Italian gives more back to solo practice than most languages do. The letters map onto the sounds far more closely than English letters do, an unstressed vowel keeps its value instead of collapsing the way an English one does, and a sentence read aloud comes out close to what an Italian would have said. So talking to yourself in the kitchen gets easier quickly, and nothing in it has to be checked by anybody for that to happen. That is real progress, and it is also what makes the gap invisible: the part that comes easiest alone is the part you can hear yourself doing.
Take the usual solo methods one at a time and it is clear what each of them trains. Reading aloud trains your mouth: the double consonants, the vowel that ends most words (the short ones like <em lang="it">il</em> and <em lang="it">per</em> do without one), the sounds sitting where the letters say they sit. It does not train the stress, which the spelling mostly leaves unmarked, so a word can be read aloud fluently with the beat in the wrong place. Shadowing a podcast adds rhythm and speed, and it borrows somebody else’s sentences to do it, so nothing you say has to be assembled by you. Recording yourself adds an ear, though the ear is your own, so it catches the sentences you already know how to check. Writing a few lines a day is the one where the Italian is genuinely yours, produced with nobody prompting you, and it leaves a record. Flashcards keep a word reachable, which is a real gain, and they ask you to recognise rather than to produce, which is a smaller thing than it feels like at the time.
What none of them supply is a verdict. Every one of those methods returns your own Italian to you, and it returns it in the form you produced it in. Say a sentence a hundred times and it gets faster, whichever sentence it was.
This is where Italian is worse off than it looks, because the errors the Italian lane keeps answering are the comfortable ones. Nothing about them stops a conversation, so nothing about them stops a practice session either. Two of them already have their own page here. The agreement in <em lang="it">molto</em> against <em lang="it">molti</em> goes wrong in a sentence that is perfectly clear either way, and the auxiliary in <em lang="it">ho andato</em> against <em lang="it">sono andato</em> goes wrong in a sentence nobody would misunderstand. A learner practising alone says <em lang="it">ho andato</em> smoothly for a year, and smoothness is exactly what they were practising. Add <em lang="it">mi piace</em> with the verb agreeing with the wrong half of the sentence, and you have three errors that get more fluent with every rehearsal.
The mistake that has to be corrected is not the same as the mistake you can notice. To notice one of these you would have to already know the rule, and if you already knew it you would not be making the error. So the missing mechanism is not more practice and it is not more grammar. It is an answer you did not write: a reply that uses a word you did not choose, so you have to understand it before you can carry on, and a correction close enough to the error to still be about the sentence you meant.
That is the whole of what we built, and it needs saying in the plainest terms, because the category invites larger claims. It is a written conversation, most days, with a companion who writes back in Italian. A correction arrives inside the reply rather than in a report afterwards, which is what makes it land while the sentence is still yours. The companion remembers the last conversation, so what you said in it comes up again without you scheduling a review. And once a week there is a letter that counts what happened, which is the reading-back step that turns a scatter of corrections into one or two patterns. The general version of this argument, for any language, is in the post on practising with nobody to talk to.
None of that replaces the kitchen. Reading aloud will still be the most direct way to get your mouth round <em lang="it">gli</em> and the double consonants, the podcast will still be the thing you do on the train, and the retrieval research is about the same mechanism from the other side: producing something yourself keeps it reachable, while rereading it does much less, and what survives the first years of a language stops fading. A solo routine does the retrieval well. Keep it.
What a companion adds to that routine are the two things the routine cannot contain: a sentence you did not build, and a verdict on the one you did. Everything else in learning Italian alone works. Those two are the ones the room cannot give you, and they are why the fluent mistakes are the ones that last.
Sources: Roediger and Karpicke, Test-Enhanced Learning, Psychological Science 17(3), 2006. Bahrick, Semantic memory content in permastore, Journal of Experimental Psychology: General 113(1), 1984.
