The Fallacy of Accident

Taxonomy: Logical Fallacy > Informal Fallacy > Accident



The word translated as "accident" is from the classical Greek of Aristotle, to whom this fallacy can be traced―see the History section. Confusingly, the common current meaning of the English word "accident" has almost nothing to do with either this fallacy or Aristotle's philosophical concept. For this reason, even though there is a fallacy often called "accident", one should ignore the English meaning of the word: the fallacy of "accident" doesn't have anything to do with car wrecks or people slipping on banana peels.

The Latin phrase "a dicto simpliciter ad dictum secundum quid", or just "dicto simpliciter" for short, is less a name and more a description of one interpretation of Aristotle's fallacy. Unfortunately, it does not describe the fallacy that I will discuss in this entry, so it too should be disregarded. I include it as an alias of the fallacy for the same reason that I include "accident", since for historical reasons one will often encounter the fallacy under that name.


Accident is one of the thirteen fallacies Aristotle discusses in his book On Sophistical Refutations, which was the first book in history on logical fallacies. Unfortunately, accident is the most difficult of the thirteen to understand. To a modern logician, the examples given by Aristotle can often be easily explained, but they don't seem to have anything in common. Aristotle seems to have thought that the examples could be explained in terms of his philosophical distinction between essential and "accidental", or non-essential, properties, but it's hard to see how to apply that distinction and Aristotle does little to explain how to do it. Instead, the examples seem to be a hodge-podge that can mostly be explained in other terms; for example, one seems to be an example of the Masked Man Fallacy.

Because of this lack of clarity, there have been multiple interpretations of this fallacy in subsequent history. Thus, in a sense, there is no one fallacy of "accident", but a number of distinct fallacies have been discussed under that name. This entry discusses one such interpretation due to its relation to recent developments in logic and artificial intelligence, but it should be noted that it has little except an historical relationship to Aristotle.


No rule is so general, which admits not some exception.


Source: Robert Burton, The Anatomy of Melancholy


Xs are normally Ys.
A is an X. (Where A is abnormal.)
Therefore, A is a Y.


Birds normally can fly.
Tweety the Penguin is a bird.
Therefore, Tweety can fly.


Consider the generalization "birds can fly" from the example. Now, it isn't true that all birds can fly, since there are flightless birds. "Some birds can fly" and "many birds can fly" are too weak, while "most birds can fly" is closer to what we mean. However, "birds can fly" is a "rule of thumb", that is, a rule that is generally true but has exceptions. The fallacy of Accident in our sense occurs when one attempts to apply such a rule to an obvious exception, such as concluding that a penguin can fly because penguins are birds and birds can fly.


Common sense is full of rules of thumb which do not hold universally, but which hold "generally" or "as a general rule", as is sometimes said. Logicians have tended to ignore rules of thumb, probably because they seem unscientifically imprecise. However, in the past couple of decades, primarily due to research in artificial intelligence, which has shown the importance of such general rules for practical reasoning, there has been growing interest in so-called "default" or "defeasible" reasoning, of which rules of thumb are a part.

The difference between rules of thumb and universal generalizations, is that the former have exceptions. For instance, flightless birds are exceptions to the rule of thumb that birds can fly. One might hope to represent this rule of thumb by the universal generalization "all non-flightless birds can fly", but even this is not correct, for flighted birds with broken wings cannot fly. One might still hope that some lengthy list of exceptions would do the trick. However, one can imagine many different scenarios in which a bird would not be able to fly: its feet are stuck in quicksand, all of the air around it has suddenly rushed into space, it has developed a phobia about flying, etc. One might then try to sum up this diversity of cases under the rubric of "untypical", or "abnormal", and say: "All typical or normal birds can fly". This is exactly what a rule of thumb is.

Rules of thumb differ from statistical generalizations such as "90% of birds can fly" in that there is no specific proportion of flighted to flightless birds that determines normality. The rule of thumb doesn't even necessarily imply that the majority of birds can fly, though it would be unusual if this didn't hold. We can imagine, for instance, that there might be so many penguins in Antarctica that the majority of birds would be flightless. However, our notion of normality applies to the familiar, everyday birds we see in our backyards, rather than "exotics" on distant continents. Clearly, then, rules of thumb are specific to a cultural and temporal context.