Voicing Assimilation in Clusters

You already know that a voiced consonant devoices at the end of a word (го́род is "GO-rat"). Voicing assimilation is the bigger rule that this is just a special case of: inside a cluster of obstruents, every consonant takes the voicing of the LAST one. The cluster decides as a group, and the deciding vote belongs to the consonant on the right. This is regressive assimilation — the influence flows backwards, from a later sound to an earlier one — and once you grasp it you stop reading Russian letter by letter and start hearing it cluster by cluster. That shift is the whole point of this page.

The core rule: the last obstruent decides the whole cluster

An obstruent is a consonant made by obstructing the airflow — the stops /p b t d k g/ and the fricatives /f v s z ʃ ʒ x/. Russian obstruents come in voiced/voiceless pairs:

When two or more obstruents sit next to each other, they cannot disagree about voicing. The cluster snaps into agreement with the rightmost obstruent: if it is voiceless, everything in front of it goes voiceless; if it is voiced, everything in front of it goes voiced. The spelling does not change — only the pronunciation does.

Direction 1: devoicing (voiced → voiceless before a voiceless consonant)

When a voiced obstruent is immediately followed by a voiceless one, the voiced consonant loses its voice.

Direction 2: voicing (voiceless → voiced before a voiced consonant)

The rule runs in the other direction too. A voiceless obstruent followed by a voiced one (other than в — see below) becomes voiced.

Two consonants that play by special rules: в and the sonorants

The clean rule above has two important refinements. Both involve consonants that are "asymmetric" — they obey assimilation from the right but do not impose it on their left.

в devoices and voices, but does NOT trigger voicing

This is the famous quirk. The letter в behaves normally as a target: it devoices before a voiceless consonant (всё → "fsyo," за́втра → "ZAF-tra"). But в is special as a trigger: a voiceless consonant in front of в does not voice. So в is the one voiced obstruent that lets a preceding voiceless consonant stay voiceless.

The historical reason is that в was once a glide-like sound (closer to /w/), patterning with the sonorants rather than the true obstruents — and it kept that one privilege. You don't need the history; you need the rule: в can be devoiced and can be voiced, but it never voices the consonant before it.

Sonorants are "transparent"

The sonorants — л м н р й — are always voiced and have no voiceless partner, so they take no part in the assimilation game. They neither devoice nor force voicing on a neighbour; a cluster looks straight through them as if they weren't there.

The big insight: assimilation crosses word boundaries

This is what most apps gloss over, and it is the heart of the matter. Russian prepositions and prefixes are pronounced as one phonological unit with the word that follows. So the assimilation rule reaches across the space between a preposition and its noun, or across the seam inside a prefixed verb. The single most striking consequence: the preposition в ("in") is pronounced "v" or "f" depending entirely on the next word's first sound — even though it is always spelled в.

So the same letter в is three different sounds — "f" in в шко́ле, "v" in в Москве́ — purely because of the word that follows it. A learner who hears "the letter в" will never get this; a learner who hears "the cluster" gets it for free. Read whole phrases, not isolated words.

Final devoicing is just this rule at the edge of a word

The final-devoicing you already learned — го́род → "GO-rat," хлеб → "khlyep," муж → "mush" — is voicing assimilation with the right-hand neighbour being silence. A voiced obstruent with nothing voiced after it has nothing to keep it voiced, so it falls to its voiceless partner. It is the same machine, running at the word edge.

This is why final-devoicing and cluster assimilation are usually learned together and summarised on the assimilation-summary page: they are one phenomenon.

Comparison with English

English assimilates voicing too, but mostly progressively (left to right) and at suffix boundaries — the plural ending agrees with the sound before it: cats /kæts/ but dogs /dɒgz/. Russian does the opposite: the influence flows backwards, and the deciding consonant is on the right. English also has no clitic-preposition fusion of the Russian kind, so the idea that the preposition "in" could be pronounced "f" or "v" by context feels alien. The fix is conceptual, not muscular: you can already devoice and voice consonants effortlessly in English; you just have to point that ability backwards and let it reach across word boundaries.

Common Mistakes

Key Takeaways

• In a cluster of obstruents, the last obstruent decides the voicing of the whole cluster (regressive assimilation).
• It works both ways: voiced → voiceless before a voiceless consonant (во́дка, всё), and voiceless → voiced before a voiced one (сде́лать, футбо́л, вокза́л).
• в is the exception as a trigger: it can itself be devoiced/voiced, but it never voices the consonant before it (твой stays "tvoy").
• Sonorants (л м н р й) are transparent — they neither assimilate nor trigger assimilation.
• Assimilation crosses word boundaries with prepositions and prefixes, so в is "f" in в шко́ле but "v" in в Москве́ — hear the cluster, not the letter.
• Final devoicing is the same rule with silence on the right; see final-devoicing and the combined assimilation-summary.