However, when taking up various inquiries, Aristotle does not evidently derive principles in any such manner, nor does he set out his conclusions in the demonstrative form recommended in the Posterior Analytics. This fact has led scholars to puzzle over Aristotle's actual methods. If his various treatises are not carried out in accordance with the method prescribed in the Posterior Analytics, what is the method being used to carry them out?
One interesting feature common to most of Aristotle's treatises suggests that something might be gained by looking closely at his treatise on dialectic, the Topics. This feature is Aristotle's frequent reviews and critiques of the views of historically prominent philosophers. Dialectic is often construed as any sort of argument proceeding from reputable opinions (endoxa). If argument proceeding from reputable opinion is what dialectic amounts to, then Aristotle's customary examinations of the positions of his predecessors are certainly dialectical. But how about Aristotle's examination of problems that follow these critical examinations? Is Aristotle's method here dialectical as well?
Some commentators answer a hearty "Yes" and have taken dialectic to be Aristotle's primary method--the recommendations of the Posterior Analytics notwithstanding. T.H. Irwin has been notable recently in his defense of what he calls 'strong dialectic' as Aristotle's primary mature method. Irwin's views are a detailed development of some themes first laid out in G. E. L. Owen's influential paper 'Tithenai ta phainomena', wherein Owen draws out the tension between the demonstrative method of the Posterior Analytics and what Owen takes to be the actual method employed by Aristotle (dialectic) in his discussion of place in the Physics. Both Irwin and Owen refer to the obscurity of Aristotle's account of epagoge in the Posterior Analytics II 19. Irwin is especially concerned that principles derived by epagoge admit of no proof or test of justification, but must be assumed true by dint of the nature of the process. Irwin takes this (and the idea of non-inferential justification generally) to be a problem for Aristotle and seeks out an alternative path to first principle, one that is capable of providing some sort of proof or justification of principles along the way. (How else could we know we've really got the right principles?) Irwin and others read a few passages from Aristotle in such a way that the passages could be construed to provide some evidence in support of the thesis that Aristotle thought that some form of dialectic could produce the principles (and with justificatory accoutrements). As Irwin writes, "Aristotle retains Plato's belief that dialectic is also a method for reaching positive conclusions; this is why he claims that it has a road to first principles (Top. 101b3-4)". 
Is Irwin correct about Aristotle on this point? That's what I'd like to ask. In this paper, I shall question whether Aristotle did in fact believe that dialectic was capable of generating first principles, or, to put it another way, whether he believed that dialectic was an epistemically fruitful philosophical method, like Plato did. I shall argue that he does not. My argument will consist of several parts. In Part One I shall briefly clarify what Aristotle took dialectic to be. In Part Two I shall very briefly examine to what extent Aristotle's treatises do in fact diverge from the method set out in the Posterior Analytics. If the divergence is less sharp than it is sometimes taken to be, one of the main motivations to seek alternative working methods in Aristotle's treatises might be diminished. Next, in Part Three, I shall examine some of the passages taken by Owen, Irwin and others to support the view that Aristotle thought dialectic could lead to principles. I will contrast these passages (along with the somewhat strained interpretations of them) with some passages in the Topics and elsewhere wherein Aristotle comments directly upon the relative epistemic powerlessness of dialectic. Lastly, in Part Four, I will take my own kick at the cat and offer an hypothesis about Aristotle's view on the function of dialectic.
I will not deny that there are difficulties in Aristotle's explicit account of the way we come to first principles. But, as I will argue, even were we to admit to the obscurity of Aristotle's account of epagoge we would still be left without sufficient reason to believe that Aristotle held that there was some alternative method--such as dialectic--for the establishment of archai.
Since the chief aim of this paper is to inquire into the powers and limits of dialectic, as Aristotle understands it, it is sensible to begin by setting out the nature of dialectical argument, as Aristotle understands it. As mentioned earlier, dialectic is often loosely construed as argument proceeding from reputable opinion. Does such a construal fully capture the essence of dialectical argument? There is reason to believe that it does not.
The Topics begins with Aristotle's statement of the treatise's purpose:
The goal of this study is to find a method with which we shall be able to construct deductions from acceptable (endoxon) premises concerning any problem that is proposed and--when submitting to argument ourselves--will not say anything inconsistent. (100a18-21)Deductions, or arguments, from accepted premises obviously plays large and important role in dialectical procedure. However, it seems that Aristotle's interest in such arguments arises within a special context within which problems are proposed and must be answered. According to Smith, "Dialectical argument is essentially argument directed at another person and proceeding through questions: the person arguing puts forward premises for acceptance or rejection by a respondent, who either accepts or rejects them."
Just about any sort of argument proceeding by means of question and answer may properly be called dialectical. Nonetheless, it seems that throughout the Topics Aristotle has in mind a specific kind of argumentative exchange in which an answerer endeavors to to defend a proposition and a questioner to undermine it. The practice is likely to have roots in Plato's Academy and seems to have been a lively and competitive affair. Topics viii gives evidence that these exchanges were bound by specific rules and were evaluated by judges. Call this competitive form of dialectic "gymnastic" dialectic.
We can better understand the function of endoxa within dialectical arguments if we understand the function they play in gymnastic dialectical exchanges. In any sort of dialectical argument, the questioner must construct an argument out of the responses elicited from the answerer by means of questions admitting of yes or no answers. As Smith notes, success in this task requires that the premises put forward by the questioner satisfy two criteria: (1) the answerer must accept the premises, and (2) the premises must logically imply the conclusion the questioner wishes to establish. The point of the Topics seems to be, in large part, to provide a method for this very purpose--the purpose of coming up with premises that satisfy these two criteria.
There seems to be two main aspects to the method. (1) The collection of premises which various types of people are likely to assent to (endoxa), and (2) the collection of topoi, or 'locations'. Something more must be said about these. Topoi seem to be general premise patterns, or simple argument forms relevant to a given desired conclusion and committed to memory. There is some reason to believe that the term refers to a type of mnemonic system pre-dating Aristotle in which premises and their relations are memorized by associating them with real or imaginary physical locations (hence 'topoi').
The strategy of dialectic then seems to be to access one's collection of topoi, and find a premise together with an argument form that will lead to a desired conclusion. But not any logically sufficient premise will do. It must be one that our dialectical opponent will concede. Thus it is necessary to have on hand an inventory of endoxa suitably classified for easy access.  If we pose a question using an endoxon premise that fits the general scheme of our topos and our opponent affirms or denies according to plan, then we can perform the operation suggested by our topos and elicit from our opponent the desired conclusion.
This is a necessarily loose and incomplete characterization of the strategy of dialectic. Nevertheless, even given such a loose characterization, the function of endoxa becomes somewhat clearer. Endoxa are often taken by interpreters of Aristotle to be a single collection of premises used to construct dialectical arguments. But then, as Smith writes,
It becomes somewhat mysterious what these endoxa are and why Aristotle should single out arguments from them for special attention. Attempts to explain this mystery have led to various accounts, some quite elaborate, of the endoxa as sources of knowledge, and to elaborate hypothesis about concomitant epistemological views presupposed by these accounts.However, on the above account, endoxa do not form a single class, but rather form various classes indexed to particular types of people. So, there is no mystery. The dialectician does not choose endoxa from a single pool consisting of the unaccountably commingled opinions of the many, the famous, those experienced in an art or science, etc., but instead chooses from one category of endoxa or another depending on which kind of endoxa his opponent appears most likely to assent to in the circumstances.
A proposition is not especially likely to be true or even probable simply because it is endoxon. But, if it is chosen from the appropriate category, it is especially likely to be believed and hence agreed to by a certain kind of person. Understood in this way--as classified strategic expedients for acquiring assent from an opponent-- the mystery about why these beliefs should be picked out for attention dissolves and endoxa look less promising as a special source of knowledge.
Finally, it should be mentioned that dialectical reasoning is not for Aristotle differentiated from demonstrative reasoning simply in virtue of the fact that the premises of dialectical deductions are endoxa rather than archai. Just as important is the fact that dialectical premises are in the form of questions, whereas demonstrative premises are in the form of assertions. Further, dialectic is of unrestricted generality whereas every natural science is specific, proceeding from premises specific to the subject matter. Aristotle stresses these features of dialectic at least as much argument from endoxa. When examining whether Aristotle's method is dialectical or not, we should take into account the presences or absence of these less well noted, but nonetheless essential features of dialectic.
When we do take into account these features, we see that Aristotle's characteristic reviews and critiques of his predecessor's arguments are dialectic only in the limited sense that some endoxa (usually the opinions of famous philosophers) are treated to a test of internal consistency. They are not dialectical in the strict sense of proceeding by question and answer (as Platonic dialogues are). Nevertheless, we can perhaps admit that Aristotle's preliminary examinations of certain endoxa are a form of dialectic. Be that as it may, the function of dialectic seems here to be merely negative. Internal inconsistency is sufficient grounds for dismissing an argument. However, it remains unclear whether or how Aristotle uses dialectic as a method for arriving at first principles and positive conclusions.
If there is reason to believe that Aristotle's demonstrative method relying on epagoge and nous is not employed in the constructive portions of the treatises, there may be reason to believe he is using some other method--perhaps even dialectical method in some hard to make out form. But before we leap to any such conclusions, it would first be necessary to rule out the presence of the demonstrative method.
So, do Aristotle's treatises in fact fail to live up to the method laid down in the Posterior Analytics? It is conceivable that we are projecting a picture of a rigorously axiomatic, deductive methodology backward onto Aristotle's account of method in APo, whereas Aristotle himself had no such picture of formalistic rigor. So when we fail to find the method of the analytics present in the scientific treatises, it is not because that method is not being employed, or because Aristotle changed his mind about method, or was inconsistent, but because we have misconceived what that method would have to look like in practice. If we were to look harder, we might indeed see the method of the analytics present in the scientific treatises, but not in the way we might suspect. If this were to turn out to be the case, then some of the motivation for attributing to Aristotle an overall dialectical method would be undermined.
First, let's examine what the trouble is supposed to be. Barnes puts the alleged conflict between the method of APo. and the scientific treatises starkly:
This, then, is the problem: on the one hand we have a highly formalised theory of scientific methodology; on the other, a practice innocent of formalisation and exhibiting rich and variegated methodological pretensions of its own. How are the two to be reconciled?Barnes own strategy of reconciliation is well known and at one point gained such wide currency that Burnyeat said that it "promises to become a new orthodoxy". Barnes argues that the method of APo. is not meant as a method for obtaining scientific results, but is merely a plan for laying out scientific results when teaching them.
However, there are difficulties with this view. If we accept it we are required to believe that Aristotle's entire scientific corpus is a collection of sketchy, preliminary investigations that draw no firm conclusions and make no claims to understanding. Barnes writes:
. . .Aristotle did not intend his treatises to be pieces of formal instruction: they are progress-reports, not text-books, and as such they need not--indeed cannot--have pedagogic form. A series of demonstrations is appropriate to the setting out of knowledge securely achieved; it is inappropriate to the sharing of tentative philosophical or scientific exploration.Prior to this passage, Barnes notes that Aristotle's works "bear all the marks of constant revision." From this fact, Barnes draws the conclusion the works are tentative and merely exploratory.
But this is a problem. It does not follow from a work's being revised that the author considered the material in the original to be merely exploratory. Revision could just as well signal a resolute change of mind about the truth on some subject. Additionally, at least some of the scientific treatises make claims about the nature of things in no uncertain terms. The confidence and resolution with which these claims are made belies a picture of tentative exploration.
Perhaps there is a plausible alternative resolution to Barnes's problem. One current suggestion is that Aristotle never intended scientific results to be set out in pure demonstrative form, but simply intended that results be capable of assuming demonst rative form given the established principles. William Wians observes that
Demonstration depends on finding the explanatory middle term. Once it has been obtained and so proper and adequate premises secured, construction of the actual proof amounts to little more than fitting together in a formal way relationships that are now obvious.Aristotle seems to say much the same in the Prior Analytics:
So if we apprehend the attributes of an object in question it will at once be in our power to readily exhibit the demonstration; for assuming that none of the true attributes of the objects concerned has been omitted in our survey, we shall be able to discover and demonstrate the proof of everything which has a proof, and to elucidate everything whose nature does not admit of proof. (46a22-27)Aristotle echoes these thoughts in the first book of the Nichomachean Ethics where he says that his method will be to "sketch it roughly and then later fill in the details," and that "anyone is capable of carrying on and articulating what has been well outlined. . . any one can add what is lacking." (1098a29-26)
These passages may support the view that Aristotle did not find it necessary to render arguments in his treatises in proper demonstrative form, but rather took the method outlined in the Posterior Analytics to describe the underlying logical character of scientific understanding. "On this view, the theory of demonstration gives a formal description of scientific practices which may themselves remain informal in their patterns of argument." As Allan Gotthelf puts it,
APo. should be understood as offering a formal description of proper science, not a requirement that proper science itself be formal. . . [I]t is much like getting the quantifiers in the right place, though here it is the "formal" operator: a formal description of science not a description of formal science. On this view, the APo. theory requires of the natural philosopher only that his exposition be puttable into the appropriate form, not that it actually be so put. . .The idea is that Aristotle's scientific writing are conditioned by considerations of the Analytics. This position has been defended in detail elsewhere. If the defenders of the "apodeictic conditioning" position are correct, then Barnes's problem of a tension between theory and practice more or less disappears. More importantly, if this view is correct, defenders of an overall dialectical method cannot simply cite the absence of formal demonstrations in the scientific treatises as evidence that Aristotle's working methods were something else, e.g., dialectical. If the " conditioning" view is correct, we should work to identify the ways in which the treatises have been conditioned by apodeictic requirements rather than search for an altogether different method.
Defe nders of a dialectical interpretation of Aristotle's method aren't simply making guesses in the dark when they claim that Aristotle attributed certain impressive powers to dialectical method. There are several passages in Aristotle's writing that can be interpreted in such a way as to lend strong support to this view. The question is whether the interpretations that lend such support are the most reasonable interpretations.
At the outset, I mentioned G.E.L. Owen's paper, "Tithenai ta Phainomena." In that paper Owen makes note of Aristotle's convention of beginning his investigations with a survey of the views of others together with the available data on the subject. Aristotle then usually proceeds to set out the puzzles (aporia) that arise when opinions are tested against each other or against the empirical data. Owen suggests that the empirical data and the prevalent opinions need not be seen as two separate bodies of evidence with disparate epistemic standings. Both can be seen to be phainomena or appearances when we see that the term is ambiguous between 'what is apparent to the senses' (empirical data) and 'what appears to us to be the case' (widespread opinions). Aristotle's method then is to begin with this unified set of appearances and then try to resolve the puzzles within the set while trying to preserve as many of the appearances as possible. But this method, Owen argues, can be assimilated to dialectic if (as Owen supposes) dialectic is "argument from endoxa" and endoxa are single class of widespread opinions. If this is the case, then it is reasonable to suppose that dialectic is the source of first premises for scientific demonstrations.
Owen hopes to support his thesis with passages from the Physics and Nicomachean Ethics. Owen writes, paraphrasing Aristotle, "'For if the difficulties are resolved and the endoxa are left standing', as Aristotle says in both the Physics and the Ethics, 'this is in itself sufficient proof.'"
Owen implies that these passages indicate a common method appearing in separate treatises which consists in "resolving the difficulties and preserving the endoxa." The further implication is that this method can be generalized to all or most of the treatises.
However, upon closer inspection of these passages, it seems that these passages make distinct and somewhat modest claims within the contexts in which they arise.
In the Physics passage Aristotle says:
We must try to conduct our search so that the essence will be given in such a way as to solve the puzzles, and what appears to be true of place will be true of it, and moreover the cause of resistance and puzzles about it will be evident. For this is how anything might be most beautifully shown. (211a7-11)We can see that Owen clearly offered us a paraphrase and not a translation. But even so, it is difficult to see how this passage says what Owen's paraphrase claims it says. The paraphrase leads us to believe that Aristotle is setting down minimum sufficient conditions for establishing a proposition --"this in itself is sufficient proof". However, as Robin Smith notes, the Physics passage concerns conditions that the "finest" or "most beautiful" proofs must satisfy. Smith writes:
These are counsels of perfection, not minimal conditions of adequacy; they go far beyond mere sufficiency to tell us what the best of all possible outcomes is. Moreover, [Aristotle] does not tell us that this is how we must conduct our search if it to issue in proof: he says that this is how we ought to conduct it. It is consistent with this demand that an adequate proof may fail to achieve some of these desiderata.Aristotle does not venture to say that this is how we may or must prove a proposition. Rather, he seems to be saying that it would indeed be a wonderful thing if we could manage to simultaneously (1) solve the puzzles, (2) leave the endoxa standing, (3) explain what causes the puzzles and, (4) explain why people get stuck on them. Such a proof would certainly be "beautifully shown, " as Aristotle says. So we should by all means "try to conduct our search" in this way, as Aristotle says. But Aristotle says nothing to suggest that these are necessary conditions of inquiry, or that these conditions can always, or even usually, be met. There are cases in which things are in fact shown, although not beautifully shown. As Smith says:
Since elsewhere (for instance Metaphysics A, NE X) he clearly thinks that sometimes our pre-philosophical opinions cannot be retained after philosophical inquiry, we ought to suppose that the outcome envisaged here will be possible only in some most happy set of circumstances.
Far more promising for those wishing to identify a dialectical methodology in Aristotle is the other passage Owen "paraphrases". The passage in from Nicomachean Ethics VII.1:
As in the other cases we must set out the appearances, and first of all go through the puzzles. In this way we must prove the common beliefs about these ways of being affected--ideally, all the common beliefs, but if not all, then most of them, and the most important. For if the objections are solved, and the common beliefs are left, it will be an adequate proof. (1145b3-8)This is far more promising than the Physics passage because Aristotle is clearly saying something about what is adequate and thus sufficient for proof. Be that as it may, Owen's "this in itself is sufficient for proof" is somewhat stronger than anything we find in the passage. It is not clear that Aristotle is here stating anything like a general condition of adequacy for all proofs. It may be pointed out that Aristotle says, "as in other cases." However, it is not obvious how this should be read. Is it "in all other cases whatsoever" or "in other cases like this one"? Or is it "as in other cases set out the appearances and go through the puzzles" and then, down further, "For, in the present case, if the puzzles are solved and the endoxa left, it will be an adequate proof"?
If he is saying something about the case at hand, then, far from saying something general about conditions of adequacy, he may be saying something like, "If we can just do this much, it will good enough for present purposes."
It is striking that in the context of the passage, Aristotle's present purpose is a discussion of the possibility of weakness of the will in which he rejects a commonly held belief. Aristotle concludes , against common opinion (and with Socrates) that, strictly speaking, no one acts akratically. Smith contends that the passage is concerned with persuasion or proving to another that something is so, not with proof in the sense of logical demonstration That helps explain why this passage precedes a rejection of a widely held belief. Aristotle is interested in what it takes to get someone to believe something they may be inclined to disbelieve. Smith puts the point well.
It is prima facie implausible that there is no such thing as weakness of will; if we are ever to persuade others of this, we must begin from their own views. However, if we can eliminate the difficulties that stand in their way of accepting it, then we will have shown them adequately that this is so.Shorn from its context, the Ethics passage does seem to prescribe a general method of inquiry. However, in context, it is plausibly read as concerning persuasion rather than proof. I am not asserting that this reading is decisive, only that it at least as likely to be correct as other readings. In that case, which interpretation is best to accept will depend on what evidence there is (other than this passage read in a certain way) for a dialectical method.
The last passage I will examine in detail might provide such evidence. Earlier, I quoted Irwin as saying that, for Aristotle, dialectic "has a road to first principles." Irwin's source is a celebrated passage in Topics I.2. Irwin himself translates it thusly:
It is useful to the philosophical sciences, because if we fully examine the puzzles on each side we will more easily see what is true or false. And it is also useful forOwen thinks this passages asserts that that dialectic "establishes" first principles. But, a "road towards" something is just a road. It is not the vehicle that takes you from where you are to where you want to go. Nevertheless, the passage does seem to be saying something rather strong about the powers of dialectic.
the first principles of each science. For we cannot say anything about them from the proper first principles of the science in question, since the first principles are prior to everything else. Hence it is necessary to discuss them through the common beliefs on each subject. And this is proper to dialectic alone, or to it more than anything else, for since it examines, it has a road towards the first principles of all disciplines. (101a34-101b4)
Naturally, the context of the passage turns out to be important for understanding just what Aristotle is saying. Chapter 1 of Book I of the Topics begins by Aristotle's stating the goal of his study (to find a method for constructing certain kinds of deductions). Chapter 2 starts by Aristotle's setting out "the number and kinds of things our study is useful for" once the goal is met. Aristotle says, "There are, then, three of these: exercise, encounters and the philosophical sciences." These are the uses of the dialectical art or method which Aristotle hopes to find.
In the previous chapter, Aristotle has just finished talking about the study of deduction generally and dialectical deduction specifically, as that is the kind that pertains to the goal of the treatise. The study of deduction generally is the study of logical consequence. It is easy to see how knowledge of rules of logical consequence would be useful for examining anything whatsoever. We can even discuss opinions about first principles, even though the principles cannot be proved or demonstrated. We can test opinions about principles for consistency, which will help us examine puzzles and "see what is true or false". This is what dialectical art is largely about--knowing how to deduce logical consequences from opinions.
Again, since the first principles of a subject are logically prior to everything else we might say about the subject, we are not going to deduce them from anything. But we can talk about the common beliefs about them and see if in the process we can clarify our understanding by eliminating inconsistent opinions and logically validating consistent ones. Skill in the dialectic art makes one particularly well suited for this sort of examination. As Aristotle recognizes, dialectic may be the only art well suited for this sort of examination.
Rather than reading "for since it examines, it has a road towards the first principles" for the last line of the passage, an alternative translation reads, "for since its ability to examine applies to the starting-points of all studies, it has a way to proceed." Aristotle has said that by means of demonstration, we cannot say anything about first principles. But, in contrast, dialectic "has a way to proceed." However, having a way to proceed does guarantee or even make likely the establishment of principles. Having a way to proceed is simply the alternative to having nothing at all to say. But there is no suggestion here that our dialectical proceedings will be anything more than "useful" in "discussing" first principles.
Indeed, it is striking how often Aristotle claims that dialectic cannot establish anything. This is not surprising, since Aristotle remarks more than once that dialectic is capable of proving opposites, i.e., constructing valid arguments for both a conclusion and its negation, which is something no science can do. In the Sophistical Refutations, Aristotle writes:
None of the arts which proves some nature is interrogative, for a deduction does not come about from both [alternatives]. But dialectic is interrogative; yet if it did prove, then it would not make questions of its first things and proper principles, at any rate, even if it did [make questions of something. (172a15-20)He makes a similar point at Posterior Analytics I.11:
Dialectic is not in this way about some definite things or some single genus. For [otherwise] it would not ask questions: it is not for the one who demonstrates to ask questions, because it is not possible to prove the same things when opposites are the case. (77a31-35)From these passage, it is clear that Aristotle has in mind a relatively stark contrast between dialectic, which asks questions, and demonstration, which does not. Further, dialectic is completely general in application. Every science in limited in scope by the finite number of principles uniquely appropriate to it. Dialectic, on the other hand, is unlimited in scope, but unable to prove anything. In the Rhetoric, Aristotle says that dialectic in in a way like rhetoric because "both arts are about certain sorts of things which are, in a way, common for all to know, not to any separate science," (1354a1-3) and are "not about any separate genus". (1335b8-9) In Sophistical Refutations Aristotle puts his point bluntly:
[T]he dialectician is not concerned with any genus, nor does he give proofs about anything, nor is he a sort of universal man. For neither are all things in some single genus, nor, if they were, would it be possible for all things to exist under the same principles. (172a13-15)How is a defender of a dialectical method in Aristotle to deal with such passages? One way is to propose a developmental thesis, wherein Aristotle comes to change his mind about the powers of dialectic. That's what T.H. Irwin does. According to Irwin, the passages quoted above are from Aristotle's early period, when Aristotle denied the possibility of a universal science. His strictures on dialectic concern what Irwin calls, "weak dialectic". Irwin agrees that weak dialectic cannot establish principles. However, Aristotle in his Metaphysics discovers a universal science, which overturns his previous doctrine that every science is limited in scope by principles appropriate to only it. Irwin argues that epagoge and nous are insufficient for the establishment of the first principles of the science of being qua being and thus something else in needed to establish them. This something else Irwin calls "strong dialectic" and points toward the discussion of the Principle of Non-Contradiction in Metaphysics Gamma as an example of the use of strong dialectic for the establishment of the most general principles.
Since this is a highly involved thesis essentially involving controversial arguments about Aristotle's development and aims, I will not try to deal with it here, for it would take us far afield. So, for present purposes, Irwin's thesis should be regarded as a live possibility.
But, besides offering a complex and controversial overall interpretation of Aristotle's development and fundamental aims, what else is left for defenders of dialectical method in Aristotle appeal to? One thing they do appeal to is the obscurity of Aristotle's account of induction in Chapter 19 of Book II of the Posterior Analytics. This is all Aristotle gives us about the way we derive first principles. And since it is less than crystal clear, some commentators are motivated to look elsewhere, usually to dialectic.
The most straightforward argument against a dialectical method for deriving principles is that Aristotle specifically says how we get them: by epagoge and nous. Whatever epagoge and nous amount to, they are not dialectic. Roger Crisp confronts such arguments by writing "Rather than play down the role of dialectic in science, however, it might be better to conclude with John Cooper that 'it is possible that Aristotle's position on this point [about epagoge and nous] is not in the end entirely coherent'". I find this statement curious.
Suppose Aristotle's account of the acquisition of arche via epagoge and nous is not in fact entirely coherent. Well, what of it? The alleged fact that Aristotle's position here is incoherent does not even begin to establish that Aristotle believed that dialectic had powers more significant than he said it did. It might indeed establish that if Aristotle's position were to be coherent, he would need to give a better account of how we get to archai. But, again, it would not begin to show that Aristotle himself thought that the strategy laid out in APo was insufficient and in need of a better alternative. And even if it was shown that he did find it insufficient, it would not begin to show that Aristotle thought that dialectic was the prime candidate for an alternative. To say that we find Aristotle's account of "intuitive induction" confusing, and that therefore there is a good chance Aristotle's method was dialectical is to say something, well, not so coherent . Our inability to clearly understand Aristotle's account of induction is not an argument that Aristotle didn't believe in it, nor is it an argument against "playing down the role of dialectic in the sciences."
If dialectic cannot establish first principles, what then is the function of dialectic in Aristotle and how does it relate to first principles and demonstration? I should like to venture a hypothesis.
In the Posterior Analytics Aristotle repeated states that the principles are better known to nature and are initially less well known to us. What is better known to us are conclusions that might be derived from principles. So, though principles have a certain objective priority, it is, as Smith says, "not only possible, but probably true that we find these first principles unconvincing or even absurd before we have acquired scientific understanding." That is, we can know what a principle of some science is, but we can fail to see it as a principle. For instance, a wise teacher might tell me that some proposition counts as a first principle in some science. But that does not mean that I see for myself that it is in fact a first principle.  In order to truly grasp principles one needs to both (1) know what propositions the principles are and (2) see for one's self that those propositions are in fact principles.
We come to every science with a head full of relevant experience and ideas. But we do not come to the study of a subject with our experiences and ideas already in good order. It is not immediately clear how all of the relevant experiences are to be integrated. And some of our ideas are confused, or false. We need to get straight before we understand principles as principles. As Aristotle says in Metaphysics Zeta, scientific education is not all that different from moral education. Just as we strive to transform our characters so that what is in itself good comes to seem to us to be good, so too do we strive to transform our minds so that what is in itself familiar comes to us to seem familiar.
Dialectic might not be capable of demonstrating anything, but it can transform our epistemic situation. As mentioned earlier, dialectic alone has a way to proceed when the subject is first principles. We can use the dialectical art to examine opinions (our own and others) about the nature and significance of principles and see which are inconsistent, which stand up to scrutiny and how they are logically related. In this way dialectic can help us to direct our attention away from irrelevant and distracting considerations and toward what really matters, thereby bringing order to our ideas and experience.
In his account of epagoge at the end of the Posterior Analytics, Aristotle uses the metaphor of a battle during a rout. Our experience ("perceptual memories many in number") is an unintegrated hodge podge, like warriors running every which way. But then, Aristotle says, one man takes stand. And then others do so as well, until a position of strength is reached. The analogy to experience go something like this. Some experience becomes particularly distinct and well comprehended. Once that point of order established, other experiences can fall into a firm place according to the nature of their relationship to the fixed point. What was once a jumble eventually becomes an integrated whole and the actual, rather than the simply apparent, relationships of priority and logical dependence are grasped.
Perhaps dialectic helps prepare us to "make a stand" by eliminating wayward notions and showing us the logical connections among the beliefs we hold. Once we get our beliefs in order, it is likely that we will capable of attending to the world in a more focused and less scattered way. Thus our subsequent experience is likely be better ordered and easier to integrate and unify. So, though dialectic cannot establish first principles, it may serve as a propadeutic to induction. By bringing one's set of beliefs into some sort of order, one may bring consistency and focus to sense awareness, which, in turn, may bring order and coherence to the experience which is the material of induction.
This hypothesis seems to me to be consistent with both the nature and limits of dialectic as Aristotle sets it out and with the admittedly difficult account of induction in the Posterior Analytics (as far as I comprehend it.) Furthermore, it does not require any necessarily problematic developmental theses.
In conclusion, I have argued that it is unlikely that Aristotle thought that dialectic was capable of generating first principles for several reasons.
First, the nature of dialectic itself makes it unlikely. Dialectic is an art concerned with valid deductions proceeding from premises classified according to what different types of people are willing to agree to contexts of question and answer. As Aristotle says, it is useful, first, in gymnastic competition, second, in argumentative encounters generally and, third, for discussing the first principles of sciences. It is capable of the latter due to its utter generality, which also accounts for its inability to prove anything.
Second, some defenders of a dialectical method in Aristotle argue that their case is helped by a disparity between the method Aristotle preaches in the Posterior Analytics and the method practiced in his treatises. But, I have argued, there is reason to believe that the theory/practice gap is merely apparent. If true, there are difficulties for those who wish to bridge the gap with a dialectical method.
Third, I argued that the passages that are most often used to support a dialectical method in Aristotle can in most cases be more plausibly interpreted as saying something else, or, at least, as saying something more limited than it is often supposed.
Lastly, I think there is a reasonably natural hypothesis about the relationship of dialectic and first principles that does not rest on sweeping developmental theories or on context-dropping interpretations of key passages. Additionally, the hypothesis does not require us to explain away or belittle Aristotle comments on epagoge and nous in the Posterior Analytics. Rather, the hypothesis attempts to identify a division of labor between dialectic and induction that is faithful to Aristotle's understanding of both.