I find that it is complementary. If I spend a lot of time reading about a subject, I can build up a great base of knowledge, but if I then later do projects involving that subject, I tend to find out that a lot of my priorities were upside down, and tend to see a lot of limitations and conditions that are hard to pick up from reading.
On the other hand, if I spend a lot of time practicing without any theory to guide me, then I find that I make some progress, but often when I later learn theory, I see huge blind spots and mistakes in my work, as well as huge shortcuts and lots of tricks that I should've applied.
I would also bet it depends heavily on how you measure it. Pure book learning probably boosts pure book scores more, while pure practice learning probably boosts practical task performance more (at first at least, until you've calibrated the book knowledge to the real world).
What a coincidence - I was just writing about this topic myself!
I think the challenge is partly that project-based learning (PBL) is under-defined and contextual, partly that it's time/money/energy/creativity-intensive, and partly that it's dependent on a pre-existing level of competence, passion and interest in the particular PBL topic.
One way of framing it is that a person has a rich, multifaceted goal:
If a person like this also is willing and able to throw a significant amount of their time and money at PBL, I think that they'll tend to see stellar results compared to the traditional trod through a textbook with occasional labs.
I'd be very curious to know if there was any successful research into predicting whether or not a particular PBL effort would succeed or fail. It seems like the same approach that's been used to evaluate the plausibility of social psychology research might be useful here. Give people a series of descriptions of PBL experiments, and ask them to predict whether or not the experiment would or would not reach significance or have a given effect size. I'd bet that people would be good at making such predictions.
Followup: I scanned the first listed paper from the systematic review. I think the challenge to arranging the prediction tournament I proposed above would be in producing descriptions of the studies.
Consider the review's approach to finding studies for inclusion.
PjBL and PBL are usually described as active, student-centred methods of instruction that encourage students to work in collaborative groups on real-world questions or challenges to promote the acquisition of higher-order thinking skills, while teachers act as facilitators of learning...
On January 23th 2020 the first author (MF) performed an electronic search on the Web of Science, PsycInfo, and ERIC entering the terms “(project based OR problem based) AND (learning OR intervention OR approach OR instruction)” into the Topic field...
The studies were only included if they met the following criteria: c1) the aim was to evaluate the effect of PjBL on content knowledge; c2) they followed a pre-post design with control group; c3) the target sample comprised students from kindergarten to grade 6; c4) they were written in English; and c5) they were peer-reviewed.
To avoid being excluded on the grounds of topic irrelevance, a study only needed to:
Note in particular that, by the definition they give, only some amount or form of encouraging and facilitating of any type of collaborative, "real-world" "questions or challenges" is required to fit their definition of PjBL. This is an extremely expansive definition.
You yourself note that the most obvious reason this review finds mixed results is that, in your words, "the studies weren't very well done." I think this is correct, but we also can chalk up the problem to:
(cross-posted from my blog, Sunday Stopwatch)
"But when am I going to use any of this?" - every single kid in school ever.
There's an idea that project-based learning is more efficient than the traditional school curriculum. It makes sense, but so did the idea that learning styles are important, and that turned out to be false.
Here are a couple of thoughts before I delve into the research:
I expect that having a project-based approach serves as a minor boost to motivation, but doesn't really make a difference regarding knowledge retention.
So, what does the research say?
This systematic review says that it's inconclusive because the studies weren't done well. And this is only for small kids. I could imagine that different approaches might have different effects depending on age.
This paper describes how biomedical engineers collaborated on some projects and concludes that project-based learning works well. Direct quote: "It is felt that the projects were successful to some extent. They certainly achieved several important learning outcomes of teamwork, ability to apply theoretical principles from multiple disciplines, effective communications, creative problem solving, and awareness of the importance of globalization especially in the biomedical engineering field." It is felt that I don't have any new insight after reading the article.
This paper tries to figure out the effects of project-based learning on "on collaborative learning, disciplinary subject learning, iterative learning, and authentic learning", none of which sound like "we made both the control and experimental group take the same quiz afterwards". And looking at the methodology - it's not. They just asked a bunch of teachers some questions. I don't consider this informative.
There's "Baran, M., & Maskan, A. (2011). The effect of project-based learning on pre-service physics teachers electrostatic achievements. Cypriot Journal of Educational Sciences, 5(4), 243-257.", but I can't access that text, only the cached abstract, which speaks in favor of project-based learning. But they had only 20 people in the treatment group, and I have no idea what the methodology even was, so... this one also fails.
This review also concludes that "a causal link between PBL instruction and positive student outcomes cannot be established with certainty".
My conclusion: we just don't know for sure. I think that having a project serves as a motivator and a framework, but if you are the type of person that doesn't need to immediately know how you'll apply the thing you're learning, then it's more or less the same. For example, if you blindly trust that logarithms will be useful for something later on, then you don't actually need a project for which logarithms would prove crucial. In both cases, you'll end up doing the same thing: reading textbooks and doing practice problems.
Please let me know if you know more about this topic!