Why are postulation (arthāpatti) and inference not the same thing?

Arthāpatti ‘postulation’ is the instrument of knowledge through which we know that Devadatta is out given that he is alive and not home. In Classical India, just like among contemporary scholars, several thinkers (especially of the Nyāya school) have tried to show that it is only a subset of inference.

Within the weekly reading group facilitated by Malcolm Keating, we are reading the section on arthāpatti of the Mānameyodaya by the Bhāṭṭa Mīmāṃsaka Nārāyaṇa Bhaṭṭa. This week, we read the part on the difference between inference and postulation according to the Prābhākaras.

It is often the case that neighbours hate each other, and Nārāyaṇa does in fact attack the members of the rival school of Prābhākara Mīmāṃsā much more violently than the Naiyāyikas. The Prābhākaras agree with the Bhāṭṭas against the Naiyāyikas that the arthāpatti is distinguished from inference, but they disagree as for why this is the case.
Nārāyaṇa uses various arguments, the last of which is that the invariable concomitance between what one knew before and what one discovers during that cognitive process is already available only in the case of inference. In other words:

  1. inference: one knows already that smoke is invariably concomitant with fire
  2. postulation: one knows only at the end of the process from “Devadatta is alive and not home” to “Devadatta is out” that “being alive while not being home” is invariably conomitant with “being out”

By contrast, the Prābhākaras use a different argument to distinguish the two, namely that the hetu ‘logical reason’ in the case of inference is firmly established, whereas it is doubted in the case of postulation. In other words:

  1. inference: one infers that there is fire on the mountain because of smoke (firmly established)
  2. postulation: one ascertains that Devadatta is out because he is possibly alive (doubted reason)

This being said, please enjoy the depiction of these positions and most of all the witty refutation of the Prābhākara one by Nārāyaṇa:

Before having ascertained that [Devadatta] is outside [his] home, the conjunction of being alive and not being home could not be known. […]

This has been said in [Kumārila’s] Bṛhaṭṭīkā*:

Therefore, the absence from home which is understood in regard to one who is present (i.e., alive) |

this [would be] the probans (if the postulation were a case of inference), but this is not seized before seizing that [the alive person] is out of his home || 143 ||

(tasmād yo vidyamānasya gṛhābhāvo ‘vagamyate | sa hetuḥ sa bahirbhāvaṃ nāgṛhītvā ca gṛhyate)

Therefore the postulation is indeed distinguished [from the inference].

By contrast the Guru (Prabhākara), who does not known this (mentioned above) tool to destroy the Naiyāyikas, has prattled on the doubt about the being alive in this case [as the trigger of postulation]: || 144 ||

“The being alive indeed has been known before as being home |

Thereafter, there must be doubt about his being alive due to the fact that one has not seen [Devadatta] in his home || 145 ||

But the doubted fact of being alive can convey the fact that he (Devadatta) is out of his home |

This is the advantage [over inference] of the postulation, that although it is doubted it conveys [something] || 146 ||

In this way, if there is a doubt regarding the fact of being alive the probans would have a doubted qualification. In this way the refutation of the status of inference is very easy for us! || 147 ||”

[Bhāṭṭa:] That is ridiculous! To elaborate:

If the fact of being alive were doubted because one has noticed that one is not home |

then the ascertainment of that (being alive) would be done, on the basis of assertions of reliable speakers and so on || 148 ||

or by means of considering signs such as the auspicious necklace (worn by married women who are not widows, so that its presence on the neck of Devadatta’s wife would be a clear sign of his being alive) of his beloved one |

But this is not desired at all [by the Prābhākara] (who in fact does not undertake any of these things). Therefore [in reality] there is no doubt at all [even in the Prābhākara’s mind] || 149 ||

Moreover, one does not seize that [Devadatta] is out on the basis of a doubtful idea that he might be alive |

for, if one doubts that he might also be dead, how could the idea that he is out [originate]? || 150 ||

“Since he is alive or not, he is out” |

who else apart from the Guru (Prabhākara) would be able to postulate that? || 151 ||”

*By the way, do you happen to know this verse from some other source? (Or do we have to imagine that Nārāyaṇa Bhaṭṭa still had access to the then lost Bṛhaṭṭīkā?)

For more on postulation, on Nārāyaṇa’s Mānameyodaya and on our reading sessions, please see this post.

(cross-posted on my personal blog)

About elisa freschi

My long-term program is to make "Indian Philosophy" part of "Philosophy". You can follow me also on my personal blog: elisafreschi.com, on Academia, on Amazon, etc.

20 Replies to “Why are postulation (arthāpatti) and inference not the same thing?”

  1. Elisa, I’ve often thought that “inference to the best explanation” or more rarely, “transcendental argument” are good translations of arthApatti, and have the benefit of familiarity for those of us trained in Western philosophical departments. Do you think that “postulation” is a better translation because it avoids the baggage of the other two?

  2. Matthew, I see your point and I would speak of “inference to the best explanation” if I had no more than a few seconds to explain it to a public trained in Western philosophy. However, to me the problem with “inference to the best explanation” is that it is
    a) theory-loaden (it may be misleading to think of Peirce while trying to read an Indian text)
    b) too much connected with an inferential *process* (while arthāpatti is instantaneous)

    As for “transcendental argument”, to me it evokes Kant and I cannot link it in any possible way to arthāpatti. How do you succeed in doing it?

    I used to like “postulation” (since it suggests that a solution is built while achieving it), but Michael Reidy showed an interesting argument against it, namely that a “postulation” is again a process constructing something, and not a swift (you can read Michael’s comment here: http://elisafreschi.com/2015/02/26/why-are-postulation-arthapatti-and-inference-not-the-same-thing/#respond and his full response at his blog, here: http://ombhurbhuva.blogspot.ie/2015/03/arthatpatti-language-goes-on-holiday.html)

    • Regarding IBE, at least, I think that it is nowadays understood as a basic kind of inference and need not be tied to a particular approach to abduction in Peirce. And it can be instantaneous too. I look outside of my window in the morning and see that the trash can is knocked over and trash strewn about. I immediately infer that racoons have toppled it over looking for food (as opposed to a prank by local kids or wind or whatever).

      I agree that the theoretical baggage of transcendental arguments with Kant is more problematic. By “more rarely” I meant that I don’t think it is as good. I wasn’t being clear.

      I think that somewhere, I saw that JN Mohanty translated arthApatti as transcendental argument and that’s why I thought of it. But I can’t think of the passage right now. It may have been in his short book *Classical Indian Philosophy*, or it may be a false memory!

      I don’t like postulation, since it sounds like something opposed to proof or decisive conclusion.

      • I have used “presumption” rather than “postulation” although I’m still wavering between the two. I think “presume” has the benefit of being relatively untheory-laden (although philosophers can taint even the most ordinary terms with theoretical significant) and getting to how I understand arthāpatti to operate. Take the Fat Devadatta sentence–it seems ordinary to hear “Fat Devadatta does not eat during the day” and reply, “Oh, I presume he eats at night.” Plus, the term has the benefit of being used by Grice in a few places for how conversational implicature works, and I think the two are closely related, though that’s not how everyone understands arthāpatti.

  3. Thanks, Matthew. Look, this is not my topic and I am not an expert, but what about the problem that the IBE achieves a lower level of probability than arthāpatti? I would not have thought of racoons and I am sure that your IBE in that case is only, say, 80% true, whereas arthāpatti seems to aim at more than that (not 100%, given that we are still within Mīmāṃsā).

    • Well, the way that Indian philosophers (at least those with whom I’m most familiar; and they are not largely mImAMsakas) tend to speak of cognitive success as “on” or “off” need not be seen as a point about probability of success but rather that factivity is built into the definition of pramANa. The probability of apparent testimony being veridical need not be 100% for testimony proper to be defined as factive.

      Incidentally, while IBE is broadly inductive, and hence does not guarantee it’s conclusion, it still can have a high degree of likelihood. I come home. The door is ajar, and the lock has been busted open. Things are scattered about and my valuables are missing. By IBE, I infer that someone (or some people) has broken in and stolen my things. Pretty strong inference.

      • Matthew, I think we discussed this topic already, but the point to me is:

        a pramāṇa is not a probabilistic tool

        It might be true or false, and this is a different part of the story, but the probability of being true is not built in in it, as it is, by contrast, in the case of IBE. arthāpatti is not just likely to be true, it is on a par with inference and sense perception (and the other instruments of knowledge). This is, in part, a pity, since I would prefer Indian epistemology to be more open to probability and less concerned with truth, but these trends remained, I am afraid, at the borders (perhaps in Sucarita Miśra?). The same problem applies to upamāna —it would be much more powerful if it did not have to be invariably valid, which paralyses its ampliative potentiality.

        • I think it is not correct to say that IBE is probabilistic. Or, at least, not all varieties of what is broadly called “IBE” are (just how to carve out the territory is disputed). See Gilbert Harman’s 1965 paper “Inference to the Best Explanation” distinguishing between enumerative induction and IBE. One can also distinguish, as many do, between abduction and IBE, where abduction is probabilistic or assessing that a claim “might” or “probably” is true, and IBE adds additional lemmas that strengthen abduction to the point where the claim is the best explanation on offer, and against this background, true. Now, you can give up those lemmas and then the conclusion becomes probable, not true, but likewise for inference and sense perception. Give up the background assumption that you are in a situation of ordinary lighting (has #thedress not taught us anything?) and then your perception is not knowledge, but it depends on how likely it is that you are in in such a situation.

          Second and relatedly, it is an important question whether there is only one arthāpatti. This point has been made in “The Heterogeneity of Arthāpatti” by Tulsi Ram Kanaujia, which argues that the different cases marshaled to support arthāpatti can lead us to different conclusions. I think this is an important question that merits careful investigation. I am not yet convinced that there are not different kinds of arthāpatti which we ought to distinguish between, even if the Mīmāṃsakas, for instance, did not.

          Finally, in light of this, The claim made about the “on/off” switch is more plausible to me in the case of Devadatta’s absence from his home than it is for the sentence “Fat Devadatta does not eat at night,” and even less so in the cases where postulation is used to adduce that language has an instrinsic capacity for meaning or that the sun has an ability to move itself across the sky. This, I think, also has to do with how we understand the background beliefs to operate in each of these cases, but this is a long enough comment for now.

          • Malcolm,

            1) I agree that the term arthāpatti might have been used to cover slightly different cases (Birgit Kellner has shown that this was the case, for instance, with abhāvapramāṇa and its later application to absent pots). This is something one need to very carefully investigate and you are very right in pointing it out. In this connection, could you provide a reference for the two cases of arthāpatti you mention, namely: “language has an instrinsic capacity for meaning” and “the sun has an ability to move itself across the sky”?

            2) I would not mix our data with the ones taken into account by Indian epistemologists. “The dress” might teach us something, but it is not legitimate to project its results back to the Indian scenario. Moreover, an Indian epistemologist would have easily explained it away as due to particular upādhis. The problem is whether a pramāṇa in a condition of nirupādhi is probabilistic or not. According to Mīmāṃsā authors, arthāpatti is not.

            3) Thanks for the important references about IBE, I realise I did not know enough about it. But perhaps other readers might be, like me, not learned enough and think that it is a choice of the best option among several available ones…thus, I would still suggest to avoid it as a translation of arthāpatti.

            4) Pointer for the reader: Michael Reidy is continuing his series on arthāpattii: http://ombhurbhuva.blogspot.co.at/2015/03/arthapatti-not-abduction.html

  4. elisa, I am considering writing a post on this topic which may make my thinking clearer. To start with, I definitely agree that translating “arthāpatti” as “inference to the best explanation” is misleading as a translation even if it turns out that its logical structure is IBE, or something similar. As I said to Matt, I’ve used “presumption” in my translation of Mukulabhaṭṭa, who talks about śrūta-arthāpatti, and I also think “postulation” is a good, relatively neutral translation which coheres with the frequent use of “kalpanā” and “ākṣipyate” with “arthāpatti.”

    As for the references to the examples I mentioned, see Kumārila’s chapter on arthāpatti in the ŚV in which he breaks out the ways arthāpatti can occur in connection to the different pramāṇa. The sun’s movement is an instance of inference first or prior. The capacity of language is an instance of postulation coming first or prior. He talks about the latter again in (I think–I need to look) the chapter on sentence meaning.

    Finally, I was being a bit humorous with the dress story and not meaning to import current thinking into Indian philosophy, but rather to make a general philosophical point about IBE and how our background beliefs are important to how we understand what’s going on. I agree it would be a described as a case of upādhi-s. My aim was to contrast situations where we can have knowledge that is not probabilistic and knowledge that is. Probably it is just a poorly-employed example.

    • Malcolm,

      I hope I did not sound too rude. My point about “the dress” was just epistemological —I appreciate the fact of keeping on thinking about contemporary problems through Sanskrit categories. Should not we be able to do it, we would have lost sight of their philosophical application.

  5. Pingback: Translating a (Sanskrit) philosophical text as a group work | elisa freschi

  6. At the end of his long commentary on the “aphorism of rasasūtra, (Nāṭyaśāstra 6.31c), Abhinavagupta in the Abhinavabhāratī articulates the distinction between rasa-gustation in “aural literature” (śravyakāvya) as compared to “visual literature” (dṛśyakāvya). He discusses the relative prominence of the textual elements (vibhāvas, anubhāvas and vyabhicārībhāvas) in a poetic text and after quoting exempla in which only one element from these three is prominent, he remarks:

    eva dvaya prādhānye côdāhāryam. kiṃtu samaprādhānya eva rasāsvādasyotkarṣaḥ. tacca prabandha eva bhavati. vastutastu daśarūpaka eva. yadāha vāmanaḥ:- “saṃdarbheṣu daśarūpakaṃ śreyaḥ. tadvicitraṃcitrapaṭavadviśeṣasākalyāt”. (Kāvyālaṃkāra Sūtrāṇi 1.3.30-31) iti tadṛūparasacarvaṇayā tu prabandhe bhāṣāveṣapravṛttyaucityādikalpanāt. tadupajīvanena mukatake. tathā ca tatra sahṛadayāḥ pūrvāparaṃucitam parikalpya īdṛgatra vaktāsminnavasare ityādi bahutaraṃ pīṭhabandharūpaṃ vidadhate. tena ye kāvyābhyāsaprāktanapuṇyādihetubalād iti (bhiḥ) sahṛdayāsteṣāṃ parimitavibhāvādyunmīlane‘pi parisphuṭa eva sākṣatkārakalpaḥ kāvyārthaḥ sphurati.ata eva teṣāṃ kāvyameva prītivyutpattikṛdanapekṣitanāṭyamapi. teṣāmapi tu nāṭya “nipatitāḥ sphuritāḥ śaśiraśmyaḥ” iti nyāyena sutarāṃ nirmalīkaraṇam. aḥrdayānām ca tadeva nairmalyādhāyi. yatra pati (prati) tā gītavādyagaṇikādayo na vyasanitāyai paryavasyañti nāṭyopalakśaṇāt.

    Mammaṭa at Kāvyaprakāśa 3:21-22 fleshes out the contextual factors hinted at by Abhinava:

    vaktṛboddhavyakākūnāṃ vākyavācyāñyasannidheḥ
    prastāvadeśakālāder vaiśiṣṭyāt pratibhājuṣām
    yo’rthasyānyārthadhīhetur vyāpāro vyaktireva sā

    What’s interesting in this passage is that Abhinava holds that sahṛdayas can by “imagining what precedes and what follows generate a considerable context even for a short verse by determining that ‘This is the utterance of a particular person on a particular occasion’ and so on” and that “For such Readers, by virtue of their sustained study of poetry and their former good deeds, the meaning of even such short, contextless texts lucidly flashes forth immediately as if directly perceived, despite the restricted unfolding of the textual elements.

    My question, learned friends, is this: how similar is this process to arthāpatti?

    • Dear Satyanarayana,

      thanks for this very interesting remark. I have a few comments and a prayer for you:
      1) I can see the similarities between the two processes, but I am afraid that epistemologists tend to be quite strict about the functioning of a given pramāṇa, so that thinking “analogically” is not enough to be doing an upamāna and having a fleshy understanding of something does not yet amount to arthāpatti —which is a pity, given that in this way much gets lost for the pramāṇas.
      2) I am inclined to think that the pramāṇas beginning with sense-perception only apply to state of affairs. Poetry is not the object of these pramāṇas, although other faculties come into play when it comes to experiencing poetry. What do you thin?
      3) The text as you reproduce it is not completely devoid of typos (apart from the lack of spaces between words, see, e.g., paryavasyañti or nāṭyopalakśaṇāt or pati (prati) tā). For the sake of readers less experienced in Abhinavagupta (me included), could you please take care of it once again? Sorry for the hassle.

      • Elisa, someone more well-versed in the history of vyākaraṇa and alaṅkāra could probably fill in the details, but as I understand it, there is a tradition of considering arthāpatti to be a figure of speech/textual matter. For instance, Kautilya mentions it in his Arthaśāstra as one of the 32 parts of a text, alongside of upamāna (dates for this could range from 150 BCE to 300 CE–I’m not sure where this section falls in Olivelle’s analysis of textual material). Also, the term appears (much later) in Appayadīkṣita’s Kuvalayānanda, where it has been taken up from the Mīmāṃsā. Gerow says that it was considered a figure before Appayya, though he doesn’t say where. This was a standard poetics textbook, so it seems that considering arthāpatti a figure was accepted by that point.

        It is also interesting to note that in his Nyāyamañjarī, Jayantabhaṭṭa, who is arguing that arthāpatti is really anumāṇa, says that Ānandavardhana’s proposal of suggestion or dhvani is also refuted, through his refutation of śrūtārthāpatti. This would indicate that he saw them as closely related, though he does not say more. And Mukulabhaṭṭa argues that a paradigmatic sentence exemplifying śrūtārthapatti (Fat Devadatta does not eat during the day) is in actuality an instance of lakṣaṇā, in his argument that dhvani is itself also lakṣaṇā.

        All of this suggests to me that there are some connections between arthāpatti and non-literal speech that are present, perhaps not picked up and made explicit by those writing major epistemological treatises, but in alaṅkāraśāstra. On my understanding of arthāpatti, that’s the right thing to think. It can begin with sense perception, but it also begins with testimony and arthāpatti itself.

          • I’m not sure I would go so far as to say this is an application of arthāpatti. I would think that should be more structural similarities — so imagining what comes before and after seems to be too much, since in arthāpatti, we are relying on prior knowledge (whether testimonial, perceptual or etc) and are doing something more constrained than just “imagining.”

            However, this is an area of interest, not expertise for me. I would like to know more about the relationship between the pramāṇas and figures in aesthetic theory (apparently there were at least 10 “quasi-pramāṇa” figures according to Gerow’s Glossary of Indian Figures of Speech, including pratyakṣa, aitihya, etc.).

  7. Dear Elisa and Malcolm, many thanks for your kind and thoughtful responses. Please accept my sincere apologies for responding belatedly. I’m neither an academic nor a scholar; I’m merely an attorney (very) interested in Sanskrit literature, and as is well known, the Law is a jealous mistress…

    Arthāpatti, Malcolm, is one of the excellences (lakṣaṇā) mentioned in the kāvyalakṣaṇā chapter of the Nāṭyaśāstra, defined at 16.161:

    arthāntarasya kathane yatrānyo’rthaḥ pratīyate
    vākyamādhuryasampannā sārthāpattirudāhṛtā

    Abhinavagupta, in the Abhinavabhāratī commentary illustrates arthāpatti with a verse from Īśvaradatta’s dhūrtaviṭasaṃvāda:

    ādaṣṭasphuritādhare bhavati yo vaktrāravinde rasaḥ
    prītiryā ca hṛtāmśuke ca jaghane kāñcīprabhoddyotite
    lakṣmīryā ca nakhakṣatāṅkuradhare pīne kapole striyo
    raktaṃ tena virajyate na hṛdayaṃ jātyantare’pi dhruvam

    Abhinava then states “atra janmāntare’api tasyā anuraktiranurvartata evetyukte janmāntareṣvāpi virāgo na prabhavati tasmādbhavabandhavimuktirna syādityarthāpattirnāma lakṣaṇaṃ.” (I’ll return to Abhinava’s views about arthāpatti infra)

    Arthāpatti thereafter does not appear as a rhetorical device from Bhāmaha to Mammaṭa till Ruyyaka’s Alaṃkārasarvasva, which is the first work to cite it (Gerow does mention this, Malcolm, at page 154, fn138 of “A Glossary of Indian Figures of Speech”). Ruyyaka defines it with the maxim of the “stick and the cakes,” daṇḍāpūpikānyāya: “atra hi ‘mūṣakakartṛkeṇa daṇḍabhakṣaṇena tatsahabhāgyapūpabhakṣaṇam arthāt siddham’ eṣa nyāyo daṇḍāpūpikāśabdenocyate. tataśca yathā daṇḍābhakṣaṇādpūpabhakṣaṇamarthāyātam tadvat kasyaścidarthasya niṣpattau sāmarthyātsamānanyāyatvalakṣaṇād yadarthāṃtaramāpatati sārthāpattiḥ.” Śobhākaramitra in the Alaṃkāraratnākara follows Ruyyaka and defines Arthāpatti as “daṇḍāpūpikayā’’patanamarthāpattiḥ.” (Ruyyaka classifies Arthāpatti into two types based on considerations of prakṛta and aprakṛta; i.e., when a prakṛta idea follows from an aprakṛta and vice-versa; Śobhākara divides it into four, adding ideas that follow from prakṛta to prakṛta and from aprakṛta to aprakṛta and then further divides these four into twelve on the consideration of the ideas being equal, greater or lesser, “sāmyam nyunatādhikyam,” and thereafter further derives these into twenty-four based on possibility and impossibility of the ideas, “sambhavāsambhavau” and all of these can be based on sound or sense, thereby yielding over a hundred!). Appayadīkṣita cites Arthāpatti as a rhetorical device in the Kuvalayānanda, significantly terming Arthāpatti as kāvyārthāpatti, defining it as “kaimutyenārthasaṃsiddhiḥ kāvyārthāpattiriṣyate” and exemplifiying it as “sā jitastvanmukhenenduḥ kā vārtā sarasīruhām.” Paṇḍitarāja Jagannātha (Appaya’s bitter opponent) in the second ānana of the Rasagaṅgādhara deletes the element of daṇḍāpūpikā from the definition of Arthāpatti, defining it as “kenacidarthena tulyanyāyatvādarthāntarasyāpattirarthāpattiḥ,” qualifying it further stating “nyāyaḥ kāraṇam.”

    The common element in the nāṭya and kāvya theorists’ definition of Arthāpatti as an alaṃkāra is that they all characterize it as a fortiori reasoning, which does not require anupapatti as one of the necessary elements. I’ll therefore beg to differ (data venia) with Gerow (page 333) who, referring to Appaya’s use of the term states that this is “the mīmāṃsaka pramāṇa ‘a fortiori.’” The basic requirement for Arthāpatti in so far as the Mīmāṃsakas (Bhāṭṭas and the Vedāntins) are concerned is anupapatti. The Vedāntaparibhāṣa divides Arthāpatti into Dṛṣṭārthāpatti and Śrutārthāpatti and further subdivides Śrutārthāpatti into Abhidhānānupapatti and Abhihitānupapatti (5.7). Abhidhānānupapatti is “anvayābhidhānānupapatti” (5.8) and Abhihitānupapatti is “vākyāvagato’rtho’nupapatti” (5.11). The Karaṇa of Arthāpatti is “idamanena vinānupapannam” (5.13). Thus, no anupapatti, no Arthāpatti. Jagannātha, specifically discussing this aspect while delineating Arthāpatti states “atra vicāryate-neyam vākyavitsaṃmatāyāmarthāpattau niviśate. āpādakasyārthasyāpatitamarthaṃ vinānupapatteratrābhāvāt…” A detailed paper on the kāvya theorists’ definition of Arthāpatti as an alaṃkāra remains a desideratum. Of all the Kāvya theorists, only Bhoja defines Arthāpatti in terms identical with the Śāstra theorists. I’ll cite Sarasvatīkaṇṭhābharaṇa 3.23.52-53:

    pratyakṣādipratīto’rtho yastathā nopapadyate
    arthāntaram ca gamayatyarthāpattim vadanti tām

    sarvapramāṇapūrvatvādekaśo’anekaśaśca sā
    pratyakṣapūrviketyādibhedaiḥ ṣoḍhā nigadyate

    Bhoja, following Kumārila’s classification of Arthāpatti in the Ślokavārtikā, conceives of pratyakṣapūrvikā Arthāpatti, anumānapūrvikā Arthāpatti, arthāpattipūrvikā arthāpatti etc.

    The question of the interrelation between Dhvani, Lakṣaṇā and Arthāpatti is quite complex, to which I’ll add the issue of Vākyaśeṣa. My thoughts on this issue oversimplify an extremely complex and prolix topic and are merely in the nature of exploratory submissions…..:) Ānandavardhana in the Dhvanyāloka speaks of only two śabdaśaktis, Abhidhā and Lakṣaṇā with Dhvani as the proposed third category. Abhinava, however, in the Locana (on Dhvanyāloka 1.4b) proposes a four-fold schema, viz. Abhidhā, Tātparya, Lakṣaṇā and Dhvani: “tena samayāpekṣā vācyāvagamanaśaktirabhidhāśaktih.tadanyathānupapattisahāyārthāvabodhanaśaktistātparyaśaktiḥ. mukhyārthabādhādisahakāryapekṣārthapratibhāsanaśaktirlakṣanāśaktiḥ. tacchaktitrayopajanitārthāvagamamulajātatatpratibhāsapavitṛitapratipattṛpratibhāsahāyārthadyotanaśaktirdhvananavyāpāraḥ…” In Abhinava’s theory of śābdabodha, the padārthas are individually signified by Abhidhā, the vākyārtha is signified by Tātparya, the syntactic incongruity amongst the individual word-meanings is cognized and resolved by Lakṣaṇā and beyond these three, in the caturtha-kakṣā is Dhvani. Commenting on the “bhama dhammia” exemplum that opens the Dhvanyāloka’s discussion on Dhvani, Abhinava posits that there is no cognition of Anvayānupapatti in this verse and thus, there’s no scope for Lakṣaṇā. Abhinava specifically holds Lakṣaṇā to be in the tṛtīya-kakṣa, after Tātparya.

    Abhinava’s śābdabodha theory thus seems to be a variety of Abhihitānvayavāda. Following Abhinava, the Dhvani theorists theorize Tātparya as the vṛtti that brings about syntactic sentential cohesion (saṃsarga) of individual word-meanings. Viśvanātha in the Sāhityadarpaṇa states “abhidhāyā ekaikapadārthabodhanavirāmādvākyārtharūpasya padārthānvayasya bodhikā tātparyanāma vṛttiḥ. tadarthaśca tātparyārthaḥ.” The problematic here is that the Bhāṭṭas (who admit Lakṣaṇā even in sentences) theorize that in the Abhihitānvayavāda view of sentence meaning, the unified cognition of the vākyārtha is by Lakṣaṇā, and not Tātparya. Even Mukulabhaṭṭa in the Abhidhāvṛittamātṛaka discussing the Abhihitānvayavāda says that here, Lakṣaṇā is believed to occur after the Vācyatva is over: “anvaye’bhihitānāṃ sā vāycayatvād ūrdhvaṃ iṣyate.” It’s Jayantabhaṭṭa in the Nyāyamañjarī who posits Tātparya as signifying the vākyārtha (The Naiyāyikas do not admit Lakṣaṇā in sentences). Abhinava holds mukhyārthabādhā as constitutive of Lakṣaṇā, which he defines as Anvyānupapatti. If, however, Lakṣaṇā is defined as being constituted by Tātparyānupapatti instead of Anvayānupapatti (as done by Annambhaṭṭa in the Dīpikā autocommentary on the Tarkasaṃgraha or by Dharmarāja in the Vedāntaparibhāṣa or by Viśvanātha Nyāya Pañcānana in the Siddhānta Muktāvalī), then Abhinava’s four-fold Dhvani schema stands refuted. To complicate matters, upon defining Tātparyānupapatti as the Lakṣaṇābīja, Lakṣaṇā itself approaches Arthāpatti. Mukulabhaṭṭa in the Abhidhāvṛittamātṛaka holds (inter alia) Lakṣaṇā to be constituted by “paramārthato’bhideyabhāvasya anupadyamānatvāt,” which can include both Anvayānupapatti as well as Tātparyānupapatti and treats the stock Arthāpatti exemplum “Pīno Devadatto divā na bhuṅkte” as an illustration of Upādāna Lakṣaṇā as well and subsumes Dhvani under Lakṣaṇā (lakṣaṇāmargāvagāhitvaṃ tu dhvaneḥ sahṛdayair nūtanatayopavarṇitasya vidyata iti diśaṃ unmīlayituṃ idaṃ atra uktam.” It might be objected that there is “anterior incongruity” in Anvayānupapatti and “posterior incongruity” in Tātparyānupapatti, but I believe this objection is insignificant. Whether Anupapatti is anterior(syntactic) or posterior (semantic), it nevertheless ultimately leads to unintelligibility and the temporal semantic lag is linguistically not material. I’ll propose a structural analogy between Anvayānupapatti/Tātparyānupapatti in Lakṣaṇā and Abhidhānānupapatti/ Abhihitānupapatti in Arthāpatti.

    I’ll beg to submit a few thoughts on Vākyaśeṣa, for which I’ll invoke Bhoja’s Śṛṅgāraprakāśa: “sarvavākyānām vidhiniṣedha paryavasāyitvātsākṣādaśrutāvapi tadupakalpanam vākyaśeṣah. tadyathā-āryanivāso’yam deśa ityukte ihaiva sthātavyam madhyāhno vartate ityukte ihaiva bhoktavyam sa corāḥ panthāḥ ityukte na gantavyam, grāhāḥ sarityasyāmityukte na snātavyamiti vākyaśeṣo bhavati…” Bhoja discusses Vākyaśeṣa in the context of elliptical utterances and theorizes that the distinction between Adyāhāra and Vākyaśeṣa is that in Adyāhāra, there is “lexical Ākāṃkṣā” and in Vākyaśeṣa there is “semantic Ākāṃkṣā:” “kaḥ punaradhyāhārvākyaśeṣayorviśeṣaḥ? śabdākāṃkṣānivartako’dhyāhāraḥ, arthākāṃkṣānivartakovākyaśeṣaḥ, iti.” He further theorizes that the “missing” Artha is got by Śrutārthāpatti: “athaiṣa pratyakṣānumānopamānaśabdaiḥ anupalabhyamānaḥ kimpramāṇaviṣayassyāt, adhyāhāravat śrutārthāpattiprameya iti.”

    I’ll now invoke Mammaṭa’s prose commentary on Kāvyaprakāśa 5.47. Mammaṭa theorizes that unlike conventional meanings which are stable and monosemic, even plain, unembellished phrases like “gatostamarkaḥ” can be polysemically suggestive, depending upon the specific relevant context: “api ca vācyorthaḥ sarvān pratipattṛn prati ekarūpa eveti niyatosau. na hi ‘gatostamarkaḥ’ ityādau vācyorthaḥ kvacid anyathā bhavati. pratīyamānas tu tattatprakarṇavaktṛpratipatrādiviśeṣasahāyatayā nānātvaṃ bhajate. tathā ca ‘gatostamarkaḥ’ ityataḥ sapatnaṃ pratyavaskandanāvasara iti abhisaraṇamupakramyatāmiti praptaprāyaste preyāniti karmakaraṇānnivartāmaha iti sāndhyo vidhirupakramyatāmiti dūraṃ mā gā iti surabhayo gṛhaṃ praveṣyantāmiti santāpodhunā na bhavatīti vikreyavastūni saṃhṛiyantāmiti nāgatodyāpi preyānityādiranavadhīrvyaṅgyorthas tatra tatra pratibhāti.”

    It might be argued that Bhoja’s examples involve Semantic Expectancy, whereas Mammaṭa’s do not. I’ll beg to submit that the linguistic structure of the examples Bhoja adduces for Vākyaśeṣa (“āryanivāso’yam deśa,” “sa corāḥ panthāḥ” etc.) is identical to the example (“gatostamarkaḥ”) that Mammaṭa adduces for Dhvani, in that both are syntactically complete sentences, but incomplete locutions. I fail to see any plausible reason why the one is said to exhibit “Semantic Expectancy” whereas the other is not. The mechanics in both these processes involves imagining a relevant context (the Upapādaka) for a contextless indefinite utterance (the Anupapanna) through Śrutārthāpatti. The prima facie distinction is that Bhoja abducts a single Upapādaka for a single Anupapadyamāna, since he believes that all utterances are pragmatically exhortative, whether or not they are so grammatically, whereas Mammaṭa abducts multiple Upapādakas for a single Anupapadyamāna. The common factor thus to Lakṣaṇā, Arthāpatti, Tātparya, Vākyaśeṣa and Dhvani is the abductive Ākāṃkṣā-based mechanics of “imagining” an enunciatory context. Abductive reasoning involves reasoning “backward” from consequent to antecedent. Therefore, Peirce calls it also “retroductive reasoning.” From a logical point of view, reasoning backward is not a valid form of inference. It is conjectural or presumptive thinking, aiming at matching pragmatic standards of plausibility, guided by the reasoner’s “guessing instinct.” However, Peirce claims (“Harvard Lectures on Pragmatism”, Collected Papers 5.188-189) that abduction is logical inference because it can be represented in “a perfect definite logical form:”

    “The surprising fact, C, is observed;
    But if A were true, C would be a matter of course.
    Hence, there is reason to suspect that A is true.
    Thus, A cannot be abductively inferred, or if you prefer the expression, cannot be abductively conjectured until its entire content is already present in the premiss, “If A were true, C would be a matter of course.”

    Peirce theorizes that Abduction involves conjecturing several hypotheses for a particular “surprising fact” and then choosing/deciding the one hypotheses that best “fits” the particular fact. This is akin to Inference to the Best Explanation (IBE). Since all the possible explanations cannot be humanly considered, I’ll submit that IBE should perhaps be Inference to the Best Available Explanation, IBAE. Peirce, like Bhoja, speaks not of literary criticism but a “real-world” Pramāṇa and thus requires one of the several conjectured hypotheses to be chosen, based on certain constraints. I’ll submit that the mechanics of this process is identical to that of Arthāpatti as a Pramāṇa. For these constraining/delimiting factors, I’ll invoke Mahimabhaṭṭa in the first chapter of the Vyaktiviveka: “sādhyasādhanabhāvaścānyoravinābhāvāvasāyakṛto’vagantavyaḥ” states “sa ca pramāṇamūlaḥ. tacca trividhaṃ. yadāhuḥ:- loko vedastathādhyātmam pramāṇam trividhaṃ smṛtaṃ.” Mahima takes these “pramāṇas” from Nāṭyaśāstra 25.120:

    loko vedastathādhyātmam pramāṇam trividhaṃ smṛtaṃ
    vedādhyātmapadārtheṣu prāyo nāṭyam pratiṣṭhitam

    He defines Loka as “lokaprasiddhārthaviṣayolokaḥ” Veda as “śāstramātraprasiddhārthaviṣayo vedaḥ” and also that “vedagrahaṇam itihāsapurāṇadharma śāstradiupalakṣaṇaṃ teṣām tanmūlatvopagamāt” and Ādhyātma as “ādhyātmikārthaviṣayaṃ adhyātmam.” I’ll beg to submit that this applies to Arthāpatti as well. Thus, while it’s theoretically possible to indulge in a “free play” of imagination to conjecture several abductively conjured scenarios why fat Devadatta might not be eating in the day-time i.e., he’s a Yogī who has mastered the physical urges of the body; he’s granted a boon of perfect health without being required to eat; he’s actually a Deva/Gandharva/Vidyādhara/Nāga in human disguise etc., the conjecture that he eats at night is more plausible from the point of view of Loka (cf. Mukula’s formulation that Lakṣaṇā is Laukiki) or conventional knowledge. In other words, from amongst several possible competing hypotheses, the one that is the most plausible is chosen as the best available explanation. Arthāpatti as a hermeneutical-interpretative tool in Kāvya is also constrained, but to a lesser degree of strictness as compared to Arthāpatti as a Pramāṇa, in that there can be several interpretative scenarios wherein there are several “best available explanations” instead of only one. If “Sāstrārthāpatti” is constrained by Loka, Veda and Ādhyātma, so is Kavyārthāpatti by the very same factors and also by certain other specific pragmatic factors, delineated by Mammaṭa at Kāvyaprakāśa 3:21-22:

    vaktṛboddhavyakākūnāṃ vākyavācyāñyasannidheḥ
    prastāvadeśakālāder vaiśiṣṭyāt pratibhājuṣām
    yo’rthasyānyārthadhīhetur vyāpāro vyaktireva sā

    Is Dhvani, therefore Arthāpatti? Jayantabhaṭṭa in the Nyāyamañjarī refutes Arthāpatti as a separate Pramāṇa and includes it under Anumāna itself. In the section on the refutation of Dhvani, he does not separately refute Dhvani, observing cryptically that Dhvani too stands refuted by the refutation of Śrūtārthāpatti (I can’t help submit that Jayanta thereby indulges in the daṇḍāpūpikānyāya!). Jayaratha, in his Vimarśinī commentary on Ruyyaka’s Alaṃkārasarvasva cites two kārikās from an anonymous author which list twelve anti-dhvani theories, one of which is that dhvani is Arthāpatti:

    tātparyaśaktirabhidhā lakṣaṇānumitī dvidhā
    arthāpattiḥ kvacittantram samāsoktyādyalaṃkṛtiḥ
    rasasya kāryatā bhogaḥ vyāpārāntarabādhanam
    dvādaśettham dhvanerasya sthithā vipratipattayaḥ

    Abhinavagupta in the Locana commentary denies that dhvani involves Śrutārthāpatti: “‘pīnacaitro divā nātti’ ityatra abhidhaiva aparyavasiteti saiva svārthānirvāhāya arthāntaram śabdāntaram vākarṣatītyanumānasya śrutārthāpatterva tārkikamīmāṃsakayoḥ na dhvaniprasaṅgaḥ ityalam bahunā.” He posits that in instances of Athāpatti, the Abhidhāśakti’s incomplete and demands a further word/meaning for completion, unlike in instances of Dhvani. However, on Abhinava’s own theory of four-fold Śābdabodha, the Abhidhāśakti exhausts itself once the Padārthas have been signified. This is the same logic on the basis of which the examples adduced by Bhoja are deemed to exhibit “Semantic Expectancy,” but I find this to be merely an excuse to side-step the issue. Dhanika, while presenting the Dhvani as the Pūrvapakṣa in the Avaloka commentary on Dhanañjaya’s Daśarūpaka mentions that the Dhvanivādins deny that Dhvani is born through Arthāpatti, since there’s no Anupapati in instances of Dhvani: na cāsāvarthāpattijanyā anupapadyamānārthāpekṣābhāvat.” Here, Anupapatti seems to be defined as Abhidhānānupapatti. If Anupapatti is also defined in terms of Abhihitānupapatti, this objection ceases to apply. Abhinava himself in the Abhinavabhāratī commentary the end of his long commentary on the Rasasūtra (Nāṭyaśāstra 6.31c, which I’ve presented in my earlier response) seems to abandon his earlier four-fold theory and presents a view analogous to the mechanics of Arthāpatti. Here’s a Mahārāsṭrī Prākrit Gāthā from the anthology Gāhāsatasaī (Gāhāsatasaī 1.4, ascribed to a Voḍisa), traditionally believed to be collected by Hāla, a king of the Sātavāhana dynasty who ruled over Pratiṣṭhana (modern-day Paiṭhaṇ, in central Maharashtra) from 20-24 C.E.:

    ua ṇiccala nippaṃdā bhisiṇīpattammi rehaï balāā
    ṇimmalamargaäbhāäṇapariṭṭhiā saṃkhasutti vva

    (paśya niścalaniḥspandā bisinīpatre rājate balākā
    nirmalamarkatabhājana paristhitā shaṅkhaśuktiriva)

    Mammaṭa (vṛtti on Kāvyaprakāśa 2.8) adduces this Gāthā as an exemplum of the “suggestiveness of suggested meaning” and offers the following dhvani reading: “atra niṣpandatvena āśvastatvam. tena ca janarahitatvam. ataḥ saṃketasthānam etad iti kayācit kiṃcitpratyucyate. athavā mithyā vadasi na tvamatrāgatobhūriti vyajyate.” The 16th century commentator Gaṅgādhara Bhaṭṭa’s headnote to his Dhvani reading of this verse reads:“‘kallolinīkānanakandarādau duḥkhāśraye cārpitacittavṛttiḥ|mṛudukramārambhaṃ abhinnadhairyaḥ ślatho’pi dīrgham ramate rateṣu.’ ityādi kāmaśāstrāddirgharamanārtha nāyakasyānyacittatām kurvatī kācidāh…” The commentary that follows attributes this Gāthā as an utterance of the Nāyikā to the Nāyaka:“tathā yadi vegavidhāraṇaparo’si tadainām balākām paśyannanyamanaskatayā cira ramasveti bhāvaḥ.”

    Mammaṭa invents a nāyikābheda enunciatory context for this gāthā. His first nāyikā (i.e., the one who hints of an assignation-spot by the pond) is the Nāyikā classified in the Nāyikābheda typology as a Parakīyā (“another’s”), one who keps her love secretive and hidden (guptā), being either unmarried (kanyakā) or married to another man (paroḍhā). This is an extremely prevalent topos in Sanskrit kāvya, viz. the topos of the asati or the kulaṭā, the wanton. The nāyikā here may be further identified as a Vācyavidagdhā (clever of speech). The Nāyaka is a paramour (upapati) and the genre of love is “love in union”, or Sambhogaśṛṅgārarasa, the Sthāyibhāva (basal sentiment) of which is Rati (eros). This is an instance of furtivus amor, or Cauryarata. Mammaṭa’s second Nāyikā is one of the 8 Avasthānāyikās of the Nāyikābheda typology, viz. the Vipralabdhānāyikā, the Nāyikā who reaches the assignation-spot, but who’s “jilted” by her lover, who doesn’t turn up. This nayika is the “experienced” jilted Nāyikā, the Prauḍhāvipralabdhānāyikā. The Nāyaka here is deceiver paramour, a Śaṭhaupapati and the genre of love is “love frustrated”, or Vipralambhaśṛṅgārarasa, the Sthāyibhāva (basal sentiment) of which is also Rati (eros).

    The enunciatory context invented by Gaṅgādhara Bhaṭṭa also relies on the Nāyikābheda typology, and also upon the Kāmaśāstra. Gaṅgādhara’s brief is that every Gāthā in the Gāhāsatasaī suggests Śṛṅgārarasa. The verse (intertext!) that Gaṅgādhara quotes in his headnote to this Gāthā is verse 5.3 from Kokkoka’s Ratirahasya, which deals with the erotic technique of Vīryastambhana, “deferral of ejaculation”. The nāyikā here is a parakīyānāyikā, the nāyaka is an upapati and the genre is Sambhogaśṛṅgārarasa. This reading is particularly piquant in so far as it posits a context where the nāyikā and the nāyaka are engaged in lovemaking proper, without any lexeme of the text actually signifying lovemaking! The rhetorical device used is Svabhāvokti.

    • Thank you for this detailed analysis–I will take a more careful look at it this weekend. In scanning over it, I am happy to see your mention of Mukulabhaṭṭa, whose work I have translated and focused on in my PhD thesis. One thing to mention is that he says that lakṣaṇā comes before, after, and both before and after the vācya, depending upon what position you take (there is a hybrid or samuccaya view which he thinks can fuse the two theories). I think that his view is probably the hybrid one, though he doesn’t make that explicit.

      The other thing I’ll say now is that your observation about arthāpatti and lakṣaṇā in Mukula (the Fat Devadatta case) is part of why I think, in agreement with you, that lakṣaṇā functions through arthāpatti and that arthāpatti can be analyzed, at least in cases involving testimonial input (though I think perhaps in others as well), as IBE. However, others do not agree with me, for various reasons.

      Again, thanks for taking the time to lay all of this out. I have said I would write a post on arthāpatti/lakṣaṇā but have gotten behind. Your remarks may spur me to do so sooner rather than later, so thank you.

Leave a Reply

Your email address will not be published. Required fields are marked *