You set yourself a small project, sometimes enlarging an existing code example, sometimes thinking about the language's strengths - e.g., if you're learning a language designed for practical extraction and reporting, you'll aim for some kind of extraction report, say, a log analyzer.
The fact is that whatever makes you pull your weight in any given language is a project. Even if the project aim is "learn me this language".
So with Scala, you might try with a simple web service producing/consuming XML packets (there are small examples for that; you can start by studying them and tying them together).
You'll want to start small because this kind of project, with a none-too-defined goal, tends to snowball; for example you might decide to develop both a web service and its client. This snowballing isn't a bad thing if you keep it under control, that is, you do not bite more than you can chew at any one time. You might be tempted to add database querying before you're sure of your "web service management" skills; if you do, you risk reaping an overwhelming crop of bugs and grow dissatisfied and disillusioned about the language and/or your own skills. Better to return mock-up data and placeholders for all the things you're not yet assured of; that will also give you a measure of how many topics you still have to cover.
This, however, is just my own point of view. You might want, or even need, to find your own way, the one that suits you best.