In plain words
One should not abandon the work born of one's nature, son of Kunti, even though it has defects; for every undertaking is covered by some defect, as fire is covered by smoke.
Word for word
One should not give up the duty born with one, O son of Kuntī, even with its fault; for all undertakings are covered by fault, as fire by smoke.
The commentators
Bhagavadgītābhāṣyam
स्वभावनियतं कर्म कुर्वाणो विषजः इव कृमिः किल्बिषं न आप्नोतीति उक्तम् ; परधर्मश्च भयावहः इति, अनात्मज्ञश्च ‘न हि कश्चित्क्षणमपि अकर्मकृत्तिष्ठति’(भ. गी. ३ । ५) इति । अतः —
सहजं सह जन्मनैव उत्पन्नम् । किं तत् ? कर्म कौन्तेय सदोषमपि त्रिगुणात्मकत्वात् न त्यजेत् । सर्वारम्भाः आरभ्यन्त इति आरम्भाः, सर्वकर्माणि इत्येतत् ; प्रकरणात् ये केचित् आरम्भाः स्वधर्माः परधर्माश्च, ते सर्वे हि यस्मात् — त्रिगुणात्मकत्वम् अत्र हेतुः — त्रिगुणात्मकत्वात् दोषेण धूमेन सहजेन अग्निरिव, आवृताः । सहजस्य कर्मणः स्वधर्माख्यस्य परित्यागेन परधर्मानुष्ठानेऽपि दोषात् नैव मुच्यते ; भयावहश्च परधर्मः । न च शक्यते अशेषतः त्यक्तुम् अज्ञेन कर्म यतः, तस्मात् न त्यजेत् इत्यर्थः ॥
किम् अशेषतः त्यक्तुम् अशक्यं कर्म इति न त्यजेत् ? किं वा सहजस्य कर्मणः त्यागे दोषो भवतीति ? किं च अतः ? यदि तावत् अशेषतः त्यक्तुम् अशक्यम् इति न त्याज्यं सहजं कर्म, एवं तर्हि अशेषतः त्यागे गुण एव स्यादिति सिद्धं भवति । सत्यम् एवम् ; अशेषतः त्याग एव न उपपद्यते इति चेत् , किं नित्यप्रचलितात्मकः पुरुषः, यथा साङ्ख्यानां गुणाः ? किं वा क्रियैव कारकम् , यथा बौद्धानां स्कन्धाः क्षणप्रध्वंसिनः ? उभयथापि कर्मणः अशेषतः त्यागः न सम्भवति । अथ तृतीयोऽपि पक्षः — यदा करोति तदा सक्रियं वस्तु । यदा न करोति, तदा निष्क्रियं तदेव । तत्र एवं सति शक्यं कर्म अशेषतः त्यक्तुम् । अयं तु अस्मिन् तृतीये पक्षे विशेषः — न नित्यप्रचलितं वस्तु, नापि क्रियैव कारकम् । किं तर्हि ? व्यवस्थिते द्रव्ये अविद्यमाना क्रिया उत्पद्यते, विद्यमाना च विनश्यति । शुद्धं तत् द्रव्यं शक्तिमत् अवतिष्ठते । इति एवम् आहुः काणादाः । तदेव च कारकम् इति । अस्मिन् पक्षे को दोषः इति । अयमेव तु दोषः — यतस्तु अभागवतं मतम् इदम् । कथं ज्ञायते ? यतः आह भगवान् ‘नासतो विद्यते भावः’(भ. गी. २ । १६) इत्यादि । काणादानां हि असतः भावः, सतश्च अभावः, इति इदं मतम् अभागवतम् । अभागवतमपि न्यायवच्चेत् को दोषः इति चेत् , उच्यते — दोषवत्तु इदम् , सर्वप्रमाणविरोधात् । कथम् ? यदि तावत् द्व्यणुकादि द्रव्यं प्राक् उत्पत्तेः अत्यन्तमेव असत् , उत्पन्नं च स्थितं कञ्चित् कालं पुनः अत्यन्तमेव असत्त्वम् आपद्यते, तथा च सति असदेव सत् जायते, सदेव असत्त्वम् आपद्यते, अभावः भावो भवति, भावश्च अभावो भवति ; तत्र अभावः जायमानः प्राक् उत्पत्तेः शशविषाणकल्पः समवाय्यसमवायिनिमित्ताख्यं कारणम् अपेक्ष्य जायते इति । न च एवम् अभावः उत्पद्यते, कारणं च अपेक्षते इति शक्यं वक्तुम् , असतां शशविषाणादीनाम् अदर्शनात् । भावात्मकाश्चेत् घटादयः उत्पद्यमानाः, किञ्चित् अभिव्यक्तिमात्रे कारणम् अपेक्ष्य उत्पद्यन्ते इति शक्यं प्रतिपत्तुम् । किञ्च, असतश्च सतश्च सद्भावे असद्भावे न क्वचित् प्रमाणप्रमेयव्यवहारेषु विश्वासः कस्यचित् स्यात् , ‘सत् सदेव असत् असदेव’ इति निश्चयानुपपत्तेः ॥
किञ्च, उत्पद्यते इति द्व्यणुकादेः द्रव्यस्य स्वकारणसत्तासम्बन्धम् आहुः । प्राक् उत्पत्तेश्च असत् , पश्चात् कारणव्यापारम् अपेक्ष्य स्वकारणैः परमाणुभिः सत्तया च समवायलक्षणेन सम्बन्धेन सम्बध्यते । सम्बद्धं सत् कारणसमवेतं सत् भवति । तत्र वक्तव्यं कथम् असतः स्वं कारणं भवेत् सम्बन्धो वा केनचित् स्यात् ? न हि वन्ध्यापुत्रस्य स्वं कारणं सम्बन्धो वा केनचित् प्रमाणतः कल्पयितुं शक्यते ॥
ननु नैवं वैशेषिकैः अभावस्य सम्बन्धः कल्प्यते । द्व्यणुकादीनां हि द्रव्याणां स्वकारणसमवायलक्षणः सम्बन्धः सतामेव उच्यते इति । न ; सम्बन्धात् प्राक् सत्त्वानभ्युपगमात् । न हि वैशेषिकैः कुलालदण्डचक्रादिव्यापारात् प्राक् घटादीनाम् अस्तित्वम् इष्यते । न च मृद एव घटाद्याकारप्राप्तिम् इच्छन्ति । ततश्च असत एव सम्बन्धः पारिशेष्यात् इष्टो भवति ॥
ननु असतोऽपि समवायलक्षणः सम्बन्धः न विरुद्धः । न ; वन्ध्यापुत्रादीनाम् अदर्शनात् । घटादेरेव प्रागभावस्य स्वकारणसम्बन्धो भवति न वन्ध्यापुत्रादेः, अभावस्य तुल्यत्वेऽपि इति विशेषः अभावस्य वक्तव्यः । एकस्य अभावः, द्वयोः अभावः, सर्वस्य अभावः, प्रागभावः, प्रध्वंसाभावः, इतरेतराभावः, अत्यन्ताभावः इति लक्षणतो न केनचित् विशेषो दर्शयितुं शक्यः । असति च विशेषे घटस्य प्रागभावः एव कुलालादिभिः घटभावम् आपद्यते सम्बध्यते च भावेन कपालाख्येन, सम्बद्धश्च सर्वव्यवहारयोग्यश्च भवति, न तु घटस्यैव प्रध्वंसाभावः अभावत्वे सत्यपि, इति प्रध्वंसाद्यभावानां न क्वचित् व्यवहारयोग्यत्वम् , प्रागभावस्यैव द्व्यणुकादिद्रव्याख्यस्य उत्पत्त्यादिव्यवहारार्हत्वम् इत्येतत् असमञ्जसम् ; अभावत्वाविशेषात् अत्यन्तप्रध्वंसाभावयोरिव ॥
ननु नैव अस्माभिः प्रागभावस्य भावापत्तिः उच्यते । भावस्यैव तर्हि भावापत्तिः ; यथा घटस्य घटापत्तिः, पटस्य वा पटापत्तिः । एतदपि अभावस्य भावापत्तिवदेव प्रमाणविरुद्धम् । साङ्ख्यस्यापि यः परिणामपक्षः सोऽपि अपूर्वधर्मोत्पत्तिविनाशाङ्गीकरणात् वैशेषिकपक्षात् न विशिष्यते । अभिव्यक्तितिरोभावाङ्गीकरणेऽपि अभिव्यक्तितिरोभावयोः विद्यमानत्वाविद्यमानत्वनिरूपणे पूर्ववदेव प्रमाणविरोधः । एतेन कारणस्यैव संस्थानम् उत्पत्त्यादि इत्येतदपि प्रत्युक्तम् ॥
पारिशेष्यात् सत् एकमेव वस्तु अविद्यया उत्पत्तिविनाशादिधर्मैः अनेकधा नटवत् विकल्प्यते इति । इदं भागवतं मतम् उक्तम् ‘नासतो विद्यते भावः’(भ. गी. २ । १६) इत्यस्मिन् श्लोके, सत्प्रत्ययस्य अव्यभिचारात् , व्यभिचाराच्च इतरेषामिति ॥
कथं तर्हि आत्मनः अविक्रियत्वे अशेषतः कर्मणः त्यागः न उपपद्यते इति ? यदि वस्तुभूताः गुणाः, यदि वा अविद्याकल्पिताः, तद्धर्मः कर्म, तदा आत्मनि अविद्याध्यारोपितमेव इति अविद्वान् ‘न हि कश्चित् क्षणमपि अशेषतः त्यक्तुं शक्नोति’(भ. गी. ३ । ५) इति उक्तम् । विद्वांस्तु पुनः विद्यया अविद्यायां निवृत्तायां शक्नोत्येव अशेषतः कर्म परित्यक्तुम् , अविद्याध्यारोपितस्य शेषानुपपत्तेः । न हि तैमिरिकदृष्ट्या अध्यारोपितस्य द्विचन्द्रादेः तिमिरापगमेऽपि शेषः अवतिष्ठते । एवं च सति इदं वचनम् उपपन्नम् ‘सर्वकर्माणि मनसा’(भ. गी. ५ । १३) इत्यादि, ‘स्वे स्वे कर्मण्यभिरतः संसिद्धिं लभते नरः’(भ. गी. १८ । ४५) ‘स्वकर्मणा तमभ्यर्च्य सिद्धिं विन्दति मानवः’(भ. गी. १८ । ४६) इति च ॥ ४८ ॥
In plain words
It has already been said: a person doing the work his own nature has fixed on him takes on no stain, like a worm born in poison. And another's dharma brings danger. And of the person who does not know the Self it was said, "no one stays, even for a moment, doing nothing." So then —
'Sahaja' means: born together with birth itself. What is 'that'? One's own action, Kaunteya — even though it is flawed, because everything is woven from the three guṇas, one should not abandon it. 'All undertakings' means whatever is set in motion — every action there is; and, given the context, that includes both one's own duty and someone else's duty. All of them, without exception, are covered by flaw, for this simple reason: everything is made of the three guṇas — the way fire is always wrapped in its own smoke. Giving up your inborn duty, your svadharma, and taking up someone else's duty instead does not free you from that flaw at all — and someone else's duty carries its own danger besides. And since no one who still lacks true knowledge can give up action altogether, that is exactly why one should never abandon it.
Is action off-limits to give up entirely, and is that the real reason not to give it up? Or is there some actual fault in giving up your inborn duty? And either way, what does that get us? Suppose it really is impossible to give up entirely — then giving it up completely would actually be a good thing, if only it could be done. True enough. But suppose someone says total renunciation simply cannot happen at all. Is that because the acting self is ceaselessly in motion by nature, the way the Sāṅkhyas say the guṇas are? Or because action itself is the only real agent, the way the Buddhists say momentary bundles flash in and out of being? Either way, total abandonment of action turns out to be impossible. There is a third view, though: when something acts, it's active; when it stops acting, that same thing is simply inactive. On this view, action really could be given up completely. Its special twist is this: the underlying thing is neither ceaselessly active nor is action itself the agent. So what is it, then? An activity that did not exist arises in an already-existing substance, and having existed, it dies away — while the pure substance itself, carrying its own latent power, just stays exactly as it is. That is the Vaiśeṣika position, and on their view that same substance is what really does the acting. Is there anything wrong with this? Yes — and here is exactly what's wrong: it flatly contradicts what the Lord himself teaches. How do we know? Because Krishna says, 'Of the unreal there is no being' (Bhagavad Gita 2.16), and so on — and the Vaiśeṣikas are saying precisely the opposite: that the unreal comes into being and the real ceases to be. If someone says, 'fine, it contradicts scripture, but so what if it's still logically sound?' — the answer is that it isn't sound at all: it contradicts every means of knowing anything. Watch what happens: if something like a compound particle is utterly non-existent before it's produced, and then, after existing a while, becomes utterly non-existent again when destroyed, then the unreal has literally become real, and the real has literally become unreal. But nothing unreal is ever actually seen coming into being, cause or no cause — a hare's horn never does, no matter what you supply it with. If, instead, things like pots are already real in substance, and a cause is needed only to bring them into visibility, that makes perfect sense. And there's more: if the unreal could become real and the real unreal, nobody could trust any act of knowing anything, ever — because the one certainty everything else depends on, that the real is real and the unreal is unreal, would itself collapse.
There is more: the Vaiśeṣikas say 'it arises' really just means a substance like a compound particle enters a relation with the existence of its own cause. Before it originates it is non-existent; afterward, once the cause gets to work, it becomes joined — by that special relation called inherence — to its own material causes, the atoms, and to existence itself. Once joined this way, the connected thing becomes 'existent, inherent in its cause.' But here is the question that has to be asked: how could something non-existent have a cause of its own at all? How could it be connected to anything? It could not — because no sound reasoning can ever make sense of a cause, or a connection to anything, belonging to the son of a barren woman.
But someone might object: 'That is not actually what the Vaiśeṣikas claim — they never say non-existence itself has a connection. For dyads and similar substances, they say, the relation of inherence in one's own cause applies only to existent things.' No, that objection fails — because the Vaiśeṣikas themselves refuse to grant that the substance exists before that connection forms. They flatly deny that a pot exists before the potter, the stick, and the wheel finish their work. And they do not think it is simply the clay, already there, taking on the shape of a pot. So, by elimination, there is nowhere else for that connection to land: on their own account, it has to belong to the non-existent thing after all.
But someone might say: even for something non-existent, a connection through inherence is not really contradictory. No — because no one ever observes such a connection for the son of a barren woman or anything like him. 'Only the prior non-existence of a pot and things like it gets connected to its own cause — not the son of a barren woman's non-existence,' the objector insists — even though both are, equally, simply non-existence. So the burden falls on them to name what actually distinguishes one non-existence from the other. Go through every recognized kind: the non-existence of one thing, of two, of everything, prior non-existence, destructive non-existence, mutual non-existence, absolute non-existence — under none of these headings can anyone show a real difference. And if there is no difference, then the pot's prior non-existence itself would have to acquire pot-hood through the potter's work, get connected to an actual potsherd, and become fit for every ordinary transaction — while that very same pot's destructive non-existence, equally non-existence, would not. That would mean none of the other non-existences are ever fit for practical use, while only prior non-existence — once it becomes a named substance like a dyad — gets to be fit for 'origination' and everything that follows from it. That makes no sense at all, because there is no difference whatsoever in being non-existence — exactly as there is no difference between absolute non-existence and destructive non-existence.
'But we never said prior non-existence turns into existence,' the Vaiśeṣika protests. Fine — then it must be the existent itself that turns into existent, the way a pot becomes a pot, or a cloth becomes a cloth. But that is just as absurd as the first claim: nothing needs to become what it already is. Even the Sāṅkhya's theory of transformation does not escape this, because it too accepts brand-new qualities coming into being and going out of it — no different, on this point, from the Vaiśeṣika. And even when the Sāṅkhya softens this to 'manifestation and disappearance' instead of outright coming-to-be, the same problem simply resurfaces: did that manifestation and disappearance themselves exist beforehand or not? Exactly the same contradiction returns. And by this very reasoning, the further idea that 'origination' and the rest are nothing but the cause rearranging itself has already been refuted too.
Rule out everything else, and this is what is left standing: Being, one Reality and no other. Ignorance then imagines it in many ways, fitting it out with birth, death, and the rest — the way one actor walks on as a hundred different characters. This is the Lord's own position, already stated in the verse, "Of the unreal there is no being." The sense of *is* never once fails us. Everything else we take to be there does.
So how, then, if the Self does not act at all, can it still be true that total renunciation of action is impossible? Whether the guṇas are real in themselves or simply imagined by ignorance, action belongs to them, not to the Self — it only ever appears in the Self because ignorance has projected it there. That is exactly why the ignorant person cannot renounce action, as it is said: 'No one is able to give up action entirely, not even for a moment' (Bhagavad Gita 3.5). But the knower is different: once knowledge has dissolved ignorance, he really can give up action completely, because whatever was only ever superimposed by ignorance simply cannot leave a trace behind. It is just like the eye-disease that makes someone see two moons — once that disease clears up, not a trace of that second moon remains. And once you see this, these words of scripture finally make sense: 'having renounced all actions in the mind' (Bhagavad Gita 5.13), 'a person devoted to his own duty finds fulfillment' (Bhagavad Gita 18.45), and 'by worshipping Him through his own work, a human being finds fulfillment' (Bhagavad Gita 18.46).
Close to the Sanskrit
It has been said that one performing action determined by his own nature does not incur stain, like a worm born in poison; and that another's dharma brings danger; and of one who does not know the Self, that "no one indeed remains even for a moment as a non-doer of action" (Bhagavadgītā 3.5). Therefore —
'Sahaja' means arisen together with birth itself. What is that? Action, O son of Kuntī, though flawed, because of being constituted by the three guṇas, one should not abandon. 'Sarvārambhāḥ': things that are undertaken are 'ārambhas', that is, all actions — this is the meaning; from the context, whichever undertakings there are — one's own duties and others' duties — all of those, since — being constituted of the three guṇas is here the cause — because of being constituted of the three guṇas, are covered by flaw, as fire by its own innate smoke. By abandoning the innate duty called svadharma, even in the performance of another's duty, one is not at all freed from flaw; and another's duty is also fear-inducing. And since it is not possible for an ignorant person to abandon action entirely, therefore one should not abandon it — this is the meaning.
Is it because action is impossible to abandon entirely that one should not abandon it? Or is it that in the abandonment of innate action a fault occurs? And what follows from this either way? If, in the first place, it is because total abandonment is impossible that innate action should not be abandoned, then it follows that total abandonment would actually be a merit — true, so it is. If, on the other hand, one says total abandonment itself is simply not tenable, is it that the person is by nature eternally active, as the guṇas are for the Sāṅkhyas? Or is activity itself the agent, as the momentarily-perishing aggregates are for the Buddhists? Under either account, the total abandonment of action is not possible. Now there is also a third position — when a thing acts, then the entity is active; when it does not act, then that same entity is inactive. This being so, it is possible to abandon action entirely. But here is the distinguishing feature of this third position: the entity is not eternally in motion, nor is activity alone the agent. What then? In a substance that already exists independently, an activity that did not exist arises, and, having existed, perishes; the pure substance itself, possessing capacity, simply remains. Thus the followers of Kaṇāda declare; and that very substance, they say, is the agent. In this position, what fault is there? This indeed is the fault: this view is contrary to the Lord's own teaching. How is this known? Because the Lord says, 'Of the non-existent there is no coming-into-being' (Bhagavad Gītā 2.16) and so on. For the followers of Kaṇāda hold that there is coming-into-being of the non-existent and non-existence of the existent — thus this view is contrary to the Lord's teaching. If one asks, 'even if contrary to the Lord's teaching, if it accords with reason, what fault remains?' — the reply is: this view does indeed have a fault, because it contradicts every means of valid knowledge. How so? If a substance such as a dyad is, prior to its origination, absolutely non-existent, and, having arisen and persisted for some time, again becomes absolutely non-existent upon destruction, then the non-existent itself becomes existent, and the existent itself becomes non-existent — non-existence becomes existence, and existence becomes non-existence. There, non-existence, said to arise before origination like a hare's horn, is said to arise depending on the inherent, non-inherent, and efficient cause. But it cannot rightly be said that non-existence thus arises and depends on a cause, because non-existent things like a hare's horn are never observed to arise. If, instead, pots and the like are already existent by nature when they 'arise', it can rightly be understood that they depend on a cause only for their manifestation. Moreover, if there could be existence of the non-existent and non-existence of the existent, then no one could have any confidence at all in the transactions of valid knowledge and its objects, because the very certainty that 'the existent is existent, the non-existent is non-existent' would become untenable.
Moreover, the Vaiśeṣikas declare that 'it arises' means, for a substance such as a dyad, the connection with the existence of its own cause. And prior to origination it is non-existent; afterward, depending on the operation of the cause, it is connected, by the relation characterized as inherence, with its own causes, the atoms, and with existence. The connected substance, being thus joined, becomes existent, inherent in its cause. There it must be asked: how could the non-existent have its own cause, or how could a connection with anything exist for it? For indeed, it is not possible, by any valid means of knowledge, to conceive of a cause of its own, or a connection with anything, for the son of a barren woman.
Objection: It is not thus that a connection is posited by the Vaiśeṣikas for non-existence. For the connection characterized by inherence in its own cause, for substances such as dyads, is said to hold only for existent things — thus the objector claims. Reply: No; because the Vaiśeṣikas themselves do not accept the substance's existence prior to the connection. For indeed, the existence of pots and the like, prior to the operation of the potter, the stick, the wheel, and so forth, is not accepted by the Vaiśeṣikas. And they do not hold that it is the clay itself that attains the form of a pot and the like. And therefore, by process of elimination, the connection is accepted by them to belong precisely to the non-existent.
Objection: Even for the non-existent, a connection characterized by inherence is not contradictory. Reply: No; because of the non-observation of any such connection for the son of a barren woman and the like. Objection continues: Only of the prior non-existence of a pot and the like does connection with its own cause occur, not of that of the son of a barren woman and the like — even though both are equally non-existence — so a distinguishing feature of non-existence must be stated. 'Non-existence of one', 'non-existence of two', 'non-existence of all', 'prior non-existence', 'destructive non-existence', 'mutual non-existence', 'absolute non-existence' — by these defining characteristics, no distinction can be shown by anyone. And, there being no distinction, the prior non-existence of the pot itself would come to acquire pot-hood by means of the potter and the rest, and would be connected with an existent thing called a potsherd, and, being connected, would become fit for all practical transactions — but not so the destructive non-existence of that very pot, even though it too is equally non-existence — so that the non-existences beginning with destructive non-existence have, in no case, fitness for practical transaction, while only prior non-existence, called a substance such as a dyad, has fitness for the transaction of 'origination' and the like — this is inconsistent; because there is no difference in respect of being non-existence, just as there is none between absolute non-existence and destructive non-existence.
Objection: We do not at all say that prior non-existence turns into existence. Reply: Then it is existence itself that turns into existence — as, for example, a pot's becoming a pot, or a cloth's becoming a cloth. This too, just like non-existence turning into existence, is contrary to valid means of knowledge. Even the Sāṅkhya's position of transformation — that too, because it accepts the origination and destruction of unprecedented properties, is not different from the Vaiśeṣika position. Even in accepting manifestation and disappearance, in determining the existence or non-existence of that manifestation and disappearance themselves, there is a contradiction with valid means of knowledge, exactly as before. By this same reasoning, the view that origination and so on are merely the configuration of the cause itself — this too is refuted.
By elimination, Being — one Reality alone — is, through ignorance, imagined in many ways, like an actor, by means of properties such as origination and destruction. This is the Lord's doctrine, stated in this verse, 'Of the non-existent there is no being' (Gītā 2.16) — because the cognition of Being does not fail, and because of the failing of the others.
How then, given that the Self is actionless, is it that the total abandonment of action is not tenable? Whether the guṇas are real entities, or whether they are imagined by ignorance, action is a property belonging to them; then it appears in the Self only as superimposed by ignorance — for this reason the ignorant person, as it is said, 'For indeed no one is able to abandon action entirely, even for an instant' (Bhagavad Gītā 3.5), cannot renounce it. The knower, however, once ignorance has ceased through knowledge, is indeed able to abandon action entirely, because there is no possibility of any residue for what was superimposed by ignorance. For indeed, of the double moon and the like, superimposed by a vision afflicted with the eye-disease timira, no residue remains even when the eye-disease departs. And this being so, this statement becomes coherent: 'having renounced all actions mentally' (Bhagavad Gītā 5.13) and so on, 'a man devoted to his own respective duty attains success' (Bhagavad Gītā 18.45), and 'having worshipped Him by his own action, a person finds success' (Bhagavad Gītā 18.46).