Which is the ontologically plausible view about artifacts such as pencils, conventionalism or realism? While conventionalism says that human conventions create artifacts as nominal objects, realism says that they are real objects which we can capture in the causal structure of the world. This paper aims to defend realism about artifacts, and to criticize conventionalism.<br>First, I describe two features of real objects, especially natural kinds: epistemologically, they are open to empirical investigation; semantically, the causal theory of reference is true of them. These features are supposed to reflect their ontological status. Second, I clarify the grounds for conventionalism: a thought experiment suggests that artifacts have the same features as nominal objects like bachelor. Semantically, the descriptive theory of reference is true of them. And in epistemological terms, they are closed to empirical investigation. Conventionalism says this reflects the ontological fact that artifacts are nominal objects.<br>Concerning artifacts, however, realism is superior to this conventionalist claim and can be defended. I discuss a case where we are engaged in the division of linguistic or epistemic labor about some artifacts. This indicates that there are some cases in which reference to artifacts should be explained in terms of the causal theory just like natural kinds, which conventionalism cannot accommodate. Then I present a general realist scheme of artifacts based on these cases. On this account, artifacts have a kind of reality which depends on their specificity in functions, origins, etc. After arguing that the intuition in the conventionalist thought experiment does not seem to be valid or reliable generally, I show that my realist scheme can safely accommodate the conventionalist intuition and explain our epistemic and linguistic practices regarding artifacts in a coherent way.