Evidential holism, which is closely associated with the works of W.V.O. Quine, maintains that evidence is only applicable to whole theories. It is endorsed in many disparate debates about analyticity, a priority, mathematical Platonism, and meaning. But in discussions about confirmation theory it receives less sympathetic attention: many philosophers of science regard evidential holism as implausible, poorly motivated and of dubious usefulness. Moreover, many of the philosophical applications of evidential holism lead to highly controversial conclusions. In order to determine how plausible holism is, I distinguish three evidentially holistic theses: holism about prediction,. holism about falsification, and holism about confirmation. I argue that Quine only subscribed to holism about prediction and falsification. I identify a strong source of motivation for following Quine: this is the idea that observation is theory-laden. If we accept that observation is theory-laden, then we have very good reasons for accepting holism about prediction and falsification. In addition, the theory-Iadenness of observation provides us with non-arbitrary grounds for restricting the scope of the holistic claims about prediction and falsification. I argue that holism about confirmation is not a consequence of the theory-ladenness of observation, and will only follow if we rely upon an untenable principle of hypothetico-deductivism. This hypothetico-deductive principle is pre-theoretically compelling, so identifying its role in a (flawed) argument for holism about confirmation provides an explanation of why so many commentators have casually assumed that evidential holism is committed to holism about confirmation. Having rejected holism about confirmation and motivated a narrower, less controversial version of evidential holism than that which is usually discussed by philosophers of science, I explore its role in three influential argumentative applications. I show that commitment to evidential holism carries no commitment to holism about meaning, or to the repudiation of a priori justification, or to mathematical Platonism.