1. Phonology

1.1 Phonemes and alphabet

Alphabet

The following table compares the forty-two Latin graphemes in the Czech alphabet with the IPA symbols that designate their sounds:
a [a] g [g] o [o] u [U]
á [aÉ] h [h] ó [oÉ] ú [uÉ]
b [b] ch [x] p [p] ů [uÉ]
c [ts] i [I] qu [k],[kv] v [v]
č [tS] í [iÉ] r [r] w [v]
d [d] j [j] ř [rˇZ] x [ks]
d, Čd, d [ď] k [k] s [s] y [I]
e [e] l [l] š [S] ý [iÉ]
é [eÉ] m [m] t [t] z [z]
ě [e] n [n] t, Čt, t [c] ž [Z]
f [f] ň, n [­]

When length is present in foreign words it is not consistently indicated in the orthography, especially when it involves [oÉ] and [uÉ]: telefon [telefoÉn]\[telefon] telephone, kultura [kUltuÉra]\[kUltUra] culture.

The graphemes i, í and y, ý both represent /I, iÉ/, the result of a historic merger of two once distinct phonemes; these letters do, however, serve as diacritics for the consonant graphemes t, d, and n. Sequences of ti, di, ni indicate /cI /, /ďI /, /­I /, in which the corresponding palatal replaces the dental, and this is true for syllables with long vowels as well. When y follows these consonant graphemes, the expected dentals are pronounced. Compare the following examples:
ty [tI] you N ti [cI] you D
dýl [diÉl] longer CCz díl [ďiÉl] part
pány [paÉnI] men Apl páni [paÉ­I] menĘNpl

When i appears after t, d, and n in foreign words, the shift to palatals does not take place:
direktor [dIrektor] director.
In combination with other consonant letters i vs. y does not indicate any phonological distinctions:
pyl [pIl] pollen pil [pIl] he drank
sýrový [siÉroviÉ] cheese adj. sírový [siÉroviÉ] sulphur adj.
bílý [biÉliÉ] white Nsgmasc bílí [biÉliÉ] white Nplanim
The vowel letters i and í are not spelled after h, ch, k, and r; and y and ý are not spelled after c, č, j, ř, š, and ž.

The grapheme ě serves a similar diacritic purpose, but it is restricted only to short syllables, and in addition to signalling the substitution of a palatal for a dental, after the labial graphemes it signals the insertion of [j] (after b, p, v, and f) and [­] (after m). Compare the following uses of e and ě:
tele [tele] calf těle [cele] body Lsg
poklade [poklade] treasure Vsg pokladě [poklaďe] treasure Lsg
týdne [tiÉdne] week Gsg týdně [tiÉd­e] weekly
hrabe [hrabe] digs hrabě [hrabje] count
fez [fez] fez harfě [harfje] harp DLsg
kape [kape] drips kápě [kaÉpje] hood
venčit [ventSIt] take out věnčit [vjentSIt] crown
mech [mex] moss měch [m­ex] bellows

The graphemes ť, Čt , ď, Čd, and ň are used to indicate [c], [ď] and [­] when they appear before a back vowel, before a consonant, or in word-final position. ť and Čt are just two different ways of representing the same letter, depending upon what sort of font or typewriter is available; the same is true for ď and Čd.
before a back vowel before consonant or word-final
ťápat or Čtápat [caÉpat] toddle
zaplaťme or zaplaČtme [zaplacme] lets pay!
ťukat or Čtukat [cUkat] tap
zaplať or zaplaČt [zaplac] pay!
ďábel or Čdábel [ďaÉbel] devil
zařiďme or zařiČdme [zarˇZIďme] lets organize!

Whereas the presence or absence of a diacritic is ignored in alphabetizing vowel letters (for instance e, é, and ě are all alphabetized as if they were a single grapheme), the presence of a diacritic gives a consonant letter its own distinct place in the alphabet, directly after the same consonant letter without the diacritic. Thus č is alphabetized after c, ř is alphabetized after r, š is alphabetized after s, and ž is alphabetized after z. However ť, Čt , ď, Čd, and ň constitute an exception to this rule, and they are alphabetized like t, d and n.

The grapheme ch is considered for all practical purposes (alphabetization, crossword puzzles, etc.) a single letter.

The graphemes g, q, w, and x appear only in foreign words.

Vowels (foreign elements are in parentheses)
short vowels long vowels
i u í ú, ů
e o é (ó)
a á
diphthongs
ij ej aj oj uj ůj (eu) (au) ou

Czech has a classic five-vowel system, consisting of short a, e, i/y, o, u and long á, é, í/ý, ó, ú/ů. Length is phonemic and in most instances indicated by an acute accent.

Long ú/ů can be written ú in initial position or after a prefix (úzký narrow, zaúčtovat calculate), but ů elsewhere (dům house Nsg, domů house Gpl, where it continues Late Common Slavic o).

Long ó occurs in foreign borrowings (próza prose) and sometimes in emphatic usage (móře lots CCz vs. moře sea). The primary (phonemic) distinction between the two vowel subsystems is one of quantity, but in the case of the high vowels short i/y vs. long í/ý and short u vs. long ú/ů there is a perceptible phonetic difference as well. Czech generally does not tolerate vowel chains; there are no vowel chains in native roots, and where two vowels meet at a morpheme boundary (when a prefix ending in a vowel is added to a stem beginning with one), a glottal stop is usually inserted to separate them: neomylně [ne?omIl­e] without error.

The diphthong ou is a native sequence (náhodou by chance), and foreign borrowings have introduced au (auto automobile) and eu (leukémie leukemia). Since the back glide [u`] is restricted to diphthongs, its existence is non-distinctive and therefore sub-phonemic, acting as an allophone of u; it does not enjoy the status of the front glide j. Czech has six diphthongs ending in j, two of which are relatively rare, aj (hraj play!) and ůj (můj my), plus the widespread ej (dělej do!), ij (šij sew!), oj (dojmový emotional), and uj (miluj love!).

In foreign words presenting vowel chains containing i, if the chain begins with i, Czech pronunciation inserts a non-orthographic j:
materiál [materIjaÉl] material,
biologie [bIjologIje] biology Nsg,
biologii [bIjologIjI] biology Asg;
if the chain ends with i followed by a consonant, it is pronounced as j:
detail [detajl] detail,
pleistocén [plejstotseÉn] pleistocene.

The liquids r and l

The liquids r and l can participate in both syllable peaks (as vowels) and slopes (as consonants). As vowels, they are only short (length is not phonemic) and occur only between non-vocalic elements (after a consonant and before another consonant or word boundary: krk neck, nesl he brought, hence there are no initial syllabic liquids).

Vocalic m is possible in two words: sedm seven and osm eight, but the usual pronunciation is [sedUm], [osUm].

Consonants

labials dentals palatals velars laryngeals obstruents:
stops p b t d t d k g h
fricatives f v s z š ž ch
affricates c č
trill+fricative ř

sonorants:
glide j
nasal m n ň
lateral l
trill r

Stops are not aspirated.

The most pervasive type of assimilation is certainly voiced vs. voiceless (described below in 1.2). These assimilations produce voiced allophones for unpaired obstruents in the chart: [dz] for c (leckdo [ledzgdo] anyone), [dZ] for č (léčba [leÉdZba] cure), and voiceless [r9ˇS] for ř (keř [ker9ˇS] bush. The remaining two unpaired obstruents, ch and h, to some extent serve as partners, despite their disparate places of articulation. But while the devoicing of h is common and regularly produces ch [x] (Bůh [buÉx] God), the voicing of ch is both less frequent and less clear in its result. It appears that the assimilative voicing of ch can, particularly after i, yield [V] (abych bylĘ[abIVbIl] so that I would) as well as [h]; alternatively there may be no voicing at all. There are two more sounds which should be added to the inventory of consonants, but are clearly subphonemic: [N], which is purely a positional variant of n before a velar (tank [taNk] tank), and the glottal stop [/] which is recommended but not obligatory before word-intial vowels and between vowels at the prefix boundary, and is not indicated in the orthography.
ß (moře [morˇZe] sea), along with its voiceless allophone [r9ˇS] (kouř [kour9ˇS] smoke) is an unusual item, and Czech is the only European language to have this sound. It is a unit phoneme, which means that it is simultaneously a dental and a palatal. ß is usually the last sound acquired by Czech children and its mispronunciation constitutes over 50% of their speech defects
(Palková 1994: 347, 350).

Also rather unusual in languages and very difficult for foreigners to pronounce are the palatal stops d and t.

F is similar to [dz] and [dZ], in that it exists in Slavic words almost entirely as a voicing partner of v and is phonemic only in foreign borrowings. It does, though, appear in at least one native root (etymologically derived from a consonant cluster), doufat hope, and the number of non-native and nativized stems (as in telefonovat telephone, trefit hit) with f is vastly larger than that of [dz] and [dZ]. In the past few centuries initial f has also figured in a number of native onomatopoetic creations, such as foukat blow, frčet buzz, and funět snort. The situation of g (as a voiced variant of k and a sound found in many foreign words, such as gól goal in sports) is similar, and this sound is also found in borrowings involving an intervocalic k in the source language, as in plakát [plagaÉt] or [plakaÉt] poster. S is likewise sometimes rendered [z] in borrowings, as in disertace [dIzertatse] dissertation (optionally spelled dizertace). The spelling of z for borrowed intervocalic s has been codified in spelling reforms for words like univerzita university and filozofie philosophy.

1.2 Phonological rules

Stress

Stress is not phonemic and always falls on the first syllable of a phonological word. However, a variety of types of clitics makes the identification of a phonological word fairly complex, and there are also some exceptions to the rule, particularly when unstressed proclitics are attached to a word (e.g., the unstressed proclitic to that in the single phonological word formed by To ’vím I know that, which is stressed on vím).

For the purposes of stress, a phonological word consists of a stress-bearing word, plus adjacent stressless words or particles. The stress-bearing words are nouns, adjectives, verbs (except auxiliaries), most adverbs, prepositions of two or more syllables, numerals, non-clitic pronouns, and some conjunctions; these words can and normally do bear stress. There are three types of stressless words which participate with the stess-bearing words in the formation of phonological words:

a) most conjunctions, the word pan Mr. (e.g., pan ’Novák Mr. Novák), some adverbs (depending on context) and some possessives and forms of the demonstrative and indefinite pronouns ten, všechen (when followed immediately by the word modified, as in všechen ’chléb all the bread) are stressless and form a phonological word with the following stressed word and whatever enclitics follow it, but remain stressless (unlike the pre-posed items in b);

b) monosyllabic prepositions and the pre-posed negative particle ne- are not intrinsically stress-bearing, but, when present, take over the stress of the following intrinsically stressed element, e.g., ’Znám ho I know him when negated becomes ’Neznám ho I dont know him, and ’čtvrtek Thursday in the presence of a preposition becomes ’ve čtvrtek on Thursday;

c) enclitic pronouns, conditional and past tense auxiliary forms follow the first stressbearing word in a clause; these act as additional unstressed syllables in the phonological word and are totally dependent (cannot appear in initial position). There are also semi-dependent enclitics, primarily to that, prepositional phrases and some other elements (including certain adverbs) that may attach to the first stressed phonological word in a clause; the stress of these items varies (to is usually unstressed, the remaining semi-dependent enclitics frequently bear their own stress).

Thus a phonological word contains the following maximal structure:
{ a)-type stressless proclitic +
   b)-type stressless word receiving stress + stress-bearing word, stressed on first syllable in absence of b) +
   c)-type stressless enclitics }
The following phonological words illustrate the various possible combinations:
a) + b) + stress-bearing word + c)
a ’nedala jsi mu ho
and not-gave AUX him it
and you didn't give it to him
a) + b) + stress-bearing word
a ’nevěděl
and not-knew
and he didn't know
a) + stress-bearing word + c)
pan ’Starý by
Mr. Starý AUX
Mr. Starý would
a) + stress-bearing word
ta ’kniha
that book
that book
b) + stress-bearing word + c)
’do školy bych
to school AUX
to school I would
b) + stress-bearing word
’na parkovišti
on parking-lot
at the parking-lot
stress-bearing word + c)
’Dal by ti to
gave AUX you it
He would give it to you

The placement of the c)-type enclitics is governed by both phonology and syntax/semantics. The default position is immediately following the first stress-bearing word in a clause, but this rule is overridden whenever it would involve inserting an enclitic between the constituents of a syntactic/semantic unit. Thus the enclitic auxiliary jsem appears only after the fourth stressbearing word in the following sentence:
Tu ’knihu ’o Čapkově ’životě a ’tvorbě jsem ’nečetl.
that book about Čapeks life and work AUX not-read
I haven't read that book about Čapeks life and work.

Glottal stop

The glottal stop [/] marks morpheme boundaries at the beginning of words and morphemes commencing with a vowel. Though not obligatory, it can serve a distinctive function (enhancing the distinction between a diphthong and a vowel chain), as in:
proudí proudí
[prou`ďiÉ] vs. [pro/UďiÉ]
root proud + ending í prefix pro + root ud + ending í
(s/he/it) gushes (s/he/it) will smoke through
(adapted from Palková 1994: 192)

Word-internally, the glottal stop appears primarily after prefixes ending in a vowel (as in the above example) or j. Word-initially, the glottal stop will devoice a voiced obstruent: v Americe [f/amerItse] in America, od Aleny [/ot/alenI] from Alena. Even speakers who do not use the glottal stop devoice these obstruents.
For speakers who use the glottal stop consistently as prescribed, all syllables have the shape of CV or CVC (where C can stand for a consonant, including the glottal stop or a consonant cluster).

Voicing phenomena

Voiced-voiceless assimilations constitute the most consistent and widespread phonotactic phenomenon in Czech, yet are not reflected in the orthography. Only obstruents participate in voicing phenomena; sonorants neither participate nor condition these assimilations (contrast: směna [sm­ena] shift vs. změna [zm­ena] change). The exceptional obstruent phoneme v participates (can be devoiced: plavky [plafkI] swimsuit), but cannot condition voicing (svoje [svoje] own, not *[zvoje]). Voicing phenomena create positional allophones also for obstruents unpaired for voicing; e.g., h and ch act as voicing partners (cf. sníh [s­iÉx], sněhu
[s­ehU] snow Nsg, GDLsg), although the latter may be realized as [V], as mentioned above in 1.1.

Regressive voicing takes place word-internally and across prefix/preposition boundaries when the last consonant in a cluster of obstruents is voiced: kdo [gdo] who, léčba [leÉdZba] cure, sbalit (prefix s + root bal) [zbalIt] pack, s bratrem [zbratrem] with brother, k dětem [gďetem]to the children.

Devoicing takes place word-finally, and devoicing is regressive word-internally and across prefix/preposition boundaries when the last consonant in a cluster of obstruents is voiceless:
sjezd [sjest] conference, tvář [tvaÉr9ĄS] cheek, podstatný (prefix pod + root stat) [potstatniÉ] essential, nad sochou [natsoxou`] above the statue.

Progressive devoicing involves only the following initial consonant clusters: sh- chř-, kř-, skř-, př-, spř-, tř-, stř-. Devoicing is regular for ř after word-initial voiceless obstruents (initial clusters of cř- and čř- do not exist): střep [str9ĄSep] shard, kříž [kr9ĄSiÉS] cross. Although progressive devoicing is the norm for sh- (shánět [sxaÉ­et] seek), there are several common words that observe the usual regressive voicing rule (shora [zhora] from upstairs).

Consonant clusters

In contrast to its avoidance of vowel strings, Czech has a remarkable tolerance for consonant clusters, permitting strings of four and even five distinct consonants: nad hřbetem [nadhrĄZbetem] over the back, s pštrosem [spStrosem] with the ostrich.
Regular assimilation in place of articulation occurs in the word-internal backing of n to [N] before velars: spánku [spaÉNkU] sleep Gsg. Note also the combination of assimilation and simplification of consonant clusters that occur across the prefix-root boundary, such as rozšířit [roSiÉrĄZIt] spread.

The remaining assimilations or simplifications are limited to more specific environments or lexical items, and some of these modifications (particularly the latter ones) are considered colloquial: the initial j- in consonant clusters is omitted (jsem [sem] I am, jdeme [deme] we are going, jméno [meÉno] name; č- dissimilates to [S] in the word čtyři [StIrĄZI] four; vzp simplifies to [sp-] in vzpomenout [spomenou`t] recall; jablko apple is pronounced [japko]; etc.

There is a tendency to simplify geminate consonants; this tendency is stronger rootinternally and weaker across morphological boundaries: panna [pana] virgin, but
poddůstojník [podduÉstoj­iÉk] sergeant. This tendency is often observed in the spelling of foreign borrowings: tenis [tenIs] tennis, komunismus [komUnIsmUs] communism.

Vowel epenthesis

An epenthetic e can be inserted between a prefix or preposition ending in -C and a following morpheme beginning in C-, however there is no good rule for predicting this, since similar and sometimes even the same environment will show forms both with and without the vowel:
vejít vs. vjet prefix v + walk prefix v + ride
enter (on foot) enter (by conveyance)
seběhnout vs. sběhnout prefix s + run prefix s + run
run down run down
Epenthesis is regular only before the oblique forms of the 1sg personal pronoun, and somewhat less regular before forms of všechen all: nade mnou above me, but nade vším / nad vším above all.
The presence of the same consonant or a voiced/voiceless counterpart makes epenthesis mandatory: ve frontě in line, beze slov without words, ze západu from the west. Epenthesis is optional when the prepositions s and z are followed by the similar consonants š and ž: se školou or s školou with the school, se ženou or s ženou with the wife.
The preposition k toward can have u as its epenthetic vowel in certain set expressions with nouns in initial p-: ku podivu to (ones) surprise; otherwise the normal e appears: ke stolu or k stolu to the table, ke mně to me. The regular morphophonemic alternation of e with a zero in morphemes is discussed in Chapter 2.

Phonology of Colloquial Czech

The primary phonological difference between literary and colloquial Czech is that the former has more fully realized the vowel changes of the Bohemian dialects presented in 0.3, producing a more balanced system for long vowels and diphthongs. Whereas Literary Czech has shifted only its long back vowels (ó > ů/ú and ú > ou), Colloquial Czech has carried out the corresponding changes in its long front vowels as well, producing é > í/ý and ý > ej. Since both etymological é and ý figure as essential components in the inflectional (and to a lesser extent derivational) morphology of Czech, these vowel changes also play a prominent role in differentiating the morphologies of the two registers of Czech. The following is an inventory of these and other significant phonological features of the spoken language. For more details, the reader is referred to Townsend 1990: 23-47 and Sgall&Hronek 1992: 30-37.

1) é > í/ý. The result is usually spelled ý after velars and r (see spelling rules above) and after t, d, n (since it does not condition the shift to palatals), but i, í elsewhere (though spellings vary after l). This change is quite consistent, regardless of position:
LCz mléko : CCz mlíko milk;
LCz nést : CCz nýst carry;
LCz dobré : CCz dobrý good
Nsgn. This vowel change is resisted in foreign borrowings and also a few native roots, which retain é even in the spoken language: malér misfortune (cf. French malheur), lék medicine.

2) ý > ej. Although the vast majority of examples involve ý as the etymological source, after c, l, s, z, and ž this change can include í > ej:
LCz cítit : CCz cejtit feel; LCz zítra : CCz zejtra tomorrow;
LCz síto : CCz sejto sieve.
This change is also reasonably consistent:
LCz mýdlo : CCz mejdlo soap;
LCz malý : CCz malej small,
although it is resisted in most nouns prefixed in vý- and some others: výběr choice, dýka dagger. Recent borrowings such as rýma cold, kýč kitsch do not reflect ý > ej, but older ones often do, as in
LCz brýle : CCz brejle glasses (cf. German Brille) and
LCz rýže : CCz rejže rice (cf. older German riÉs, modern German Reis);
however, the paucity of examples makes it hard to state this as a rule.

3) ú > ou in initial position. Whereas literary Czech avoids this change in initial position, it is occasionally implemented in the spoken language:
LCz úřad : CCz ouřad office;
LCz úzký : CCz ouzký narrow.
This change, however, is stylistically marked, inconsistently realized, and is gradually dying out.

4) V-prothesis before initial o-. Initial o- tends to develop a prothetic back glide in the spoken language, maximally realized as v, although this is sometimes avoided when a v follows the initial o:
LCz ocas : CCz vocas tail, but only ovoce fruit and ovšem of course (and note also the LCz word otec father which is replaced in CCz by tatínek or táta rather than receiving prothetic v-).
Although not entirely consistent, v-prothesis is very common and can even include some foreign words:
LCz olej : CCz volej oil.
It can also occur at the preposition/prefix boundary:
LCz do očí : CCz do vočí into eyes,
LCz poodejít : CCz povodejít step aside.
In contrast, the (etymologically) prothetic j in initial je is dropped in colloquial Czech in the LCz word ještě : CCz eště still, but this is an isolated example.

5) Length adjustments. The spoken register can both shorten vowels that are long in literary Czech, and lengthen vowels that are short. Shortening of long vowels is fairly common when it involves í > i and ů > u in desinences: LCz prosím :
CCz prosim I ask; please,
LCz stolů : CCz stolu tables
Gpl; however shortening of other vowels and outside of desinences is sporadic and optional:
LCz říkat : CCz říkat/řikat say,
LCz spoléhat : CCz spoléhat/spolehat.
Note that shortening can also take place after é > í/ý, producing i/y:
LCz novém: CCz novým/novym new
Lsgm/n. Vowel lengthening is less common and largely restricted to particular words and expressive contexts:
LCz dveře: CCz dvéře door,
LCz nahoře : CCz nahóře upstairs.

6) Consonant adjustments. Consonants in certain clusters tend to be deleted both wordinitially and word-internally, particularly in high-frequency words:
LCz když : CCz dyž when; if,
LCz džbán : CCz žbán pitcher,
LCz který : CCz kerej which,
LCz vezmu : CCz vemu I take,
LCz zvláštní : CCz zvlášní or zláštní peculiar.
On occasion more drastic truncations are observed:
LCz nějaký : CCz ňákej some,
LCz člověk : CCz čék person.
Word-finally spoken Czech regularly drops the final lĘin masculine past forms:
LCz vedl : CCz ved [vet] he led.
Initial s followed by a stop may appear as š in foreign words:
LCz student : CCz študent.
Other consonantal adjustments are restricted to specific lexical items:
LCz ženská : CCz žencká female,
LCz sahat : CCz šahat touch.
The status of most of these consonant adjustments is less a reflection of the norms of CCz than a result of the dynamics of rapid and repeated pronunciation.