It is perhaps too readily assumed that in the domain of religious phenomena there is a law by which these phenomena are bounded and conditioned. Assuming such a law to exist, the attempts to formulate it will be directed in a greater or less degree by preference. For religious phenomena, by which is here meant the outward manifestations of religions, cannot be examined and classified, without a comprehensive knowledge of the religions themselves. And if, as a French writer has contended, “the man who would write the history of a religion must believe it no longer but must have believed it once,” it follows that few persons, even in this versatile age, can claim to be proficient in more than three or four religions. From which it also follows that lack of knowledge must be supplied by fertility of imagination or by the exercise of preference on the part of him who employs the comparative method in order to discover the law. And yet, it is only by eliminating this personal element and by confining our attention to material which is neither inaccurate nor defective that we can hope to arrive at the truth. It must be confessed that the rough and ready generalisations with which we are so familiar in this connection and the lack of care which is taken in gathering and sifting the materials upon which they are based, almost lead us to despair of useful results. The attempt to evolve a law from insufficient data is like an attempt to measure volume in terms of two dimensions or like an attempt to classify animals without an intimate knowledge of them. A salamander has four legs and a tail: so has a sheep. A zoology based on these criteria alone would not carry us very far. The biologist might kindly step in with his law of evolution and say some soothing words respecting their common origin, but we should leave off where we began and know no more of those animals than we did at the start, namely, that they each have four legs and a tail.[2]