This is the Hungarian consonantal system, using symbols from the International Phonetic Alphabet (IPA).
^* It is debated whether the palatal consonant pair consists of stops or affricates. They are considered affricates by Tamás Szende, head of the department of General Linguistics at PPKE, and stops by Mária Gósy, research professor, head of the Department of Phonetics at ELTE. The reason for the different analyses is that the relative duration of the friction of /c/ (as compared to the duration of its closure) is longer than those of the stops, but shorter than those of the affricates. /c/ has the stop-like nature of having a full duration no longer than those of other (voiceless) stops such as /p, t, k/ but, considering the average closure time in relation to the friction time of the consonants, its duration structure is somewhat closer to those of the affricates.
Almost every consonant may be geminated, written by doubling a single letter grapheme: ⟨bb⟩ for [bː], ⟨pp⟩ for [pː], ⟨ss⟩ for [ʃː] etc., or by doubling the first letter of a grapheme cluster: ⟨ssz⟩ for [sː], ⟨nny⟩ for [ɲː], etc.
The phonemes /d͡z/ and /d͡ʒ/ can appear on the surface as geminates: bridzs [brid͡ʒː] ('bridge (the card game)'). (For the list of examples and exceptions, see Hungarian dz and dzs.)
Hungarian orthography does not use háčky or any other consonant diacritics like the surrounding Slavic languages. Instead, the letters c, s, z are used alone (/t͡s/, /ʃ/, /z/) or combined in the digraphs cs, sz, zs (/t͡ʃ/, /s/, /ʒ/), while y is used only in the digraphs ty, gy, ly, ny as a palatalization marker to write the sounds /c͡ç/, /ɟ͡ʝ/, /j/ (formerly /ʎ/), /ɲ/.
The most distinctive allophones are:
/j/ becomes [ç] if between a voiceless obstruent and a word boundary (e.g. lopj [lopç] 'steal').
/j/ becomes [ʝ] e.g. between voiced obstruents, such as dobj be [dobʝ bɛ] 'throw (one/some) in'
/h/ may become [ɦ] between two vowels (e.g. tehát [tɛɦaːt] 'so'), [ç] after front vowels (e.g. ihlet [içlɛt] 'inspiration'), and [x] word-finally after back vowels (e.g. doh [dox] 'musty') if it isn't deleted (which it often is; e.g. méh [meː] 'bee').According to Gósy, it becomes [x] (rather than [ç]) in words such as pech, ihlet, technika ('bad luck, inspiration, technology/technique'), while it becomes postvelar fricative in words such as doh, sah, jacht, Allah, eunuch, potroh.
/h/ becomes [xː] when geminated, in certain words: dohhal [doxːɒl] ('with blight'), peches [pɛxːɛʃ] ('unlucky').
Hungarian has seven pairs of corresponding short and long vowels. Their phonetic values do not exactly match up with each other, so ⟨e⟩ represents /ɛ/ and ⟨é⟩ represents /eː/; likewise, ⟨a⟩ represents /ɒ/ while ⟨á⟩ represents /aː/. For the other pairs, the short vowels are slightly lower and more central, and the long vowels more peripheral:
/i, y, u/ are phonetically near-close [ɪ, ʏ, ʊ].
/eː/ has been variously described as close-mid [eː] and mid [e̞ː].
/ɛ/ and the marginal /ɛː/ are phonetically near-open [æ, æː], but they may be somewhat less open [ɛ, ɛː] in other dialects.
/ø/ is phonetically mid [ø̞].
/aː/ and the marginal /a/ are phonetically open central [äː, ä].
The sound marked by ⟨a⟩ is considered to be [ɒ] by Tamás Szende and [ɔ] by Mária Gósy. Gósy also mentions a different short /a/ that contrasts with both /aː/ and /ɒ/, present in a few words like Svájc ('Switzerland'), svá ('schwa'), advent ('advent'), hardver ('hardware', this usage is considered hyperforeign), and halló (used when answering the phone; contrasting with haló 'dying', and háló 'web').
There are two more marginal sounds, namely the long /ɛː/ as well as the long /ɒː/. They are used in the name of the letters E and A, which are pronounced /ɛː/ and /ɒː/, respectively.
Although not found in Budapest, some dialects contrast three mid vowels /e/ (written ⟨ë⟩ in some works, but not in the standard orthography), /ɛ/, and /eː/. Thus mentek could represent four different words: mëntëk [mentek] ('you all go'), mëntek [mentɛk] ('they went'), mentëk [mɛntek] ('I save'), and mentek [mɛntɛk] ('they are exempt'). In Budapest, the first three collapse to [mɛntɛk], while the latter one is unknown, having a different form in the literary language (mentesek).
As in Finnish, Turkish, or Mongolian, vowel harmony plays an important part in determining the distribution of vowels in a word. Hungarian vowel harmony classifies the vowels according to front vs. back assonance and rounded vs unrounded for the front vowels. Excluding recent loanwords, Hungarian words have either only back vowels or front vowels due to these vowel harmony rules.
While [i], [iː], [ɛ], and [eː] are all front unrounded vowels, they are considered to be "neutral vowels" in Hungarian vowel harmony. Therefore if a word contains back vowels, neutral vowels may appear alongside them. However if only neutral vowels appear in a stem, the stem is treated as though it is of front vowel assonance and all suffixes must contain front vowels.
Vowel harmony in Hungarian is most notable when observing suffixation. Vowel harmony must be maintained throughout the entire word, meaning that most suffixes have variants. For example, the dative case marker [nɒk] vs. [nɛk]. Stems that contain back vowels affix back vowel suffixes, and stems that contain only front vowels affix front vowel suffixes. However, the front vowel stems distinguish rounded vs. unroundedness based on the last vowel in the stem. If the last vowel is front and rounded, it takes a suffix with a front rounded vowel; otherwise it follows the standard rules. While suffixes for most words have front/back vowel variants, there are not many that have rounded/unrounded variants, indicating that this is a rarer occurrence.
One is able to observe the distinction when looking at the plural affix, either [-ok] (back), [-ɛk] (front unrounded), or -ök [-øk] (front rounded).
As can be seen above, the neutral vowels are able to be in both front and back vowel assonance words with no consequence.
However, there are about fifty monosyllabic roots that only contain [i], [iː], or [eː] that all take a back vowel suffix instead of the front vowel suffix.
These exceptions to the rule are hypothesized to have originated from the roots originally having contained a phoneme no longer present in modern Hungarian, the unrounded back vowel /ɨ/, or its long counterpart /ɨː/. It is theorized that while these vowels merged with /i/ or /iː/, less commonly /eː/ or /uː/, the vowel harmony rules sensitive to the backness of the original sound remained in place. The theory finds support from etymology: related words in other languages generally have back vowels, often specifically unrounded back vowels. For example, nyíl 'arrow' (plural nyíl-ak) corresponds to Komi ньыл /nʲɨl/, Southern Mansi /nʲʌːl/.
The overall characteristics of the consonant assimilation in Hungarian are the following:
Assimilation types are typically regressive, that is the last element of the cluster determines the change.
In most cases, it works across word boundaries if the sequence of words form an "accentual unity", that is there is no phonetic break between them (and they bear a common phase stress). Typical accentual units are:
attributes and qualified nouns, e.g. hideg tél [hidɛk‿teːl] ('cold winter');
adverbs and qualified attributes, e.g. nagyon káros [nɒɟoŋ‿kaːroʃ] ~ [nɒɟon‿kaːroʃ] ('very harmful');
verbs and their complements, e.g. nagyot dob [nɒɟod‿dob] ('s/he throws long toss'), vesz belőle [vɛz‿bɛløːlɛ] ('take some [of it]').
There are obligatory, optional and stigmatized types of assimilation.
The palatal affricates behave like stops in assimilation processes. Therefore in this section, they will be treated as stops, including their IPA notations [ɟ] and [c].
In a cluster of consonants ending in an obstruent, all obstruents change their voicing according to the last one of the sequence. The affected obstruents are the following:
In obstruent clusters, retrograde voicing assimilation occurs, even across word boundaries:
[3] /v/ is unusual in that it undergoes assimilation but doesn′t cause voicing, e.g. hatvan ('sixty') is pronounced [hɒtvɒn] not *[hɒdvɒn]. Voicing before [v] occurs only in south-western dialects, though it is stigmatized.
Similarly, /h/ causes devoicing, but never undergoes voicing in consonant clusters. e.g. dohból [doxboːl] 'from (the) musty smell'.
Other than a few foreign words, morpheme-initial /d͡z/ doesn't occur (even its phonemic state is highly debated), therefore it is hard to find a real example when it induces voicing (even alapdzadzíki is forced and not used colloquially). However, the regressive voice assimilation before [d͡z] does occur even in nonsense sound sequences.
Nasals assimilate to the place of articulation of the following consonant (even across word boundaries):
only [ŋ] precedes a velar consonant (e.g. hang [hɒŋɡ], 'voice'), [ɱ] precedes a labiodental consonant (e.g. hamvad [hɒɱvɒd], 'smoulder'), and [m] precedes bilabial consonants.
[m] before labial consonants /p b m/: színpad [siːmpɒd] ('stage'), különb [kylømb] ('better than'), énmagam [eːmːɒɡɒm] ('myself');
[ɱ] before labiodental consonants /f v/: különféle [kyløɱfeːlɛ] ('various'), hamvas [hɒɱvɒʃ] ('bloomy');
[ɲ] before palatal consonants /c͡ç ɟ͡ʝ ɲ/: pinty [piɲc] ('finch'), ángy [aːɲɟ] ('wife of a close male relative'), magánnyomozó [mɒɡaːɲːomozoː] ('private detective');
[ŋ] before velar consonants /k ɡ/: munka [muŋkɒ] ('work'), angol [ɒŋɡol] ('English');
Nasal place assimilation is obligatory within the word, but optional across a word or compound boundary, e.g. szénpor [seːmpor] ~ [seːnpor] ('coal-dust'), nagyon káros [nɒɟoŋ‿kaːroʃ] ~ [nɒɟon‿kaːroʃ] ('very harmful'), olyan más [ojɒm‿maːʃ] ~ [ojɒn‿maːʃ] ('so different').
Voiceless sibilants form a voiceless geminate affricate with preceding alveolar and palatal stops (d [d], gy [ɟ], t [t], ty [c]):
Clusters ending in sz [s] or c [t͡s] give [t͡sː]: metszet [mɛt͡sːɛt] 'engraving, segment', ötödször [øtøt͡sːør] 'for the fifth time', négyszer [neːt͡sːɛr] 'four times', füttyszó [fyt͡sːoː] 'whistle (as a signal)'; átcipel [aːt͡sːipɛl] 's/he lugs (something) over', nádcukor [naːt͡sːukor] 'cane-sugar'.
Clusters ending in s [ʃ] or cs [t͡ʃ] give [t͡ʃː]: kétség [keːt͡ʃːeːɡ] 'doubt', fáradság [faːrɒt͡ʃːsaːɡ] 'trouble', egység [ɛt͡ʃːeːɡ] 'unity', hegycsúcs [hɛt͡ʃːuːt͡ʃ] 'mountain-top'.
Two sibilant fricatives form a geminate sibilant fricative; the assimilation is regressive as usual:
sz [s] or z [z] + s [ʃ] gives [ʃː]: egészség [ɛɡeːʃːeːɡ] 'health', község [køʃːeːɡ] 'village, community';
sz [s] or z [z] + zs [ʒ] gives [ʒː]: vadászzsákmány [vɒdaːʒːaːkmaːɲ] 'hunter′s game'; száraz zsömle [saːrɒʒ‿ʒømlɛ] 'dry bread roll';
s [ʃ] or zs [ʒ] + sz [s] gives [sː]: kisszerű [kisːɛryː] 'petty', rozsszalma [rosːɒlmɒ] 'rye straw';
s [ʃ] or zs [ʒ] + z [z] gives [zː]: tilos zóna [tiloz‿zoːnɒ] 'restricted zone', parázs zene [pɒraːz‿zɛnɛ] 'hot music'.
Clusters zs+s [ʃː], s+zs [ʒː], z+sz [sː] and sz+z [zː] are rather the subject of the voice assimilation.
If one of the two adjacent sibilants is an affricate, the first one changes its place of articulation, e.g. malacság [mɒlɒt͡ʃːaːɡ], halászcsárda [hɒlaːʃt͡ʃaːrdɒ] 'Hungarian fish restaurant'. Sibilant affricate-fricative sequences like /t͡ʃʃ/ are pronounced the same as geminate affricate [t͡ʃː] during normal speech.
Sibilant assimilation can be omitted in articulated speech, e.g. to avoid homophony: rozsszalma [rosːɒlmɒ] ~ [roʃsɒlmɒ] 'rye straw' ≠ rossz szalma [ros‿sɒlmɒ] 'straw of bad quality', and rossz alma [rosː‿ɒlmɒ] 'apple of bad quality' as well.
NB. Letter cluster szs can be read either as sz+s [ʃː], e.g. egészség [ɛɡeːʃːeːɡ] 'health', or as s+zs [ʒː], e.g. liszteszsák [listɛʒːaːk] 'bolting-bag' depending on the actual morpheme boundary. Similarly zsz is either zs + z [zː], e.g. varázszár [vɒraːzːaːr] 'magic lock', or z + sz [sː], e.g. házszám [haːsːaːm] 'street-number'; and csz: cs + z [d͡ʒz] ~ c + sz [t͡ss]. Moreover, single digraphs may prove to be two adjacent letters on morpheme boundary, like cs: cs [t͡ʃ] ~ c + s [t͡ʃʃ]; sz: sz [s] ~ s + z [zː], zs: zs [ʒ] ~ z + s [ʃː].
Combination of a "palatalizable" consonant and a following palatal consonant results in a palatal geminate. Palatalizable consonants are palatal ones and their non-palatal counterparts: gy [ɟ] ~ d [d], l [l] ~ j [j], n [n] ~ ny [ɲ], ty [c] ~ t [t].
Full palatal assimilation occurs when the ending palatal consonant is j [j]: nagyja [nɒɟːɒ] 'most of it', adja [ɒɟːɒ] 's/he gives it'; tolja [tojːɒ] 's/he pushes it'; unja [uɲːɒ] 's/he is bored with it', hányja [haːɲːɒ] 's/he throws it'; látja [laːcːɒ] 's/he sees it', atyja [ɒcːɒ] 'his/her father'. — Cluster lyj [jː] is a simple orthographic variant of jj [jː]: folyjon [fojːon] 'let it flow'.
Partial assimilation takes place if an alveolar stop (d, t) is followed by palatal is gy [ɟ], ty [c]: hadgyakorlat [hɒɟːɒkorlɒt] 'army exercises', nemzetgyűlés [nɛmzɛɟːyːleːʃ] 'national assembly'; vadtyúk [vɒcːuːk] 'wild chicken', hat tyúk [hɒc‿cuːk] 'six hens'.
Some sources report that alveolars stops change into their palatal counterparts before ny [ɲ]: lúdnyak [luːɟɲɒk] 'neck of a goose', átnyúlik [aːcɲuːlik] 'it extends over'. The majority of the sources don't mention this kind of assimilation.
When the first consonant is nasal, the partial palatal assimilation is a form of the nasal place assimilation (see above).
The full palatal assimilation is an obligatory feature in the standard Hungarian: its omission is stigmatized and it is considered as a hypercorrection of an undereducated person. Partial palatal assimilation is optional in articulated speech.
Long consonants become short when preceded or followed by another consonant, e.g. folttal [foltɒl] 'by/with (a) patch', varrtam [vɒrtɒm] 'I sewed'.
The middle alveolar stops may be omitted in clusters with more than two consonants, depending on speed and articulation of speech: azt hiszem [ɒs‿hisɛm] ~ [ɒst‿hisɛm] 'I presume/guess', mindnyájan [miɲːaːjɒn] 'one and all', különbség [kylømpʃeːɡ] ~ [kylømʃeːɡ] 'difference'. In morpheme onsets like str- [ʃtr], middle stops tends to be more stable in educated speech, falanxstratégia [fɒlɒnʃtrɒteːɡiɒ] ~ [fɒlɒŋkʃtrɒteːɡiɒ] ~ [fɒlɒŋksʃtrɒteːɡiɒ] 'strategy based on phalanxes'.
/l/ assimilates to [r] before /r/ (e.g. balra [ˈbɒrːɒ], 'to the left').
/l/ also tends to be omitted between a preceding vowel and an adjacent stop or affricate rapid speech, causing the lengthening of the vowel or diphthongization (e.g. volt [voːt] 'was', polgár [ˈpoːɡaːr] 'citizen'). This is considered non-standard.
Standard Hungarian allows (prefers) hiatus between adjacent vowels. However some optional dissolving features can be observed:
An optional weak glide [j̆] may be pronounced within a word (or a compound element) between two adjacent vowels if one of them is i [i], e.g. fiaiéi [ˈfiɒieːi] ~ [ˈfij̆ɒj̆ij̆eːj̆i] ('the ones of his/her sons'). This, however, is rarely transcribed.
Adjacent identical short vowels other than a and e may be pronounced as the corresponding long vowel, e.g. zoológia [ˈzo.oloːɡiɒ] ~ [ˈzoːloːɡiɒ] ('zoology').
Two adjacent i’s are always pronounced as single short [i] in the word endings, e.g. Hawaii [ˈhɒvɒi]. This reduction is reflected in the current orthography when the adjective-forming suffix -i is added to a noun ending in i. In this case suffix -i is omitted also in writing. e.g. Lenti (a placename) + -i → lenti 'of Lenti'.
The stress is on the first syllable of the word. The articles a, az, egy, and the particle is are usually unstressed.
