Apart from [iɛ] and [uɔ], there are also vowel+glide sequences [ɛɪ̯,æɪ̯,aɪ̯,iu̯,ɨu̯,au̯], which are very common. Rarer sequences include [uɪ̯], [ɔɪ̯] and [ɔu̯], with the last one occurring only in onomatopoeias and loanwords. Phonemically, they are all sequences of two phonemes, rather than proper diphthongs. In some dialects, [iu̯] and [ɨu̯] fall together as [ɛu̯]. [au̯] can also merge with [ɔu̯] as [ɔu̯].[4]
^The alveolo-palatals occur only in some non-standard dialects, with the fricatives /ɕ,ʑ/ being more common than the affricates /t͡ɕ,d͡ʑ/.[6]
^/f/ occurs only in some loanwords, such as ortografeja transl. orthography. In other loanwords, it is replaced by /p/, as in kopejs transl. coffee.[7]
^/x/ occurs only in some loanwords, such as tehnologeja transl. technology. In other loanwords, it is replaced by /k/, as in kokejs transl. hockey.[7]
^ ab/v,vʲ/ are traditionally classified as approximants [ʋ,ʋʲ] which phonetically may be fricatives [v,vʲ].[7]
^ abFor the approximants /ɪ̯/ and /ʊ̯/ the contrast between their hard and soft versions is phonetically realized as a contrast between retraction and advancement. The "soft" counterparts of /ɪ̯/ and /ʊ̯/ are /i̯/ and /u̯/.[8][9]Brejdak (2006, p. 198-199) considers those to have phonemic status and argues that the contrast between them and the plain /ɪ̯/ and /ʊ̯/ corresponds to the soft/hard contrast. However, Nau (2011, p. 13) considers the phonemic status of /u̯/ and especially /i̯/ (which he transcribes with a non-IPA symbol ⟨jʲ⟩) as questionable. If the difference is considered to be allophonic, the resulting single palatal approximant can be transcribed with ⟨j⟩ and the corresponding labio-velar with ⟨w⟩. In this article, the difference is assumed to be phonemic.
There are two phonemic tonal accents in Latgalian, which appear only on long syllables, i.e. those with a long vowel, a diphthong, or a sequence of a short vowel and a sonorant. These are falling (also called level) and broken (also called sharp). However, there are only a handful of minimal (or near-minimal) pairs, such as [rɛ̀ɪ̯tʲ] 'swallow' and [rɛ̂ɪ̯t] 'tomorrow', both written reit.[9]
Phonetically, both of the tonal accents are falling; the falling accent is realized as an even decrease in intensity and pitch, whereas the broken accent is realized as a sudden decrease in intensity and pitch.[9]