Experimental, quasi-experimental and nonexperimental

Classification here is simpler than students expect once you know the two questions to ask. Did the researcher control the intervention, deciding who received what? And was assignment to groups random? Both yes gives you a true experiment. Manipulation without randomisation gives you a quasi-experimental design — the researcher still assigns the intervention, but to intact groups, a pre-existing unit, or by some non-random rule. Neither gives you nonexperimental research, where the researcher observes and measures what already varies. Applying those two questions to a paper's methods section resolves most cases quickly, and stating them explicitly in your post shows the classification is principled rather than guessed. Two questions, asked in order, resolve almost every study you will meet in this library. Where they do not, the methods section is usually too vague to classify, which is itself a finding worth reporting rather than a reason to guess.

The reason the distinction matters, which is the part worth saying, is causal inference. Randomisation is what makes groups comparable on everything you did not measure, so a difference afterwards can be attributed to the intervention. Without it, any difference could reflect why those groups differed to begin with, which is why quasi-experimental studies invest heavily in matching, statistical adjustment and pre-intervention measurement. Nonexperimental designs cannot establish causation at all, though they are often the only ethical or feasible option — you cannot randomly assign people to smoke. When choosing your three examples, prefer papers whose methods sections state the design explicitly and describe allocation, because papers that merely say participants were divided into groups will not let you demonstrate anything, and evaluating the effectiveness of a design is the peer-response half of this assignment rather than an optional extra. Judge fitness for the question asked, not prestige of the design.