The Complete Aristotle (eng.)

First,
then, we must speak of the methods of overthrowing a view
universally, because such are common to both universal and
particular problems, and because people more usually introduce
theses asserting a predicate than denying it, while those who argue
with them overthrow it. The conversion of an appropriate name which
is drawn from the element ‘accident’ is an extremely precarious
thing; for in the case of accidents and in no other it is possible
for something to be true conditionally and not universally. Names
drawn from the elements ‘definition’ and ‘property’ and ‘genus’ are
bound to be convertible; e.g. if ‘to be an animal that walks on two
feet is an attribute of S’, then it will be true by conversion to
say that ‘S is an animal that walks on two feet’. Likewise, also,
if drawn from the genus; for if ‘to be an animal is an attribute of
S’, then ‘S is an animal’. The same is true also in the case of a
property; for if ‘to be capable of learning grammar is an attribute
of S’, then ‘S will be capable of learning grammar’. For none of
these attributes can possibly belong or not belong in part; they
must either belong or not belong absolutely. In the case of
accidents, on the other hand, there is nothing to prevent an
attribute (e.g. whiteness or justice) belonging in part, so that it
is not enough to show that whiteness or justice is an attribute of
a man in order to show that he is white or just; for it is open to
dispute it and say that he is white or just in part only.
Conversion, then, is not a necessary process in the case of
accidents.
    We must also define the errors that occur in problems. They are
of two kinds, caused either by false statement or by transgression
of the established diction. For those who make false statements,
and say that an attribute belongs to thing which does not belong to
it, commit error; and those who call objects by the names of other
objects (e.g. calling a planetree a ‘man’) transgress the
established terminology.
<
    div id="section20" class="section" title="2">
2
    Now one commonplace rule is to look and see if a man has
ascribed as an accident what belongs in some other way. This
mistake is most commonly made in regard to the genera of things,
e.g. if one were to say that white happens (accidit) to be a
colour-for being a colour does not happen by accident to white, but
colour is its genus. The assertor may of course define it so in so
many words, saying (e.g.) that ‘Justice happens (accidit) to be a
virtue’; but often even without such definition it is obvious that
he has rendered the genus as an accident; e.g. suppose that one
were to say that whiteness is coloured or that walking is in
motion. For a predicate drawn from the genus is never ascribed to
the species in an inflected form, but always the genera are
predicated of their species literally; for the species take on both
the name and the definition of their genera. A man therefore who
says that white is ‘coloured’ has not rendered ‘coloured’ as its
genus, seeing that he has used an inflected form, nor yet as its
property or as its definition: for the definition and property of a
thing belong to it and to nothing else, whereas many things besides
white are coloured, e.g. a log, a stone, a man, and a horse.
Clearly then he renders it as an accident.
    Another rule is to examine all cases where a predicate has been
either asserted or denied universally to belong to something. Look
at them species by species, and not in their infinite multitude:
for then the inquiry will proceed more directly and in fewer steps.
You should look and begin with the most primary groups, and then
proceed in order down to those that are not further divisible: e.g.
if a man has said that the knowledge of opposites is the same, you
should look and see whether it be so of relative opposites and of
contraries and of terms signifying the privation or presence of
certain states, and of contradictory terms. Then, if no clear
result be reached so far in these cases, you should again divide
these until you come to those that are not further divisible, and
see (e.g.) whether it be so of just deeds and unjust, or of the
double and the half, or of blindness and sight, or of being and
not-being: for if in any case it be shown that the knowledge of
them is not the same we shall have demolished the problem.
Likewise, also, if the predicate belongs in no case. This rule is
convertible for both destructive