According to Aristotle, there are three types of knowledge:
1) Perception of particulars; knowledge given by the senses
2) Experience; memories arrived at by induction based upon experience, e.g. this is a chair and so I can sit on it
3) Scientific; knowledge both of what is the case, and why it is the case
When deducing from experiental knowledge, the premises can be true but the conclusions false. For example, I might conclude from my experience of chairs that all chairs I sit on will take my weight. But this is not necessarily true, and cannot be thought of as scientific knowledge since we have no reason to believe the conclusion, no reason why it should be the case.
An example of deduction that would yield scientific knowledge:
1) Lights that are relatively near the earth do not twinkle
2) All visible planets are relatively near the earth
All visible planets do not twinkle
An example of deduction that wouldn't yield scientific knowledge:
1) All visible planets do not twinkle
2) Every light that does not twinkle is a relatively near light
Every visible planet is a relatively near light
Example two is unscientific because we have no reason to accept the first assumption without reference to the second assumption and the conclusion. With example one, although the two assumptions aren't self-explanatory, they don't assume each other nor the conclusion, and so the conclusion is scientific.
All scientific knowledge can therefore be reduced to certain first principles that are self-explanatory. For example:
Q) why are cows deficient in teeth?
A) because they have four stomachs to digest the food
Q) why do they have four stomacs?
A) because they are ruminants
The final answer is self-explanatory. To be a ruminant is part of the essence of a 'cow', and so no further explanation is needed.
Objection: Do we not have scientific knowledge of contingent and particular matters of fact? Does the theory of coherence, in contrast to the foundationalist picture offered by Aristotle's theory of scientific knowledge, suggest that we can have scientific knowledge without necessarily needing to dervive that knowledge from first principles? For example, can we treat the theory of evolution as scientific knowledge? According to Aristotle, it is merely experiential knowledge, conjecture. But then science must either concern itself equally with conjecture and that which Aristotle considers scientific knowledge, or admit that it knows very little indeed.
Objection: furthermore, Aristotle's conception of scientific knowledge negates the use of interpolation and extrapolation, of quantitative prediction, since the only reason we have to beleive some knowlege based upon those methods is experience. Whilst it is true that science generally seeks more reliable explanations than those that are by nature experiential, it again seems to be an essential part of science that Aristotle would neglect.
Objection: Can we really find self-explanatory first principles for all of science? It seems that often we find even our first principles require other first principles to be explained, and that we tend to generalise explanations until a first principle that coheres can be found. Gravity is a good example here, since it itself is not self-explanatory, but we treat it as a first principle in order that we might understand almost all physics, except that which attempts to explain gravity of course.
In general, regardless of the problems specific to Aristotle's division between experiential and scientific knowledge, one can draw out a general method of learning from his distinctions:
Perception --induction--> Experience --dialectic--> scientific knowledge
And finally, Aristotle tackles the question of metaphysics, and how that is different from physics and the other sciences. He concludes that there is a special kind of scientific knowledge, wisdom, that concerns itself with the most universal causes, and that wisdom is that which metaphysics supposedly yields.