Polish phonology

Consonants

The Polish consonant system is more complicated; its characteristic features include the series of affricates and palatal consonants that resulted from four Proto-Slavic palatalizations and two further palatalizations that took place in Polish and Belarusian.

Phonemes

The consonant phonemes of Polish are as follows:

The tongue shape of the postalveolar sounds is similar to the shape postalveolar approximant (one of the realizations of the English phoneme, see also Pronunciation of English /r/). The alveolo-palatals are pronounced with the body of the tongue raised to the hard palate but a greater area of the front of the tongue is raised close to the hard palate compared to the English palato-alveolar sounds. The series are known as "rustling" (szeleszczące) and "soughing" (szumiące) respectively; the equivalent alveolar series (⟨s, z, c, dz⟩) is called "hissing" (syczące).

Polish contrasts affricates and stop–fricative clusters by the fricative components being consistently longer in clusters than in affricates. Stops in clusters may have either a plosive release accompanied by a weak aspiration or a fricated release (as in an affricate) depending on the rate of speech and individual speech habits.

• czysta ('clean' fem.) vs. trzysta or ('three hundred').
• dżem ('jam') with vs. drzem or ('take a nap' imper. sing.).

Both realizations of stop-fricative clusters are considered correct and typically respelled as tsz, d-ż and czsz, dżż respectively in normative descriptions of Polish pronunciation. The distinction is lost in colloquial pronunciation in south-eastern Poland both being realized as simple affricates as in some Lesser Polish dialects. According to Sawicka (1995:150), Dunaj (2006:170), such a simplification is allowed in the standard language variety only before another consonant or before a juncture, e.g. trzmiel or ('bumblebee'), patrz or ('look', imper. sing.).

For the possibility of an additional velar fricative for ⟨h⟩, see § Dialectal variation below. On the same grounds as for Sawicka (1995:146) gives a phonemic status for speakers who have in their system.

Allophones

• are palatalized before .
• have a labiodental allophone , which occurs before labiodental consonants (as in symfonia 'symphony' or konfiguracja 'configuration'). Before fricatives, orthographic nasal consonants m, n may be realized as nasal approximants , analogous to below. This occurs in loanwords, and in free variation with the typical consonantal pronunciation (e.g. instynkt 'instinct').
• are denti-alveolar except before and postalveolar consonants. They are pronounced with the tip of the tongue very close or touching to the upper front teeth and partially the front of the alveolar ridge. In western and northern Poland, is maintained in native words across a morpheme boundary in nk, e.g. sionka ('a small hallway') contrasts with siąka ('((s)he sniffs') . In other parts of Poland, the contrast is neutralized towards , i.e. is used for both. In foreign words ⟨nk, ng⟩ represent .
• are palatalized laminal alveolar before in recent borrowings. They are pronounced with the blade of the tongue very close or touching the alveolar ridge.
• are apical alveolar before apical postalveolar while is apical alveolar before .
• can be assimilated to affricates before , before and before .
• can be assimilated to before and to before .
• are variously described as apical postalveolar or as (laminal) flat postalveolar. They are articulated with a flat, retracted tongue body, the tongue tip being raised and the entire blade moved up and back behind the corner of the alveolar ridge. A recent study shows that and the release of are predominantly alveolar, while the place of articulation of the stop in varies between alveolar and postalveolar. This agrees with characterizations of as alveolar in older sources. They may be described as retroflex to indicate that they are not palatalized laminal postalveolar . Strictly speaking, this is at odds with the narrower definition of retroflex consonants as subapical, in which the tongue curls back and its underside becomes the active articulator. Occasionally, were used in a similar vein.
• become palatalized laminal postalveolar before in recent loanwords.
• are alveolo-palatal . They are articulated with the blade of the tongue behind the alveolar ridge and the body of the tongue raised toward the palate. Before fricatives, is usually realized as a nasalized palatal approximant , for example, państwo ('state/country') , Gdańsk .
• are velar . Before fricatives and word-finally, is realized as nasalized velar approximant . According to Sawicka (1995:127–128, 136), this allophone is non-labialized .
• is primarily velar ; it has the strongest friction before consonants , weaker friction before vowels and weakest friction intervocalically, where it may be realized as glottal (this variant "may appear to be voiced"). has a voiced allophone , which occurs whenever is followed by a voiced obstruent (even across a word boundary), in accordance with the rules given under § Voicing and devoicing below. For example, klechda 'legend, myth' is , dach ('roof') is , but dach domu ('roof of the house') is .
• before are postpalatal . If are acknowledged as phonemes they are as well, but their distribution is limited to contexts before . A postpalatal allophone of appears only in front of .
• is apical alveolar and becomes denti-alveolar before a following denti-alveolar consonant . A palatalized laminal or alveopalatal is used before .
• is apical alveolar. It has been traditionally classified as a trill , with a tap supposedly only occurring as an allophone or in fast speech. However, more recent studies show that is predominantly realized as a tap , sometimes as an approximant or a fricative, but almost never as a trill. One study found that in an intervocalic context a trilled occurs in less than 3% of cases, while a tapped occurred in approximately 95% of cases. Another study by the same researcher showed that in a postconsonantal position, is realized as a tapped in 80–90% of cases, while trilled occurs in just 1.5% of articulations. A palatalized laminal tap is used before in recent loanwords.
• is a palatal approximant . According to Rocławski (1976:123), is reduced and very short after consonants before vowels, for example miasto ('city') , piasek ('sand') .
• is a velar approximant . According to Wierzchowska (1976:123), is most commonly non-labialized ; a labialization being typical only before . A palatalized allophone before is given by Sawicka (1995:128).
• The approximants may be regarded as non-syllabic vowels when they are not followed by a vowel. For example, raj ('paradise') , dał ('he gave') , autor ('author') .
• are regularly devoiced after a voiceless obstruent and optionally after a voiced obstruent which was devoiced. For example, wiatr ('wind') is pronounced , while kadr ('a frame") can be pronounced or . (See § Voicing and devoicing below.)

Distribution

Polish, like other Slavic languages, permits complex consonant clusters, which often arose from the disappearance of yers (see § Historical development above). Polish can have word-initial and word-medial clusters of up to four consonants, whereas word-final clusters can have up to five consonants. Examples of such clusters can be found in words such as bezwzględny ('unconditional' or 'heartless', 'ruthless'), źdźbło ('blade of grass'), wstrząs ('shock'), and krnąbrność ('disobedience'). A popular Polish tongue-twister (from a verse by Jan Brzechwa) is W Szczebrzeszynie chrząszcz brzmi w trzcinie ('In Szczebrzeszyn a beetle buzzes in the reeds').

For the restrictions on combinations of voiced and voiceless consonants in clusters, see § Voicing and devoicing below. Unlike languages such as Czech, Polish does not have syllabic consonants: the nucleus of a syllable is always a vowel.

The consonant is restricted to positions adjacent to a vowel. It also cannot precede y. (For other restrictions on consonants appearing before i or y, see § Distribution above.)

Voicing and devoicing

Polish obstruents (stops, affricates and fricatives) are subject to voicing and devoicing in certain positions. This leads to neutralization of voiced/voiceless pairs in those positions (or equivalently, restrictions on the distribution of voiced and voiceless consonants). The phenomenon applies in word-final position and in consonant clusters.

In Polish consonant clusters, including across a word boundary, the obstruents are all voiced or all voiceless. To determine (based on the spelling of the words) whether a given cluster has voiced or voiceless obstruents, the last obstruent in the cluster, excluding w or rz (but including ż), should be examined to see if it appears to be voiced or voiceless. The consonants n, m, ń, r, j, l, ł do not represent obstruents and so do not affect the voicing of other consonants; they are also usually not subject to devoicing except when surrounded by unvoiced consonants. Some examples follow (click the words to hear them spoken):

• łódka ('boat'), → before the voiceless k
• kawka ('jackdaw'), → before the voiceless k
• także ('also'), → before the voiced ż
• jakby ('as if'), → before the voiced b
• krzak ('bush'), → ; rz does not determine the voicing of the cluster
• odtworzyć ('to replay'), → & → ; w does not determine the voicing of the cluster
• dach domu ('roof of the house'), → ; the rule still applies across a word boundary

In some dialects of Wielkopolska and the eastern borderlands, remains voiced after voiceless consonants.

The above rule does not apply to sonorants: a consonant cluster may contain voiced sonorants and voiceless obstruents, as in król , wart , słoń , tnąc .

Utterance-finally, obstruents are pronounced voiceless. For example, the in bóg ('god') is pronounced , and the in zajazd ('inn') represents . If followed by a word beginning with an obstruent then the above cluster rules apply across morpheme boundaries. When the second word begins with a sonorant the voicing of any preceding word-final obstruent varies regionally. In western and southern Poland, final obstruents are voiced (voicing pronunciation) if the following word starts with a sonorant (here, for example, the in brat ojca 'father's brother' would be pronounced as ). On the other hand, they are voiceless (devoicing pronunciation) in eastern and northern Poland ( is pronounced ). This rule does not apply to prepositional clitics w, z, bez, przez, nad, pod, od, przed which are always voiced before sonorants.

Hard and soft consonants

Multiple palatalizations and some depalatalizations that took place in the history of Proto-Slavic and Polish have created quite a complex system of what are often called "soft" and "hard" consonants. These terms are useful in describing some inflection patterns and other morphological processes, but exact definitions of "soft" and "hard" may differ somewhat.

"Soft" generally refers to the palatal nature of a consonant. The alveolo-palatal sounds ⟨ń, ś, ź, ć, dź⟩ are considered soft, as normally is the palatal ⟨j⟩. The ⟨l⟩ sound is also normally classed as a soft consonant: like the preceding sounds, it cannot be followed by ⟨y⟩ but takes ⟨i⟩ instead. The palatalized velars , and might also be regarded as soft on this basis.

Consonants not classified as soft are dubbed "hard". However, a subset of hard consonants, ⟨c, dz, sz, ż/rz, cz, dż⟩, often derive from historical palatalizations (for example, ⟨rz⟩ usually represents a historical palatalized ⟨r⟩) and behaves like the soft consonants in some respects (for example, they normally take ⟨e⟩ in the nominative plural). These sounds may be called "hardened" or "historically soft" consonants.

The historical palatalized forms of some consonants have developed in Polish into noticeably different sounds: historical palatalized t, d, r have become the sounds now represented by ⟨ć, dź, rz⟩ respectively. Similarly palatalized ⟨s, z, n⟩ became the sounds ⟨ś, ź, ń⟩. The palatalization of labials has resulted (according to the main phonological analysis given in the sections above) in the addition of , as in the example pies just given. These developments are reflected in some regular morphological changes in Polish grammar, such as in noun declension.

Phonological status of palatalized consonants

In some phonological descriptions of Polish, however, consonants, including especially the labials m, p, b, f, w, are regarded as occurring in "hard" and "soft" pairs. In this approach, for example, the word pies ('dog') is analyzed not as but as , with a soft . These consonants are then also analyzed as soft when they precede the vowel (as in pić 'to drink'). Unlike their equivalents in Russian, these consonants cannot retain their softness in the syllable coda (when not followed by a vowel). For example, the word for "carp" has the inflected forms karpia, karpie etc., with soft (or , depending on the analysis), but the nominative singular is karp, with a hard .

Similar considerations lead to two competing analyses of palatalized velars. In Sawicka (1995:146–47), all three palatalized velars are given phonological status on the grounds of their distribution and minimal contrasts between , , and , , e.g. giełda ('stock market'), magiel ('laundry press ') but giętki ('flexible'), higiena ('hygiene'). Phonemes , and do not occur before where they are separated by a distinct e.g. kiosk ('kiosk'), filologia ('philology'), Hiob ('Job'). A system with and but without is given by Rocławski (1976:86), Wiśniewski (2007:187), Jassem (2003:103) and Ostaszewska & Tambor (2000:135). This analysis is based on an assumption that there is actually no but only as chie, hie occur only in loanwords. However, a decomposed palatalization of kie, gie i.e. , in all contexts is a predominant pronunciation in contemporary Polish. Based on that, a system without palatalized velars is given by Strutyński (2002:73), Rocławski (2010:199) and Osowicka-Kondratowicz (2012:223). In such a system palatalized velars are analyzed as , and before and , and before other vowels. This is the main analysis presented above.

The consonants t, d, r (and some others) can also be regarded as having hard and soft forms according to the above approach, although the soft forms occur only in loanwords such as tir ('large lorry'; see TIR). If the distinction is made for all relevant consonants, then y and i can be regarded as allophones of a single phoneme, with y following hard consonants and i following soft ones (and in initial position).

Glottal stop

In more contemporary Polish, a phonetic glottal stop may appear as the onset of a vowel-initial word (e.g. Ala ). It may also appear following word-final vowels to connote particular affects; for example, nie ('no') is normally pronounced , but may instead be pronounced or in a prolonged interrupted . This intervocalic glottal stop may also break up a vowel hiatus, even when one appears morpheme-internally, as in poeta ('poet') or Ukraina ('Ukraine') . A relatively new phenomenon in Polish is the expansion of the usage of glottal stops. In the past, initial vowels were pronounced with an initial voiceless glottal fricative (so that Ala was pronounced ), pre-iotation (so that igła ('needle') was pronounced ), or pre-labialization (so that oko 'eye' was pronounced ).

Dialectal variation

In some Polish dialects (found in the eastern borderlands and in Upper Silesia) there is an additional voiced velar fricative , represented by the letter ⟨h⟩. It may be actually a voiced glottal fricative [] for some speakers, especially word-finally. In most varieties of Polish, both ⟨h⟩ and ⟨ch⟩ represent .

Some eastern dialects also preserve the velarized dental lateral approximant, , which corresponds to in most varieties of Polish. Those dialects also can palatalize to in every position, but standard Polish does so only allophonically before and . and are also common realizations in native speakers of Polish from Lithuania, Belarus and Ukraine.

Rocławski (1976) notes that students of Polish philology were hostile towards the lateral variant of ⟨ł⟩, saying that it sounded "unnatural" and "awful". Some of the students also said that they perceived the lateral ⟨ł⟩ as a variant of ⟨l⟩, which, he further notes, along with the necessity of deciding from context whether the sound meant was or , made people hostile towards the sound. On the other hand, some Poles view the lateral variant with nostalgia, associating it with the elegant culture of interwar Poland.

In the Masurian dialects and some neighboring dialects, mazurzenie occurs: postalveolar merge with the corresponding dentals unless is spelled ⟨rz⟩ (a few centuries ago, it represented a fricative trill similar to Czech ⟨ř⟩, distinct from ; only the latter sound occurs in modern Polish, with the exception of a small number of dialects).

Stress

The predominant stress pattern in Polish is penultimate: the second-last syllable is stressed. Alternating preceding syllables carry secondary stress: in a four-syllable word, if the primary stress is on the third syllable, there will be secondary stress on the first.

There must be a syllable for each written vowel except when the letter i precedes another vowel (in that case the i represents either , or palatalization of the preceding consonant, or both, depending on analysis; see Polish orthography and the above). Also, the letters u and i sometimes represent only semivowels after another vowel, as in autor ('author'); these semivowels mostly occur in loanwords (so not in native nauka 'science, the act of learning', for example, nor in nativized Mateusz 'Matthew').

Some loanwords, particularly from classical languages, have the stress on the antepenultimate (third-last) syllable. For example, fizyka () ('physics') is stressed on the first syllable. That may lead to a rare phenomenon of minimal pairs differing only in stress placement: muzyka 'music' vs. muzyka – genitive singular of muzyk 'musician'. When further syllables are added at the end of such words through suffixation, the stress normally becomes regular: uniwersytet (, 'university') has irregular stress on the third (or antepenultimate) syllable, but the genitive uniwersytetu () and derived adjective uniwersytecki () have regular stress on the penultimate syllables. Over time, loanwords tend to become nativized to have a penultimate stress.

Another class of exceptions to the usual stress pattern is verbs with the conditional endings -by, -bym, -byśmy etc. Those endings are not counted in determining the position of the stress: zrobiłbym ('I would do') is stressed on the first syllable and zrobilibyśmy ('we would do') on the second. According to prescriptive grammars, the same applies to the first and second person plural past tense endings -śmy, -ście although this rule is often ignored in colloquial speech (so zrobiliśmy 'we did' is said to be correctly stressed on the second syllable, although in practice it is commonly stressed on the third as zrobiliśmy). The irregular stress patterns in the presence of these verb endings are explained by the fact that the endings are detachable clitics rather than true verbal inflections: for example, instead of kogo zobaczyliście? ('whom did you see?') it is possible to say kogoście zobaczyli? – here kogo retains its usual stress (first syllable) in spite of the attachment of the clitic. Reanalysis of the endings as inflections when attached to verbs causes the different colloquial stress patterns.

Some common word combinations are stressed as if they were a single word. That applies in particular to many combinations of preposition plus a personal pronoun, such as do niej ('to her'), na nas ('on us'), przeze mnie ('because of me'), all stressed on the bolded syllable of the preposition.