Dear Madhav,Thanks for your rich remarks.
The non-change of śrī to śry in order to preserve it honorific function you discussed in your article "Contextualizing the eternal language: features of priestly sanskrit" (In Ideology and Status of Sanskrit, Leiden 1996: 401-436, section 3: "Use of honorific śrī"), where you refer back to your earlier 1993 article "Inscriptional Evidence for Honorific Śrī" in Indo-Aryan." In: Perspectives in Indian Aesthetics and Literature, Essays in honour of the Late Prof. Dr. G.ed. by Saroj Deshpande and Maneesha Dikshit: 254-278. Pune: Dastane Ramchandra.
Your evaluation in 1996 that "Thus, the pragmatic need to maintain the recognizable identity of this item overrides the grammatical requirement of making the sandhi" is no doubt appropriate for the priestly sanskrit you are dealing with in that article. However, if unsandhied junctures occur in the prose of a grammarian, such consideration would become rather awkward. Instead of making the grammatical rules optional, one could perhaps conclude that it is the condition saṁhitāyām which is not fulfilled (leaving the grammatical sandhi rules that depend on that condition intact); that the prose text, imagined as a spoken text, contains a "light" pause between subordinate clause and main clause etc., or between explicandum and gloss (even if there is no daṇḍa).
The use or non-use or partial use of sandhi in grammatical rules within Pāṇini's AA which you cite is again an important but very different domain. According to Joshi and Roodbergen (1991, p. 2), "Pāṇini's sūtra-language differs from ordinary, literary Skt in this that the rules followed in this type of Skt are not necessarily applied in the sūtra-language also." According to the analysis of Joshi and Roodbergen (ib.), the sūtra "vṛddhirādaīc" consists of only two words, "vṛddhir" and the samāhāra dvandva "ād-aīc"; and for a samāhāra compound ending in -c Pāṇini gives the samāsānta-suffix ṬaC (AA 5.4.106), as in vāc+tvac+ṬaC vāktvacam “aggregate of speech and skin”. However, in the sūtra-language this is not applied, in a “concern to keep the expression clear and unambiguous” (commentators’ solution: this rule is anitya).
Regarding commentarial prose, I got one reaction off-list which referred me to a really excellent multifaceted discussion of the problematic (without suggesting a symbol to use in the case of probably significant cases of non observing sandhi).
It is p. 12-17 in Lo Turco, Bruno, ed. 2019. Mokṣopāya-Ṭīkā of Bhāskarakaṇṭha: The Fragments of the Nirvāṇaprakaraṇa. Part II: Critical Edition. Studia Indologica Universitatis Halensis 14. Halle (Saale): Universitätsverlag Halle-Wittenberg.All best,Jan HoubenOn Mon, 13 Jul 2026 at 02:34, Madhav Deshpande <mmdesh@umich.edu> wrote:Several cases of at least incomplete sandhi occur in Pāṇini. In the rule "vr̥ddhirādaīc", the final c does not turn into k, as it does in vāc > vāk. But in the rule "updeśe 'j anunāsika it", the c of ac at least turns into j due to sandhi, but not the changes of c > k, and k > g. Pāṇini does not follow his rule of the final c > k in such places, because he has aK and aC as two separate Pratyāhāras and he needs to keep them separate. In one of my old articles, I had pointed to some inscriptional usages of śrī like śrī + ananta not undergoing sandhi, because changing śrī to śry would harm its honorific function. Similar exceptional sandhis are noted by Pāṇini [omāṅoś ca], where the sandhi of śivāya + om results into śivāyom and not śivāyaum. Similar exceptional sandhis occur for pādapūraṇa [so 'ci lope cet pādapūraṇam]. Several phenomena coming under the condition "apādādau" may be seen as metrical exceptions to normal phenomena.Madhav M. DeshpandeProfessor Emeritus, Sanskrit and LinguisticsUniversity of Michigan, Ann Arbor, Michigan, USASenior Fellow, Oxford Center for Hindu StudiesAdjunct Professor, National Institute of Advanced Studies, Bangalore, India[Residence: Campbell, California, USA]On Sun, Jul 12, 2026 at 6:59 AM Jan E.M. Houben via INDOLOGY <indology@list.indology.info> wrote:Dear List members,As sandhi at the juncture of two words is expected in correct sanskrit, its absence could point to carelessness of the author or scribe, but also (rather) to the intention to make a quoted form stand out more clearly, as in pacyase iti, or to indicate a "weak" clause boundary, for which a da.n.da was felt to be too "strong". I'm sure this problem has been dealt with in various transcriptions and editions of inscriptions and manuscripts. Is there any best practice regarding such places where sandhi is not observed, intentionally or erroneously? Especially in the former case the editor has to avoid "hypercorrection". Any widely accepted symbol (which could by itself lead to confusion by suggesting that the symbol represents something in the source)? Or adding a (clumsy) remark in the text, "sandhi not applied", or in the apparatus each time? (In Epigraphia indica volumes it appears commonly in notes "sandhi is not observed here", e.g. EI 37 p. 83, 84, 304.) A plus-sign, pacyase+iti, would probably at first sight suggest that an existing sandhi in the witness (manuscript / inscription) is undone by the editor.Best regards,Jan Houben--Jan E.M. Houben
Directeur d'Études, Professor of South Asian History and Philology
Sources et histoire de la tradition sanskrite
École Pratique des Hautes Études (EPHE, Paris Sciences et Lettres)
Sciences historiques et philologiques
Groupe de recherches en études indiennes (EA 2120)
johannes.houben [at] ephe.psl.eu
_______________________________________________
INDOLOGY mailing list
INDOLOGY@list.indology.info
https://list.indology.info/mailman/listinfo/indology
--Jan E.M. Houben
Directeur d'Études, Professor of South Asian History and Philology
Sources et histoire de la tradition sanskrite
École Pratique des Hautes Études (EPHE, Paris Sciences et Lettres)
Sciences historiques et philologiques
Groupe de recherches en études indiennes (EA 2120)
johannes.houben [at] ephe.psl.eu