Some of them have moons, that
serve to give them light in the absence of the sun, as our
moon does to us. They are all, in their motions, subject to
the same law of gravitation as the earth is. From all this
similitude it is not unreasonable to think that these planets
may, like our earth, be the habitation of various orders of
living creatures. There is some probability in this conclusion
from analogy.[1]

The argument from analogy is sometimes said to range through all
degrees of probability from certainty to zero. But this is true only
if we take the word analogy in its loosest sense for any kind of
resemblance. If we do this, we may call any kind of argument an
argument from analogy, for all inferences turn upon resemblance. I
believe that if I throw my pen in the air it will come down again,
because it is like other ponderable bodies. But if we use the word in
its limited logical sense, the degree of probability is much nearer
zero than certainty. This is apparent from the conditions that
logicians have formulated of a strict argument from analogy.

1. The resemblance must be preponderating. In estimating the value of
an argument from analogy, we must reckon the points of difference
as counting against the conclusion, and also the points in regard
to which we do not know whether the two objects agree or differ. The
numerical measure of value is the ratio of the points of resemblance
to the points of difference _plus_ the unknown points. Thus, in the
argument that the planets are inhabited because they resemble the
earth in some respects and the earth is inhabited, the force of the
analogy is weakened by the fact that we know very little about the
surface of the planets.

2. In a numerical estimate all circumstances that hang together as
effects of one cause must be reckoned as one. Otherwise, we might make
a fallaciously imposing array of points of resemblance. Thus in Reid's
enumeration of the agreements between the earth and the planets, their
revolution round the sun and their obedience to the law of gravitation
should count as one point of resemblance. If two objects agree in _a_,
_b_, _c_, _d_, _e_, but _b_ follows from _a_, and _d_ and _e_ from
_c_, the five points count only as two.

3. If the object to which we infer is known to possess some property
incompatible with the property inferred, the general resemblance
counts for nothing. The moon has no atmosphere, and we know t