Connected Speech and Assimilation

There is a gap between the Swedish you read and the Swedish you hear, and it trips up advanced learners who can parse a newspaper but freeze when a Swede orders coffee. The gap is connected speech: in running speech, words do not keep the tidy shapes the spelling gives them. Final consonants fall off the most common little words, sounds at word boundaries merge into single new sounds, and whole syllables vanish. None of this is sloppy or substandard — it is how every native speaker, including a news anchor, actually talks. This page covers the regular processes so that you can both understand fast Swedish and avoid the giveaway of pronouncing every letter. The retroflex sounds these processes lean on are detailed on Retroflex Consonants, the melody that rides over connected speech is on Sentence Intonation, and the regional spread of these reductions is on Regional Differences in Everyday Speech.

The big one: function words lose their final consonant

The most pervasive reduction in spoken Swedish is the dropping of final consonants in high-frequency function words. These are the words you say hundreds of times a day — pronouns, the conjunction "and," common prepositions — and in connected speech they shrink. Learn the reduced forms as the default, not as exceptions.

The two that matter most are det → "de" and jag → "ja", because they are in almost every sentence. The final -t of det and the final -g of jag are simply not pronounced in normal speech; the careful, every-letter version sounds like a foreigner reading aloud or a teacher dictating. Likewise och is almost always just "å" [ɔ] — Swedes write och and say å so habitually that learners often hear "å" and cannot find it in the dictionary, because the word on the page is och.

Retroflexion across word boundaries

Swedish has a sandhi process where an /r/ followed by a coronal consonant — /t d n l s/ — fuses into a single retroflex sound, the tongue curled back. You meet this inside words first (kort [kɔʈ], bord [buːʈ], barn [bɑːɳ]), but the crucial point for connected speech is that it crosses word boundaries. The /r/ at the end of one word and the /d/, /t/, /n/, /l/ or /s/ at the start of the next merge into one retroflex consonant.

The textbook case is är du "are you," where r + d becomes a single retroflex phrase comes out roughly [ɛ‿ɖʉ], "ä-du," with no separate r and no separate d — one curled-back sound joins the two words.

This is why fast Swedish sounds like it has no spaces: the retroflex sandhi literally erases the seam between an r-final word and a coronal-initial one. (Note: this works only in alveolar-/r/ varieties — Southern/Scanian Swedish, with its back uvular r, does not retroflex this way. See the regional page.)

Vad gör du? — careful versus connected

It helps to take one ordinary question and watch it collapse. Vad gör du? "What are you doing?" is, letter by letter, [vɑːd jœːr dʉː]. In real speech almost nothing of that survives intact:

• vad loses its -d → "va" [va]
• gör keeps its sje-less initial but its r reaches forward
• du merges with the preceding r via retroflexion

The result is something like [va ˈjœːɖʉ] — "va görru" — three written words pronounced as essentially two compressed beats. A learner who says all four consonants ([vɑːd jœːr dʉː], clean and separated) is perfectly intelligible but unmistakably non-native.

Assimilation and elision in fast speech

Beyond the function-word drops, ordinary fast speech assimilates and deletes more freely:

• Nasal assimilation: a final /n/ takes the place of a following stop. en gång "once" tends toward [eŋ ɡɔŋ]; utan mig "without me" → [ˈʉːtam mɛj], the n of utan becoming m before the m.
• Cluster simplification: awkward consonant runs at boundaries shed a member. och sedan "and then" collapses toward [ɔ ˈseːn] — och to "å," sedan to "sen."
• Whole-word contraction of frequent collocations: sådan/sådant/sådana → sån/sånt/såna; någon/något/några → nån/nåt/nåra; de/dem → dom. These are so universal that the contracted spellings (sån, nån, dom) appear in print dialogue.

Should you imitate this — or just understand it?

A fair question for a learner: do you have to drop these consonants, or just recognise them? The honest answer is layered. Understanding the reductions is non-negotiable — if you cannot decode det → de, och → å, vad gör du → va görru, you will not follow normal conversation, full stop. Producing them is a spectrum: the function-word reductions (de, ja, å, me, dom, nån, sån) are so universal that using the careful forms instead actually sounds more foreign, so adopt those. The faster assimilations and elisions you can let come naturally as your speech speeds up; forcing them prematurely can sound like caricature. Aim first to understand everything and to reduce the four or five workhorses, and let the rest follow.

Common Mistakes

Key Takeaways

• Final consonants drop from high-frequency function words: det → de, jag → ja, och → å, med → me, är → ä. Learn these reduced forms as the default; the spelled-out versions sound stilted.
• Retroflexion crosses word boundaries: r + /t d n l s/ at a seam fuses into one retroflex sound — är du → äru, var du → va-du — which is why fast Swedish has no audible gaps. (Alveolar-r varieties only; not Southern/Scanian.)
• One ordinary question shows it all: Vad gör du? → "va görru", and Det är bra → "de ä bra."
• Fast speech also assimilates nasals and contracts collocations: de/dem → dom, något → nåt, sådant → sånt, sedan → sen.
• Understand all of it; produce at least the universal function-word reductions — using the careful forms is itself a non-native marker.