Grassmann’s Law as weak phonetic change

1. According to Ohala (1989; 1993), dissimilation processes, especially involving non-adjacent segments, are highly unnatural, in the sense that the frequently invoked principles of speech production are unable to explain such a type of phonetic change. Rather, it represents the outcome of the listener’s mis-application of corrective processes. Acoustic-perceptual cues of some features (glottalization, pharyngalization, retroflexion, nasalization, aspiration) spread over beyond the immediate “hold” of the segment they are distinctive on. Therefore, the listener can interpret the original presence of one of these features on the segment in which it was distinctive as it were the outcome of an erroneous speaker’s production; consequently, it undoes such a feature through a mechanism of hypercorrection. For example, an original form *phath- could be misinterpreted by the listener as the result of an erroneous speech production, originated by the spread of the second segment [+aspirated] feature (h-) on the first one (*p-). 2. Such an explanation could shed new light on the well-known dissimilation process of Ancient Greek, the so-called Grassmann’s Law (= GL), the phonetic change in virtue of which in an original diaspirate root a (regressive) dissimilation process takes place: τίθησι < *θίθησι, τριχός < θριχός (cf. nom. θρίξ), τρέφω < *θρέφω and so on. Greek inscriptions from different dialectal areas show in some cases the preservation of the (presumably) original diaspirate roots (see e.g. Att. hέχει, καθέχει; καταθιθέναι; θρεφθε‚ς etc.); furthermore, diaspirate non original roots are also attested, in which original forms such as *path-/*phat- are both represented as phath- (see e.g. Att. hαρ[ιθμόν; Ἀνθίλοχος; θρόφος; ἐνθαῦθα; χαλχί[ον] etc.). Such evidence points the speaker’s entropic trend to spread over the laryngeal specifications beyond the affected segment. In contrast with this, dissimilated forms such path- from *phath, with a regular application of GL, represent the listener’s syntropic trend, which aims to (hyper)correct all the diaspirate forms (both the original ones and those “wrongly” produced in the speech). 3. The interplay between the entropic speech production and the syntropic listener’analysis produces a long term syncronic variation between diaspirate and dissimilated forms, until the dissimilated forms imposes the template path- as a rule. The delay with which this type of linguistic change spreads out in the lexicon could be due to the lack of an articulatory basis of the dissimilatory processes, which are rather grounded on the listener’s inability to parse the acoustic signal appropriately. The same motivations could explain the likely retention of the original diaspirate roots in Mycenaean and in Homeric Greek, where both (-)h- and aspirated voiceless obstruents seem not to undergo dissimilation. 4. Lastly, dissimilation as listener’s mis-perception could enlighten the instability in the dissimilation processes in some compounds (e.g. ἐκεχειρία alongside ἐχέθυμος and ἐχέφρων), where GL is anything but systematic. One could bring into play the presence of a morpheme-boundary in these forms between the two [+aspirated] segments as an actual obstacle towards the listener’s effort to (hyper-)correct through a dissimilatory process.