness (unlike other languages, wherein intervocalic "s" is voiced to /z/; cf. "causal", "casa"[ITA], "Kaserne"[GER]); thus, πάσα|pass (n.)=*pása*pasa, σήμα|sign, signal=*síma*shma. Σ/σ is rarely found before *n*, *n*, *r*, *l*, *l*, and when it does, it generally remains voiceless: the cases where this happens are (almost?) always foreign words or compound words, where the prefix ends in Σ/σ and the stem begins with one of the aforementioned consonants: Ισλάμ|Islam=*islám*islam, δυσνόητος|abstruse=*δisnóitos*dusnohtos, δυσλεξία|dyslexia=*δisleksía*duslejia, εισροή|inflow, inrush=*isroí*eisroh (but Ισραήλ|Israel=*izraíl*israhl).I cannot think of any examples with *n* or *l*, but, if there are any, they should follow those with *n* and *l*. However, I have also heard people pronounce Σ/σ as voiced in those cases, which means that pronunciation as *zn*, *zl*, *zr* is also acceptable. In brief, the above are summarised in the following table:

Voicing of Σ/σ
σβ σγ σδ σμ σν σρ σλ σV (V=vowel) σC (C=voiceless consonant: θ, κ, π, τ, φ, χ)
*zv* *zγ* *zδ* *zm* *sn* {*zn*} *sr* {*zr*} *sl {*zl*}* *sV* *sC*

A strange case of possible assimilation of voice is the pronunciation of the consonant cluster ΣΠΡ/σπρ in the verb σπρώχνω|I push(*sproχno*sprwxvw) and its derivatives (e.g., σπρωξιά|push (n.), σπρωξίδι|jostle (n.)), which in some parts of Greece is pronounced as *zbróχno* (voicing of both Σ/σ and Π/π). The origin of this voicing is obscureIt is unlikely that the voice comes from the last consonant of the cluster, Ρ/ρ, since it does not affect other similar clusters, such as ΣΚΡ/σκρ and ΣΤΡ/στρ, either word-initially or word-internally. It should also be noted that σπρώχνω is the only purely Greek word that begins with this cluster; hence, a possible source of the voicing could be the influence of a proclitic (e.g., τον σπρώχνω|I push him) or assimilation of another particle's voiced sound (e.g., μην σπρώχνεις|do not push), whence it could have spread to all forms of the verb. Equally obscure is also its etymology (the "official" etymology from ancient προωθώ with "deployment of pre-positional σ" sounds too far-fetched), which could also shed light upon this strange issue (if it were, for example, (ε)μπρος+ωθώ(?)→σμπρο+... and then weird choice of orthography in the Middle Ages). and, since it is purely idiomaticAs a Macedonian, I also grew up thinking that the voiced pronunciation was the "right" and only one, not even noticing the alternative pronunciation used, e.g., on TV, until an Athenian friend of mine objected to this "peculiarity" and drew my attention to the very orthography, at which time I became aware of the different way the word was pronounced by many people around me. and not mainstream, it would not be worth even mentioning, if this epichoric pronunciation were not presented as a feature of standard Greek.

AY/αυ, ΕΥ/ευ, ΗΥ/ηυ
Another case of regressive assimilation of voice is the above mentioned pronunciation of Υ/υ in AY/αυ, ΕΥ/ευ, ΗΥ/ηυ as *v* or *f* depending on the following sound being voiced (vowel, nasal, liquid, *b*, *v*, *d*, *δ*, *g*, *γ*, *g*, *γ*, *z*, *dz*) or voiceless (*p*, *f*, *t*, *θ*, *k*, *χ*, *k*, *χ*, *s*, *ts*, Ø) respectively.

Voicing of XY/xυ (X/x: Α/α, Ε/ε, Η/η)
xυ+ α, β, γ, δ, ε, ζ, η, ι, λ, μ, ν, ο, ρ, ωθ, κ, ξ, π, σ, τ, φ, χ, ψ
= *xf*

Assimilation of Place

This involves mainly the pronunciation of the velars and the nasals depending on the following sound (hence, assimilation of place in Greek is always progressive).

When any of the velar sounds (*k*, *g*, *χ*, *γ*) precedes a front vowel (*e* or *i*), it is converted to the corresponding palatal (*k*, *g*, *χ*, *γ*) or palatalised.Note that this concerns only the pronunciation since the orthography remains the same for both velars and palatals (allophony).The reason is simply that it is difficult to place the (inflexible mid-back part of the) tongue at two different points (velum for *k*, *g*, *χ*, *γ* and palate for *e*, *i*) within the same syllable, as already explained before. Thus, κοίτα|look!=*kíta*koita, γύρνα|turn!=*γírna*gurna, γκέτο|ghetto=*géto*, αρχαίος|ancient=*arχéos*arxaios, etc.

Palatalisation of the velars before *i* (when followed by a consonant C) and *e*
κιC, κηC, κυC, κειC, κοιC→*ki* κε, και→*ke*
γιC, γηC, γυC, γειC, γοιC→*γi* γε, γαι→*γe*
χιC, χηC, χυC, χειC, χοιC→*χi* χε, χαι→*χe*
γκιC, γκηC, γκυC, γκειC, γκοιC→*gi*, *ngi* γκε, γκαι→*ge*, *nge*
nasal+velar+*i*(followed by consonant C)/*e*When a palatalised velar is preceded by a nasal (and this is always a case of the velar nasal ** rendered as Γ/γ), the nasal is also palatalised (to *n*). As pre-nasalised *k* and *γ* are rendered as ΓΚ/γκ and ΓΓ/γγ respectively, the pronunciation is the same as for these digraphs, as presented above. Note also that the digraph ΓΚ/γκ may also stand for both the velar (*g* or *g*) alone or for nasal+velar (*g* or *ng*).
γγιC, γγηC, γγυC,γγειC, γγοιC→*i*, *ngi* γγε, γγαι→*e*, *nge*
γχιC, γχηC, γχυC, γχειC, γχοιC→*i* γχε, γχαι→*e*

It is often maintained (cf. also ALLE87, p. 140, for the alleged pronunciation of Greek in the 15th-16th century AD) that *n* and *l* undergo a similar palatalisation (*n*+*i*→*ni*, *l*+*i*→*li*) in front of *i* (ALLE87 alleges that this is so even in front of *e*). This is only true when another vowel follows the *i* in the same syllable (a phenomenon described later), but when *i* is the only vowel in the syllable (and, hence, is not followed by another vowel) such a pronunciation is purely idiomaticThe palatalisation of *n* and *l* in the syllables *ni* and *li* is often described as a feature of Peloponnesian speech (particularly of the area around Patras), but I have encountered persons from all over Greece, from Crete to Thrace, Central Macedonia and Western Continental Greece, that have this tendency. Unlike the palatalisation of the velars, there is no physiological reason for the conversion of the dental/alveolar *n* and *l* to palatals, because the mid-front part of the tongue is more flexible and capable of accommodating a dental/alveolar consonant and a palat vowel in succession. My personal suspicion used to be the Slavic influence, but this is more of a hunch than an opinion substantiated by actual evidence and, even though it might explain the phenomenon in the Peloponnese and Central Macedonia (regions of historical presence of Slavs), it cannot be easily extrapolated to the other regions. (as regards *i*, because *e* alone has no effect on *n* and *l*) and not representative of mainstream Greek, wherein *ni* and *li* are only possible if *n* or *l* are followed by two consecutive *i*'s in the same syllable (e.g., νιοί|(idiomatic form of νέοι=)young (pl. nomin.)=*ní*nioi, παλιοί|old, ancient (pl. nomin.)=*palí*palioi) and are very rare.

The assimilation of place of the nasals is similar to the aforementioned conversion con→com before the labials. In principle, the only nasals allowed before any of the basic consonants, are those of the same place of articulation (thus, only *m* before *p*, *b*, *f*, *v*; only *n* before *t*, *d*, *θ*, *δ*; only **, rendered as Γ/γ, before *k*, *g*, *χ*, *γ*). Word-internal phonotactics have taken care of this conversion in Greek original words, which is reflected in the orthography, e.g. αμφορά|amphora=*amforá*amfora, Κόρινθος|Corinth=*kórinθos*korinqos, άγχος|stress, anxiety=*áχos*agxos, etc. The actual conversion is more evident when words come together to form compound words, wherein if the last letter of the first constituent is a nasal (the only possibility being Ν/ν, since the only consonants that are allowed to end Greek words are Ν/ν and Σ/ς), it assimilates to the first letter of the second constituent: συν|with, together+μορφώνομαι|I shape/form myself=συμμορφώνομαι|I conform (*sin*+*morfónome*=*simorfónome*summorfwnomai), εν|in+γράφω|I scribe, I write=εγγράφω|I inscribe (*en*+*γráfo*=*eγráfo*eggrafw). This assimilation also applies to Ψ/ψ, ΤΣ/τσ and Ξ/ξ, which comprise *p*, *t*, *k* as their first constituent and are, therefore, only preceded by *m*, *n* and ** respectively (thus, συμψηφίζω|compensate?sumyhfizw).

Assimilation of (final) Ν/ν before labials and velars
ν+π→μπ ν+β→μβ ν+φ→μφ ν+μ→μμ ν+ψ→μψ
ν+κ→γκ ν+γ→γγ ν+χ→γχ (ν+ξ→γξ)This is the theory, but in practice it is doubtful whether this is anywhere attested; the only example I have found (συν+ξύλο→σύξυλος) is clearly an exception!

Assimilation of Manner

The only case of assimilation of manner that I can think of is similar to the aforementioned one (con→col, con→cor) for English. Thus, when a prefix ending in Ν/ν (and this is mainly, if not exclusively, either εν or συν) is affixed to a word starting with liquid (Ρ/ρ or Λ/λ), the prefix's final Ν/ν is converted to that letter; thus συν|with, together+λαβή|grip=συλλαβή|syllable, συν|with, together+ροή|flux, flow (n.)=συρροή|conflux, confluence. As the converted-to sounds are continuants and since there is no gemination in Greek, the final Ν/ν appears to be lost in the pronunciation. The final Ν/ν of the prefix συν is assimilated also before the sibilant continuants Σ/σ and Ζ/ζ, but orthography is peculiar: Ν/ν is converted to Σ/σ only if it is followed by Σ/σ+vowel, e.g., συν|with, together+σωρεύω|heap (v.)=συσσωρεύω|accumulate, pile upsusswreuw, otherwise it is dropped, e.g., συν|with, together+στροφή|turn (v.), rotation=συστροφή|contortionsustrofh, συν|with, together+ζω|live=συζώ|live together, cohabitsuzw.

Assimilation of (final) Ν/ν before liquids and sibilants
ν+λ→λλ ν+ρ→ρρ συν+σV→συσσV συν+σC→συσC συν+ζ→συζ


Dissimilation is the reverse phenomenon of assimilation, wherein two neighbouring sounds that share the same property are differentiated with respect to this property. In Greek, the property involved in dissimilation is that of sustainability of sound, namely a conversion of a stop to a continuant or vice versa.

voiceless stop+voiceless stop→voiceless continuant+voiceless stop: this mainly involves the clusters ΠΤ/πτ and ΚΤ/κτ, which are "dissimilated" to ΦΤ/φτ and ΧΤ/χτ respectively, e.g., επτά|seven(=*eptá*epta)→εφτά(=*eftá*efta), οκτώ|eight(=*októ*oktw)→οχτώ(=*oχtó*oxtw).

voiceless continuant+voiceless continuant→voiceless continuant+voiceless stop: this is the counterpart of the previous conversion and involves the clusters ΦΘ/φθ and ΧΘ/χθ, which are "dissimilated" to ΦΤ/φτ and ΧΤ/χτ respectively, e.g., φθάνω|I reach(=*fθáno*)→φτάνω(=*ftáno*), χθες|yesterday(=*χθés*)→χτες(=*χtés*).

Σ/σ+voiceless continuant→Σ/σ+voiceless stop: this is chiefly a question of ΣΘ/σθ and ΣΧ/σχ being converted to ΣΤ/στ and ΣΚ/σκ respectively, e.g., αισθάνομαι|I feel(=*esθánome*aisqanomai)→αιστάνομαι(=*estánome*aistanomai), άσχημος|ugly (masc. nomin.)(=*ásχimos*asxhmos)→άσκημος(=*áskimos*askhmos); I cannot identify any case of ΣΦ/σφ being dissimilated to ΣΠ/σπ.The reverse transformation (ΣΠ/σπ→ΣΦ/σφ) seems to have taken place in the case of σπόγγος|sponge→σφουγγάρι(=*sfugári*).

voiceless continuant+Σ/σ→voiceless stop+Σ/σ: etymologically, none of the voiceless continuants, Φ/φ, Θ/θ, Χ/χ, precedes Σ/σ in a Greek word; so, this is exclusively a question of Υ/υ being used with its consonantal value (which is always *f* before Σ/σ), namely ΑΥΣ/αυσ→ΑΨ/αψ, ΕΥΣ/ευσ→ΕΨ/εψ, e.g., έκαυσα|I burned(=*ékafsa*ekausa)→έκαψα(=*ékapsa*ekaya), πίστευσα|I believed(=*pístefsa*pisteusa)→πίστεψα(=*pístepsa*pisteya).

nasal+voiced continuant→nasal+voiced stop: the clusters ΜΒ/μβ and ΝΔ/νδ may be dissimilated to ΜΠ/μπ (*mb*) and ΝΤ/ντ (*nd*) respectively, particularly if Ρ/ρ is following, e.g., άνδρας|man(=*ánδras*)→άντρας(=*ándras*), γαμβρός|groom(=*γamvrós*)→γαμπρός(=*γambrós*); as already seen, their velar counterpart, ΓΓ/γγ, is almost always dissimilated (*g* instead of *γ*).

Even though dissimilation has been at times described as a phonetic lawcf. BLAS82, p. 88, "weder Spirans mit Spirans noch Tenuis mit Tenuis ist der [neugriechischen] Sprache gemäß. Ebenso verträgt das σ weder vor noch nach sich gern die Spirans|neither spirant [i.e., continuant or "fricative"] with spirant nor tenuis [i.e., voiceless stop] with tenuis is compliant with the [modern Greek] language. Likewise σ does not tolerate a spirant either before or after it"; also CARA95, p. 151-152, "Another teacher of Greek thought that the Greeks had changed the pronunciation of certain letters, as for example, they pronounced 'p' as 'f' and cited as instance the word epta, (= 'seven'), which he thought Modern Greeks pronounced as efta.", the truth is that it is entirely optional and that both the dissimilated and non-dissimilated forms are acceptable as correct, albeit sometimes one more common than the other (e.g., γαμπρός is almost universally used instead of γαμβρός; on the other hand, άμβλωση|abortion=*ámvlosi* is never dissimilated to άμπλωση=*ámblosi*amblwsh).

Dissimilation (preferred form in bold)
πτ→φτ κτ→χτ φθ→φτ χθ→χτ σθ→στ σχ→σκ αυσ→αψ ευσ→εψ μβ→μπ(=*mb*) νδ→ντ(=*nd*) γγ→γγ(=*g* or *ng*)


This is probably the most prominent characteristic of Greek. It involves the conversion of unstressed prevocalic *i* (close front vowel) to a palatal continuant consonant (*γ* or *χ*) and its potential "incorporation" into the preceding consonant, if this is possible.

Theory: As already mentioned, diphthongs (succession of two vowels accommodated in one syllable) most often involve at least one close vowel (*i*, *u*), which can be either the first or the second element of the diphthong (opening diphthong and closing diphthong respectively) and is generally considered a semivowel, rendered in English as "w" or "y", e.g., wall≈*l*, yell≈*l*These are as good approximations as one may find in English for the opening diphthongs. In fact, the initial vowel is felt by native speakers to be a consonant (as evident from "a wall" and "a yell" instead of "an wall" and "an yell", which would be the correct form if they were considered vowels, cf. "an eye"); hence the term semivowel. and cow≈*káu*, boy≈*bói* for the two kinds of diphthongs respectively.

Greek practice: Since semivowels are alien to the (mainstream) Greek phonology (at least as leading sounds in a diphthong), wherever there "ought" to be a semivowel, the tendency in Greek is to convert it to a (real) consonant. Since (possibly due to orthography and/or etymology) *u* does not form diphthongs in Greek, this is a question of (prevocalic) *i* being converted to the corresponding (i.e., palatal) continuant *γ*, or to *χ* (the latter due to regressive assimilation of voice)Most likely the pronunciation of Υ/υ in the digraphs AY/αυ, ΕΥ/ευ, ΗΥ/ηυ is the result of a similar consonantisation *u*→*v* (or *f* due to regressive assimilation of voice).. Thus, if unstressed *i* precedes another vowel in the same word, it may form a diphthong with it.This is the second source of unclarity about the pronunciation of Greek. The formation of a diphthong (and the subsequent consonantisation) is sometimes up to the speaker and sometimes it is dictated by semantics (e.g., the word άδεια means "permission/vacation/leave" if pronounced *áδia*adeia, i.e., in diaeresis, and "empty (fem. sing., neut. plur.)" if pronounced *áδia*→*áδγa*adeia2, i.e., diphthongally with a subsequent consonantisation). If it does, the Greek tendency is to convert this *i* from a vowel all the way (i.e., past any semivowel or approximant "zone") to the closest consonant, which is the palatal continuant *γ*.

Word-initially: when there is consonantisation, the spelling is also modified to indicate it: ιώτα|iota (the letter)=*ióta*→*ta*→*γóta*=γιώταgiwta, ιατρός|doctor=*iatrós*→*iatrós*→*γatrós*=γιατρόςgiatros, υιός|son=*iós*→*s*→*γós*=γιοςgiosΙωάννης|Johannes, John=*ioánis*→*ioánis*→*γoánis*→*γánis*(with elision of *o*)=Γιάννηςgiannhs; the formation of a diphthong and the subsequent consonantisation of *i* are prevented when *i* is stressed, since stress is a property of (pure) vowels, e.g., ίαση|cure is always pronounced *íasi*iash; sometimes consonantisation is also prevented when it may lead to confusion, e.g., ιερός|sacred(=*ierós*ieros) is never pronounced as *γerós*geros(=γερός|sturdy, healthy, strong) and ιός|virus=*iós*ios is never consonantised to *γós*(=γιος|son).

Intervocalic: although it should not differ from the consonantisation of word-initial *i*, intervocalic *i* is rarely consonantised and that mainly in colloquial speech, orthography again indicating this through the use of γι(=*γ*); thus, from singular (βάιον→)βάι|(branch of) palm tree (hence the name of the palm forest of Crete) is formed the plural βάια=*váia*→*váia*(essentially a triphthong like "via")→*váγa*)=βάγιαbagiaThat the cause of this plural form is the consonantisation of intervocalic *i* is evident when considering that stress transfered over the *i* due to declension prevents the consonantisation, e.g., Κυριακή των Βαΐων|Palm Sunday.; other examples include Τραϊανός|Trajan=*traianós*→*traianós*→*traγanós*=Τραγιανός, the adjective μαγιάτικος|of/relating to May=*maγátikos*magiatikos from Μάιος|May(=*máios* or *máios*maios, but never *máγos*!), the rendering of Maya, Meyer, Bayern (where Y/y is rather a semivowel) as Μάγια=(*máγa*), Μάγιερ(=*máγer*), Μπάγερν=(*báγern*), etc.

After a voiced consonant (other than *g*, *γ*, *m*, *n*, *l*): when *i* is consonantised, it retains its *γ* value; thus, βιάσου!|hurry up!=*vγásu*biasou, δυο|two=*δγó*duo, καρπούζια|watermelons=*karpúzγa*karpouzia, ψάρια|fish (pl.)=*psárγa*yaria, also κάμπια|caterpillar=*kámbγa*, δόντια|teeth=*δóndγa*dontia, Τζια|(colloquial name of the Greek island) Kea=*dzγá*tzia; sometimes there is uncertainty about whether the orthography needs to reflect the consonantisation or not, e.g., καινούργιος and καινούριος are both acceptable spellings of the Greek word for "new" and are both pronounced *kenúrγos*; there is no consonantisation of *i* after word-initial Ρ/ρ, e.g., ρυάκι|runnel=*riáki*ruaki.

After a voiceless consonant (other than *k*, *χ*): when *i* is consonantised, it is de-voiced to *χ* due to progressive assimilation of (non-)voice (actually, to avoid regressive assimilation of voice); thus, αγκάθια|thorns=*agáθχa*agkaqia, αμάξια|car(t)s=*amáksχa*amajia, πιάνο|piano=*pχano*piano, σιωπή|silence=*sχopí*siwphφωτιά|fire=*fotχá*fwtia, καρφιά|nails=*karfχá*karfia, προσόψια|face towels=*prosópsχa*prosoyia, also παπούτσια|shoes=*papútsχa*papoutsia.

After a velar: in that case, the consonantised (as *γ* after *g* and *γ* or *χ* after *k* and *χ*) *i* is "incorporated" into the velar, producing the corresponding palatalI cannot clearly identify this phenomenon as any kind of assimilation; it rather appears as if the "palatal property" of the (palatal) vowel *i* is transfered to (or assimilated by) the velar.; thus, κιόσκια|kiosks=*kóskia*kioskia, χιόνι|snow=*χóni*xioni, Γκιώνα|(mount) Giona=*góna*gkiwna, γεια|hello (lit. health)=*γá*geia.

After *n* or *l*: as these are the only other (i.e., non-velar) consonants that can be palatalised, the effect of a consonantised *i* on them is the same as in the previous case, namely it is "incorporated" into the previous consonant to produce the palatals *n* and *l*; thus, νιάτα|youth=*náta*niata, λιακάδα|sunshine=*lakáδa*liakada.It is evident that this is the only case of occurrence of the palatals *n* and *l* and, since a consonantised *i* is necessary for their production, these sounds are referred to in Greek as "νι" and "λι" (similarly, the palatals that correspond to the velars, namely *k*, *g*, *γ* and *χ*, are referred to as "κι", "γκι", "γι" and "χι" respectively); that this is insufficient to describe the actual sounds is evident from the transcription of the Serbian (basketball-player) name Paspalj=*páspal* as Πάσπαλι(=*páspali*), Πάσπαλιε(=*páspale*) or Πάσπαλ=(*páspal*), none of which represents the Serbian pronunciation, even though the sound does exist in Greek; unfortunately, Greek lacks a special character for indicating the consonantised *i* and its effect on preceding consonants (cf. the Germano-Slavic convention of writing Anja=*ána*, Banja Luka=*bána lúka* and Ljubljana=*lublána*).

After *m*: in this singular case, the consonantised *i* (*γ*) assimilates the nasality of *m* and is converted into the palatal nasal *n*; thus μια|one (fem.)=*mná*mia.

Consonantisation of prevocalic unstressed *i*  (V=any vowel)
Word-initially (→*γ*)*Ø|nothing/beginning of wordiV**γV*
Intervocalic (→*γ*)*ViV**VγV*
Post-voiced (→*γ*)*biV**bγV*
Post-voiceless (→*χ*)*piV**pχV*
Post-velar (→palatalised velar)*kiV**kV*
After *n* (→*n*)*niV**nV*
After *l* (→*l*)*liV**lV*
After *m* (→*mn*)*miV**mnV*



Syllabification relates to the rules governing the division of words into syllables, which are more reasonable in Greek (and in almost every language!) than in English. For indicating the separate syllables, the hyphen ("-") is used. The nucleus of each syllable is a vowel (meaning a single vowel letter, a diphthong or a vowel digraph), each vowel sound in a word defining a separate syllable; so much is clear. However, what needs to be clarified is what happens with the consonants around them.

Unlike English and many other languages, Greek syllabification does not take into account any etymologies, but merely the succession of vowels and consonants within a word, even if it is a compound word. The following rules apply:

  1. A single intervocalic consonant (letter) belongs to the syllable of the following vowel, e.g., έ-να|one (neut. nomin.)=*é-na*, έ-ξι|six=*é-ksi*.
  2. Any word-initial consonants belong to the syllable of the following vowel, e.g., μί-α|one (fem. nomin.)=*mí-a*, τρί-α|three (neut. nomin.)=*trí-a*.
  3. Word-final consonants (mainly Ν/ν, Σ/ς) belong to the syllable of the previous vowel, e.g., έ-νας|one (masc. nomin.)=*é-nas*, τρεις|three (masc./fem. nomin.)=*trís* (one syllable).
  4. A cluster of two or more consecutive word-internal consonants (that is consonant letters) is assigned to the syllable of the following vowel, if a Greek word begins with the combination of the first two; otherwise, the first consonant is assigned to the previous syllable and the remaining ones to the following syllable, e.g., ε-πτά|seven=*e-ptá*, εν-νέ-α|nine=*e-né-a*From the difference between the written and the ALPAG forms, it is evident that syllabification does not follow pronunciation (in writing the first syllable is εν, while in pronunciation it is plainly *e*)., αρ-χή|beggining, start=*ar-xí*, ά-σπρο|white (neut. nomin.)=*á-spro*, συγ-χω-ρώ|I forgive=*si-χo-ró*.

The last one is certainly the most difficult to apply, especially for non-native speakers. Thus, the following table is a useful reference:

Allowable (+) and Non-Allowable (-) Word-Initial Consonant Pairs (C1C2)
C1\C2 Β/β Γ/γ Δ/δ Ζ/ζ Θ/θ Κ/κ Λ/λ Μ/μ Ν/ν Ξ/ξ Π/π Ρ/ρ Σ/σ Τ/τ Φ/φ Χ/χ Ψ/ψ
Β/β - + + - - - + - - - - + - - - - -
Γ/γ - - + - - + + - + - - + - - - - -
Δ/δ - - - - - - - - - - - + - - - - -
Ζ/ζ - - - - - - - - - - - - - - - - -
Θ/θ - - - - - - + - + - - + - - - - -
Κ/κ - - - - - - + - + - - + - + - - -
Λ/λ - - - - - - - - - - - - - - - - -
Μ/μ - - - - - - - - + - + - - - - - -
Ν/ν - - - - - - - - - - - - - + - - -
Ξ/ξ - - - - - - - - - - - - - - - - -
Π/π - - - - - - + - + - - + - + - - -
Ρ/ρ - - - - - - - - - - - - - - - - -
Σ/σ + + - - + + - + - - + - - + + + -
Τ/τ - - - + - - - + - - - + + - - - -
Φ/φ - - - - + - + - - - - + - + - - -
Χ/χ - - - - + - + - + - - + - + - - -
Ψ/ψ - - - - - - - - - - - - - - - - -


Every word has no more than one stressed syllable. Unlike English and German that develop a secondary stress in long words, the one-stressed-syllable-per-word rule holds even for very long words with many syllables, e.g., ορνιθοσκαλίσματα|chicken scratches, scrawls, scribbles=*orniθoskalísmata*. The stressed syllable has to be one of the last three, known as the rule of "trisyllabotony".

Some monosyllabic or disyllabic words (articles, particles, prepositions, pronouns, consunctions) are unstressed in the course of normal speech and they usually adhere to one of the stressed words around them, usually the word to which they refer. These unstressed words, although written separately, are pronounced essentially as one unit with the stressed word (e.g., ο άγγελος|the angel=*oángelos*oaggelos, δικός μου|mine (lit. my own)=*δikózmu*dikosmou) and are known as clitics. If the unstressed word precedes the stressed one (first example), it is known as proclitic, if it follows (second example), it is named enclitic. Proclitics do not have any effect on the accentuation of the main word, but when the word is augmented by an (unstressed) enclitic, the stressed syllable may be caused to move before the three last syllables of the word-enclitic compound. In that case, the rule of trisyllabotony imposes a second (not secondary) stress on the word's last syllable (the ultima); thus, το άλογο|the horse=*toáloγo*toalogo, but το άλογό μου|my horse=*toáloγómu*toalogomou.

Final Consonants

Greek words (excluding foreign loans that have not undergone "Hellenisation") may end either in a vowel or in Ν/ν or Σ/ς (word-final lower-case Σ has a different form, namely ς). In relics of the ancient language, Ξ/ξ, Ψ/ψ or Ρ/ρ may also serve as final consonants (e.g., άπαξ|once=*ápaks*, Άραψ|Arab=*áraps*, ρήτωρ|orator=*rítor*)Κ/κ was also possible in a limited number of words, e.g, the particle οὐκ|no, not and the preposition ἐκ|from, out of..

In some cases (particularly in proclitics), word-final Ν/ν may be dropped if it is not compatible with the first consonant of following word (this includes all liquids, nasals, sibilants and possibly all continuants, e.g., τον λαγό|the rabbit (acc.)→το λαγό) or assimilated by it (but only in pronunciation and never in writing; thus, τον πατέρα|the father (acc.)=*tombatéra*tonpatera).

A similar behaviour, as regards assimilation with a following consonant, is exhibited by word-final Σ/ς both in proclitics (e.g., της μπανάνας|of the banana=*tizbananas*thsmpananas, but not so commonly) and before enclitics (e.g., οι μπανάνες μας|our bananas=*ibanánezmas*oimpananesmas).


Go to previous page Go to main page Go to next page