Czech final devoicing
At the end of a Czech word, voiced consonants become voiceless. led (ice) is pronounced "let." dub (oak) is "dup." muž (man) is "muš." This is automatic and absolute — every native speaker does it without thinking.
The voicing pairs
Each voiced consonant has a voiceless partner — same place and manner of articulation, just with the vocal cords off. At word ends, Czech swaps voiced for voiceless:
| Voiced (writing) | Voiceless (sound) | Example |
|---|---|---|
| b | p | dub → "dup" (oak) |
| d | t | led → "let" (ice) |
| ď | ť | loď → "loť" (boat) |
| g | k | gang → "gank" |
| h | ch /x/ | sníh → "sních" (snow) |
| v | f | lev → "lef" (lion) |
| z | s | mráz → "mrás" (frost) |
| ž | š | muž → "muš" (man) |
Why this exists
Producing a voiced consonant requires your vocal cords to vibrate. At a word boundary your speech apparatus relaxes — the vocal cords are about to stop entirely for a pause, and they tend to stop a bit early. Many languages have this pattern: German, Russian, Polish, Dutch, Turkish, and others all do final devoicing. English doesn't, which is why English speakers find it counter-intuitive.
Why English speakers miss it
In English, you can hear a clear difference between "bat" and "bad" at word end. The d in "bad" is voiced — your vocal cords keep humming through it. This is unusual cross-linguistically, but it's what we're trained to do.
When you read led aloud and produce the d as English-style voiced /d/, you sound foreign even though every individual sound is correct. Czechs hear "led" pronounced with voiced d as a distinct accent marker — it's one of the most reliable L2 tells.
How to fix it
Override the spelling instinct
When reading aloud, automatically swap the final letter for its voiceless partner in your head: see led → produce let. After a few hundred reps it stops needing conscious effort.
Pair-drill the obvious cases
Take a word and its plural or genitive form (where the voiced consonant returns because it's no longer word-final): led "ice" / ledu "of ice." Singular is "let," plural keeps the voiced d. Practicing the pair makes the rule visceral.
Drill phrases
Tap to hear native pronunciation. Listen for the voiceless final consonant.
- led — IPA: /lɛt/ · Ice · Final devoicing
- chléb — IPA: /xlɛːp/ · Bread · Final devoicing
- dub — IPA: /dup/ · Oak · Final devoicing
- lev — IPA: /lɛf/ · Lion · Final devoicing
- muž — IPA: /muʃ/ · Man · Final devoicing
- nůž — IPA: /nuːʃ/ · Knife · Final devoicing
- mráz — IPA: /mraːs/ · Frost · Final devoicing
- had — IPA: /ɦat/ · Snake · Final devoicing
- sníh — IPA: /sɲiːx/ · Snow · Final devoicing
- krev — IPA: /krɛf/ · Blood · Final devoicing
Related: voicing assimilation
Final devoicing has a cousin called regressive voicing assimilation — when consonants meet within a word, voicing spreads from the second to the first. So kdo "who" is pronounced /ˈɡdo/ (k voices to g because of the following voiced d), and vsadit "to bet" is /ˈfsaɟɪt/ (v devoices to f because of the following voiceless s).
Final devoicing is the special case where the "consonant" you're devoicing against is silence. The general principle: in Czech, voiced and voiceless consonants want to match.