I get asked a lot about learning languages, so I have a few comments about it here. Hopefully I can awaken you from some dogmatic slumbers about language.
This is hard for people to understand because I think most monolingual people think that languages are just different word lists that people use. As a result, 101 students will manually look up every word in the dictionary to translate. This actually increases the mental load of learning a language because people have the idea that to speak it, they have to think of something in English, then translate the sentence word by word, then say that. What a pain.
So what is a language if not words? It really is a set of constraints as to how words can go together: what order they go in when modifying each other, but also languages are morphology. Verb endings and tenses and such are literally the most important part of a sentence. If you don't have a productive and reflexive use of verbs, you are literally just going to be reciting nouns you know like a monkey.
This is actually why I recommend people learning Romance languages or German to use Michel Thomas's audio. Thomas doesn't lecture at all about what he's doing, but he focused only on using verbs and building up basic expressions from the bottom up until it's understood reflexively by students. To actually learn any language, this is more or less what you are going to have to mentally do anyway in the process.
I would say it's actually possible to fluently speak a language knowing only about 50 words. If you understand the "grammar" of a language, you can basically get by anywhere anytime with a couple dozen words only. What words you don't know can easily be figured out, but you can't wing it with grammar and you can't wing it with morphology.
Granted, the same is true of programming "languages" as well, weirdly enough. No one would think "knowing a [computing] language" means just knowing all the function and variable names. The important thing is knowing the syntax of how you put functions (loosely verbs) and variables (loosely nouns) together. After all, variable names are always different and functions can be easily invented too or called from some obscure library. Someone who knows a language is someone who can use its syntax to produce novel expressions. If you take a Python script, replace its functions with C functions, it's still Python, just calling a bunch of undefined functions. People can only get away with even sort of believing this in the domain of human languages if you just don't know enough and end up assuming that all languages just work the same.
So really when you learn a language, you can't look at it as new words, but new patterns of speech that interconnect in a logical way.
Speaking fluently in that language means being able to use and combine its basic constructions into complex thoughts put in words. This is why I'm really against "translating in your head." If you're doing that, you're not actually using the language. You're teaching yourself a silly English-word-replacement game. I know it's very hard for word-thinkers not to think in words, but if you can't stop doing that for a second, you're not going to be able to learn a new language.
There's this lazy idea that somehow if you passively sit around and watch people using a language this will somehow endow you with the ability to flexibly produce a language in the same way you see others using it. People want to believe it because they want to be able to watch TV or play a cell phone game like Duolinguo or valueless Rosetta Stone-like software and somehow gain competence in a language.
It's not going to happen ever. Learning to play a boring computer game using words from a different languages is not the same as learning to speak the language.
You might say of "just listening to speech" that "that's what children do," but that's not true at all. Children try pretty hard to participate and understand conversation. They sometimes have a desperate personal need to understand each passing sentence and hear the language they are trying to learn for hours a day for years. You watching some forgettable movie in the background as you play with your phone don't.
If you want to know if you are actually learning a language, ask yourself that. People are weirdly afraid against actually thinking through things and making new expressions in other languages when that's exactly how you learn them.
A lot of language nerds love to email me about their Anki cards or their harebrained schemes for mass-memorizing words as if they're an Asian studying for a chemistry test. Given what I've said about "learning words," you can guess my opinion on that. Once people abandon the lazy route, sometimes they take up the via dolorosa: the route of suffering and assume that training themselves like a Pavlovian dog will help them become fluent in a language.
In reality, the only question that matter is: "Are you actually thinking?" Are you actually going through the mental process of creating new sentences in a new language?
When I was learning Latin obviously I had no Latin-speaking friends and could barely get my hands on anything Latin-related. But after I learned the basics of the language I started thinking in it constantly. First that starts in my always implicitly translating English song lyrics or ads in my head into Latin. That's actually difficult if you're dealing with something modern and idiomatic. Not as bad with church songs. As time goes on, I would overtly remember things in Latin sentences instead of English. If I mumbled something under my breath I would make sure it was Latin. At all points in time, I was thinking about how the language was structured and what it meant to produce sentences in it.
The sad fact is that most people who "learn" languages in school treat them as advanced cross-word puzzle like games where they don't actually think in the language, but have hilarious mnemonic devices in their head for relating what they want to say in English with something in the language they're learning.
If anything, you should become worse at translating the further you go on and the more independently you can stand on your own in another language.
Latin is a good example. I can read and comprehend Latin very well, but if asked to translate what I'm reading, I find that more and more difficult the better I read Latin. Now it's easy for me to report the meaning of a passage, but phrase-by-phrase translation is something you have to think through because Latin and English are structurally very different. This isn't just word order, but even how a Latin speaker approaches expressions and the kinds of phrases they use can translate only very delicately into English.
The problem nearly doesn't exist between English and Spanish, which are basically the same language. I'm sure someone who only knows Spanish will feel like English and Spanish have many differences, but in the context of other languages, like Latin or Chinese or Japanese, it's hard not to view English and Spanish as having basically the same kind of syntax 95% of the time. That actually goes for most modern European languages.
Every language has its own set of phonological rules that determine what particular sounds are said how and where. Phonological rules give us "our accents." When someone speaks English in an accent, they are really just speaking English using the phonological constraints of whatever language they're more familiar with. If they speak English competently, there's at least some extent to which they are abandoning their native phonological rules.
When you first start learning a language, you might read something aloud and say "I sound stupid." This is because your natural way of speaking is obviously to say everything with an accent consistent with English. You can probably remember the apathetic jock in Spanish class or whatever who religiously pronounced every Spanish word he mindlessly read with an almost intentionally non-Spanish accent.
To actually speak another language is to adopt the phonological tendencies and even the prosodic and tonal traits of that language. When you initially do that, you will probably sound very stupid to yourself since violating phonological rules you're familiar with always sounds wrong. If you do overcome that illusion of felt stupidity, you won't sound stupid when it counts. If you refuse to improve your accent immediately and from the beginning you will sound like an utter moron forever.
There's actually a trick too: when you imitate a foreign accent, you are actually implicitly adopting the phonological rules of their language that you have noticed in real life. My suggestion is when you are starting out, read the other language in what you'd guess would be a stereotypical accent of the person speaking the language. If your imitation is good, you're speaking their language without an accent.
That reminds me.
There's an idea in academic and clinical linguistics as well as popular culture that children have a magical plasticy of the brain that makes them uniquely good at learning languages. This is supposed to be the reason why children learn languages "fast" and adults don't. I think this is a myth. You don't have to send me all the "proof" about this (don't worry, the Universities of Georgia and Arizona would've failed me totally if I hadn't seen it for my linguistics degrees there). I sort of assumed that this was true for years, but on further thought, I think it's just a conspiracy of irrelevant data and copes... or at least, it's not nearly as true as people pretend it is: adults are just about as capable of learning languages in most senses.
After all, think about it, children actually take several years to function in a language, which is often much longer than an adult that knows what he's doing. The Michel Thomas style tapes which I alluded to above are good at giving an adult a passable diving-board for a language in about 8 hours. It can be done. You can also give an adult a crash-course in phonology and articulatory phonetics that will make it easy to understand and with practice produce the sounds children take years to master.
The motivation of a child and adult are utterly different. A language-less child has lots of reasons to invest most of his mental life in attention to language. Apathetic adults don't.
What I really get sick of is doomer adults who cope with their laziness by talking about how hard it is to learn a language as an adult. Many adults still learn languages all the time. There is some circumstantial evidence that infants cue into some acoustic cues and other things quicker than adults, but I think in most cases we're just looking at infants semi-consciously honing in on what details they've acknowledged to be linguistically relevant. In reality, developed humans have huge institutional and intellectual advantages to learn.