From Filologia Neotestamentaria 8 (1995), pp. 151-185. Here is only the ending paragraphs (and without footnotes), but you can read it also complete.
There seems to be but one course to take: to abandon the Erasmian pronunciations and to return to the Greek pronunciation. This is "a scientific demand and a practical desideratum", to use a phrase coined by a great New Testament scholar in another connection, and that for the following reasons:
1. The Erasmian claim to pronounce Greek in a scientific way, that is, in the ancient Greek fashion, is beset by insuperable difficulties.
First, it is common knowledge that no-one can learn to pronounce a foreign language by merely reading books in that language or consulting dictionaries, even such as are provided with phonetic helps. One must expose oneself constantly to the sounds of that language by listening to and trying to imitate native speakers. And even then it will be extremely difficult to learn to pronounce the language as the natives do, if the learner is older than eight years of age. In the case of ancient Greek we have no longer the possibility to hear Socrates or Plato, let alone the ability as grown-up students of Greek to imitate its correct pronunciation.
Second, it becomes immediately incumbent upon the Erasmians that they apply to the texts of each particular period the pronunciation that was current at the time. Thus, Homer should be pronounced with the pronunciation that was used in his time, Plato and Aristotle with the V-Ivth c. Athenian pronunciation (which was undergoing important changes), the New Testament with a pronunciation that was practically identical with the Modern Greek pronunciation, and the Church Fathers in the Modern Greek way.
Third, four and one half centuries of trying to establish the scientific nature of the Erasmian pronunciation has led to results that are demonstrably false, or that have failed to convince the theorists themselves. To illustrate this I will quote a few passages from one of the more recent defences of Erasmianism, Allen’s Vox Graeca. Practical difficulties in "distinguishing the voiceless unaspirated plosives from the aspirated, both in speaking and hearing" lead Allen to bypass the Erasmian pronunciation at these points and to counsel "pronouncing the aspirated plosives in the Byzantine manner" (i.e. Modern Greek)! (p. 27). On p. 35 a certain pronunciation is recommended not on scientific grounds, but "on practical grounds"! On p. 57 "any degree of aspiration that may have existed here can be ignored by the modern reader". When on p. 73 he cannot make up his mind, he recommends a certain course because "if we are wrong, at least we shall be doing nothing worse than, say, pronouncing Aeschylus as Demosthenes might have done; whereas, if we adopt the other alternative, we may be giving an author a pronunciation which he had never received in antiquity"! This revealing admission is most telling, but one also wonders why in the light of this Erasmians still persist in pronouncing e.g. the New Testament (even from their point of view) in an anachronistic way? On p. 83 the conclusions to which his study has led him are not good enough for recommendation, so he counsels "the simplest solution seems to be one which is in fact quite widely adopted, namely to anticipate developments by two or three centuries"! We may therefore, ask, Why not substitute the entire concoction by what we know to have been the pronunciation "two or three centuries" later, i.e. practically Modern Greek? With regard to the notorious "musical accent" of ancient Greek, Allen says on p. 118: "The author has listened to a number of recordings, recent and less recent, of attempted tonal [i.e. musical] recitation of ancient Greek, and, whilst some are less objectionable or ridiculous than others, has found none of them convincing". After such a confession, which is tantamount to a total failure by Erasmians to tell us how the so-called ancient Greek musical accent sounded, one would have expected the author to recommend the so-called Hellenistic stress-accent, (which still lives in Modern Greek). But nothing of the kind. The author goes on: "The carefully considered advice is therefore given, albeit reluctantly, not to strive for a tonal rendering, but rather to concentrate one’s efforts on fluency and accuracy in other aspects of the language". In the light of the above admissions the inevitable question arises: Just what is the point of persisting in pronunciations in which even their supporters and theorists have lost confidence?
If it is so clear then that the pronunciation (in the strict sense, not only of the value of the various letters, but also of the sound quality) of Homer and of classical antiquity is, in the absence of magnetic tape-recordings, for ever lost to us and beyond the possibility of recovery or reconstruction, is it not, in that case, historically and scientifically more honest and correct to pronounce the language according to its own natural and historical development, rather than to impose upon it foreign sounds imported from other sister or rather "niece" languages within the Indo-European family? If only one pronunciation is to be used in pronouncing all these types of writing —coming as they do from a time span of 1200 years and more, during which period the pronunciation in fact evolved— then surely the Greek pronunciation (whose roots go back to the Vth and Ivth c. B.C.), is the only legitimate candidate, not the artificial construct of Erasmus.
2. The Greek pronunciation of Greek is a sine qua non for Textual Criticism. The manuscript tradition is full of errors that were often the inevitable consequence of the double tradition —the living language and historical orthography— exactly the same type of errors that we find in the Attic inscriptions of classical times. The Greek pronunciation is the key to many variants and must be made the basis for a correct evaluation of their origin as well as their solution.
3. There is also the pragmatic issue. Pronouncing Greek in the Greek way will facilitate scholarly contact with Greece. Moreover, it will open an avenue with the starting-point of a little knowledge of New Testament Greek (or even classical Greek) to enter the wealth of Byzantine and Modern Greek, which are the direct descendants of Hellenistic and New Testament Greek. In this way New Testament Greek will cease to be treated as an island with its attendant misconstructions; it will be seen as part of a greater living unity, the Greek language, Greek thought, and the Greek literature as a whole. This will not fail to enrich the scientific study of the New Testament, which for too long has been deprived of inestimable insights by its persistent adhesion to the error of Erasmus.
From: The error of Erasmus and un-greek pronunciations of Greek, by Chrys C. Caragounis, Filologia Neotestamentaria 8 (1995), pp. 151-185.