Aristotle addresses a number of the criticisms that can be leveled against poetry. First among these is the accusation that the events depicted are impossible. This criticism can fall under two categories. Less grave describes the event if the impossibility arises from a lack of technical knowledge on the part of the poet. For instance, he may describe a horse galloping with both front legs thrown forward, not realizing that horses do not move like this. More grave describes the situation if the impossibility arises from the poet's inability to give an accurate description of something he knows quite well.
Aristotle answers that, often, impossible events—such as Homer's description of Achilles' pursuit of Hector in the Iliad —serve to heighten the astonishment and excitement of the story. When the poet can achieve similar effects while staying within the realm of possibility, however, this route should be preferred. Aristotle lays out the general principle that a poet should always aim for a convincing impossibility in favor of an unconvincing possibility.
Further, not all poetry is meant to describe things as they are. Some poets describe things as they ought to be, and others write to accord themselves with popular opinion rather than realism. For instance, Sophocles claimed that while Euripides portrayed people as they are, he portrayed them as they ought to be. Other poets stay true to popular myths rather than realism when depicting the gods.
As for events that are not impossible but merely improbable, the poet must show either that they accord with opinion or that the events are not as improbable as they may seem.
Aristotle also discusses contradictions the poet might make in language, but this discussion is very difficult to follow without a knowledge of ancient Greek. Basically, Aristotle suggests that what may at first seem to be a contradiction in language may result from a metaphorical usage or some other poetic device.
While many errors are excusable or explainable, Aristotle asserts that the only excuse for an improbable plot or unattractive characterization is if they are necessary or are put to good use. Otherwise, they should be avoided at all costs.
In Chapter 26, Aristotle addresses the question of which is the higher form, tragedy or epic poetry. The argument in favor of epic poetry is based on the principle that the higher art form is less vulgar and addressed toward a refined audience. Tragedy is performed before large audiences, which results in melodramatic performances or overacting to please the crowds. Epic poetry is more cultivated than tragedy because it does not rely on gesture at all to convey its message.
Aristotle answers this argument by noting that the melodrama and overacting are faults of the performance and not of the tragic poet himself. The recital of epic poetry could similarly be overdone without reflecting poorly on the poet. Further, not all movement is bad—take dance, for instance—but only poorly executed movement. Also, tragedy does not need to be performed; it can be read, just like epic poetry, and all its merits will still be evident.
Further, he advances several reasons for considering tragedy superior. First, it has all the elements of an epic poem and has also music and spectacle, which the epic lacks. Second, simply reading the play without performing it is already very potent. Third, tragedy is shorter, suggesting that it is more compact and will have a more concentrated effect. Fourth, there is more unity in tragedy, as evidenced by the fact that a number of tragedies can be extracted from one epic poem.
There are some seeming contradictions in Aristotle's view regarding impossible or improbable events. On the one hand, he claims that they can enhance a story by making it more astonishing. He warns that they can strain a story's credibility if overdone, but he does seem to applaud their prudent application. On the other hand, Aristotle is firmly insistent on the unity of plot, which demands that events be connected by a probable or necessary causal sequence. How, then, can improbable, or even impossible, events be an acceptable part of this sequence? In Chapter 24, Aristotle asserts that a story should never contain improbable events. If a plot would be ruined by removing these improbable events, then that just reflects poorly on the plot. If the improbable events can be removed, then it is absurd to include them in the first place.
A clue to solving this problem lies in a claim Aristotle makes just before the passage alluded to in Chapter 24, and again near the end of Chapter 25: a convincing impossibility is preferable to an unconvincing possibility. The key, it seems, is not so much that the sequence of the plot be true to life but that it be plausible. When Aristotle condemns improbable events, he is primarily concerned with events in the plot that seem out of place. Provided the plot maintains its own internal logic, it can get away with depicting the improbable.
We might link this discussion of plot to what Aristotle says about inconsistency in character: a character may behave inconsistently provided he is consistent in his inconsistency. That is, we should be able to perceive an internal logic that drives the character to irrational behavior. Similarly, a plot may be improbable provided it is convincing in its improbabilities. All good science fiction writers know that they can depict the improbable provided they do so in a consistent and convincing manner.
Aristotle's argument in Chapter 26 that tragedy is superior to epic poetry comes in three waves. First, he lists all the arguments given in favor of epic poetry. Second, he cancels all these arguments out, mostly by showing that they are leveled against the performance of tragedy rather than anything in the genre itself. Third, he lists the advantages that tragedy has over epic poetry, which can be boiled down to two main points: (1) tragedy has all the elements of epic poetry and then some, and (2) tragedy is more condensed and so has a more concentrated effect.
These two points are quite valid when we bear in mind that both tragedy and epic poetry aim at arousing the emotions of pity and fear. Music and spectacle can certainly add to emotional effect, which gives tragedy an edge that epic poetry lacks. Also, if the effect of tragedy is more concentrated, it can provide a more powerful emotional punch. Abraham Lincoln's Gettysburg Address is so powerful partly because it is so short: there are no boring bits, and the effect is immediate. We might say the same thing about the brevity of the Poetics itself: it's a far better read than lengthy manuals on literary theory.
On the other hand, we might question Aristotle's dismissal of the arguments in favor of epic poetry. Granted, they are all directed against the performance of tragedy while Aristotle is more interested in the poetry itself. But we might ask to what extent the performance can be distinguished from the poetry. That is, if there is better epic poetry around than tragedy, what meaning is there in arguing that tragedy is an inherently better genre? For instance, we could make a number of arguments in favor of comic books as a genre. Just as tragedy has all the elements of epic poetry and then some, comics have all the elements of prose fiction (words) and then some (they have pictures as well). Comics are also usually much shorter than books, meaning that they should be able to back a more concentrated punch. There are many more arguments we could make in favor of comics as a genre, but the fact remains that very few comics approach the sophistication or quality of a good novel. This should not reflect poorly on comics as a genre, but it might lead us to question how valuable it is to praise a genre in the abstract without looking at the products of that genre.
Of course, the fact is, Greek tragedy has produced a number of masterworks, and posterity suggests that no Greek epic poet after Homer approached the great tragedians in terms of quality. But this seems to be more of an argument in favor of the tragedies that have been written rather than favoring the genre in the abstract.