Aristotle has just explained the various kinds of deductions. There is dialectic deductions and demonstrative deductions. Fallacies look like deductions but are not ones. He continues in Topics, Book 1.
Further, besides all the deductions we have mentioned there are the fallacies that start from the premisses peculiar to the special sciences, as happens (for example) in the case of geometry and its sister sciences. For this form of reasoning appears to differ from the deductions mentioned above- the man who draws a false figure reasons from things that are neither true and primitive, nor yet reputable. For he does not fall within the definition: he does not assume opinions that are received either by everyone or by the majority or by the wise that is to say, by all, or by most, or by the most reputable of them but he conducts his deduction upon assumptions which, though appropriate to the science in question, are not true- for he effects his fallacy either by describing the semicircles wrongly or by drawing certain lines in a way in which they should not be drawn.
Sometimes a deduction uses special arguments found in particular kinds of disciplines. Fallacies also exist that look like these special kinds of arguments. Just like other fallacies, these ones do not reason from what is true and primitive or from what is commonly believed. That means that these kinds of arguments are not deductions. If you do not argue on the basis of common opinion, then you must be arguing from the truth. If you do not argue starting with the truth, then all that is left is fallacies. In geometry, for example, this will involve mistakes such as drawing line incorrectly or describing semi-circles wrongly.
Very few people believe that Aristotle was wrong here. However, it is important to remember that Aristotle is describing deductive reasoning here. Later on, Aristotle will mention another form of reasoning.
Now that Aristotle has finished his argument, he will explain just what these arguments are intended to prove. Only then can they be properly evaluated.